Chapter IV

Mixers

Frequency Conversion, Image Rejection, Noise Figure, Nonlinear Models, and Diode & Transistor Mixer Circuits

Outline — click chapter to expand
4.1 Mixer Characteristics
4.2 Mixer Models & Circuits
Practice Problems
Appendix — Interactive

4.1 Mixer Characteristics

Section Overview

What a Mixer Is
A mixer is a three-port device built around a nonlinear or time-varying element that performs frequency conversion. Practical RF/microwave mixers exploit the nonlinearity of a diode or transistor. The output spectrum contains the sum and difference of the two input frequencies and their harmonics, so filtering is required to select the desired component. Credit for the invention is generally given to Armstrong, who first used vacuum-tube mixers to shift a received signal to an intermediate frequency.
電路符號 — 圓圈內一個 ×(⊗)The circuit symbol — a × inside a circle (⊗)回路記号 — 円の中に × を一つ(⊗)
RF IF LO
通用混波器符號The general mixer symbol汎用ミキサ記号

圓圈內一個乘號 $\otimes$(×)代表A multiplication sign $\otimes$ (×) inside a circle denotes an 円の中の乗算記号 $\otimes$(×)は理想乘法器 (multiplier)ideal multiplier理想乗算器:$v_{IF}\propto v_{RF}\times v_{LO}$。三個埠 = 兩輸入(RF、LO)+ 一輸出(IF)。: $v_{IF}\propto v_{RF}\times v_{LO}$. Three ports = two inputs (RF, LO) plus one output (IF).:$v_{IF}\propto v_{RF}\times v_{LO}$。三つのポート = 二つの入力(RF、LO)+ 一つの出力(IF)。

此符號This symbol この記号は不指定內部拓樸says nothing about the internal topology内部トポロジーを指定しません——無論裡面是 single-ended、single-balanced、double-balanced (ring)、二極體或電晶體 mixer,在系統方塊圖上都畫成這個符號。要分辨是哪一種,得看內部電路(二極體數量、是否平衡、有無 hybrid),而不是這個方塊。— whether the mixer inside is single-ended, single-balanced, double-balanced (a ring), diode or transistor, it is drawn as this same symbol on a system block diagram. To tell which it is you must look at the internal circuit (the number of diodes, whether it is balanced, whether there is a hybrid), not at this block.——内部がシングルエンド・単平衡・二重平衡(リング)であっても、ダイオードでもトランジスタのミキサでも、システムブロック図では同じこの記号で描かれます。どれであるかを見分けるには、このブロックではなく内部回路(ダイオードの数、平衡かどうか、ハイブリッドの有無)を見る必要があります。

⚠ 別跟⚠ Do not confuse it with an ⚠ これを加法器/合成器 (combiner)adder / combiner加算器/合成器 搞混:圓圈內是 $+$ 號才是相加,$\times$ 才是混波相乘。 : a $+$ inside the circle means addition, whereas $\times$ means mixing (multiplication). と混同しないでください:円の中が $+$ なら加算、$\times$ なら混合(乗算)です。

Port Convention — Up vs. Down Conversion

Mixer TypeInput PortsOutput PortFrequency Relation
Up-converting IF (Intermediate Frequency)  +  LO (Local Oscillator) RF $f_{RF} = f_{LO} \pm f_{IF}$
Down-converting RF  +  LO (Local Oscillator) IF $\color{red}{\boldsymbol{f_{IF} = |f_{RF} - f_{LO}|}}$

Important Mixer Characteristics

CharacteristicWhat It MeasuresWhere Discussed
Conversion gain / lossPower ratio from input port to converted output port§4.1.3
Noise figure (SSB / DSB)Added noise; SSB vs. DSB depends on signal location relative to LO§4.1.4
Port-to-port isolationLO leakage to RF/IF, RF leakage to IF, etc.§4.1.5
Port VSWR & DC offsetMatch quality and self-mixing offset at the IF§4.1.5
Linearity — IP2, IP3Intermodulation distortion; second- and third-order intercept points§4.1.6
Image frequency rejectionSpurious response at $f_{LO} \mp f_{IF}$ that aliases onto the desired IF§4.1.2
Reading Roadmap
§4.1 defines the mixer figures of merit (conversion gain, NF, isolation, linearity, image frequency). §4.2 builds up small-signal linear → small-signal nonlinear → large-signal / switching models, then walks through the diode mixer family (single-ended → single-balanced → double-balanced → image-reject) and transistor mixers.

4.1.1 Three-Port Overview

A mixer is a three-port device that uses a nonlinear or time-varying element to achieve frequency conversion. Practical mixers rely on the nonlinearity of a diode or transistor.

  • Up-converting mixer: inputs are IF and LO; output is RF.
  • Down-converting mixer: inputs are RF and LO; output is IF.

The output contains a spectrum of frequencies at the sum and difference of all harmonics of the two inputs — filtering selects the desired component.

Key Mixer Characteristics
Conversion gain/loss · Noise figure · Port-to-port isolation · Port VSWR · DC offset · Linearity (IP2, IP3)

4.1.2 Frequency Conversion

(a) Up-converting mixer Local oscillator fLO Mixer fRF = fLO ± fIF IF oscillator fIF f 0 fIF fLOfIF fLO fLO+fIF (b) Down-converting mixer RF oscillator fRF Mixer fIF = fRF ± fLO Local oscillator fLO f fRFfLO fLO fRF fRF+fLO
Block diagrams and output spectra: (a) up-converting mixer with IF and LO inputs producing $f_{LO} \pm f_{IF}$; (b) down-converting mixer with RF and LO inputs producing $f_{RF} \pm f_{LO}$.

Up-Converting Mixer

The output is the product of LO and IF signals:

Up-converting output $$v_{RF}(t) = K\cos(2\pi f_{LO}t)\cos(2\pi f_{IF}t) = \frac{K}{2}\bigl[\cos 2\pi(f_{LO}+f_{IF})t + \cos 2\pi(f_{LO}-f_{IF})t\bigr]$$
  • $f_{LO}+f_{IF}$: upper sideband (USB)
  • $f_{LO}-f_{IF}$: lower sideband (LSB)
  • DSB signal contains both; SSB retains only one (produced by filtering or image-reject mixer)

Down-Converting Mixer

Down-converting — desired IF $${\color{red}{\boldsymbol{f_{IF} = |f_{RF} - f_{LO}|}}}$$

Image Frequency

For a receiver with $f_{RF} = f_{LO} + f_{IF}$, the image frequency is:

Image frequency $$f_{IM} = f_{LO} - f_{IF} \quad\Leftrightarrow\quad f_{RF} - f_{IM} = 2f_{IF}$$

A signal arriving at $f_{IM}$ down-converts to the same IF as the desired signal — it is indistinguishable at the IF stage. An image-reject filter (passband at $f_{RF}$, stopband at $f_{IM}$) is placed before the mixer to suppress this response.

f |V| fIM fRF fLO fIF fIF Desired: fRF = fLO + fIF Image: fIM = fLOfIF Separation = 2fIF
超外差架構 (Superheterodyne Architecture)Superheterodyne architectureスーパーヘテロダイン構成

定義:Definition:定義:超外差 (super-heterodyne) 是一種接收機架構,將天線接收的射頻 (RF) 訊號透過 the superheterodyne is a receiver architecture in which the radio-frequency (RF) signal from the antenna is multiplied by a local oscillator (LO) in a スーパーヘテロダインは、アンテナで受信した高周波(RF)信号を混頻器mixerミキサ與本振 (LO) 相乘,下變頻 (down-convert) 到一個 and downconverted to a で局部発振器(LO)と乗算し、ダウンコンバージョンして固定的中頻 (IF)fixed intermediate frequency (IF)固定の中間周波数(IF),之後在這個固定中頻處進行高選擇性濾波、放大、與解調。由 Edwin Armstrong 於 1918 年提出。, where highly selective filtering, amplification and demodulation are then performed. Proposed by Edwin Armstrong in 1918.に変換する受信機アーキテクチャです。その固定中間周波数で高選択度のフィルタリング・増幅・復調を行います。1918 年に Edwin Armstrong が提案しました。

核心概念:The core idea:核心となる考え方:

  • 每個輸入頻段都被搬移到every input band is translated to どの入力帯域も同一個中頻 $f_{IF}$the same intermediate frequency $f_{IF}$同じ中間周波数 $f_{IF}$。要選不同的電台或頻道,只需要調整 LO 頻率,IF 段的硬體完全不變。. To select a different station or channel you need only retune the LO; the IF hardware is entirely unchanged.へ移されます。異なる放送局やチャネルを選ぶには LO 周波数を調整するだけでよく、IF 段のハードウェアはまったく変わりません。
  • 窄頻、高 Q 的選擇性濾波器只在固定的 IF 處設計一次,比直接在不斷變動的 RF 上做選擇性容易非常多。A narrowband, high-Q selectivity filter is designed once at a fixed IF, which is vastly easier than achieving that selectivity directly at a continually changing RF.狭帯域・高 Q の選択フィルタは固定の IF で一度設計するだけでよく、絶えず変動する RF で直接その選択度を実現するよりはるかに容易です。
  • 主要增益由 IF 放大器提供 → RF 段不需要高增益、節省功耗。Most of the gain is provided by the IF amplifier → the RF section needs no high gain, saving power.利得の大部分は IF 増幅器が受け持ちます → RF 段に高利得は不要となり、消費電力を節約できます。

典型訊號鏈:A typical signal chain:典型的な信号チェーン:天線 → RF 預選/影像抑制濾波器 → LNA → 混頻器 (與 LO 相乘) → IF 濾波器 → IF 放大器 → 解調器 (Demodulator)。 antenna → RF preselect / image-reject filter → LNA → mixer (multiplying with the LO) → IF filter → IF amplifier → demodulator.アンテナ → RF プリセレクト/イメージ除去フィルタ → LNA → ミキサ(LO と乗算) → IF フィルタ → IF 増幅器 → 復調器。

主要缺點 — 影像頻率 (image frequency):The main drawback — the image frequency:主な欠点 — イメージ周波数:對於 $f_{RF} = f_{LO} + f_{IF}$ 的設計,另一個位於 $f_{IM} = f_{LO} - f_{IF}$ 的訊號也會被混頻到相同的 IF,在 IF 端 for a design with $f_{RF} = f_{LO} + f_{IF}$, a signal at $f_{IM} = f_{LO} - f_{IF}$ is also mixed to the same IF and is $f_{RF} = f_{LO} + f_{IF}$ の設計では、$f_{IM} = f_{LO} - f_{IF}$ にある別の信号も同じ IF に変換され、IF 側では無法區分indistinguishable区別がつきません。因此 there. Hence the 。したがって必須在混頻器前先以 RF 預選濾波器 (image-reject filter) 抑制鏡像頻段image band must be suppressed by an RF preselect (image-reject) filter ahead of the mixerミキサの前段で RF プリセレクト(イメージ除去)フィルタによりイメージ帯域を抑圧する必要があります;這也是為什麼 IF 不能選得太低 (兩者間隔 = $2f_{IF}$,IF 越低代表預選濾波器要越窄越陡)。; and this is also why the IF cannot be chosen too low (the two are separated by $2f_{IF}$, so a lower IF demands a narrower, steeper preselect filter).。これは IF を低く選びすぎられない理由でもあります(両者の間隔は $2f_{IF}$ であり、IF が低いほどプリセレクトフィルタは狭く急峻でなければなりません)。

Antenna RF Preselect / Image-Reject LNA LO IF Filter (narrow, fixed) IF Amp Demodulator baseband fRF fRF fIF fIF

English version.

Definition. The superheterodyne architecture is a receiver topology in which the incoming RF signal is multiplied (mixed) with a local-oscillator (LO) tone and down-converted to a fixed intermediate frequency (IF), where all of the channel filtering, amplification, and demodulation take place. Invented by Edwin Armstrong in 1918, it remains the dominant architecture for high-selectivity radios.

Why it wins.

  • Every received channel is translated to the same IF — to retune, you change only $f_{LO}$, while every component after the mixer stays put.
  • The narrow, high-$Q$, high-selectivity filter only has to be designed once, at a single fixed frequency. Doing the same selectivity at a sliding RF would be impractical.
  • Most of the receiver gain is at IF, so the RF front end can be low-gain and low-power.

Signal chain. Antenna → RF preselect / image-reject filter → LNA → mixer (× LO) → IF filter → IF amplifier → demodulator → baseband.

The catch — the image. If the desired channel sits at $f_{RF} = f_{LO} + f_{IF}$, then a signal at $f_{IM} = f_{LO} - f_{IF}$ will mix down to the same IF and become indistinguishable from the wanted signal. A preselect / image-reject filter must therefore precede the mixer. This is also why $f_{IF}$ can't be too low: the image sits only $2f_{IF}$ away from the wanted channel, so a lower IF demands a steeper, narrower RF preselector.

Lecture Example — IS-54 Image Frequency

Problem: IS-54 digital cellular receive band: 869–894 MHz, first IF = 87 MHz. Find the two possible LO ranges. Using the upper LO range, find the image frequency range. Does it fall in the receive passband? Why is the upper LO range preferred?

Step 1 — LO frequency ranges ($f_{LO} = f_{RF} \pm f_{IF}$): $$f_{LO,\text{upper}} = (869\text{–}894) + 87 = 956\text{–}981 \text{ MHz}$$ $$f_{LO,\text{lower}} = (869\text{–}894) - 87 = 782\text{–}807 \text{ MHz}$$
Step 2 — Image frequency (upper LO): $$f_{IM} = f_{LO} - f_{IF} = (956\text{–}981) + 87 = 1043\text{–}1068 \text{ MHz}$$ This is outside the 869–894 MHz receive band. ✓
Step 3 — Tuning ratio comparison:
$$\text{Upper LO:}\quad \frac{2(981-956)}{981+956} = \frac{50}{1937} \approx 0.0258$$ $$\text{Lower LO:}\quad \frac{2(807-782)}{807+782} = \frac{50}{1589} \approx 0.0315$$
Smaller tuning ratio → narrower VCO tuning range → simpler, more stable design. Upper LO is preferred.

4.1.3 Conversion Gain/Loss

Conversion gain / loss $$G_C = 10\log_{10}\frac{P_{\text{IF,out}}}{P_{\text{RF,in}}}\;(\text{dB}), \qquad L_C = -G_C = 10\log_{10}\frac{P_{\text{RF,in}}}{P_{\text{IF,out}}}\;(\text{dB})$$
  • Diode mixers: $L_C \approx 4$–$7$ dB (inherent conversion loss)
  • Transistor mixers: can achieve positive $G_C$ (conversion gain)
  • $L_C$ is minimized at an optimal LO power level (typically 0–10 dBm for diode mixers)
為什麼 $L_C$ 對 LO power 是 U 型曲線?Why is $L_C$ a U-shaped curve against LO power?なぜ $L_C$ は LO パワーに対して U 字曲線になるのか?
LC LO Power PLO,opt LO too weak g(t) 弱weak g(t)g(t) 弱 LO too strong Rs / Cj 主導 dominates が支配的 LC,min Conversion loss 由 The conversion loss is set by 変換損失は兩個相互競爭的機制two competing mechanisms二つの競合するメカニズム決定,兩者交會處就是最佳 LO drive。 , and their crossover is the optimum LO drive. で決まり、両者が交差する点が最適な LO ドライブです。

左側 — LO too weak ($L_C$ 偏高)Left — LO too weak ($L_C$ high)左側 — LO が弱すぎる($L_C$ が高い)

二極體幾乎沒在切換。time-varying conductance $g(t) = dI/dV$ 在 $f_{LO}$ 的 Fourier 分量非常小,$f_{IF}$ 的 mixing product 也就很弱;二極體幾乎呈線性 → The diodes are barely switching. The Fourier component of the time-varying conductance $g(t) = dI/dV$ at $f_{LO}$ is very small, so the mixing product at $f_{IF}$ is weak too; the diode behaves almost linearly → ダイオードはほとんどスイッチングしていません。時変コンダクタンス $g(t) = dI/dV$ の $f_{LO}$ におけるフーリエ成分が非常に小さいため、$f_{IF}$ のミキシング積も弱くなります。ダイオードはほぼ線形にふるまい → 沒有 frequency conversion 能力no frequency-conversion capability周波数変換能力がありません.

谷底 — Optimum LO drive (Schottky 約 0–10 dBm)The bottom — the optimum LO drive (about 0–10 dBm for a Schottky)谷底 — 最適な LO ドライブ(ショットキーで約 0–10 dBm)

LO 振幅夠大,每個週期將二極體用力推到 ON/OFF。$g(t)$ 趨近於 square-wave,於 $f_{LO}$ 的基頻分量 — 也就是真正負責 RF→IF 轉換的那一項 — 達到The LO amplitude is large enough to drive the diode hard ON/OFF every cycle. $g(t)$ approaches a square wave and its fundamental component at $f_{LO}$ — the very term responsible for the RF→IF conversion — reaches its LO 振幅が十分大きく、毎周期ダイオードを強く ON/OFF させます。$g(t)$ は方形波に近づき、$f_{LO}$ における基本波成分 — まさに RF→IF 変換を担う項 — は最大值maximum最大値。$L_C$ 在此最低。. $L_C$ is at its lowest here.です。$L_C$ はここで最小になります。

右側 — LO too strong ($L_C$ 反而上升)Right — LO too strong ($L_C$ rises again)右側 — LO が強すぎる($L_C$ が逆に上昇)

  • Series resistance $R_s$ 主導:The series resistance $R_s$ takes over:直列抵抗 $R_s$ が支配的になります: 二極體被強迫進入深度導通後,bulk $R_s$ 變成主要的損耗元件。power 被消耗在 $R_s$,不是轉到 IF。 once the diode is forced into deep conduction, the bulk $R_s$ becomes the dominant loss element. Power is dissipated in $R_s$ rather than converted to the IF. ダイオードが深い導通に押し込まれると、バルクの $R_s$ が主要な損失要素になります。電力は IF に変換されるのではなく $R_s$ で消費されます。
  • Junction capacitance $C_j$ 短路 RF:The junction capacitance $C_j$ shorts out the RF:接合容量 $C_j$ が RF を短絡します: 重 forward bias 縮小 depletion region → $C_j$ 上升 → RF 訊號從 nonlinear junction 旁邊漏掉。 heavy forward bias shrinks the depletion region → $C_j$ rises → the RF signal leaks past the nonlinear junction. 強い順方向バイアスは空乏層を縮小させ → $C_j$ が上昇 → RF 信号が非線形接合の脇を通って漏れます。
  • Harmonics & intermod 增加:Harmonics & intermodulation increase:高調波 & 相互変調が増加します: 更多 LO power 被導向 higher-order products ($2f_{LO}\pm f_{RF}$、$3f_{LO}\pm f_{RF}$ ...),而非想要的 $f_{IF}$。 more of the LO power is diverted into higher-order products ($2f_{LO}\pm f_{RF}$, $3f_{LO}\pm f_{RF}$, …) rather than the wanted $f_{IF}$. LO パワーの多くが、目的の $f_{IF}$ ではなく高次の積($2f_{LO}\pm f_{RF}$、$3f_{LO}\pm f_{RF}$ …)へ振り向けられます。
  • Port-to-port isolation 變差:Port-to-port isolation degrades:ポート間アイソレーションが劣化します: LO leakage 增加,後級被 compress,實際 IF gain 下降。 LO leakage increases, the following stages compress, and the actual IF gain falls. LO 漏れが増え、後段が圧縮され、実際の IF 利得が低下します。

整體圖像:The overall picture:全体像:存在一個甜蜜點 — switching 夠強,$g(t)$ 在 $f_{LO}$ 有強分量,但 $R_s$/$C_j$ 損耗尚未接管 — 那就是 U 型曲線的最低點。 there is a sweet spot — switching hard enough that $g(t)$ has a strong component at $f_{LO}$, but before the $R_s$/$C_j$ losses take over — and that is the bottom of the U-shaped curve.スイートスポットが存在します — スイッチングが十分強く $g(t)$ が $f_{LO}$ に強い成分をもち、かつ $R_s$/$C_j$ の損失がまだ支配的でない領域 — そこが U 字曲線の最低点です。

Conversion loss versus LO power — output spectrum and L_C curve
Lecture slide 9 — output spectrum after mixing and the characteristic $L_C$-vs-LO-power bowl.
Slide 9 — text transcript
  • One factor that strongly affects conversion loss is the local-oscillator power level. In a diode mixer, minimum conversion loss often occurs for LO powers between 0 and 10 dBm.
  • This power level is large enough that accurate characterization of mixer performance often requires nonlinear analysis.
  • Right plot: $L_C$ versus LO power — a bowl-shaped curve with a unique optimum.(轉換損耗)Right plot: $L_C$ versus LO power — a bowl-shaped curve with a unique optimum.右のグラフ:$L_C$ 対 LO パワー — 唯一の最適点をもつ椀型の曲線(変換損失)

4.1.4 Noise Figure — DSB vs SSB

Noise figure depends on whether the input is a DSB or SSB signal because both sidebands contribute to the IF noise regardless of which carries the desired signal.

DSB vs SSB Noise Figure
For a DSB signal both sidebands carry signal power $\Rightarrow$ higher input SNR $\Rightarrow$ lower noise figure.
For a SSB signal only one sideband carries signal, but noise still folds in from both bands: $$\boxed{F_{SSB} = 2\,F_{DSB}} \quad\Longleftrightarrow\quad NF_{SSB} = NF_{DSB} + 3\text{ dB}$$
推導:為何 DSB 噪聲指數比 SSB 小 3 dBDerivation: why the DSB noise figure is 3 dB below the SSB one導出:なぜ両側波帯(DSB)の雑音指数は片側波帯(SSB)より 3 dB 小さいのか

DSB signal: $v_{DSB}(t) = A\cos(\omega_{LO}+\omega_{IF})t + A\cos(\omega_{LO}-\omega_{IF})t$

After mixing with LO and low-pass filtering: $v_{IF}(t) = AK\cos(\omega_{IF}t)$

Input signal power: $S_i = A^2$ (from both sidebands). Output power: $S_o = A^2K^2/2$. $N_i = kT_0B$.

SSB signal: $v_{SSB}(t) = A\cos(\omega_{LO}+\omega_{IF})t$ only

After mixing: same IF amplitude $AK/2$, but $S_i = A^2/2$ (half the DSB input power).

Both cases generate the same added noise $N_{added}$ in the mixer. Since $S_i^{SSB} = S_i^{DSB}/2$ and $S_o$, $N_o$ are identical:

$$F_{SSB} = \frac{S_i^{SSB}/N_i}{S_o/N_o} = 2 \cdot \frac{S_i^{DSB}/N_i}{S_o/N_o} = 2F_{DSB}$$

The physical reason: DSB signal uses both spectral slots that contribute noise, so each watt of input is more "efficiently" used.

Practical Note
Diode mixer noise figures are typically 5–8 dB. For a DSB measurement $F_{DSB}$, the SSB noise figure is 3 dB higher. System link budgets for SSB (single-channel) signals must use $NF_{SSB}$.

4.1.5 Port-to-Port Isolation, VSWR, and DC Offset

Port-to-Port Isolation

Isolation is the attenuation (dB) from any input port to any other port. High isolation numbers are desirable.

  • LO-to-RF isolation is most critical in receivers: LO leakage at the RF port is radiated by the antenna and must meet FCC limits. Typical spec: 20–40 dB.
  • Isolation depends mainly on internal device balance and external transformer/hybrid performance.
  • Higher applied signal levels worsen isolation.

Port VSWR

RF and IF port VSWRs affect conversion gain and bandwidth. High LO port VSWR wastes available LO power (underdrive degrades performance).

DC Offset

Port-to-port isolation determines dc offset. Critical in direct-conversion (homodyne) receivers and phase detectors — dc offset can corrupt the baseband signal.

LO self-mixing and interferer self-mixing paths
Two self-mixing paths in a homodyne receiver — (left) LO leakage through the LNA back to the mixer RF port; (right) a strong RF interferer leaking forward into the LO port.
How does the dc offset arise? — self-mixing mechanism(自混頻產生 DC offset 的機制)How does the dc offset arise? — the self-mixing mechanismDC オフセットはどのように生じるのか — セルフミキシングのメカニズム

In a direct-conversion (homodyne) receiver the LO sits at the same frequency as the desired RF. Two leakage paths cause a signal to mix with itself, producing a DC term at the mixer output:

(1) LO self-mixing(LO 自混頻).(1) LO self-mixing.(1) LO セルフミキシング。 Because LO-to-RF isolation is finite, a portion of the LO leaks backward through the mixer to the LNA input, reflects off the antenna / LNA / matching components, and returns to the mixer's RF port together with the wanted signal:

$$v_{RF}(t) = v_{sig}(t) + \alpha\cos(\omega_{LO}t+\phi), \qquad v_{LO}(t)=\cos\omega_{LO}t$$

Multiplying in the mixer:

$$v_{IF}(t) = K\,v_{RF}v_{LO} = K\,v_{sig}\cos\omega_{LO}t + \underbrace{\tfrac{K\alpha}{2}\cos\phi}_{\text{DC offset}} + \tfrac{K\alpha}{2}\cos(2\omega_{LO}t+\phi)$$

The $\cos\phi$ term sits exactly at $f_{IF}=0$ — directly on top of the desired baseband signal. The $2\omega_{LO}$ term is easily low-pass filtered out.

(2) Interferer self-mixing(干擾源自混頻).(2) Interferer self-mixing.(2) 干渉波のセルフミキシング。 A large in-band interferer at the RF port can also leak forward to the LO port (poor RF-to-LO isolation). It then mixes with itself, producing another DC term proportional to its instantaneous power $|v_{int}|^2$. Because the interferer envelope varies in time, this becomes a time-varying DC offset — much worse than a static one, since it sits inside the signal bandwidth and cannot simply be ac-coupled away.

Why the mixer can't escape it. Both mechanisms require only finite isolation between ports. Even at 40 dB isolation, a 0 dBm LO produces a $-40$ dBm leakage that, after self-mixing, can easily exceed the wanted signal at the baseband input of the ADC.

Mitigations. Increase LO-to-RF / RF-to-LO isolation (better balance, baluns, layout); use ac-coupling for the slow component; servo-loop dc cancellation; or avoid homodyne entirely (low-IF / superheterodyne).

Three port-to-port isolation paths in a mixer (RF-IF, LO-RF, LO-IF)
Lecture slide 12 — the three port-to-port leakage paths; LO-to-RF is the most critical (FCC radiation limit).
about isolation of ports
  • Port-to-port isolation is the attenuation (dB) between a signal input at any port and its measured level at any other port. Higher numbers are better.
  • Isolation depends mainly on internal device balance and external transformer/hybrid performance. Higher applied signal levels worsen isolation.
  • In a receiver, LO power coupled through the mixer to the RF port is radiated by the antenna. Because such signals interfere with other services and users, the FCC sets stringent limits on radiated power.
  • Therefore, high LO-to-RF isolation is always required in a receiver mixer; typical spec is 20–40 dB.
  • LO-to-IF leakage is hard to filter when the frequencies are close.

4.1.6 Linearity

  • Dynamic range = $P_{1\text{dB}} - P_{\min,\text{discernible}}$ (dB)
  • IP2: output power where second-order spurs $(f_1 \pm f_2 \pm f_{LO})$ equal desired output
  • IP3: output power where third-order spurs $(2f_1 \pm f_2 \pm f_{LO})$ equal desired output
Intermodulation distortion in a mixer — IM1 and IM3 in the IF band
Intermodulation distortion in a mixer — two RF tones at $f_1$ and $f_2$ plus their 3rd-order intermods at $2f_1{-}f_2$ and $2f_2{-}f_1$ are all mixed with $f_{LO}$ and translated into the IF band, where the desired tones (IM$_1$) sit between the unwanted IM$_3$ spurs.
Mixer Spurious Products
Because the LO frequency shifts all products, mixer IM spurs appear at $(mf_{RF} + nf_{LO})$ for integers $m$, $n$. The most troublesome are IP2 (even-order) and IP3 (odd-order). Balanced mixer topologies cancel even-order harmonics.
How do balanced mixers cancel even-order products?(平衡式混頻器如何消除偶次諧波)How do balanced mixers cancel even-order products?平衡ミキサはどのように偶数次の積を打ち消すのか

A balanced mixer drives two identical nonlinear devices (diodes or transistors) with the same RF but anti-phase LO (180° hybrid / balun), then subtracts their output currents. Expand each device's $i$–$v$ characteristic as a power series:

$$i(v) = a_1 v + a_2 v^2 + a_3 v^3 + \cdots$$

With anti-phase drive, $v_1 = +v$ and $v_2 = -v$. Apply the series to each branch:

$$\begin{aligned} i_1 &= a_1(+v) + a_2(+v)^2 + a_3(+v)^3 + a_4(+v)^4 + \cdots = +a_1 v + a_2 v^2 + a_3 v^3 + a_4 v^4 + \cdots \\ i_2 &= a_1(-v) + a_2(-v)^2 + a_3(-v)^3 + a_4(-v)^4 + \cdots = -a_1 v + a_2 v^2 - a_3 v^3 + a_4 v^4 + \cdots \end{aligned}$$

The differential output current is

$$i_o = i_1 - i_2 = 2\bigl(a_1 v + a_3 v^3 + a_5 v^5 + \cdots\bigr)$$

Every even-order term — $a_2 v^2$, $a_4 v^4$, $\dots$ — appears with the same sign in both branches and cancels in the subtraction. Only odd-order terms double up. The same parity argument applies to the products $v_{RF}v_{LO}$: with anti-phase LO, the wanted $v_{RF}v_{LO}$ (odd in $v_{LO}$) survives, while LO-feedthrough $v_{LO}$ and LO² products are suppressed by the cancellation.

Consequences. (i) LO-to-IF feedthrough vanishes (huge isolation improvement). (ii) Even-order IM products — $2f_1$, $2f_2$, $f_1\pm f_2$ — are rejected, so IP2 is much higher. (iii) DC offset from LO² self-mixing is also cancelled. The cancellation is only as good as the balance between the two devices and the symmetry of the hybrid/balun — typical real-world rejection: 20–40 dB.

Why odd-order products survive. They are odd functions of $v$, so reversing the sign of $v$ also reverses the sign of the current; subtracting two oppositely-signed currents adds them. This is exactly why IP3 (third-order) remains the dominant linearity limit even in balanced mixers.

4.1.7 Semiconductor Devices & Diode vs Transistor Comparison

Devices Used as Mixers

  • Passive: Schottky barrier diode — lowest junction capacitance, highest frequency, lowest noise
  • Active: BJT, HBT, MOSFET, MESFET, HEMT — transconductance nonlinearity enables conversion gain
Property Diode (Passive) Transistor (Active)
Conversion Inherent loss (4–7 dB) Conversion gain possible
Noise figure Lower (5–8 dB) Higher, wider range
Linearity (IP3) Better Lower
Operating frequency Higher (small $C_j$)
$C_j$ = junction capacitance of the diode; small $C_j$
→ less RF shunting(旁路/分流)at high frequency→ less RF shunting at high frequency→ 高い周波数での RF の分流が減少
Lower
LO power required High Low
Integration Needs external filters/hybrids Monolithic IC capable
Typical application Base stations Mobile devices, ICs

4.2 Mixer Models and Circuits

4.2.1 Small-Signal Linear Model

In the linear approximation, the mixer is modeled as a multiplier with constant $K$ (conversion gain):

Linear mixer model $$v_{IF}(t) = K \cdot v_{RF}(t)\cdot v_{LO}(t) = \frac{K\,V_{RF}V_{LO}}{2}\bigl[\cos(\omega_{RF}-\omega_{LO})t + \cos(\omega_{RF}+\omega_{LO})t\bigr]$$

Low-pass filtering selects the desired IF at $\omega_{IF} = |\omega_{RF} - \omega_{LO}|$. This model ignores nonlinear effects and is only valid for small signals.

Detailed derivation — $v_{IF}(t) = K\,v_{RF}\,v_{LO} \;\Rightarrow\; \tfrac{K}{2}V_{RF}V_{LO}[\cos(\omega_{RF}-\omega_{LO})t + \cos(\omega_{RF}+\omega_{LO})t]$(推導步驟)Detailed derivation — $v_{IF}(t) = K\,v_{RF}\,v_{LO} \;\Rightarrow\; \tfrac{K}{2}V_{RF}V_{LO}[\cos(\omega_{RF}-\omega_{LO})t + \cos(\omega_{RF}+\omega_{LO})t]$詳細な導出 — $v_{IF}(t) = K\,v_{RF}\,v_{LO} \;\Rightarrow\; \tfrac{K}{2}V_{RF}V_{LO}[\cos(\omega_{RF}-\omega_{LO})t + \cos(\omega_{RF}+\omega_{LO})t]$

Step 0 — Inputs. Model RF and LO as pure tones:

$$v_{RF}(t) = V_{RF}\cos\omega_{RF}t, \qquad v_{LO}(t) = V_{LO}\cos\omega_{LO}t$$

Step 1 — Linear mixer assumption. The ideal mixer is a memoryless multiplier with conversion constant $K$:

$$v_{IF}(t) \;=\; K\cdot v_{RF}(t)\cdot v_{LO}(t)$$

Step 2 — Substitute the tones. Plug in the cosines:

$$v_{IF}(t) = K\,(V_{RF}\cos\omega_{RF}t)(V_{LO}\cos\omega_{LO}t) = K\,V_{RF}V_{LO}\,\cos\omega_{RF}t\,\cos\omega_{LO}t$$

Step 3 — Product-to-sum identity. Use the trigonometric identity

$$\boxed{\;\cos A\,\cos B \;=\; \tfrac{1}{2}\bigl[\cos(A-B) + \cos(A+B)\bigr]\;}$$

Where does this come from? Add the two angle-sum formulas:

$$\begin{aligned} \cos(A-B) &= \cos A\cos B + \sin A\sin B\\ \cos(A+B) &= \cos A\cos B - \sin A\sin B\\ \hline \cos(A-B) + \cos(A+B) &= 2\cos A\cos B \end{aligned}$$

Divide both sides by 2 to get the boxed identity. The $\sin A\sin B$ terms cancel — that cancellation is the whole reason a multiplier produces only sum and difference frequencies, with no $\omega_{RF}$ or $\omega_{LO}$ leakage in the ideal case.

Step 4 — Apply with $A=\omega_{RF}t$, $B=\omega_{LO}t$.

$$\cos\omega_{RF}t\,\cos\omega_{LO}t \;=\; \tfrac{1}{2}\bigl[\cos(\omega_{RF}-\omega_{LO})t \;+\; \cos(\omega_{RF}+\omega_{LO})t\bigr]$$

Step 5 — Collect. Combine with the $K\,V_{RF}V_{LO}$ prefactor from Step 2:

$$\color{#b44a2f}{\boldsymbol{v_{IF}(t) \;=\; \frac{K\,V_{RF}V_{LO}}{2}\bigl[\cos(\omega_{RF}-\omega_{LO})t \;+\; \cos(\omega_{RF}+\omega_{LO})t\bigr]}}$$

Step 6 — Interpret. The output contains exactly two spectral components:

  • Difference frequency $\omega_{IF} = |\omega_{RF}-\omega_{LO}|$ — the wanted IF (down-conversion).
  • Sum frequency $\omega_{RF}+\omega_{LO}$ — the image of multiplication, rejected by the IF low-pass filter.

Both tones have amplitude $\tfrac{1}{2}K V_{RF}V_{LO}$. The factor of $\tfrac{1}{2}$ is unavoidable — it is the price of representing a real multiplier with cosines (half the energy goes to each sideband).


中文版(推導步驟)Step-by-step derivation導出の手順

  1. 輸入訊號:The input signals:入力信号:令 $v_{RF}=V_{RF}\cos\omega_{RF}t$、$v_{LO}=V_{LO}\cos\omega_{LO}t$。 let $v_{RF}=V_{RF}\cos\omega_{RF}t$ and $v_{LO}=V_{LO}\cos\omega_{LO}t$.$v_{RF}=V_{RF}\cos\omega_{RF}t$、$v_{LO}=V_{LO}\cos\omega_{LO}t$ とします。
  2. 線性混頻器假設:The linear-mixer assumption:線形ミキサの仮定:輸出為兩訊號之乘積乘上常數 $K$,即 $v_{IF}=K\,v_{RF}v_{LO}$。 the output is the product of the two signals times a constant $K$, i.e. $v_{IF}=K\,v_{RF}v_{LO}$.出力は二つの信号の積に定数 $K$ を掛けたもの、すなわち $v_{IF}=K\,v_{RF}v_{LO}$ です。
  3. 代入並提出振幅:Substitute and factor out the amplitudes:代入して振幅をくくり出す:得到 $K V_{RF}V_{LO}\cos\omega_{RF}t\cos\omega_{LO}t$。 giving $K V_{RF}V_{LO}\cos\omega_{RF}t\cos\omega_{LO}t$.$K V_{RF}V_{LO}\cos\omega_{RF}t\cos\omega_{LO}t$ が得られます。
  4. 套用積化和差公式:Apply the product-to-sum identity:積和公式を適用する:$\cos A\cos B=\tfrac12[\cos(A-B)+\cos(A+B)]$,此公式由兩個 $\cos(A\pm B)$ 展開式相加 ($\sin A\sin B$ 項互相抵消) 而得 — 這正是理想乘法器只產生和頻與差頻、不會洩漏 $\omega_{RF}$、$\omega_{LO}$ 本身的根本原因。$\cos A\cos B=\tfrac12[\cos(A-B)+\cos(A+B)]$, which follows from adding the two expansions of $\cos(A\pm B)$ (the $\sin A\sin B$ terms cancelling) — and this is precisely why an ideal multiplier produces only the sum and difference frequencies and never leaks $\omega_{RF}$ or $\omega_{LO}$ themselves.$\cos A\cos B=\tfrac12[\cos(A-B)+\cos(A+B)]$。この公式は $\cos(A\pm B)$ の二つの展開式を足し合わせる($\sin A\sin B$ の項が相殺する)ことで得られます — これこそ理想乗算器が和周波と差周波のみを生成し、$\omega_{RF}$ や $\omega_{LO}$ 自体を漏らさない根本的な理由です。
  5. 合併常數:Combine the constants:定数をまとめる:得到最終形式 $v_{IF}=\tfrac{K V_{RF}V_{LO}}{2}[\cos(\omega_{RF}-\omega_{LO})t+\cos(\omega_{RF}+\omega_{LO})t]$。 giving the final form $v_{IF}=\tfrac{K V_{RF}V_{LO}}{2}[\cos(\omega_{RF}-\omega_{LO})t+\cos(\omega_{RF}+\omega_{LO})t]$.最終形 $v_{IF}=\tfrac{K V_{RF}V_{LO}}{2}[\cos(\omega_{RF}-\omega_{LO})t+\cos(\omega_{RF}+\omega_{LO})t]$ が得られます。
  6. 濾波:Filtering:フィルタリング:後級 LPF 保留差頻 $\omega_{IF}=|\omega_{RF}-\omega_{LO}|$(即所要的 IF),濾掉和頻 $\omega_{RF}+\omega_{LO}$。 the following LPF keeps the difference frequency $\omega_{IF}=|\omega_{RF}-\omega_{LO}|$ (the wanted IF) and removes the sum frequency $\omega_{RF}+\omega_{LO}$.後段の LPF は差周波 $\omega_{IF}=|\omega_{RF}-\omega_{LO}|$(目的の IF)を残し、和周波 $\omega_{RF}+\omega_{LO}$ を除去します。
Small-signal linear mixer model
Linear mixer model — RF and LO are multiplied (gain $K$), producing sum and difference tones at $\omega_{RF}\pm\omega_{LO}$. The low-pass filter keeps the difference (IF) and rejects the sum.

4.2.2 Small-Signal Nonlinear Model

Single-diode mixer schematic with DC bias choke
Single-diode mixer — RF and LO drive the diode together with DC bias $V_{DC}$ supplied through an RF choke; the output current $i(t)$ is taken at the IF port.
Diode I–V characteristic and small-signal sign convention
Diode I–V characteristic ($I = I_s(e^{V/V_T}-1)$) with the sign convention $V^+ \to V^-$, $I$ flowing into the anode. The strong nonlinearity above $V \gtrsim V_T$ is what enables frequency mixing; the Taylor series below expands $I(V)$ about the DC bias point.

The diode current is expanded in a Taylor series about the DC bias point $v_{DC}$, with the combined signal $v_{RF}+v_{LO}$ as the perturbation. Keeping terms through 4th order:

Taylor series — diode current $$i(v_{DC} + v_{RF} + v_{LO}) = i_{DC} + a_0(v_{RF}+v_{LO}) + a_1(v_{RF}+v_{LO})^2 + a_2(v_{RF}+v_{LO})^3 + a_3(v_{RF}+v_{LO})^4$$
where the coefficients are derivatives evaluated at $v_{DC}$: $$a_0 = \frac{di}{dv}\bigg|_{v_{DC}},\quad a_1 = \frac{1}{2}\frac{d^2i}{dv^2}\bigg|_{v_{DC}},\quad a_2 = \frac{1}{6}\frac{d^3i}{dv^3}\bigg|_{v_{DC}},\quad a_3 = \frac{1}{24}\frac{d^4i}{dv^4}\bigg|_{v_{DC}}$$
完整推導:Taylor 展開 → IF 頻率成分 → 轉換電壓增益 $G_v$Full derivation: Taylor expansion → the IF frequency component → the conversion voltage gain $G_v$完全な導出:テイラー展開 → IF 周波数成分 → 変換電圧利得 $G_v$

Step 1 — Taylor expansion of the diode I–V. Assuming the small-signal condition $v_{DC} \gg v_{RF}+v_{LO}$, expand $i(v)$ about $v_{DC}$:

$$\begin{aligned} i(v) &= i(v_{DC}) + \frac{di}{dv}\bigg|_{v_{DC}}\!(v_{RF}+v_{LO}) + \frac{1}{2!}\frac{d^2 i}{dv^2}\bigg|_{v_{DC}}\!(v_{RF}+v_{LO})^2 \\ &\quad + \frac{1}{3!}\frac{d^3 i}{dv^3}\bigg|_{v_{DC}}\!(v_{RF}+v_{LO})^3 + \frac{1}{4!}\frac{d^4 i}{dv^4}\bigg|_{v_{DC}}\!(v_{RF}+v_{LO})^4 + \cdots \end{aligned}$$

Define the differential conductances at the DC bias point:

$$\color{#b44a2f}{\boldsymbol{g_{DC} = \frac{di}{dv}\bigg|_{v_{DC}}, \quad g'_{DC} = \frac{d^2 i}{dv^2}\bigg|_{v_{DC}}, \quad g''_{DC} = \frac{d^3 i}{dv^3}\bigg|_{v_{DC}}, \quad g'''_{DC} = \frac{d^4 i}{dv^4}\bigg|_{v_{DC}}}}$$

so the series becomes

$$i(v) = i_{DC} + g_{DC}(v_{RF}+v_{LO}) + \tfrac{1}{2}g'_{DC}(v_{RF}+v_{LO})^2 + \tfrac{1}{6}g''_{DC}(v_{RF}+v_{LO})^3 + \tfrac{1}{24}g'''_{DC}(v_{RF}+v_{LO})^4 + \cdots$$

Keeping terms up to 4th order and factoring out one $(v_{RF}+v_{LO})$ from the perturbation, the slide rewrites this as

$$i(v) \approx i_{DC} + (v_{RF}+v_{LO})\bigl[a_0 + a_1(v_{RF}+v_{LO}) + a_2(v_{RF}+v_{LO})^2 + a_3(v_{RF}+v_{LO})^3\bigr]$$

with $a_0 = g_{DC}$, $\;a_1 = \tfrac{1}{2}g'_{DC}$, $\;a_2 = \tfrac{1}{6}g''_{DC}$, $\;a_3 = \tfrac{1}{24}g'''_{DC}$.


Step 2 — collect frequency components at $\omega_{IF} = |\omega_{RF}-\omega_{LO}|$. Substitute $v_{RF} = V_{RF}\cos\omega_{RF}t$ and $v_{LO} = V_{LO}\cos\omega_{LO}t$ into the expansion. Three product types generate a tone at $\omega_{IF}$:

  • the $a_1$ term gives the wanted product $V_{LO}V_{RF}\cos(\omega_{RF}-\omega_{LO})t$ directly;
  • the $a_3$ term gives $\tfrac{3}{2}V_{LO}V_{RF}^3$ and $\tfrac{3}{2}V_{RF}V_{LO}^3$ contributions at $\omega_{IF}$ (from the trig identity $\cos^2 A\cos B = \tfrac{1}{2}\cos B + \tfrac{1}{4}\cos(2A\pm B)$);
  • even-order ($a_0$, $a_2$) terms do not produce a tone at $\omega_{IF}$ — they create DC, $2\omega_{RF}$, $2\omega_{LO}$ and $\omega_{RF}\pm 2\omega_{LO}$.
$$i_{IF}(t) = a_1 V_{LO}V_{RF}\cos(\omega_{RF}-\omega_{LO})t + \tfrac{3}{2}a_3 V_{LO}V_{RF}^3\cos(\omega_{RF}-\omega_{LO})t + \tfrac{3}{2}a_3 V_{RF}V_{LO}^3\cos(\omega_{RF}-\omega_{LO})t$$

Combine and factor:

$$i_{IF}(t) = V_{LO}V_{RF}\!\left[a_1 + \tfrac{3}{2}a_3\!\left(V_{RF}^2 + V_{LO}^2\right)\right]\cos(\omega_{RF}-\omega_{LO})t$$

Step 3 — conversion voltage gain. Across an IF load $Z_L$ the output voltage is $v_{IF} = i_{IF}Z_L$. Define $G_v = v_{IF}/v_{RF}$ and read off the amplitude of the $\cos(\omega_{RF}-\omega_{LO})t$ component:

$$G_v = \frac{V_{LO}V_{RF}Z_L\!\left[a_1 + \tfrac{3}{2}a_3(V_{RF}^2+V_{LO}^2)\right]}{V_{RF}} = V_{LO}Z_L\!\left[a_1 + \tfrac{3}{2}a_3\!\left(V_{RF}^2 + V_{LO}^2\right)\right]$$

Four physical factors in $G_v$ (matching the labeled arrows on slide 18):

Symbol Name Meaning
$V_{LO}$ LO pump factor larger LO drives the diode harder through its nonlinear region → bigger conversion
$Z_L$ Load effect IF voltage is $i_{IF}Z_L$ — gain scales linearly with the IF load impedance
$a_1$ Linear gain factor half the second derivative of $I$–$V$ at bias — the dominant small-signal conversion
$\tfrac{3}{2}a_3(V_{RF}^2{+}V_{LO}^2)$ Gain-compression factor ($a_3 < 0$) fourth-derivative correction; subtracts from $a_1$ as RF or LO grow → $G_v$ falls (1-dB compression, IP3)

Why only odd-order matters for $\omega_{IF}$. A product $\cos^m\omega_{RF}t\cdot\cos^n\omega_{LO}t$ produces $\omega_{RF}-\omega_{LO}$ only when $m+n$ is odd with $m=n=1$ at lowest order, or with both odd at higher order — that is why $a_0$ (constant) and $a_2$ (cubic) contribute nothing at $\omega_{IF}$.

Collecting frequency components at $\omega_{IF} = |\omega_{RF} - \omega_{LO}|$:

IF current — nonlinear $$i_{IF}(t) = V_{LO}V_{RF}\!\left[a_1 + \frac{3}{2}a_3\!\left(V_{RF}^2 + V_{LO}^2\right)\right]\cos(\omega_{RF}-\omega_{LO})t$$

Small-signal limit. When $V_{RF}, V_{LO}$ are small enough that the cubic correction $\tfrac{3}{2}a_3(V_{RF}^2+V_{LO}^2) \ll a_1$, the bracket collapses to $a_1 = \tfrac{1}{2}g'_{DC}$ and the IF current simplifies to a clean two-tone product:

IF current — small-signal approximation $$i_{IF}(t) \approx V_{LO}V_{RF}\cdot\tfrac{1}{2}g'_{DC}\cdot\cos\omega_{IF}t \quad{\scriptstyle\textcolor{gray}{\text{(for small }V_{RF}\text{ and }V_{LO}\text{)}}}$$

This linear form is what we use whenever the mixer can be treated as a simple multiplier — the entire nonlinear behavior collapses into one number, the differential trans-conductance $g'_{DC} = d^2i/dv^2\big|_{v_{DC}}$. Compression and IM3 only re-appear once the cubic term is restored.

Conversion voltage gain $$G_v = V_{LO}Z_L\!\left[a_1 + \frac{3}{2}a_3\!\left(V_{RF}^2 + V_{LO}^2\right)\right]$$

Since $a_3 < 0$ for a diode, the bracketed term decreases with increasing $V_{RF}$ or $V_{LO}$ — this is gain compression.

物理直覺:LO 泵浦如何影響轉換增益Physical intuition: how the LO pump affects conversion gain物理的直感:LO ポンプが変換利得に与える影響

$G_v$ 的 $a_1$ 項代表 The $a_1$ term of $G_v$ represents $G_v$ の $a_1$ 項は線性轉換linear conversion線形変換:LO 電壓乘上 RF 電壓,生成 IF。$a_3$ 項是三階非線性修正,對於二極體 $a_3 < 0$,所以當 $V_{RF}$ 或 $V_{LO}$ 增大時,增益下降(增益壓縮)。: the LO voltage multiplies the RF voltage to generate the IF. The $a_3$ term is the third-order nonlinear correction, and for a diode $a_3 < 0$, so as $V_{RF}$ or $V_{LO}$ grows the gain falls (gain compression).:LO 電圧が RF 電圧に掛かって IF を生成します。$a_3$ 項は三次の非線形補正で、ダイオードでは $a_3 < 0$ であるため、$V_{RF}$ や $V_{LO}$ が大きくなると利得が低下します(利得圧縮)。

注意 $V_{LO}$ 扮演「泵浦(pump)」的角色——它決定 $a_1$ 係數的效率,也決定了壓縮點的位置。$V_{LO}$ 增大時,$G_v$ 整體增大(線性項主導),但 $a_3 V_{LO}^2$ 也增大,加快壓縮。Note that $V_{LO}$ plays the role of the “pump” — it sets the efficiency of the $a_1$ coefficient and also fixes where the compression point lies. As $V_{LO}$ grows, $G_v$ grows overall (the linear term dominating), but $a_3 V_{LO}^2$ grows too, hastening compression.$V_{LO}$ が「ポンプ(pump)」の役割を担う点に注意してください——それが $a_1$ 係数の効率を決め、圧縮点の位置も決めます。$V_{LO}$ が大きくなると $G_v$ は全体として大きくなります(線形項が支配的)が、$a_3 V_{LO}^2$ も大きくなり圧縮が早まります。

4.2.3 OP1dB and OIP3 in a Mixer

The derivations are identical to the amplifier case (Ch. 2 / Ch. 3), with the LO pump replacing the bias as the key parameter. The 4th-order Taylor expansion in a mixer maps to the 3rd-order expansion in an amplifier for the purpose of computing intercept points.

Output 1-dB Compression Point

OP1dB $$OP1dB = \frac{a_1^3 V_{LO}^2 Z_L}{3|a_3|}\cdot(10^{0.05}-1)\cdot 10^{0.1} \approx 0.0863\;\frac{a_1^3 V_{LO}^2 Z_L}{3|a_3|}\quad(\text{W})$$

Output Third-Order Intercept

OIP3 and input intercept voltage $$V_{IP3} = \sqrt{\frac{2a_1}{3|a_3|}}\;(\text{V}), \qquad OIP3 = \frac{a_1^3 V_{LO}^2 Z_L}{3|a_3|}\;(\text{W})$$
OIP3 vs OP1dB relationship $$OIP3\;(\text{dBm}) = OP1dB\;(\text{dBm}) + 10.64\;\text{dB}$$
IM3 power from two-tone test $$P_{IM3}\;(\text{dBm}) = 3\,P_{IM1}\;(\text{dBm}) - 2\cdot OIP3\;(\text{dBm})$$
完整推導:兩音測試 → $V_{IP3}$、$OIP3$、$OIP3 - OP_{1\text{dB}} = 10.64$ dB、$P_{IM3} = 3P_{IM1} - 2\,OIP3$Full derivation: the two-tone test → $V_{IP3}$, $OIP3$, $OIP3 - OP_{1\text{dB}} = 10.64$ dB, $P_{IM3} = 3P_{IM1} - 2\,OIP3$完全な導出:ツートーン試験 → $V_{IP3}$、$OIP3$、$OIP3 - OP_{1\text{dB}} = 10.64$ dB、$P_{IM3} = 3P_{IM1} - 2\,OIP3$
Two-tone test — IM1 and IM3 in RF and IF bands
Two-tone test setup — two RF tones at $f_1$, $f_2$ together with their close-in IM$_3$ spurs at $2f_1{-}f_2$ and $2f_2{-}f_1$ are mixed with $f_{LO}$ and translated into the IF band, where the desired IM$_1$ pair sits between the unwanted IM$_3$ pair.

Step 1 — two-tone setup. Drive the mixer with two RF tones at $\omega_{RF1}$ and $\omega_{RF2}$ (spaced by one channel bandwidth, i.e. $|\omega_{RF1}-\omega_{RF2}|$ is small) plus the LO:

$$\begin{aligned} v(t) &= v_{DC} + v_{RF1}(t) + v_{RF2}(t) + v_{LO}(t) \\ &= v_{DC} + V_{RF}\cos\omega_{RF1}t + V_{RF}\cos\omega_{RF2}t + V_{LO}\cos\omega_{LO}t \end{aligned}$$

Both RF tones have the same amplitude $V_{RF}$ (the "sweep parameter" we vary in the two-tone test).

Step 2 — identify the two relevant IF tones. Substitute $v(t)$ into the Taylor series of the diode current and collect the angular-frequency components at

  • $\omega_{IF}^{(1)} = |\omega_{RF1}-\omega_{LO}|$ — the desired down-converted tone (call it IM$_1$);
  • $\omega_{IF}^{(3)} = |2\omega_{RF1}-\omega_{RF2}-\omega_{LO}|$ — the closest 3rd-order intermod (call it IM$_3$). Note this falls inside the IF channel because the RF tones are only one bandwidth apart.
$$\begin{aligned} i_{IM_1}(t) &= a_1 V_{LO}V_{RF}\cos(\omega_{RF1}-\omega_{LO})t \\ i_{IM_3}(t) &= \tfrac{3}{2}a_3 V_{LO}V_{RF}^{\,3}\cos(2\omega_{RF1}-\omega_{RF2}-\omega_{LO})t \end{aligned}$$

The factor $\tfrac{3}{2}$ comes from $\cos^2\omega_{RF1}t\cdot\cos\omega_{RF2}t\cdot\cos\omega_{LO}t$ and the trig identity $\cos^2 A\cos B\cos C$ producing one of its eight mixing products at $|2A-B-C|$.

Step 3 — output powers into the IF load $Z_L$. Using $P = \tfrac{1}{2}V^2/Z_L$ for a sinusoid of amplitude $V$ across $Z_L$ — here written as $P = \tfrac{1}{2}V^2 Z_L$ in the slide's convention (current-source representation, $V$ being the current amplitude times $Z_L$):

$$P_{IM_1}(V_{RF}) = \tfrac{1}{2}\bigl(V_{LO}V_{RF}\,a_1\bigr)^{2}Z_L, \qquad P_{IM_3}(V_{RF}) = \tfrac{1}{2}\bigl(\tfrac{3}{2}V_{LO}V_{RF}^{\,3}\,a_3\bigr)^{2}Z_L$$

Step 4 — intercept condition. The (extrapolated) input intercept voltage $V_{IP3}$ is where $P_{IM_1}$ and $P_{IM_3}$ would meet:

$$P_{IM_1}(V_{RF}=V_{IP3}) = P_{IM_3}(V_{RF}=V_{IP3}) \;\;\Longrightarrow\;\; (V_{LO}V_{IP3}\,a_1)^2 = \bigl(\tfrac{3}{2}V_{LO}V_{IP3}^{\,3}\,a_3\bigr)^2$$

Cancelling $V_{LO}^2$ and $V_{IP3}^2$ on both sides, then taking the positive root:

$$a_1^2 = \tfrac{9}{4}a_3^2\,V_{IP3}^{\,4} \;\;\Longrightarrow\;\; V_{IP3} = \sqrt{\frac{2a_1}{3|a_3|}}\;(\text{V})$$

Step 5 — output intercept $OIP3$. Plug $V_{RF}=V_{IP3}$ back into $P_{IM_1}$:

$$OIP3 = P_{IM_1}(V_{IP3}) = \tfrac{1}{2}\bigl(V_{LO}\,a_1\bigr)^2 V_{IP3}^{\,2}\,Z_L = \tfrac{1}{2}V_{LO}^2 a_1^2 \cdot \frac{2a_1}{3|a_3|}\cdot Z_L = \frac{a_1^3 V_{LO}^2 Z_L}{3|a_3|}\;(\text{W})$$

Step 6 — relate $OIP3$ to $OP_{1\text{dB}}$. From §4.2.3, $OP_{1\text{dB}} = 0.0863\cdot \dfrac{a_1^3 V_{LO}^2 Z_L}{3|a_3|}$, so

$$\frac{OIP3}{OP_{1\text{dB}}} = \frac{1}{0.0863} \approx 11.59 \;\;\Longrightarrow\;\; OIP3\,(\text{dBm}) - OP_{1\text{dB}}\,(\text{dBm}) = 10\log_{10}(11.59) \approx 10.64\;\text{dB}$$

Step 7 — the "slope-3" rule. Take the ratio $P_{IM_3}/P_{IM_1}^{\,3}$ and use $OIP3$ to clean it up:

⚠ need check $$\frac{P_{IM_3}(V_{RF})}{\bigl(P_{IM_1}(V_{RF})\bigr)^{3}} = \frac{\tfrac{1}{2}\bigl(\tfrac{3}{2}V_{LO}V_{RF}^{\,3}a_3\bigr)^2 Z_L}{\bigl(\tfrac{1}{2}(V_{LO}V_{RF}a_1)^2 Z_L\bigr)^3} = \frac{9\,a_3^{\,2}}{Z_L^{\,2}\,V_{LO}^{\,4}\,a_1^{\,6}} = \left(\frac{1}{OIP3}\right)^{2}$$

Convert to dBm by taking $10\log_{10}$ of both sides:

$$\boxed{P_{IM_3}\,(\text{dBm}) = 3\,P_{IM_1}\,(\text{dBm}) - 2\,OIP3\,(\text{dBm})}$$

In a $P_{out}$–$P_{in}$ log–log plot this says the IM$_3$ tone rises three times faster than the wanted IM$_1$ tone (slope 3 vs slope 1). The two extrapolated lines meet at $OIP3$ — this is the geometric construction used to measure $OIP3$ from a single two-tone sweep.

Key Result — Cascade Analysis
A mixer can be treated as an amplifier in cascaded system analysis. Use Friis-style cascade formulas with $G_C$ as the gain and the mixer's $IP3$ / $NF$ as given parameters.
Practice Example 1 — Mixer Cascade

Problem: A down-converting receiver chain has: LNA ($G=15$ dB, $NF=1.5$ dB, $OIP3=+20$ dBm), followed by a mixer ($L_C=6$ dB, $NF_{SSB}=8$ dB, $OIP3=+10$ dBm). Find the cascaded NF and OIP3 referred to the LNA input.

Click to reveal solution

Cascaded NF (Friis formula, linear ratios; $F = 10^{NF/10}$):

$$F_{LNA} = 10^{1.5/10} \approx 1.413, \quad F_{mixer} = 10^{8/10} \approx 6.310, \quad G_{LNA} = 10^{15/10} \approx 31.62$$ $$F_{cascade} = F_{LNA} + \frac{F_{mixer}-1}{G_{LNA}} = 1.413 + \frac{5.310}{31.62} \approx 1.413 + 0.168 = 1.581$$ $$NF_{cascade} = 10\log_{10}(1.581) \approx 2.0\;\text{dB}$$

Cascaded OIP3 (referred to each stage's output):

$$\frac{1}{IIP3_{in}} = \frac{1}{IIP3_{LNA}} + \frac{G_{LNA}}{IIP3_{mixer}}$$ $$IIP3_{LNA} = OIP3_{LNA} - G_{LNA} = 20 - 15 = +5\;\text{dBm}$$ $$IIP3_{mixer} = OIP3_{mixer} - G_{mixer} = 10 - (-6) = +16\;\text{dBm}$$ $$\frac{1}{IIP3_{cascade}} = \frac{1}{10^{0.5}} + \frac{10^{1.5}}{10^{1.6}} \approx \frac{1}{3.162} + \frac{31.62}{39.81}$$

Converting to power ratios (mW): $IIP3_{LNA} = 3.162$ mW, $IIP3_{mix} = 39.81$ mW.

$$\frac{1}{IIP3_{casc}} = \frac{1}{3.162} + \frac{31.62}{39.81 \cdot 31.62} = 0.3162 + 0.0251 = 0.3413 \;\text{mW}^{-1}$$ $$IIP3_{casc} \approx 2.93\;\text{mW} \approx +4.7\;\text{dBm}$$

4.2.4 Single-Ended Diode Mixer

The simplest diode mixer: one diode, RF and LO combined in a diplexer, output low-pass filtered.

IF current — single-ended $$i_{IF}(t) = V_{LO}V_{RF}\!\left[a_1 + \frac{3}{2}a_3(V_{RF}^2+V_{LO}^2)\right]\cos\omega_{IF}t \approx \frac{1}{2}g'_{DC}\,V_{LO}V_{RF}\cos\omega_{IF}t \quad(\text{small signal})$$
Disadvantages
Poor input impedance match; limited RF-to-LO isolation; no cancellation of LO harmonics or noise. Used mainly as a reference / simplest design.
Single-ended diode mixer schematic — diplexer, DC block, RF/LO chokes, LPF
Lecture slide 22 — single-ended diode mixer: RF/LO diplexer + DC bias + RF choke + low-pass IF filter.
IF current — single-ended diode mixer $$i_{IF}(t) = V_{LO}V_{RF}\!\left[a_1+\tfrac{3}{2}a_3(V_{RF}^2+V_{LO}^2)\right]\cos\omega_{IF}t \;\approx\; \tfrac{1}{2}g'_{DC}V_{LO}V_{RF}\cos\omega_{IF}t \quad{\scriptstyle\textcolor{gray}{\text{(small }V_{RF},\,V_{LO}\text{)}}}$$
Single-ended diode mixer
  • The single-ended diode mixer contains only one diode in single-ended connection — the simplest type of diode mixer.
  • RF and LO inputs are combined in a diplexer, which superimposes the two voltages to drive the diode.
  • The AC output of the diode in a down-converter is passed through a low-pass filter to provide the desired IF output voltage.
  • Note: isolation and reflection are bad — the diode is hard to impedance-match.
為什麼輸出只剩下 RF × LO 的乘積?(給文組生的直觀解釋) Why is the output left with only the product of RF × LO? (an intuitive explanation for non-specialists) なぜ出力には RF × LO の積だけが残るのか?(専門外の人向けの直感的な説明) ▶ 點此展開▶ Click to expand▶ クリックして展開

白話結論:The plain-language conclusion:平たく言えば:把 RF 訊號(你想聽的那個遙遠電台)和 LO 訊號(電路自己產生的「基準頻率」) feed the RF signal (the distant station you want to hear) and the LO signal (the “reference frequency” the circuit generates itself) RF 信号(聞きたい遠方の放送局)と LO 信号(回路自身が生成する「基準周波数」)を同時餵進一顆「會把訊號弄歪的二極體」into a diode that distorts signals, both at once同時に「信号を歪ませるダイオード」へ入れる,二極體的非線性會無意間幫你做出「乘法」。乘出來的結果包含一個新的、頻率較低的訊號 — 那就是後級電路想要的 IF。後面再用一個「只通過低頻」的濾波器把雜七雜八的高頻全部丟掉,剩下的就只有「RF × LO 的乘積項」。, and the diode's nonlinearity inadvertently performs a multiplication for you. The product contains a new, lower-frequency signal — the IF that the following stages want. A filter that “passes only low frequencies” then throws away all the assorted high-frequency clutter, and what remains is just the “RF × LO product term”.と、ダイオードの非線形性が思いがけず「乗算」を行ってくれます。その積には新しい、より低い周波数の信号が含まれます — それが後段の回路が欲しがる IF です。続いて「低い周波数だけを通す」フィルタで雑多な高周波をすべて捨てれば、残るのは「RF × LO の積の項」だけです。


「會平方就等於會乘法」— 這到底怎麼回事? “Being able to square is the same as being able to multiply” — how does that work? 「平方できることは乗算できることと同じ」— これはどういうことか? ▶ 點此展開▶ Click to expand▶ クリックして展開

二極體本身The diode itself ダイオード自身は並沒有要做乘法has no intention of multiplying乗算をするつもりなどありません。它只會做一件事:「. It does only one thing: “。ダイオードがすることはただ一つ:「把輸入電壓平方一下(再加一點)square the input voltage (plus a bit more)入力電圧を平方する(さらに少し加える)」。但只要它會平方,把兩個訊號「加起來再餵進去」,乘法就”. But so long as it can square, adding two signals together and feeding them in makes the multiplication 」。しかし平方さえできれば、二つの信号を「足してから入れる」だけで乗算は自動掉出來fall out automatically自動的に出てきます。為什麼?關鍵就在這個小學代數恆等式:. Why? The key is this piece of elementary algebra:。なぜでしょうか。鍵はこの初等代数の恒等式です:

$$(a+b)^2 \;=\; a^2 \;+\; \boxed{2ab} \;+\; b^2$$

看到中間那個 See that middle 真ん中の $2ab$ 了嗎?它就是「a 乘以 b」 — 但它不是我們主動算出來的,它是 term? That is “a times b” — but it is not something we set out to compute; it is の項が見えますか?それが「a 掛ける b」です — しかしそれは私たちが意図して計算したものではなく、平方展開時免費送的a free gift of expanding the square平方を展開したときに無料で付いてくるもの。所以只要有一個元件會把東西平方,把 $a+b$ 餵進去,輸出就會「順便」夾帶 $a \times b$ 這項。. So as long as a component squares things, feeding it $a+b$ makes the output carry the $a \times b$ term “along for the ride”.です。したがって、素子が平方さえすれば、$a+b$ を入力するだけで出力は $a \times b$ の項を「ついでに」含むことになります。

用具體數字試一遍:Try it with actual numbers:具体的な数値で試してみましょう:

假設某一瞬間 $v_{RF} = 3$ V、$v_{LO} = 2$ V,diplexer 把它們加起來:$v = 3 + 2 = 5$ V 送進二極體。Suppose at some instant $v_{RF} = 3$ V and $v_{LO} = 2$ V, and the diplexer adds them: $v = 3 + 2 = 5$ V goes into the diode.ある瞬間に $v_{RF} = 3$ V、$v_{LO} = 2$ V であり、ダイプレクサがそれらを足すとします:$v = 3 + 2 = 5$ V がダイオードに入ります。

二極體做平方:$5^2 = 25$。這 25 是怎麼組成的?The diode squares it: $5^2 = 25$. How is that 25 made up?ダイオードはこれを平方します:$5^2 = 25$。この 25 は何で構成されているでしょうか。

$$25 = \underbrace{9}_{v_{RF}^2 = 3^2} + \underbrace{12}_{2 v_{RF} v_{LO} = 2 \cdot 3 \cdot 2} + \underbrace{4}_{v_{LO}^2 = 2^2}$$

中間那個 Inside that middle 真ん中の 12 = 2 × 3 × 2 裡面, term, の項の中に、明明白白地包含 3 × 2 這個乘積the product 3 × 2 is plainly present3 × 2 という積がはっきり含まれています。二極體沒有任何「乘法電路」,但只要它會平方、又有人把兩個訊號加起來餵給它,乘積就一定會出現在輸出裡 — 這是純粹的代數,躲都躲不掉。. The diode has no “multiplier circuit” whatsoever, but so long as it squares and somebody adds two signals before feeding them in, the product must appear in the output — pure algebra, and there is no escaping it.。ダイオードには「乗算回路」など一切ありませんが、平方さえでき、誰かが二つの信号を足してから入力すれば、積は必ず出力に現れます — 純粋な代数であり、逃れようがありません。

類比 — 為什麼「面積」會自動產生「交互項」:An analogy — why “area” automatically produces a “cross term”:類比 — なぜ「面積」が自動的に「交差項」を生むのか:

想像一塊地,長 $= a$,寬 $= b$,面積就是 $a \times b$。現在把長和寬都「加」起來放在一條線上($a+b$),然後算這條線的「平方面積」 $(a+b) \times (a+b)$,你會發現一個邊長 $(a+b)$ 的正方形可以切成四塊:Imagine a plot of land of length $a$ and width $b$, whose area is $a \times b$. Now lay the length and the width end to end along a single line ($a+b$) and compute the “square area” of that line, $(a+b) \times (a+b)$: a square of side $(a+b)$ can be cut into four pieces:長さ $a$、幅 $b$ の土地を想像してください。その面積は $a \times b$ です。ここで長さと幅を一本の線上に並べて足し($a+b$)、その線の「平方の面積」 $(a+b) \times (a+b)$ を計算すると、一辺 $(a+b)$ の正方形は四つの部分に切り分けられます:

  • 左上:$a \times a = a^2$(純 a 區)top left: $a \times a = a^2$ (the pure-a region)左上:$a \times a = a^2$(純粋な a の領域)
  • 右下:$b \times b = b^2$(純 b 區)bottom right: $b \times b = b^2$ (the pure-b region)右下:$b \times b = b^2$(純粋な b の領域)
  • 右上 + 左下:兩塊 $a \times b$ 的長方形 $\Rightarrow 2ab$(這就是「乘積」項!)top right + bottom left: two $a \times b$ rectangles $\Rightarrow 2ab$ (there is your “product” term!)右上 + 左下:$a \times b$ の長方形が二つ $\Rightarrow 2ab$(これが「積」の項です!)

那兩塊長方形的存在The existence of those two rectangles is その二つの長方形の存在は是幾何強迫geometrically compulsory幾何学的に強制されたもの的 — 你只要把兩個邊放在同一邊長上去算面積,就一定會冒出「a 和 b 互相當邊」的長方形區域。這跟我們對二極體做的事一模一樣:只要把兩個訊號加在一起再「平方」(算面積),就一定會跑出「兩者相乘」的成分。— put two lengths along the same side and compute the area, and you inevitably get rectangular regions with “a and b as each other's sides”. This is exactly what we do to a diode: add two signals together and “square” them (compute an area), and a component that is “the two multiplied together” must appear.です — 二つの辺を同じ辺の上に置いて面積を計算すれば、必ず「a と b が互いの辺になる」長方形の領域が現れます。これはダイオードに対して行っていることとまったく同じです:二つの信号を足して「平方」する(面積を計算する)と、「両者の積」という成分が必ず現れます。

關鍵領悟:The key insight:重要な気づき:

線性元件(普通電線、線性放大器)只會做「A linear component (an ordinary wire, a linear amplifier) can only do “線形素子(普通の電線や線形増幅器)は「$\alpha \cdot v$」這種「把訊號等比例放大」的動作 — 你把 $a+b$ 餵進去,它丟出 $\alpha a + \alpha b$,”, scaling a signal proportionally — feed it $a+b$ and out comes $\alpha a + \alpha b$, with 」という「信号を比例的に拡大する」動作しかできません — $a+b$ を入れれば $\alpha a + \alpha b$ が出てくるだけで、永遠是兩個訊號各走各的、不會互相干涉the two signals forever going their separate ways and never interacting二つの信号は常に別々のままで、互いに干渉しません。所以線性元件. So a linear component 。したがって線形素子は絕對做不出乘法can never produce a multiplication絶対に乗算を生み出せません.

但二極體不一樣 — 它的轉移函數包含 $v^2$(甚至 $v^3$、$v^4$)。一旦有 $v^2$ 在,把「兩個訊號的和」代進去做平方,依照代數定律,A diode is different — its transfer function contains $v^2$ (and even $v^3$, $v^4$). Once $v^2$ is present, substituting “the sum of two signals” and squaring makes しかしダイオードは違います — その伝達関数は $v^2$(さらには $v^3$、$v^4$)を含みます。$v^2$ が存在する以上、「二つの信号の和」を代入して平方すれば、代数の法則により$2ab$ 這個乘積項必然出現the product term $2ab$ appear as a matter of algebraic law$2ab$ という積の項が必然的に現れます。二極體沒有意圖要做乘法,但它「會平方」這件事本身已經足夠把乘法「順便」做出來。. The diode has no intention of multiplying, but the mere fact that it squares is already enough to do the multiplication “on the side”.。ダイオードに乗算の意図はありませんが、「平方できる」という事実だけで乗算を「ついでに」行うには十分なのです。

所以工程師需要的是「會平方」的元件,不是「會乘法」的元件。So what an engineer needs is a component that squares, not one that multiplies.つまり技術者に必要なのは「平方できる」素子であって、「乗算できる」素子ではありません。市面上沒有便宜又好用的純「乘法器晶片」可以在 GHz 跑,但便宜可靠的二極體(或 FET 的非線性區)到處都是。混頻器就是利用這個小聰明 — 找一個會把訊號弄歪的元件,靠代數恆等式把它變成乘法器。 There is no cheap, convenient pure “multiplier chip” on the market that runs at GHz, but cheap and reliable diodes (or the nonlinear region of a FET) are everywhere. A mixer exploits exactly this little trick — find a component that distorts a signal and let an algebraic identity turn it into a multiplier.市場には GHz で動作する安価で使いやすい純粋な「乗算器チップ」はありませんが、安価で信頼できるダイオード(や FET の非線形領域)はどこにでもあります。ミキサはまさにこの小さな工夫を利用します — 信号を歪ませる素子を見つけ、代数の恒等式によってそれを乗算器に変えるのです。

那為什麼二極體會「平方」?它又不是計算機 So why does a diode “square”? It is not a calculator ではなぜダイオードは「平方」するのか?計算機でもないのに ▶ 點此展開▶ Click to expand▶ クリックして展開

核心答案:The core answer:核心的な答え:二極體並沒有「真的在做平方」這個動作。它只是 the diode is not “really performing a squaring operation” at all. It is simply ダイオードは「本当に平方という演算を行っている」わけではありません。それは単に一條彎彎的曲線a curved line一本の曲がった曲線(電流 vs 電壓的關係不是直線)。而(the relation between current and voltage is not a straight line). And (電流と電圧の関係が直線ではない)にすぎません。そして任何一條彎曲線,在數學上都可以拆成「直線 + 一點點平方 + 一點點立方 + …」這些簡單元素的疊加any curved line can, mathematically, be decomposed into a superposition of simple elements: “a straight line + a bit of a square + a bit of a cube + …”どんな曲線も数学的には「直線 + 少しの平方 + 少しの立方 + …」という単純な要素の重ね合わせに分解できます — 這就是泰勒展開(Taylor expansion)的精神。「會平方」是「會彎」的副產品,不是二極體刻意做的事。 — that is the spirit of the Taylor expansion. “Being able to square” is a by-product of “being curved”, not something the diode does deliberately. — これがテイラー展開の精神です。「平方できる」は「曲がっている」ことの副産物であり、ダイオードが意図的に行っていることではありません。

1. 二極體的真實面貌:一條指數曲線。1. What a diode really is: an exponential curve.1. ダイオードの正体:一本の指数曲線。

物理告訴我們,二極體的電流 $i$ 跟電壓 $v$ 的關係是「指數型」:Physics tells us that a diode's current $i$ is related to its voltage $v$ exponentially:物理によれば、ダイオードの電流 $i$ は電圧 $v$ と指数関数的な関係にあります:

$$i(v) \;=\; I_s\!\left(e^{v/V_T} - 1\right)$$

其中 $V_T \approx 26$ mV 是熱電壓。這條曲線在 $v < 0$ 時幾乎平貼地面(電流近 0),在 $v > 0$ 時急遽上升,整體呈現「where $V_T \approx 26$ mV is the thermal voltage. The curve hugs the axis for $v < 0$ (essentially zero current) and rises steeply for $v > 0$, giving an overall “ここで $V_T \approx 26$ mV は熱電圧です。この曲線は $v < 0$ ではほぼ軸に張り付き(電流はほぼ 0)、$v > 0$ では急激に立ち上がり、全体として「愈大愈陡the bigger it gets the steeper it gets大きくなるほど急峻になる」的彎曲特性 — 跟下面 I–V 圖看到的形狀完全一致。” character — exactly the shape you see in the I–V plot below.」という曲がり方の特徴を示します — 下の I–V 図に見られる形とまったく同じです。

2. 把指數曲線「展開」 — 平方項自然出現。2. Expand the exponential curve — and the square term appears of its own accord.2. 指数曲線を展開する — 平方項が自然に現れます。

高中數學教過:指數函數 $e^x$ 可以展成無限級數:As taught in school mathematics, the exponential $e^x$ expands as an infinite series:高校数学で習うとおり、指数関数 $e^x$ は無限級数に展開できます:

$$e^x = 1 + x + \tfrac{x^2}{2!} + \tfrac{x^3}{3!} + \tfrac{x^4}{4!} + \cdots$$

把 $x = v/V_T$ 代進去再乘上 $I_s$,扣掉 $-1$,得到:Substituting $x = v/V_T$, multiplying by $I_s$ and subtracting the $-1$ gives:$x = v/V_T$ を代入し、$I_s$ を掛け、$-1$ を引くと、次が得られます:

$$i(v) \;=\; I_s\!\left[\,\frac{v}{V_T} \;+\; \boxed{\tfrac{1}{2}\!\left(\frac{v}{V_T}\right)^{\!2}} \;+\; \tfrac{1}{6}\!\left(\frac{v}{V_T}\right)^{\!3} \;+\; \cdots\,\right]$$

看見那個被框起來的 $v^2$ 項了嗎?這就是「平方」的來源 — 它不是電路特別設計的,See that boxed $v^2$ term? That is where the “square” comes from — it is not something the circuit was designed to produce; 枠で囲まれた $v^2$ の項が見えますか?これが「平方」の出どころです — 回路が特別に設計して作り出したものではなく、它是指數函數本身「自帶」的成分it is a component the exponential function carries within itself指数関数そのものが「もともと備えている」成分です。指數函數 = 「常數 + 一次 + 平方 + 立方 + …」的無窮疊加。. An exponential = an infinite superposition of “a constant + a linear term + a square + a cube + …”.。指数関数 = 「定数 + 一次 + 平方 + 立方 + …」の無限の重ね合わせです。

3. 更一般地:任何「彎曲線」都包含平方項。3. More generally: every curved line contains a square term.3. より一般に:どんな曲線も平方項を含みます。

這個現象不只是指數函數的專利。This is not the exclusive property of the exponential.この現象は指数関数だけの専売特許ではありません。任何一條平滑的彎曲線,在工作點附近都可以展開成多項式Any smooth curve can be expanded as a polynomial about the operating point滑らかな曲線はどれも、動作点の近傍で多項式に展開できます(這就是泰勒定理):(that is Taylor's theorem):(これがテイラーの定理です):

$$f(v) \;\approx\; f(v_0) \;+\; f'(v_0)\,(v-v_0) \;+\; \tfrac{1}{2}f''(v_0)\,(v-v_0)^2 \;+\; \cdots$$

只要 $f''(v_0) \neq 0$(也就是曲線在 $v_0$ 是「真的彎」、不是直線),As long as $f''(v_0) \neq 0$ (i.e. the curve really is curved at $v_0$ rather than straight), $f''(v_0) \neq 0$ でありさえすれば(つまり曲線が $v_0$ で直線ではなく「本当に曲がって」いれば)、平方項一定存在a square term must exist平方項は必ず存在します。所以「會弄歪訊號的元件」≈「會平方的元件」是普遍成立的,不限於二極體 — FET 的次臨界區、BJT 的指數區、甚至任何有飽和特性的放大器,. So “a component that distorts a signal” ≈ “a component that squares” holds quite generally, and not only for diodes — the subthreshold region of a FET, the exponential region of a BJT, even any amplifier with saturation characteristics, 。したがって「信号を歪ませる素子」≈「平方する素子」は広く成り立ち、ダイオードに限りません — FET のサブスレッショルド領域、BJT の指数領域、さらには飽和特性をもつあらゆる増幅器も、都有平方項,因此都能當混頻器用all have a square term and can therefore all be used as mixersいずれも平方項をもち、したがってミキサとして使えます.

4. 視覺直觀:「直線 vs 彎線」決定有沒有平方。4. A visual intuition: “straight vs curved” decides whether there is a square term.4. 視覚的な直感:「直線 vs 曲線」が平方項の有無を決めます。

  • 直線A straight line直線($i = G v$):只有一次項,沒有平方項 ⇒ 線性 ⇒ 不能混頻。($i = G v$): only a linear term and no square ⇒ linear ⇒ cannot mix.($i = G v$):一次項のみで平方項なし ⇒ 線形 ⇒ 混合できません。
  • 拋物線A parabola放物線($i = a_2 v^2$):只有平方項 ⇒ ($i = a_2 v^2$): only a square term ⇒ a ($i = a_2 v^2$):平方項のみ ⇒ 完美perfect完全な的混頻器(理論上理想的「平方律」混頻器)。 mixer (the theoretically ideal “square-law” mixer).ミキサ(理論上理想的な「平方則」ミキサ)。
  • 指數曲線An exponential curve指数曲線(二極體):有一次、平方、立方、…(a diode): linear, square, cubic and so on —(ダイオード):一次・平方・立方などが全部allすべて都有,但平方項夠強就足以混頻;其他項會變成不想要的失真(IM3 等)。 present; the square term alone is strong enough to mix, while the others become unwanted distortion (IM3 and the like).存在します。平方項が十分強ければ混合には足り、他の項は不要な歪み(IM3 など)になります。

實際的二極體不是「純粹的平方器」,但它的 I–V 曲線在 bias 點附近的彎度,A real diode is not a “pure squarer”, but the curvature of its I–V characteristic around the bias point 実際のダイオードは「純粋な平方器」ではありませんが、バイアス点付近の I–V 特性の曲がり具合が提供了足夠多的平方成分supplies enough square content十分な平方成分を供給する,足以做出可用的混頻效果。 to give usable mixing.ため、実用的な混合効果が得られます。設計師選擇 bias 點時,就是在挑「平方項夠強、立方項夠弱」的甜蜜點In choosing the bias point, the designer is picking the sweet spot where “the square term is strong enough and the cubic term weak enough”設計者がバイアス点を選ぶとき、それは「平方項が十分強く、立方項が十分弱い」スイートスポットを選んでいるのです — 這也是 conversion gain 和 linearity(IP3)之間 trade-off 的物理根源。 — and this is the physical root of the trade-off between conversion gain and linearity (IP3). — これが変換利得と直線性(IP3)のトレードオフの物理的な根源でもあります。

一句話:In one sentence:一言でいえば:二極體不是計算機 — 它沒有在「執行平方運算」。它只是一條彎曲的 I–V 曲線。但「彎曲」這件事,在數學上展開後一定包含 $v^2$ 這個成分,於是這顆元件就在它毫不知情的情況下,順便提供了平方功能 — 而平方功能又(如前一個區塊所述)順便提供了乘法功能。整條「彎 ⇒ 平方 ⇒ 乘法 ⇒ 差頻」的因果鏈,全是 A diode is not a calculator — it is not “performing a squaring operation”. It is merely a curved I–V characteristic. But “being curved”, expanded mathematically, must contain a $v^2$ component, and so the device supplies a squaring function entirely without knowing it — and a squaring function (as the previous section showed) in turn supplies a multiplication. The whole causal chain “curved ⇒ square ⇒ multiply ⇒ difference frequency” is ダイオードは計算機ではありません — 「平方演算を実行している」わけではなく、単に曲がった I–V 特性にすぎません。しかし「曲がっている」ことは数学的に展開すれば必ず $v^2$ という成分を含み、この素子はそれと知らぬまま平方の機能を提供します — そして平方の機能は(前の節で述べたとおり)ついでに乗算の機能を提供します。「曲がる ⇒ 平方 ⇒ 乗算 ⇒ 差周波」という因果の連鎖はすべて數學恆等式逼出來的forced by algebraic identities数学の恒等式が強制したもの,不需要任何聰明的電路設計。, and requires no clever circuit design at all.であり、巧妙な回路設計は一切必要ありません。

小結語:In summary:まとめ:這就是為什麼 RF 工程師永遠在找「彎得剛剛好」的元件。彎太少(接近線性)→ 混頻效率差;彎太多(高次項太強)→ 失真嚴重。二極體、JFET、Gilbert cell 裡的差動對 BJT,都是這場「彎度調校」的不同解。 This is why RF engineers are forever looking for components that are “curved just right”. Too little curvature (nearly linear) → poor mixing efficiency; too much (the higher-order terms too strong) → severe distortion. Diodes, JFETs and the differential BJT pair in a Gilbert cell are all different answers to the same problem of “tuning the curvature”.RF 技術者が「ちょうどよく曲がった」素子を探し続けるのはこのためです。曲がりが少なすぎる(線形に近い)→ 混合効率が悪い。曲がりが大きすぎる(高次項が強すぎる)→ 歪みが深刻。ダイオード、JFET、Gilbert cell の差動 BJT 対は、いずれも同じ「曲がり具合の調整」という問題に対する異なる答えです。


1. 先講「相加」和「相乘」的差別。1. First, the difference between “adding” and “multiplying”.1. まず「足す」と「掛ける」の違いについて。

  • 如果你只是把兩首歌「If you merely “二つの曲を単に「add足す」在一起播放,你會同時聽到兩首歌 — 兩個原始頻率還是各自獨立存在,” two songs together and play them, you hear both songs at once — the two original frequencies still exist independently and 」だけで再生すると、二つの曲が同時に聞こえます — 元の二つの周波数はそれぞれ独立に存在したままで、不會冒出新的音高no new pitch appears新しい音高は現れません.
  • 但如果你把兩首歌「But if you “しかし二つの曲を「multiply掛ける」起來(每一個時刻,把兩邊的音量瞬間相乘),就會出現新的「拍頻(beat)」— 也就是兩個頻率的” the two songs (multiplying the two instantaneous volumes together at every moment), a new “beat” appears — that is, the 」と(各瞬間に両者の音量を掛け合わせると)、新しい「うなり(beat)」が現れます — つまり二つの周波数のPoor and and.
  • 鋼琴調音時的「咚咚咚」拍頻就是這個現象:兩個接近但不相同的音同時響時,你的耳朵聽到的是兩者頻率差所造成的響度起伏。混頻器要做的就是這件事 — 把「頻率差」當成輸出。The “thump-thump-thump” beat you hear when tuning a piano is exactly this phenomenon: when two nearby but unequal notes sound together, what your ear hears is a loudness fluctuation at the difference of their frequencies. A mixer does precisely this — it takes the “frequency difference” as its output.ピアノを調律するときに聞こえる「ドンドンドン」といううなりがまさにこの現象です:近いけれども一致しない二つの音が同時に鳴ると、耳が聞くのは両者の周波数差による音量の揺らぎです。ミキサがするのはまさにこれ — 「周波数の差」を出力とすることです。

2. 二極體做的不是「加」,而是「弄歪」。2. What a diode does is not “add” but “distort”.2. ダイオードがするのは「足す」ことではなく「歪ませる」ことです。

普通的電線是「線性」元件 — 你給它兩個訊號相加,它就乖乖把兩個訊號相加丟出來。An ordinary wire is a “linear” element — give it two signals added together and it obediently spits out the two signals added together.普通の電線は「線形」素子です — 二つの信号を足して与えれば、素直に二つの信号を足したものを出力します。什麼新頻率都不會出現。No new frequency appears whatsoever.新しい周波数は一切現れません。

但二極體(Diode)But a diode しかしダイオードは不是線性的is not linear線形ではありません。它的電流不是跟電壓成正比,而是大約跟「電壓的指數」成比例($i \approx I_s\,e^{v/V_T}$)。對它做泰勒展開,會展成這樣:. Its current is not proportional to voltage but roughly proportional to “the exponential of the voltage” ($i \approx I_s\,e^{v/V_T}$). Expanding it as a Taylor series gives:。その電流は電圧に比例せず、およそ「電圧の指数」に比例します($i \approx I_s\,e^{v/V_T}$)。これをテイラー展開すると次のようになります:

$$i(v) \;=\; a_0 + a_1 v + a_2 v^2 + a_3 v^3 + \cdots$$

關鍵就在 The key lies in 鍵となるのは$a_2 v^2$ 這個平方項the square term $a_2 v^2$$a_2 v^2$ という平方項。它代表二極體會把輸入電壓「平方」 — 而平方就是乘法的特例($v \times v$)。. It means the diode “squares” the input voltage — and squaring is just a special case of multiplication ($v \times v$).です。これはダイオードが入力電圧を「平方」することを意味します — そして平方は乗算の特殊な場合にすぎません($v \times v$)。

3. 把 RF + LO 一起餵進去,平方項就會自動產生「乘積」。3. Feed RF + LO in together and the square term automatically produces the “product”.3. RF + LO を一緒に入れると、平方項が自動的に「積」を生みます。

RF 和 LO 在 diplexer 那裡已經被加在一起送進二極體,所以二極體看到的是 $v = v_{RF} + v_{LO}$。代進平方項 $a_2 v^2$:The RF and LO have already been added together at the diplexer before reaching the diode, so what the diode sees is $v = v_{RF} + v_{LO}$. Substituting into the square term $a_2 v^2$:RF と LO はダイオードに届く前にダイプレクサですでに足し合わされているため、ダイオードが見るのは $v = v_{RF} + v_{LO}$ です。平方項 $a_2 v^2$ に代入すると:

$$a_2 (v_{RF}+v_{LO})^2 \;=\; a_2\,v_{RF}^2 \;+\; \underbrace{2 a_2\,v_{RF} v_{LO}}_{\text{← 這就是「乘積」項!}} \;+\; a_2\,v_{LO}^2$$ $$a_2 (v_{RF}+v_{LO})^2 \;=\; a_2\,v_{RF}^2 \;+\; \underbrace{2 a_2\,v_{RF} v_{LO}}_{\text{← there is the “product” term!}} \;+\; a_2\,v_{LO}^2$$ $$a_2 (v_{RF}+v_{LO})^2 \;=\; a_2\,v_{RF}^2 \;+\; \underbrace{2 a_2\,v_{RF} v_{LO}}_{\text{← there is the “product” term!}} \;+\; a_2\,v_{LO}^2$$

就跟國中代數的 $(a+b)^2 = a^2 + 2ab + b^2$ 一模一樣。中間那個 $2ab$ 項,就是「a 乘以 b」 — 它的存在Exactly the same as the schoolroom algebra $(a+b)^2 = a^2 + 2ab + b^2$. That middle $2ab$ term is “a times b” — and its existence 中学代数の $(a+b)^2 = a^2 + 2ab + b^2$ とまったく同じです。真ん中の $2ab$ 項が「a 掛ける b」であり、その存在は完全是平方展開後自動跑出來的falls out automatically from expanding the square平方を展開すると自動的に出てくるもの,不是電路有什麼特別設計。, not from any special circuit design.であり、回路に特別な設計があるからではありません。

4. 這個乘積項裡藏著「差頻」(IF)。4. Hidden inside that product term is the difference frequency (the IF).4. この積の項の中に「差周波」(IF) が隠れています。

當 $v_{RF}=V_{RF}\cos\omega_{RF}t$、$v_{LO}=V_{LO}\cos\omega_{LO}t$ 時,乘積 $v_{RF}v_{LO}$ 用三角恆等式展開:With $v_{RF}=V_{RF}\cos\omega_{RF}t$ and $v_{LO}=V_{LO}\cos\omega_{LO}t$, expanding the product $v_{RF}v_{LO}$ with a trigonometric identity:$v_{RF}=V_{RF}\cos\omega_{RF}t$、$v_{LO}=V_{LO}\cos\omega_{LO}t$ のとき、積 $v_{RF}v_{LO}$ を三角恒等式で展開すると:

$$\cos\omega_{RF}t\,\cos\omega_{LO}t = \tfrac{1}{2}\bigl[\cos(\omega_{RF}-\omega_{LO})t + \cos(\omega_{RF}+\omega_{LO})t\bigr]$$

於是這個乘法產生了So this multiplication produces こうしてこの乗算は兩個全新的頻率two entirely new frequenciesまったく新しい二つの周波数 — 一個是「 — one the “を生み出します — 一つは「差頻difference frequency差周波」$f_{RF}-f_{LO}$(就是想要的 IF),另一個是「” $f_{RF}-f_{LO}$ (the IF we want) and the other the “」$f_{RF}-f_{LO}$(これが欲しい IF)、もう一つは「和頻sum frequency和周波」$f_{RF}+f_{LO}$。” $f_{RF}+f_{LO}$.」$f_{RF}+f_{LO}$ です。

5. 低通濾波器只放差頻過去。5. The low-pass filter lets only the difference frequency through.5. 低域通過フィルタは差周波だけを通します。

電路最後加了一個「The circuit finishes with a “回路の最後には「Low-pass filter(低通濾波器)low-pass filter低域通過フィルタ」— 顧名思義,只讓低頻通過,高頻全部擋掉。原始的 RF、LO、和它們的和頻通通是高頻會被擋住;而「差頻」(IF)通常設計得很低,順利通過。” — as the name says, it passes low frequencies and blocks all the high ones. The original RF and LO and their sum frequency are all high and get blocked; the “difference frequency” (the IF) is usually designed to be low and passes through.」が置かれています — 名前のとおり低い周波数だけを通し、高い周波数はすべて阻止します。元の RF・LO とそれらの和周波はいずれも高周波なので阻止され、「差周波」(IF)は通常低く設計されているため通過します。

所以你在 IF output 端看到的訊號,雖然數學上是 $v_{RF} \times v_{LO}$ 整個乘積,但被濾波器只挑出其中的「差頻」那一塊。這就是為什麼整個電路的So although mathematically the signal at the IF output is the whole product $v_{RF} \times v_{LO}$, the filter has picked out only the “difference frequency” part of it. That is why the したがって IF 出力で見える信号は、数学的には $v_{RF} \times v_{LO}$ という積全体ですが、フィルタによってそのうちの「差周波」の部分だけが取り出されています。これが回路全体の等效行為equivalent behaviour等価な動作等同於一顆「乘法器」。 of the whole circuit is that of a “multiplier”.が「乗算器」と同じになる理由です。


一句話總結:In one sentence:一言でまとめると:二極體因為「不夠誠實」而會把訊號弄歪(產生 $v^2$、$v^3$ 等項),代數上 $(v_{RF}+v_{LO})^2$ 自動跑出 $2 v_{RF} v_{LO}$ 這個「乘積」,乘積在頻譜上等於「差頻 + 和頻」,再讓低通濾波器把和頻丟掉 — 剩下來的就是 RF 和 LO 相乘所產生的差頻 IF。整個過程沒有任何「特別設計的乘法器晶片」,只有「把訊號弄歪 + 濾掉不要的頻率」這兩個動作而已。 Because a diode is “not honest enough” it distorts the signal (producing $v^2$, $v^3$ and so on); algebraically $(v_{RF}+v_{LO})^2$ automatically yields the “product” $2 v_{RF} v_{LO}$; in the spectrum that product is “the difference plus the sum frequency”; and a low-pass filter then throws the sum away — leaving the difference-frequency IF produced by multiplying RF and LO. Nowhere in the process is there a “specially designed multiplier chip”, only the two actions “distort the signal” and “filter out what you do not want”.ダイオードは「正直すぎない」ために信号を歪ませ($v^2$、$v^3$ などの項を生じ)、代数的には $(v_{RF}+v_{LO})^2$ から自動的に $2 v_{RF} v_{LO}$ という「積」が出てきます。この積はスペクトル上では「差周波 + 和周波」に等しく、低域通過フィルタで和周波を捨てれば — 残るのは RF と LO の乗算で生じた差周波の IF です。この過程のどこにも「特別に設計された乗算器チップ」はなく、「信号を歪ませる」と「不要な周波数を濾し取る」という二つの動作があるだけです。

類比:An analogy:類比:你不用「乘法器」也能算 $a \times b$ — 只要知道 $(a+b)^2 - (a-b)^2 = 4ab$,把「先平方再相減」這兩個簡單動作組合起來就能做出乘法。混頻器就是用「先讓二極體做平方、再讓濾波器選頻率」這兩個簡單動作,組合出「RF × LO」這個本來看起來很神奇的乘法效果。 You can compute $a \times b$ without a “multiplier” — just know that $(a+b)^2 - (a-b)^2 = 4ab$ and combine the two simple actions “square, then subtract”. A mixer uses the two simple actions “let the diode square, then let the filter select a frequency” to compose the seemingly magical “RF × LO” multiplication.「乗算器」がなくても $a \times b$ は計算できます — $(a+b)^2 - (a-b)^2 = 4ab$ を知っていれば、「まず平方し、次に引く」という二つの簡単な動作を組み合わせるだけで乗算ができます。ミキサも「まずダイオードに平方させ、次にフィルタに周波数を選ばせる」という二つの簡単な動作で、一見不思議な「RF × LO」の乗算を作り出しています。

What is the "RF choke"?(RF choke / 射頻阻流圈 的作用)What is the "RF choke"?「RF choke」とは何か?(RF チョーク/高周波阻止コイルの働き)

An RF choke is an inductor sized so its reactance is large at the RF/LO frequency but small at DC:

$$X_L = \omega L \;\gg\; Z_0 \quad\text{at RF / LO}, \qquad X_L \to 0 \quad\text{at DC}$$

So it acts like an open circuit to RF and a short to DC — exactly what a bias-feed line needs.

In the slide-22 schematic the choke sits between the $+V_{DC}$ supply and the diode's anode node. Two jobs:

  • Pass the DC bias current ($I_{DC}$) to the diode unimpeded — sets the operating point on the I–V curve where $g'_{DC}$ is large enough for good conversion.
  • Block RF / LO from leaking up into the supply, which would otherwise be a low-impedance ground path that shorts the signal at the diode.

Combined with a series DC-blocking capacitor (the "DC block" you see on the RF and IF sides — which does the opposite: passes RF, blocks DC), the choke forms a bias tee: a three-port network that separates the DC bias path from the RF/IF signal path without disturbing either.

Rule of thumb: at $f = 1$ GHz a $100$ nH choke gives $|X_L| \approx 628\;\Omega \gg 50\;\Omega$, while at DC its $0\,\Omega$ resistance lets the bias current through freely.


中文版(射頻阻流圈 RF choke)The RF chokeRF チョーク

RF choke(射頻阻流圈)就是一個電感,其電感值 $L$ 經過挑選,使其在 RF/LO 頻率下的阻抗很大,但在 DC 時幾乎為零:is simply an inductor whose inductance $L$ is chosen so that its impedance at RF/LO frequencies is large while at DC it is essentially zero:(高周波チョーク)とは単なるインダクタで、そのインダクタンス $L$ は RF/LO 周波数でのインピーダンスが大きく、DC ではほぼ零になるように選ばれます:

$$X_L = \omega L \;\gg\; Z_0 \quad\text{(在 RF / LO)}, \qquad X_L \to 0 \quad\text{(在 DC)}$$ $$X_L = \omega L \;\gg\; Z_0 \quad\text{(at RF / LO)}, \qquad X_L \to 0 \quad\text{(at DC)}$$ $$X_L = \omega L \;\gg\; Z_0 \quad\text{(at RF / LO)}, \qquad X_L \to 0 \quad\text{(at DC)}$$

因此它對 RF 而言像是 So to RF it looks like an したがって RF に対してはopen circuit(開路)open circuit開放(open circuit),對 DC 而言像是 and to DC like a のように見え、DC に対してはshort(短路)short短絡(short) — 這正是 bias-feed(偏壓饋入線)所需要的特性。 — exactly the property a bias feed requires.のように見えます — これはまさにバイアス供給に必要な特性です。

在 slide 22 的電路中,RF choke 位於 $+V_{DC}$ 供電端與二極體陽極之間,扮演兩個角色:In the slide 22 circuit, the RF choke sits between the $+V_{DC}$ supply and the diode anode, playing two roles:slide 22 の回路では、RF チョークは $+V_{DC}$ の電源とダイオードのアノードの間に位置し、二つの役割を果たします:

  • 讓 DC bias current 通過Letting the DC bias current throughDC バイアス電流を通す:$I_{DC}$ 不受阻擾地流到二極體,設定其工作點(讓 $g'_{DC}$ 足夠大以獲得良好的 conversion gain)。: $I_{DC}$ flows unimpeded to the diode, setting its operating point (making $g'_{DC}$ large enough for good conversion gain).:$I_{DC}$ は妨げられずにダイオードへ流れ、その動作点を設定します(良好な変換利得を得るために $g'_{DC}$ を十分大きくします)。
  • 阻擋 RF / LO 訊號Blocking the RF / LO signalsRF / LO 信号を阻止する:避免射頻訊號漏到電源供應端 — 否則電源會變成一個低阻抗對地路徑,把訊號在二極體節點短路掉。: preventing RF from leaking into the power supply — otherwise the supply would become a low-impedance path to ground, shorting the signal out at the diode node.:高周波が電源側へ漏れるのを防ぎます — さもないと電源が低インピーダンスの接地経路となり、ダイオードのノードで信号を短絡してしまいます。

配合串聯的 Combined with the series 直列に入れた DC-blocking capacitor(DC 阻隔電容,作用相反:通 RF、阻 DC),choke + cap 就形成所謂的 (the DC-blocking capacitor, which does the opposite: passing RF and blocking DC), the choke and capacitor form what is called a (DC 阻止コンデンサ。働きは逆で、RF を通し DC を阻止します)と組み合わせると、チョークとコンデンサはいわゆる bias tee(偏壓 T 形網路)bias teeバイアスティー(bias tee),可將 DC bias 路徑與 RF/IF 訊號路徑互相獨立而不互相干擾。, which keeps the DC bias path and the RF/IF signal path independent of each other.を形成し、DC バイアス経路と RF/IF 信号経路を互いに独立に保ちます。

數字直覺:A feel for the numbers:数値の感覚:在 $f = 1$ GHz 下,$L = 100$ nH 的 choke 提供 $|X_L| \approx 628\;\Omega \gg 50\;\Omega$(系統參考阻抗)— 對 RF 等同於開路;而 DC 時其等效電阻幾乎為 $0\,\Omega$,bias current 暢通無阻。 at $f = 1$ GHz an $L = 100$ nH choke presents $|X_L| \approx 628\;\Omega \gg 50\;\Omega$ (the system reference impedance) — an open circuit as far as RF is concerned; while at DC its equivalent resistance is essentially $0\,\Omega$ and the bias current flows freely.$f = 1$ GHz では、$L = 100$ nH のチョークは $|X_L| \approx 628\;\Omega \gg 50\;\Omega$(システム基準インピーダンス)を示します — RF から見れば開放です。一方 DC では等価抵抗はほぼ $0\,\Omega$ であり、バイアス電流は妨げなく流れます。

What is the "Diplexing coupler"?(Diplexer / 雙工器 的作用)What is the "Diplexing coupler"?「Diplexing coupler」とは何か?(ダイプレクサ/双方向合波器の働き)

A diplexer(雙工器,又稱雙頻合路器)A diplexerダイプレクサ是一個三埠的頻率選擇網路,作用是把兩個位於不同頻段的訊號 is a three-port frequency-selective network whose job is to は三ポートの周波数選択ネットワークで、異なる周波数帯にある二つの信号を合併(combine)combine合成(combine) 到一條共用線上,或反過來把共用線上的訊號 two signals in different frequency bands onto a single shared line, or conversely to して一本の共用線に載せる、あるいは逆に共用線上の信号を分離(split)split分離(split) 到兩個輸出埠。在單端二極體混頻器中,它把 RF 與 LO 同時送進同一個 diode,又能避免兩者在輸入端互相干擾。 the signals on a shared line into two output ports. In a single-ended diode mixer it delivers the RF and the LO to the same diode simultaneously while keeping the two from interfering at the input.して二つの出力ポートへ振り分ける働きをします。シングルエンドのダイオードミキサでは、RF と LO を同時に同じダイオードへ送りつつ、両者が入力端で干渉しないようにします。

→ 點此跳到 slide-22 schematic 並讓 Diplexing coupler 區塊高亮閃爍→ Click to jump to the slide-22 schematic and make the Diplexing coupler block flash→ クリックして slide-22 の回路図へ移動し、Diplexing coupler のブロックを点滅表示

內部結構:Internal structure:内部構造:兩個互補的濾波器共用一個 common port。 two complementary filters sharing a common port.二つの相補的なフィルタが一つの common port を共有します。

common port (to diode) high-pass filter passes f_RF low-pass filter passes f_LO / f_IF port A (RF band) port B (LO / IF band) f_RF + f_LO f_RF f_LO / f_IF
Diplexer 內部結構 — 一個 common port 透過 high-pass filter 接 port A(RF 頻段),透過 low-pass filter 接 port B(LO / IF 頻段)。每個頻段只走自己的濾波器路徑,因此 insertion loss 接近 0 dB。The internal structure of a diplexer — a common port connected to port A (the RF band) through a high-pass filter and to port B (the LO / IF band) through a low-pass filter. Each band travels only its own filter path, so the insertion loss is close to 0 dB.ダイプレクサの内部構造 — 一つの common port が高域通過フィルタを介して port A(RF 帯域)に、低域通過フィルタを介して port B(LO / IF 帯域)に接続されます。各帯域は自分のフィルタ経路だけを通るため、挿入損失は 0 dB に近くなります。
  • $f_{RF}$ 從 RF 埠進入時,遇到 LO 側的 low-pass filter(在 $f_{RF}$ 為阻帶)會被反射回去,因此只能從 common port 輸出。When $f_{RF}$ enters at the RF port it meets the LO side's low-pass filter (a stopband at $f_{RF}$) and is reflected, so it can only leave through the common port.$f_{RF}$ が RF ポートから入ると、LO 側の低域通過フィルタ($f_{RF}$ では阻止帯域)に出会って反射されるため、common port からしか出られません。
  • $f_{LO}$ 同理,只會從 common port 流出,不會洩漏到 RF 埠。$f_{LO}$ likewise can only flow out of the common port and does not leak into the RF port.$f_{LO}$ も同様に common port からのみ流れ出し、RF ポートへは漏れません。
  • 在 common port 兩個頻段共存於同一條線上,一同驅動 diode。At the common port the two bands coexist on the same line and drive the diode together.common port では二つの帯域が同じ線上に共存し、一緒にダイオードを駆動します。

為什麼不用 power combiner?Why not use a power combiner?なぜ power combiner を使わないのか?

合波器類型Type of combiner合波器の種類 每路 insertion lossInsertion loss per path経路あたりの挿入損失 RF↔LO isolation 原理Principle原理
Wilkinson / 電阻式 combinerWilkinson / resistive combinerWilkinson / 抵抗式コンバイナ 至少 3 dBat least 3 dB少なくとも 3 dB(功率對分)(the power is split in two)(電力が二分される) 中等moderate中程度 純粹相加 → pure addition → 純粋な加算 → 一半功率half the power電力の半分耗散在另一埠 is dissipated in the other portが他方のポートで消費される
Diplexer(雙工器)Diplexerダイプレクサ 幾乎 0 dBalmost 0 dBほぼ 0 dB(每頻段只走自己的濾波器)(each band travels only its own filter)(各帯域は自分のフィルタだけを通る) 高(由濾波器 rejection 決定)high (set by the filter rejection)高い(フィルタの除去量で決まる) 利用頻率選擇性合波using frequency-selective combining周波数選択的な合波を利用

混頻器使用 diplexer 而非 combiner 的原因:Why a mixer uses a diplexer rather than a combiner: ミキサが combiner ではなく diplexer を使う理由:不能浪費 3 dB 的 LO driveyou cannot afford to waste 3 dB of LO drive3 dB の LO ドライブを無駄にできないため — 否則 conversion gain 下降、NF 變差。 — it would lower the conversion gain and worsen the NF. — 変換利得が下がり、NF が悪化してしまいます。


在 slide 22 的位置:Its place in slide 22:slide 22 での位置:

  • RF input 與 LO input 進入 diplexer 後,the RF input and the LO input enter the diplexer and are RF input と LO input が diplexer に入った後、合併到同一條訊號線merged onto a single signal line同一の信号線に統合され送往 diode。 to the diode.diode へ送られます。
  • 下游緊接 Immediately downstream is the 下流にすぐ続くのが DC block(串聯電容):通過合併後的 RF/LO,但阻擋 DC bias 經此回灌到 diplexer。(a series capacitor): it passes the combined RF/LO but stops the DC bias from feeding back into the diplexer.(直列コンデンサ):合成後の RF/LO は通しますが、DC bias がここを経て diplexer へ逆流するのを阻止します。
  • 再之後是 After that, the その後に続くのが RF choke 把 DC bias 送到 diode anode(如本章前面說明)。 delivers the DC bias to the diode anode (as explained earlier in this chapter). が DC bias を diode のアノードへ送ります(本章の前半で説明したとおり)。

實作方式:How it is built:実装方法:常用 lumped 元件(電感 + 電容)或 distributed 結構(如四分之一波長 stub),在 RF/LO 頻段差異夠大時容易設計;若兩者頻率太接近,diplexer 的濾波器斜率就必須做得很陡,這也是混頻器設計時 IF 不能太低的原因之一。 usually with lumped elements (inductors and capacitors) or a distributed structure (such as quarter-wave stubs), which is easy to design when the RF and LO bands are sufficiently far apart; if the two frequencies are too close, the diplexer's filter skirts must be very steep — one of the reasons the IF cannot be too low in a mixer design.通常は lumped 素子(インダクタとコンデンサ)または分布定数構造(1/4 波長スタブなど)を用います。RF と LO の帯域が十分に離れていれば設計は容易ですが、二つの周波数が近すぎる場合は diplexer のフィルタのスカートを非常に急峻にしなければならず、これがミキサ設計で IF を低くしすぎられない理由の一つです。

4.2.5 Single-Balanced Diode Mixer

Two single-ended mixers combined via a 3-dB hybrid. The choice of hybrid (90° vs 180°) determines the trade-off between input match and isolation.

Single-balanced diode mixer — RF and LO summed in 3-dB hybrid, two diodes, LPF
Lecture slide 23 — single-balanced diode mixer block diagram (3-dB hybrid + two diodes + LPF).
Single-balanced diode mixer
  • A single-balanced diode mixer combines two identical single-ended mixers with a 3-dB 90° or 180° hybrid.
  • 90° hybrid → better input match at the RF port.
  • 180° hybrid → better RF-to-LO isolation; rejects even-order LO harmonics.
  • Single-balanced mixers generally reject LO AM noise.
  • Require higher LO power than single-ended (must drive two diodes).
  • Diode-input voltages: $v_1 = \text{RF}(\Delta 90°) + \text{LO}(\Delta 180°)$, $v_2 = \dots$ (per the hybrid's S-matrix).
解釋:什麼是「LO AM 噪聲」?為何單平衡混頻器能抑制它?Explanation: what is “LO AM noise”, and why does a single-balanced mixer suppress it?解説:「LO AM 雑音」とは何か?なぜ単平衡ミキサはそれを抑制できるのか?

AM = Amplitude Modulation(振幅調變)AM = amplitude modulationAM = Amplitude Modulation(振幅変調) — 指訊號的 — the component in which the signal's — 信号の振幅(包絡)amplitude (its envelope)振幅(包絡線)隨時間變化的成分。 varies with time.が時間とともに変化する成分です。

在本振(LO)的脈絡下:In the context of a local oscillator:局部発振器(LO)の文脈では:

  • 理想的 LO 是一個純單頻訊號:An ideal LO is a pure single-frequency signal: 理想的な LO は純粋な単一周波数信号です:振幅固定、頻率固定fixed amplitude and fixed frequency振幅一定・周波数一定.
  • 實際的 LO 會帶有兩種雜訊邊帶: A real LO carries two kinds of noise sideband: 実際の LO は 2 種類の雑音側波帯を伴います:
    • AM 噪聲(振幅雜訊)AM noise (amplitude noise)AM 雑音(振幅雑音):包絡上的小幅抖動 → $v_{LO}(t) = [V_0 + n_{AM}(t)]\cos(\omega_{LO}t)$: small jitter on the envelope → $v_{LO}(t) = [V_0 + n_{AM}(t)]\cos(\omega_{LO}t)$:包絡線上の微小な揺らぎ → $v_{LO}(t) = [V_0 + n_{AM}(t)]\cos(\omega_{LO}t)$
    • PM 噪聲(相位雜訊)PM noise (phase noise)PM 雑音(位相雑音):相位/頻率抖動 → $v_{LO}(t) = V_0\cos[\omega_{LO}t + \phi_n(t)]$: phase/frequency jitter → $v_{LO}(t) = V_0\cos[\omega_{LO}t + \phi_n(t)]$:位相/周波数の揺らぎ → $v_{LO}(t) = V_0\cos[\omega_{LO}t + \phi_n(t)]$

為什麼混頻器能抑制 AM 噪聲?Why does a mixer suppress AM noise?なぜミキサは AM 雑音を抑制できるのか? 切換式混頻器(二極體環、Gilbert cell)的 LO 通常被驅動到 In a switching mixer (a diode ring, a Gilbert cell) the LO is usually driven into スイッチング型ミキサ(ダイオードリング、Gilbert cell)では、LO は通常飽和切換saturated switching飽和スイッチング狀態 — LO 此時的角色是「開關」而非線性乘法器。一旦 LO 的振幅大到足以把開關完全切到飽和區,振幅上的小幅波動就— its role there is as a switch, not a linear multiplier. Once the LO amplitude is large enough to drive the switch fully into saturation, small fluctuations in amplitude 状態まで駆動されます — このとき LO の役割は「スイッチ」であり、線形乗算器ではありません。LO の振幅がスイッチを完全に飽和させるほど大きくなると、振幅の微小な変動は不會影響切換的瞬間no longer affect the instant of switchingスイッチングの瞬間に影響しなくなり,因此不會轉換到 IF。但是相位雜訊會直接改變切換時刻,所以 , and so are not converted to the IF. Phase noise, however, directly alters the switching instant, so 、そのため IF へ変換されません。しかし位相雑音はスイッチングの時刻を直接変えるため、PM 噪聲會穿透到 IFPM noise passes straight through to the IFPM 雑音はそのまま IF へ通過します.

單平衡拓樸更上一層:The single-balanced topology goes one better:単平衡トポロジーはさらに一歩進みます: 兩個二極體都看到同樣的 LO 包絡 $v_n(t)$,AM 噪聲在兩條路徑上是 both diodes see the same LO envelope $v_n(t)$, so the AM noise is a 両方のダイオードが同じ LO 包絡線 $v_n(t)$ を見るため、AM 雑音は 2 つの経路で共模(common-mode)common-mode同相(common-mode)訊號;經反向二極體對相減($+a_1$ vs. $-a_1$)後,AM 噪聲產生的乘積只剩 RF/LO 基頻成分(如 $v_n(t)\sin\omega_{LO}t$),全部落在 IF 帶外,被 LPF 濾掉。 signal on the two paths; after subtraction by the anti-parallel diode pair ($+a_1$ vs. $-a_1$), the products generated by the AM noise are left only with RF/LO fundamental components (such as $v_n(t)\sin\omega_{LO}t$), all of which fall outside the IF band and are removed by the LPF.信号となります。逆並列ダイオード対による減算($+a_1$ vs. $-a_1$)の後、AM 雑音が生む積には RF/LO の基本波成分($v_n(t)\sin\omega_{LO}t$ など)しか残らず、いずれも IF 帯域外に落ちて LPF で除去されます。

口訣:A mnemonic:覚え方: 切換式混頻器 → a switching mixer スイッチング型ミキサ → 抑制 AM、放行 PMsuppresses AM and passes PMAM を抑制し、PM を通す;單平衡進一步把殘留 AM 也對消掉。; a single-balanced topology then cancels the residual AM as well.。単平衡トポロジーはさらに残留 AM も相殺します。

解釋:方框中那個「X」是什麼?3 dB 90° hybrid 是用什麼電路實現的?Explanation: what is that “X” in the box? What circuit realizes a 3 dB 90° hybrid?解説:枠の中のあの「X」は何か?3 dB 90° hybrid はどんな回路で実現されるのか?

方框內的「X」只是The “X” inside the box is merely a 枠内の「X」は単なる示意符號schematic symbol図記号,代表四個埠之間的交叉耦合關係, denoting the cross-coupling between the four ports; であり、4 つのポート間の交差結合を表すもので、並不是真的有一條斜線把訊號交叉走there is no actual diagonal line carrying signals across実際に信号が斜めの線を通って交差しているわけではありません。實際的 3 dB 90° hybrid(正交耦合器,quadrature coupler)有幾種常見的物理實現方式:. A real 3 dB 90° hybrid (a quadrature coupler) has several common physical realizations:。実際の 3 dB 90° hybrid(直交結合器、quadrature coupler)には、いくつかの一般的な物理的実現方法があります:

1 2 3 4 Z₀/√2 , λ/4 Z₀/√2 , λ/4 Z₀ λ/4 Z₀ λ/4 Branch-line coupler 支線耦合器(最常見,Pozar 課本標配)Branch-line coupler (the commonest, standard fare in Pozar)ブランチライン結合器(最も一般的、Pozar の定番)
四條 λ/4 微帶線組成方環;水平串聯臂 $Z_0/\sqrt{2}\approx35.4\,\Omega$,垂直分支臂 $Z_0=50\,\Omega$。窄頻(~10–20% BW),印製板上最容易做。A square ring of four λ/4 microstrip lines; the horizontal series arms are $Z_0/\sqrt{2}\approx35.4\,\Omega$ and the vertical branch arms $Z_0=50\,\Omega$. Narrowband (~10–20% BW) and the easiest to make on a printed board.4 本の λ/4 マイクロストリップ線路で正方形のリングを構成し、水平の直列アームは $Z_0/\sqrt{2}\approx35.4\,\Omega$、垂直の分岐アームは $Z_0=50\,\Omega$ です。狭帯域(~10–20% BW)で、プリント基板上で最も作りやすい構造です。
1 2 3 4 虛線 = 跨接 bond wiresDashed lines = crossing bond wires破線 = 交差する bond wire Lange coupler 交指型耦合器(寬頻、MMIC 常用)Lange coupler (broadband, common in MMICs)Lange 結合器(広帯域、MMIC で一般的)
多條平行微帶指狀線,用 bond wire 交錯連接 → 在窄間距下達到 3 dB 緊耦合。Several parallel microstrip fingers interleaved and connected by bond wires → achieving tight 3 dB coupling at narrow spacings.複数の平行なマイクロストリップ指状線を bond wire で交互に接続 → 狭い間隔で 3 dB の密結合を実現します。頻寬可達一個 octave 以上The bandwidth can exceed an octave帯域幅は 1 octave 以上に達することもあり,MMIC/單晶 IC 中最常用。, making it the most common choice in MMICs and monolithic ICs.、MMIC やモノリシック IC で最もよく使われる選択肢です。
L L C C C C L L 1 2 3 4 Lumped-element LC 集總元件等效(低頻/CMOS IC)Lumped-element equivalent (low frequency / CMOS IC)集中定数素子による等価回路(低周波 / CMOS IC)
用電感與電容組成等效 π 或 T 網路,模擬 λ/4 線的相位行為。在 $<5$ GHz 的 CMOS RFIC 中常用,因為 λ/4 微帶線太大而不切實際。Inductors and capacitors forming an equivalent π or T network that mimics the phase behaviour of a λ/4 line. Common in CMOS RFICs below $<5$ GHz, where a λ/4 microstrip line would be impractically large.インダクタとコンデンサで等価な π 型または T 型回路網を構成し、λ/4 線路の位相特性を模擬します。$<5$ GHz 以下の CMOS RFIC でよく用いられます。λ/4 マイクロストリップ線路では大きすぎて現実的でないためです。

其他常見實現:Other common realizations:その他の一般的な実現方法:

  • 耦合線方向耦合器(coupled-line directional coupler)Coupled-line directional coupler結合線路方向性結合器(coupled-line directional coupler):兩條平行傳輸線在 λ/4 長度內互相耦合。在微帶線上很難做到 3 dB 緊耦合,所以常做成 : two parallel transmission lines coupled over a λ/4 length. Tight 3 dB coupling is hard to achieve in microstrip, so it is often built as a :2 本の平行伝送線路が λ/4 の長さで互いに結合します。マイクロストリップでは 3 dB の密結合が難しいため、多くの場合 broadside-coupled stripline(上下兩層)或上面的 (two stacked layers) or as the (上下 2 層)として、あるいは上記の Lange coupler.
  • RC-CR polyphase filter:CMOS RFIC 產生本振正交(I/Q)訊號時的常見作法 — 用 RC 與 CR 兩個分壓網路天生產生 ±45° 相位差,組合即得 90° 差。: the usual way of generating quadrature (I/Q) local-oscillator signals in a CMOS RFIC — an RC and a CR divider naturally produce ±45° of phase shift, and combining them gives a 90° difference.:CMOS RFIC で局部発振器の直交(I/Q)信号を生成する際の一般的な方法です — RC と CR の分圧回路がそれぞれ ±45° の位相シフトを自然に生じ、両者を組み合わせると 90° の差が得られます。
  • 波導實現Waveguide realizations導波管による実現:Riblet short-slot 耦合器、branch-guide 耦合器(毫米波/雷達系統)。: the Riblet short-slot coupler and the branch-guide coupler (in millimetre-wave and radar systems).:Riblet short-slot 結合器、branch-guide 結合器(ミリ波・レーダーシステム)。Magic-T 是 180° hybrid,不是 90°。A magic-T is a 180° hybrid, not a 90° one.Magic-T は 180° hybrid であり、90° ではありません。

本課重點:The point for this course:本講義での要点: Pozar 課本與本投影片提到的「3 dB 90° hybrid」幾乎都默認為 the “3 dB 90° hybrid” mentioned in Pozar and in these slides is almost always taken to be the Pozar と本スライドで言及される「3 dB 90° hybrid」は、ほぼ常に branch-line coupler(左圖)。它的 S 矩陣就是前面推導用的那個 — 而電路的「X」符號只是抽象示意,實際內部是四條 λ/4 微帶線組成的方環。(left-hand figure). Its S-matrix is the one used in the derivation above — and the circuit's “X” symbol is purely an abstraction; inside it is really a square ring of four λ/4 microstrip lines.(左図)を指します。その S 行列が前述の導出で使ったものです — 回路の「X」記号は純粋な抽象表現であり、内部は実際には 4 本の λ/4 マイクロストリップ線路からなる正方形リングです。

S-matrix(無損、匹配,埠號照原圖編號)S-matrix (lossless, matched, ports numbered as in the original figure)S-matrix(無損失・整合、ポート番号は原図に準拠)

$$[S] = \frac{-1}{\sqrt{2}}\begin{bmatrix} 0 & j & 1 & 0 \\ j & 0 & 0 & 1 \\ 1 & 0 & 0 & j \\ 0 & 1 & j & 0 \end{bmatrix}$$

由 port 1 進入的訊號 → port 2 得到 $-j/\sqrt{2}$($-90°$、A signal entering at port 1 → port 2 receives $-j/\sqrt{2}$ ($-90°$, port 1 から入った信号 → port 2 は $-j/\sqrt{2}$ を受け取り($-90°$、半功率half power半分の電力)、port 3 得到 $-1/\sqrt{2}$($-180°$、), port 3 receives $-1/\sqrt{2}$ ($-180°$, )、port 3 は $-1/\sqrt{2}$ を受け取り($-180°$、半功率half power半分の電力)、port 4 隔離。), and port 4 is isolated.)、port 4 はアイソレートされます。

90° Hybrid Version

The S-matrix for the 90° hybrid routes RF and LO to diodes with the phases:

$$v_1 = \frac{1}{\sqrt{2}}[-V_{RF}\sin\omega_{RF}t - V_{LO}\cos\omega_{LO}t], \quad v_2 = \frac{1}{\sqrt{2}}[-V_{RF}\cos\omega_{RF}t - V_{LO}\sin\omega_{LO}t]$$
90° Hybrid Key Property
Reflected signals from mismatched diodes cancel at the RF port $\Rightarrow$ perfect input match at RF even with diode mismatch. Isolation between RF and LO depends on diode matching.

📖 逐元素詳細解析Element-by-element walkthrough要素ごとの詳細解説(含 branch-line 結構、相位推導、$V_\Gamma^1, V_\Gamma^2$ 是怎麼來的、$i_o$ 三角恆等式逐步展開等): (covering the branch-line structure, the phase derivation, where $V_\Gamma^1, V_\Gamma^2$ come from, the step-by-step trigonometric expansion of $i_o$, and more): (branch-line 構造、位相の導出、$V_\Gamma^1, V_\Gamma^2$ の由来、$i_o$ の三角関数展開の各ステップなどを含む):slide26_90deg_single_balanced_mixer_explained.html →

完整推導:90° hybrid 單平衡混頻器的 IF 輸出與 LO AM 噪聲抑制Full derivation: the IF output of a 90° hybrid single-balanced mixer and its suppression of LO AM noise完全な導出:90° hybrid 単平衡ミキサの IF 出力と LO AM 雑音の抑制

Step 1 — RF and LO inputs.

$$v_{RF}(t) = V_{RF}\cos\omega_{RF}t, \qquad v_{LO}(t) = V_{LO}\cos\omega_{LO}t$$

Step 2 — 90° hybrid scattering matrix. Ports: 1 = RF in, 2 = diode 1, 3 = diode 2, 4 = LO in.

$$[S] = \frac{-1}{\sqrt{2}}\begin{bmatrix} 0 & j & 1 & 0 \\ j & 0 & 0 & 1 \\ 1 & 0 & 0 & j \\ 0 & 1 & j & 0 \end{bmatrix}$$

Step 3 — voltages across the two diodes. Diode 1 sees RF shifted by $-90°$ and LO by $-180°$; diode 2 sees RF shifted by $-180°$ and LO by $-90°$:

$$\begin{aligned} v_1(t) &= \tfrac{1}{\sqrt{2}}\bigl[V_{RF}\cos(\omega_{RF}t-90°) + V_{LO}\cos(\omega_{LO}t-180°)\bigr] \\ &= \tfrac{1}{\sqrt{2}}\bigl[V_{RF}\sin\omega_{RF}t - V_{LO}\cos\omega_{LO}t\bigr] \\[6pt] v_2(t) &= \tfrac{1}{\sqrt{2}}\bigl[V_{RF}\cos(\omega_{RF}t-180°) + V_{LO}\cos(\omega_{LO}t-90°)\bigr] \\ &= \tfrac{1}{\sqrt{2}}\bigl[-V_{RF}\cos\omega_{RF}t + V_{LO}\sin\omega_{LO}t\bigr] \end{aligned}$$

Step 4 — add LO AM noise. Because both diodes carry an LO component of equal amplitude, AM noise $v_n(t)$ rides on both equally. Keep only the quadratic ($a_1$) term of the diode I–V (the linear small-signal nonlinear model). The two diodes are reversed (anti-parallel pair), so diode 2 contributes with $-a_1$:

$$\begin{aligned} i_o(t) &= i_1(t) + i_2(t) \\ &\approx a_1\bigl[v_1(t)+v_n(t)\bigr]^2 + (-a_1)\bigl[v_2(t)+v_n(t)\bigr]^2 \\ &\approx a_1\bigl[v_1^2(t) - v_2^2(t)\bigr] + 2a_1 v_n(t)\bigl[v_1(t)-v_2(t)\bigr] \quad{\scriptstyle\textcolor{gray}{\text{(common }v_n^2\text{ cancels)}}} \end{aligned}$$

Step 5 — expand $v_1^2 - v_2^2$. Using $\sin^2 x = \tfrac{1}{2}(1-\cos 2x)$, $\cos^2 x = \tfrac{1}{2}(1+\cos 2x)$, and $\sin A\cos B = \tfrac{1}{2}[\sin(A+B)+\sin(A-B)]$:

$$\begin{aligned} i_o(t) &\approx \frac{-a_1}{2}\Bigl[V_{RF}^{\,2}\cos 2\omega_{RF}t + \underbrace{2V_{RF}V_{LO}\sin(\omega_{RF}-\omega_{LO})t}_{\text{wanted IF term}} - V_{LO}^{\,2}\cos 2\omega_{LO}t\Bigr] \\ &\quad + \sqrt{2}\,a_1 v_n(t)\bigl[V_{RF}(\sin\omega_{RF}t+\cos\omega_{RF}t)-V_{LO}(\cos\omega_{LO}t+\sin\omega_{LO}t)\bigr] \end{aligned}$$

Step 6 — apply the low-pass filter. The $2\omega_{RF}$, $2\omega_{LO}$, and the $v_n\cdot(\text{RF/LO fundamental})$ products all fall outside the IF band and are rejected. The desired IF tone survives:

$$i_{IF}(t) \approx -a_1 V_{RF}V_{LO}\sin(\omega_{RF}-\omega_{LO})t = -\tfrac{1}{2}g'_{DC}\,V_{RF}V_{LO}\sin\omega_{IF}t$$

Conclusion. Two important properties fall out of the same algebra:

  • The IF tone has twice the amplitude of a single-ended mixer (anti-phase pair adds constructively), yet only a $\sqrt{2}$ LO voltage per diode is needed.
  • The AM-noise term $2a_1 v_n(t)[v_1-v_2]$ contains only RF/LO fundamentals → it is filtered out by the LPF. The 90° single-balanced topology thus rejects LO AM noise that would otherwise translate to the IF band.

中文摘要Summaryまとめ

在 90° hybrid 拓樸下,兩二極體接成 In the 90° hybrid topology the two diodes are connected as an 90° hybrid トポロジーでは、2 つのダイオードは 反向對anti-parallel pair逆並列対,將 RF/LO 分別以 $-90°/-180°$ 與 $-180°/-90°$ 相位差注入兩個二極體。經過二次非線性 $a_1$ 項並相加(含反向極性 $-a_1$)後,二倍頻 $2\omega_{RF}$、$2\omega_{LO}$ 與 LO AM 噪聲 $v_n(t)$ 與 RF/LO 基頻的乘積都落在 LPF 帶外,被濾除掉;只剩下 $-\tfrac{1}{2}g'_{DC}V_{RF}V_{LO}\sin\omega_{IF}t$ 這個 IF 訊號。, injecting the RF and LO into them with phase differences of $-90°/-180°$ and $-180°/-90°$ respectively. After passing through the second-order nonlinear $a_1$ term and being summed (with the reversed polarity $-a_1$), the second harmonics $2\omega_{RF}$ and $2\omega_{LO}$, together with the products of the LO AM noise $v_n(t)$ with the RF/LO fundamentals, all fall outside the LPF passband and are filtered out; what remains is just the IF signal $-\tfrac{1}{2}g'_{DC}V_{RF}V_{LO}\sin\omega_{IF}t$.として接続され、RF と LO をそれぞれ $-90°/-180°$ と $-180°/-90°$ の位相差で注入します。2 次の非線形 $a_1$ 項を通り、(反転極性 $-a_1$ を含めて)加算された後、2 倍波 $2\omega_{RF}$ と $2\omega_{LO}$、および LO AM 雑音 $v_n(t)$ と RF/LO 基本波との積はすべて LPF の通過帯域外に落ちて除去され、残るのは IF 信号 $-\tfrac{1}{2}g'_{DC}V_{RF}V_{LO}\sin\omega_{IF}t$ だけです。LO 的 AM 噪聲因此自動被抑制The LO's AM noise is therefore suppressed automaticallyしたがって LO の AM 雑音は自動的に抑制されます — 這是單平衡混頻器相對於單端混頻器最重要的優點之一。 — one of the most important advantages of a single-balanced mixer over a single-ended one. — これはシングルエンドミキサに対する単平衡ミキサの最も重要な利点の一つです。

180° Hybrid Version

RF drives the $\Delta$ port, LO drives the $\Sigma$ port:

$$v_1 = \frac{1}{\sqrt{2}}[-V_{RF}\sin\omega_{RF}t - V_{LO}\sin\omega_{LO}t], \quad v_2 = \frac{1}{\sqrt{2}}[-V_{RF}\sin\omega_{RF}t + V_{LO}\sin\omega_{LO}t]$$
180° Hybrid Key Property
Reflected signals cancel at the LO port $\Rightarrow$ perfect RF-to-LO isolation even with diode mismatch. Input match at RF depends on diode matching. Also rejects even-order LO harmonics.

📖 逐元素詳細解析Element-by-element walkthrough要素ごとの詳細解説(含 rat-race ring 結構、Σ/Δ 相位推導、$V_\Gamma^{RF}=-\Gamma V_{RF}$ 與 $V_\Gamma^{LO}=0$ 完整代數、與 90° 版本的鏡像對照表等): (covering the rat-race ring structure, the Σ/Δ phase derivation, the full algebra of $V_\Gamma^{RF}=-\Gamma V_{RF}$ and $V_\Gamma^{LO}=0$, a mirror-image comparison table against the 90° version, and more): (rat-race ring 構造、Σ/Δ 位相の導出、$V_\Gamma^{RF}=-\Gamma V_{RF}$ と $V_\Gamma^{LO}=0$ の完全な代数、90° 版との鏡像対照表などを含む):slide28_180deg_single_balanced_mixer_explained.html →

Common Properties of Single-Balanced Mixers

  • Both configurations suppress LO AM noise at the IF output (the AM noise from LO appears equally in $v_1$ and $v_2$ and cancels at the summed output)
  • Require higher LO power than single-ended (drive two diodes)
  • Better isolation and match than single-ended mixers

4.2.6 Double-Balanced Diode Mixer (Ring Mixer)

Combines two single-balanced mixers using four diodes in a ring configuration and two hybrids or transformers (baluns). Provides good isolation between all three ports.

Double-balanced diode ring mixer — 180° hybrid + 4-diode ring + LO balun
Lecture slide 29 — double-balanced ring mixer: RF enters a 180° hybrid (yellow, top-left); LO drives a balun (yellow, bottom-right) feeding the four-diode ring. IF is taken from the center tap of the LO balun. Hybrid + balun together force the symmetry that cancels even-order products on all three ports.

🔵 Rat-race coupler 深入解說An in-depth explanation of the rat-race couplerRat-race coupler の詳細解説(拓樸與設計參數、完整 [S]-matrix 與 even-odd mode 推導、$\lambda/8$ / $3\lambda/8$ stub 的互動式 Smith chart、半電路 ABCD 推導): (topology and design parameters, the full [S]-matrix with an even-odd mode derivation, an interactive Smith chart for the $\lambda/8$ / $3\lambda/8$ stubs, and the half-circuit ABCD derivation): (トポロジーと設計パラメータ、完全な [S]-matrix と even-odd mode による導出、$\lambda/8$ / $3\lambda/8$ スタブのインタラクティブな Smith chart、半回路 ABCD 導出):slide29_double_balanced_ring_mixer_interactive.html#howRF →

🎮 互動式動畫解說Interactive animated walkthroughインタラクティブなアニメーション解説(拖曳波形游標、觀察四顆二極體在 LO 正負半週的切換、rat-race 內部 $\lambda/4$ 與 $3\lambda/4$ 結構、balun 中心抽頭如何抑制 LO、含完整 [S]-matrix 與 even-odd mode 推導): (drag the waveform cursor, watch the four diodes commutate over the positive and negative LO half-cycles, see the $\lambda/4$ and $3\lambda/4$ structure inside the rat-race and how the balun centre tap suppresses the LO, with the full [S]-matrix and even-odd mode derivation): (波形カーソルをドラッグし、LO の正負半周期で 4 つのダイオードが切り替わる様子、rat-race 内部の $\lambda/4$ と $3\lambda/4$ の構造、バラン中心タップがどのように LO を抑制するかを確認できます。完全な [S]-matrix と even-odd mode の導出を含む):slide29_double_balanced_ring_mixer_interactive.html →

  • Rejects even-order harmonics and IM products from both RF and LO signals
  • Generally achieves higher IP3 than single-ended or single-balanced mixers
  • Conversion loss is very low
  • Requires higher LO power (two diodes conduct on each half-cycle)
  • Input match at RF port may be less than ideal
Walk-through of slide 29(黃色標示處)Walk-through of slide 29 (the highlighted parts)slide 29 の解説(黄色で示した箇所)
  • 180° hybrid(左上黃框):The 180° hybrid (yellow box, top left):180° hybrid(左上の黄色枠):把 RF 訊號分成 $\Delta$(差)和 $\Sigma$(和)兩路,分別驅動 ring 的兩個對角節點。$\Sigma$ 埠用 $Z_0$ 終端吸收 — 保證 RF 是平衡差動進入 ring。 splits the RF signal into a $\Delta$ (difference) and a $\Sigma$ (sum) path, driving two diagonally opposite nodes of the ring. The $\Sigma$ port is absorbed in a $Z_0$ termination — guaranteeing that the RF enters the ring as a balanced differential signal. RF 信号を $\Delta$(差)と $\Sigma$(和)の 2 経路に分け、リングの対角にある 2 つのノードを駆動します。$\Sigma$ ポートは $Z_0$ 終端で吸収されます — これにより RF が平衡差動信号としてリングに入ることが保証されます。
  • 4-diode ring(右上):The 4-diode ring (top right):4-diode ring(右上):四顆二極體首尾相連成環。LO 在每個半週只導通兩顆對角的 diode,等效為一個由 LO 控制的「機械開關」(switching mixer)。 four diodes connected head to tail in a ring. On each half-cycle the LO turns on only two diagonally opposite diodes, making it equivalent to a “mechanical switch” controlled by the LO (a switching mixer). 4 つのダイオードが首尾を接続してリングを構成します。LO は各半周期で対角の 2 つのダイオードだけを導通させ、LO で制御される「機械的スイッチ」(スイッチング型ミキサ)と等価になります。
  • LO balun(右下黃框):The LO balun (yellow box, bottom right):LO バラン(右下の黄色枠):把單端 LO 訊號轉成「正負相位」的差動驅動進入 ring 的另一對節點;IF 由 balun 中心抽頭取出,並接到地。 converts the single-ended LO signal into a differential “positive and negative phase” drive for the other pair of ring nodes; the IF is taken from the balun centre tap, which is also connected to ground. シングルエンドの LO 信号を「正相・逆相」の差動駆動に変換し、リングのもう一対のノードへ入力します。IF はバランの中心タップから取り出され、そこは同時に接地されています。
  • 三個埠(RF / LO / IF)The three ports (RF / LO / IF) are 3 つのポート(RF / LO / IF)は幾何上完全互相平衡geometrically perfectly balanced against one another幾何学的に完全に互いに平衡しており,所以任何「LO 進、RF 出」或「RF 進、LO 出」的偶次項(自混頻、DC offset、$2f_{LO}$、$f_1 \pm f_2$)都會在 balun / hybrid 的對稱結構中相消。, so any even-order term of the “LO in, RF out” or “RF in, LO out” kind (self-mixing, DC offset, $2f_{LO}$, $f_1 \pm f_2$) cancels in the symmetric structure of the balun and hybrid.、そのため「LO 入力・RF 出力」や「RF 入力・LO 出力」といった偶次項(セルフミキシング、DC offset、$2f_{LO}$、$f_1 \pm f_2$)は、バランやハイブリッドの対称構造の中で相殺されます。

🔄 Balun 運作說明How the balun worksバランの動作解説(center-tapped transformer balun 如何把 LO 分成 +LO 與 −LO、為何同一個中心抽頭同時就是 IF 取出點,以及它為何天生就能抑制 LO): (how a centre-tapped transformer balun splits the LO into +LO and −LO, why that same centre tap is simultaneously the IF take-off point, and why it suppresses the LO inherently): (center-tapped transformer バランが LO を +LO と −LO に分ける仕組み、なぜ同じ中心タップが同時に IF の取り出し点になるのか、そしてなぜ本質的に LO を抑制できるのか):slide29_double_balanced_ring_mixer_interactive.html#howLO →

4.2.7 Image Reject Diode Mixer

Uses a 90° hybrid at the RF input and another at the IF output to separate USB and LSB into distinct IF output ports.

Down-Conversion (Image Separation)

For RF input $v_{RF}(t) = V_U\cos(\omega_{LO}+\omega_{IF})t + V_L\cos(\omega_{LO}-\omega_{IF})t$:

為什麼 $V_U \neq V_L$?(image 與 desired 振幅不同)Why is $V_U \neq V_L$? (the image and the desired signal have different amplitudes)なぜ $V_U \neq V_L$ なのか?(image と desired は振幅が異なる)

$V_U$(在 $\omega_{LO}+\omega_{IF}$)與 $V_L$(在 $\omega_{LO}-\omega_{IF}$)是$V_U$ (at $\omega_{LO}+\omega_{IF}$) and $V_L$ (at $\omega_{LO}-\omega_{IF}$) are $V_U$($\omega_{LO}+\omega_{IF}$)と $V_L$($\omega_{LO}-\omega_{IF}$)は兩個完全獨立的訊號,不是同一個訊號的兩個邊帶two entirely independent signals, not two sidebands of the same signal完全に独立した 2 つの信号であり、同一信号の 2 つの側波帯ではありません。兩者都落在 LO 兩側、相距 $\omega_{IF}$,混波後都降到同一個 IF $\omega_{IF}$;其中一個是. Both sit either side of the LO, $\omega_{IF}$ away, and after mixing both come down to the same IF $\omega_{IF}$; one of them is the 。両者は LO の両側に $\omega_{IF}$ だけ離れて位置し、ミキシング後はどちらも同じ IF $\omega_{IF}$ に落ちます。その一方が想要的訊號 (desired)desired signal所望信号(desired),另一個是 and the other the であり、もう一方が鏡像 (image)imageイメージ(image).

它們來自天線在They come from whatever the antenna happens to receive at これらはアンテナが兩個不同 RF 頻率two different RF frequencies2 つの異なる RF 周波数上各自收到的東西(可能一個是目標電台、另一個是別的干擾源),彼此不相干,所以振幅各自獨立,$V_U \neq V_L$ 是常態。(one perhaps the target station, the other some other interferer), and being unrelated their amplitudes are independent, so $V_U \neq V_L$ is the norm.でそれぞれ受信したものであり(一方は目的の局、他方は別の妨害源かもしれません)、互いに無関係なので振幅も独立しており、$V_U \neq V_L$ が普通です。

故意用不同符號,是為了在數學上把兩者Different symbols are used deliberately so that the two can be あえて異なる記号を使うのは、数学上で両者を分開追蹤tracked separately別々に追跡する:普通混波器無法區分它們(都變成 $\omega_{IF}$)。只有保留 $V_U,V_L$ 標記,下式才能顯示 image-reject mixer 如何靠相位干涉把 USB 與 LSB 送到 in the mathematics: an ordinary mixer cannot distinguish them (both become $\omega_{IF}$). Only by keeping the labels $V_U,V_L$ can the expression below show how an image-reject mixer uses phase interference to send the USB and the LSB to ためです:通常のミキサは両者を区別できません(どちらも $\omega_{IF}$ になります)。$V_U,V_L$ という表記を保ってはじめて、下の式は image-reject mixer が位相干渉によって USB と LSB を不同輸出埠different output ports異なる出力ポート,留下 desired、丟掉 image。若一開始就令 $V_U=V_L$,就看不出「哪一項被分離掉」了。, keeping the desired signal and discarding the image. Had we set $V_U=V_L$ from the start, it would be impossible to see “which term gets separated out”.へ送り、desired を残して image を捨てる様子を示せます。最初から $V_U=V_L$ としてしまうと、「どの項が分離されるのか」が見えなくなります。

IF hybrid outputs — image reject mixer $$v_1(t) = \frac{KV_{LO}V_L}{2}\cos\omega_{IF}t \quad\text{(LSB → port 1)}, \qquad v_2(t) = \frac{KV_{LO}V_U}{2}\sin\omega_{IF}t \quad\text{(USB → port 2)}$$

USB and LSB are routed to separate IF ports — one can be arbitrarily selected as the desired signal while the other (image) is discarded.

Up-Conversion (SSB Modulator)

Operating in reverse, the image reject mixer acts as an SSB modulator: the IF input is up-converted to either USB or LSB depending on which IF port is driven.

Practical Limitation
A good 90° hybrid at the IF frequency is required. When $f_{IF}$ is very low (e.g., a few MHz), fabricating a compact 90° hybrid is difficult — this limits image rejection performance.
Image-reject mixer — 90° RF hybrid + two mixers + two LPF + 90° IF hybrid → LSB/USB
Lecture slide 30 — image-reject mixer routes USB and LSB to separate IF output ports via 90° hybrids at RF and IF.

📖 逐元素詳細解析Element-by-element walkthrough要素ごとの詳細解説(含完整 5-step 小訊號分析、為何 USB 跟 LSB 會分別跑到不同輸出埠的相位干涉機制、實務限制、反向用作 SSB modulator 等): (with the full 5-step small-signal analysis, the phase-interference mechanism by which USB and LSB end up at different output ports, the practical limitations, running it in reverse as an SSB modulator, and more): (完全な 5 ステップの小信号解析、USB と LSB が異なる出力ポートへ分かれる位相干渉の仕組み、実用上の制限、逆方向に SSB modulator として使う方法などを含む):slide30_image_reject_mixer_explained.html →

Slide 30 — text transcript

When $f_{IF}$ is very low, an image-rejection filter cannot achieve the required selectivity; a mixer-based image-reject scheme becomes important.

  • Down-conversion: the image reject diode mixer can isolate the RF LSB and RF USB into separate output IF signals. One can arbitrarily select either the IF LSB or IF USB as the desired signal and reject the other.
  • Up-conversion (SSB modulator): the IF signal is up-converted and delivered to either the RF LSB or RF USB port. Either SSB can be selected for transmission.
  • Image reject diode mixers have reasonably good properties in input match, isolation, conversion loss, and third-order intercept.
  • A practical difficulty is fabricating a good 90° hybrid at the IF frequency, especially when $f_{IF}$ is low.

📝 HW7:Passive Mixer Comparison Table — Discuss📝 HW7: Passive Mixer Comparison Table — Discuss📝 HW7:Passive Mixer Comparison Table — Discuss

問題The question問題

下表整理了五種被動 mixer 拓樸在四個關鍵指標上的相對表現(依據 Pozar slide 32)。請討論:The table below summarizes the relative performance of five passive mixer topologies against four key metrics (following Pozar slide 32). Discuss:下表は 5 種類の受動 mixer トポロジーについて、4 つの主要指標に対する相対的な性能をまとめたものです(Pozar slide 32 に基づく)。次の点を議論してください:

  1. 為什麼每個拓樸在某些指標上比其他拓樸好/差?Why is each topology better or worse than the others on particular metrics?なぜ各トポロジーは特定の指標で他より優れ、あるいは劣るのか?
  2. 哪一種 mixer 整體最好?為什麼實務上仍會選用其他拓樸?Which mixer is best overall? And why are the other topologies still chosen in practice?総合的にどの mixer が最良か?また、なぜ実務では他のトポロジーも選ばれるのか?
  3. 每個拓樸的「賣點」是什麼?什麼時候會被選用?What is each topology's “selling point”? When would it be chosen?各トポロジーの「売り」は何か?どのようなときに選ばれるのか?
Mixer Type Number of Diodes RF Input Match RF-LO Isolation Conversion Loss Third-Order Intercept (IP3)
Single-ended 1 Poor Fair Good Fair
Balanced (90°) 2 Good Poor Good Fair
Balanced (180°) 2 Fair Excellent Good Fair
Double-balanced 4 Poor Excellent Excellent Excellent
Image reject 2 or 4 Good Good Good Good

圖例:Key: 凡例:Poor  Fair  Good  Excellent

討論(逐項分析)Discussion (item by item)議論(項目ごとの分析)

① Single-ended(單端混頻器)① Single-ended① Single-ended(シングルエンドミキサ)
  • Match:PoorMatch: PoorMatch:Poor — RF 與 LO 透過 — the RF and LO are coupled to a single diode through — RF と LO を同一條傳輸線the same transmission line同一の伝送線路耦合到一顆二極體,沒有對稱結構吸收反射 → 反射高、VSWR 差。, with no symmetric structure to absorb reflections → high reflection and poor VSWR.を通して 1 つのダイオードに結合しており、反射を吸収する対称構造がありません → 反射が大きく VSWR が悪くなります。
  • RF-LO Isolation:FairRF-LO isolation: FairRF-LO Isolation:Fair — 只靠 RF 端的 bandpass filter 過濾掉 LO 漏出,沒有結構性對消,隔離度普通(典型 10–20 dB)。 — it relies solely on a bandpass filter at the RF port to keep the LO from leaking out, with no structural cancellation, so the isolation is ordinary (typically 10–20 dB). — RF ポートの帯域通過フィルタだけで LO の漏れを抑えており、構造的な相殺がないためアイソレーションは平凡です(典型的に 10–20 dB)。
  • Conversion Loss:GoodConversion loss: GoodConversion Loss:Good — 單顆二極體就能完成混頻,傳輸效率高(典型 6–8 dB)。 — a single diode performs the mixing, so the transmission efficiency is high (typically 6–8 dB). — 1 つのダイオードで変換を行うため伝送効率が高くなります(典型的に 6–8 dB)。
  • IP3:FairIP3: FairIP3:Fair — 一顆二極體承擔全部訊號,飽和門檻低 → 失真早。 — a single diode carries the whole signal, so the saturation threshold is low and distortion sets in early. — 1 つのダイオードが全信号を担うため飽和のしきい値が低く、歪みが早く現れます。
  • 賣點:Selling point:売り:結構最簡單,零件最少, the simplest structure with the fewest components; the usual architecture for 構造が最も単純で部品点数が最少であり、低成本毫米波 mixerlow-cost millimetre-wave mixers低コストのミリ波 mixer 常用此架構。 . でよく使われる構成です。
② Balanced (90°) — 用 90° hybrid② Balanced (90°) — using a 90° hybrid② Balanced (90°) — 90° hybrid を使用
  • Match:GoodMatch: GoodMatch:Good ✅ — 90° hybrid 的 ✅ — the ✅ — 90° hybrid の反射對消reflection-cancelling反射相殺性質:兩顆 diode 不匹配時反射回 RF 埠的訊號正好對消($V_\Gamma^{RF}=0$)。 property of the 90° hybrid: when the two diodes are mismatched, the signals reflected back to the RF port cancel exactly ($V_\Gamma^{RF}=0$).特性により、2 つの diode が不整合でも RF ポートへ戻る反射信号がちょうど相殺します($V_\Gamma^{RF}=0$)。
  • RF-LO Isolation:PoorRF-LO isolation: PoorRF-LO Isolation:Poor ⚠ — 90° hybrid 把 ⚠ — the 90° hybrid “kicks” all the ⚠ — 90° hybrid は反射reflection反射都「踢」到 LO 埠,所以 LO 埠看到的反射 = $|\Gamma|\cdot V_{RF}$ → RF 直接漏到 LO。 over to the LO port, so what the LO port sees is $|\Gamma|\cdot V_{RF}$ → the RF leaks straight into the LO.をすべて LO ポートへ「蹴り出す」ため、LO ポートが見る反射は $|\Gamma|\cdot V_{RF}$ となり → RF が直接 LO へ漏れます。
  • Conversion Loss:GoodConversion loss: GoodConversion Loss:Good — 兩顆二極體分擔訊號,效率合理(典型 6–7 dB)。 — two diodes share the signal, giving reasonable efficiency (typically 6–7 dB). — 2 つのダイオードが信号を分担し、効率は妥当です(典型的に 6–7 dB)。
  • IP3:FairIP3: FairIP3:Fair — 跟 single-ended 差不多(沒有偶次項對消的好處)。 — much the same as single-ended (there is no benefit of even-order cancellation). — single-ended とほぼ同等です(偶次項相殺の恩恵はありません)。
  • 賣點:Selling point:売り:RF 端永遠完美匹配,適用於對 input return loss 要求嚴格的接收機前端(VSWR < 1.2)。 the RF port is always perfectly matched, making it suitable for receiver front ends with strict input return-loss requirements (VSWR < 1.2).RF ポートが常に完全に整合するため、input return loss の要求が厳しい受信機フロントエンドに適します(VSWR < 1.2)。
③ Balanced (180°) — 用 180° hybrid (rat-race)③ Balanced (180°) — using a 180° hybrid (rat-race)③ Balanced (180°) — 180° hybrid (rat-race) を使用
  • Match:FairMatch: FairMatch:Fair — 180° hybrid 的反射回到 RF 埠 → RF 匹配取決於二極體匹配。 — the 180° hybrid returns the reflection to the RF port → the RF match depends on how well the diodes are matched. — 180° hybrid では反射が RF ポートへ戻るため → RF の整合はダイオードの整合度に依存します。
  • RF-LO Isolation:ExcellentRF-LO isolation: ExcellentRF-LO Isolation:Excellent ✅✅ — 180° hybrid 的 Σ/Δ 結構天生讓 ✅✅ — the Σ/Δ structure of the 180° hybrid inherently means the ✅✅ — 180° hybrid の Σ/Δ 構造により、本質的にLO 端LO portLO ポート看不到 RF 反射($V_\Gamma^{LO}=0$);隔離度 25–40 dB。 never sees the RF reflection ($V_\Gamma^{LO}=0$); isolation is 25–40 dB.は RF の反射を見ません($V_\Gamma^{LO}=0$)。アイソレーションは 25–40 dB です。
  • Conversion Loss:GoodConversion loss: GoodConversion Loss:Good — 同樣兩顆 diode,效率合理。 — again two diodes, with reasonable efficiency. — 同じく 2 つの diode を使い、効率は妥当です。
  • IP3:FairIP3: FairIP3:Fair — 沒有完全的偶次項對消(只有單向)。 — there is no complete even-order cancellation (only in one direction). — 偶次項の完全な相殺はありません(片方向のみ)。
  • 賣點:Selling point:売り:LO 漏到 RF 嚴重的場合(例如 LO 功率很大、本振合成器離 RF 帶很近時),180° 提供天然的 RF-LO 隔離。 where LO leakage into the RF is severe (a very high LO power, say, or a synthesizer close to the RF band), the 180° version provides natural RF-LO isolation.LO が RF へ漏れることが深刻な場合(LO 電力が非常に大きい、シンセサイザが RF 帯域に近いなど)、180° 版は本質的な RF-LO アイソレーションを提供します。
④ Double-balanced — 用 hybrid + balun + 4 顆 diode (ring)④ Double-balanced — a hybrid + a balun + 4 diodes (a ring)④ Double-balanced — hybrid + バラン + 4 つの diode(リング)
  • Match:PoorMatch: PoorMatch:Poor ⚠ — Double-balanced 的對稱性是為了 ⚠ — the symmetry of a double-balanced mixer is designed for ⚠ — Double-balanced の対称性は隔離isolationアイソレーション而設計,不是為了 match。實際 VSWR 通常 ~2:1,需要外加 matching circuit。, not for matching. In practice the VSWR is typically ~2:1 and an external matching circuit is needed.のために設計されており、整合のためではありません。実際の VSWR は典型的に ~2:1 で、外付けの整合回路が必要です。
  • RF-LO Isolation:ExcellentRF-LO isolation: ExcellentRF-LO Isolation:Excellent ✅✅ — 三埠互相隔離 25–50 dB,是所有拓樸中最好的(hybrid + balun + 4 顆 diode 的雙重對稱性)。 ✅✅ — 25–50 dB of mutual isolation between all three ports, the best of any topology (the double symmetry of hybrid + balun + 4 diodes). ✅✅ — 3 つのポート間で 25–50 dB の相互アイソレーションが得られ、全トポロジー中で最良です(hybrid + バラン + 4 つの diode による二重の対称性)。
  • Conversion Loss:ExcellentConversion loss: ExcellentConversion Loss:Excellent ✅ — 4 顆 diode 同時工作,每顆只承擔 1/2 訊號 → 高功率容量、低失真。典型 5–6 dB。 ✅ — all 4 diodes work together and each carries only half the signal → high power handling and low distortion. Typically 5–6 dB. ✅ — 4 つの diode が同時に動作し、それぞれ信号の半分しか担わないため → 高い電力耐量と低歪みが得られます。典型的に 5–6 dB です。
  • IP3:ExcellentIP3: ExcellentIP3:Excellent ✅✅ — 偶次 IMeven-order IM偶次 IM($2f_1$, $2f_2$, $f_1\pm f_2$)($2f_1$, $2f_2$, $f_1\pm f_2$) ($2f_1$, $2f_2$, $f_1\pm f_2$)天生對消cancels inherentlyが本質的に相殺され,IP3 比其他拓樸高 5–10 dB。, giving an IP3 5–10 dB higher than the other topologies.、IP3 は他のトポロジーより 5–10 dB 高くなります。
  • 賣點:Selling point:売り:最高線性度、最佳隔離the highest linearity and the best isolation最高の直線性と最良のアイソレーション。專業級接收機標配 — Mini-Circuits 的 ADE 系列、Marki Microwave、Hittite 的雷達/通訊 mixer 都採此架構。. Standard equipment in professional receivers — Mini-Circuits' ADE series, Marki Microwave and Hittite's radar/communications mixers all use this architecture.。プロ用受信機の標準装備です — Mini-Circuits の ADE シリーズ、Marki Microwave、Hittite のレーダー/通信用 mixer はいずれもこの構成を採用しています。
  • 代價:The cost:代償:LO 功率需求高(要驅動 4 顆 diode)、製造成本高、需要外加 matching。 a high LO power requirement (4 diodes to drive), high manufacturing cost, and the need for external matching.LO 電力の要求が高く(4 つの diode を駆動する必要がある)、製造コストが高く、外付けの整合が必要です。
⑤ Image reject — 用兩個 90° hybrid + 兩個 mixer⑤ Image reject — two 90° hybrids plus two mixers⑤ Image reject — 2 つの 90° hybrid と 2 つの mixer を使用
  • 所有指標都 GoodGood on every metricすべての指標が Good — 沒有特別差也沒有特別好。 — nothing especially bad and nothing especially good. — 特に悪い点も特に良い点もありません。
  • 獨有功能Its unique capability独自の機能:把 USB 跟 LSB 分到: it separates the USB and the LSB into :USB と LSB を不同輸出埠different output ports異なる出力ポート → 不需要 RF 預選濾波器即可抑制 image。 → suppressing the image without needing an RF preselect filter.へ分離します → RF プリセレクトフィルタなしでイメージを抑制できます。
  • 賣點:Selling point:売り:當 IF 很低、RF 預選濾波器難做時的 the IF が非常に低く、RF プリセレクトフィルタが実現困難な場合の唯一選擇only option唯一の選択肢。常見於 GSM/LTE 收發機、雷達、頻譜分析儀。 when the IF is very low and an RF preselect filter is impractical. Common in GSM/LTE transceivers, radar and spectrum analysers.です。GSM/LTE トランシーバ、レーダー、スペクトラムアナライザでよく用いられます。
  • 代價:The cost:代償:需要兩個 90° hybrid(RF 端 + IF 端)。IF hybrid 在低 IF 下難做,最終 two 90° hybrids are needed (one at RF, one at IF). The IF hybrid is hard to build at low IF, so the final 90° hybrid が 2 つ必要です(RF 側と IF 側)。IF hybrid は低い IF では作りにくく、最終的なimage rejection 約 20–30 dBimage rejection is about 20–30 dBイメージ除去は約 20–30 dB(需 trim 才能達 40+ dB)。(trimming is needed to reach 40+ dB).(40+ dB に達するには trim が必要)。

整體比較與選擇指南Overall comparison and a selection guide全体比較と選択の指針

為什麼 Double-balanced 不是「絕對最好」?Why isn't double-balanced “absolutely the best”?なぜ Double-balanced は「絶対的に最良」ではないのか?
看表可能會以為 double-balanced 表現最好(三個 Excellent),但實際上有幾個取捨: The table might suggest double-balanced performs best (three Excellents), but in practice there are several trade-offs: 表を見ると double-balanced が最も優れている(Excellent が 3 つ)ように思えますが、実際にはいくつかのトレードオフがあります:
  • LO 功率高High LO powerLO 電力が高い — 要驅動 4 顆 diode 同時工作,典型 LO power +7~+17 dBm(其他拓樸只需 +0~+10 dBm) — driving 4 diodes together needs a typical LO power of +7 to +17 dBm (other topologies need only +0 to +10 dBm) — 4 つの diode を同時に駆動するため、典型的な LO 電力は +7~+17 dBm(他のトポロジーは +0~+10 dBm で足ります)
  • RF match 差Poor RF matchRF の整合が悪い — 需要外加 matching network,PCB 面積大 — an external matching network is required, taking PCB area — 外付けの整合回路網が必要で、PCB 面積を取ります
  • 製造成本高High manufacturing cost製造コストが高い — 需要兩個 hybrid/balun + 4 顆精密配對的 diode — two hybrids/baluns plus 4 precisely matched diodes — 2 つの hybrid/バランと 4 つの精密に選別された diode が必要です
  • 低頻製造難Hard to build at low frequencies低い周波数では作りにくい — 整顆 balun 在低 GHz 段就很大顆 — the whole balun becomes bulky in the low-GHz range — 低 GHz 帯ではバラン全体が大きくなります
所以 system 選 mixer 不是看「指標總分」,而是看 So a system does not choose a mixer by “total score” but by したがってシステムが mixer を選ぶ基準は「指標の総合点」ではなく、應用需求的優先序the priorities of the applicationアプリケーションの優先順位. です。

應用場景對照Matching applications to topologiesアプリケーションとトポロジーの対応

應用場景Applicationアプリケーション首選拓樸Preferred topology推奨トポロジー理由Rationale理由
毫米波雷達 (60–100 GHz),低成本Millimetre-wave radar (60–100 GHz), low costミリ波レーダー (60–100 GHz)、低コストSingle-ended結構簡單、製造便宜、半導體 die 面積最小Simple structure, cheap to make, the smallest semiconductor die area構造が単純・製造が安価・半導体 die 面積が最小
VSWR 嚴苛的 LNA 後級A stage after an LNA with strict VSWR requirementsVSWR 要求が厳しい LNA の後段Balanced 90°RF return loss 永遠 > 20 dBRF return loss always > 20 dBRF return loss が常に > 20 dB
LO 功率大、LO 漏 RF 嚴重High LO power with severe LO leakage into the RFLO 電力が大きく、LO の RF への漏れが深刻Balanced 180°RF-LO 隔離 25–40 dB(天生)RF-LO isolation of 25–40 dB (inherently)RF-LO アイソレーション 25–40 dB(本質的に)
通訊接收機(最終級 mixer)Communication receivers (the final mixer stage)通信受信機(最終段の mixer)Double-balancedIP3 高、所有埠隔離好、$2f$ spur 低High IP3, good isolation between all ports, low $2f$ spursIP3 が高く、全ポート間のアイソレーションが良好で、$2f$ スプリアスが低い
IF 很低(< 100 MHz)+ image 抑制A very low IF (< 100 MHz) plus image suppressionIF が非常に低い(< 100 MHz)+ イメージ抑制Image rejectRF 預選濾波器難做時的唯一方案The only option when an RF preselect filter is impracticalRF プリセレクトフィルタが実現困難なときの唯一の方法
頻譜儀第一級下變頻The first downconversion of a spectrum analyserスペクトラムアナライザの初段ダウンコンバージョンImage reject 或 Double-balancedImage reject or double-balancedImage reject または Double-balanced需要寬頻 + 高線性度 + image 抑制Wide bandwidth + high linearity + image suppression are all needed広帯域 + 高い直線性 + イメージ抑制がすべて必要
三句話結論Three sentences to conclude三文でのまとめ
  1. 沒有「最好」的 mixerThere is no “best” mixer「最良の」mixer は存在せず,只有「最適合 application 的 mixer」 — 五個指標彼此 trade-off。, only “the mixer best suited to the application” — the five metrics trade off against one another.、あるのは「アプリケーションに最も適した mixer」だけです — 5 つの指標は互いにトレードオフの関係にあります。
  2. 從結構複雜度看:In order of structural complexity: 構造の複雑さの順に:single-ended → balanced → double-balanced,越複雜越能對稱對消某些不要的訊號,但, where greater complexity cancels more of the unwanted signals by symmetry, but 。複雑になるほど対称性によって不要な信号をより多く相殺できますが、不會does notすべての指標が同時に同時改善所有指標。 improve every metric at once.改善されるわけではありません。
  3. Image-reject 是唯一能在電路結構中分離 USB/LSB 的拓樸;其他都需要靠濾波器。 is the only topology that separates USB from LSB in the circuit structure itself; all the others need a filter. は回路構造そのもので USB と LSB を分離できる唯一のトポロジーです。他はすべてフィルタが必要です。

4.2.8 Large-Signal Model

Start from the device current $i(v)$ driven by $v = v_{LO}+v_{DC}+v_{RF}$. When the LO dominates — $v_{LO} \gg v_{DC}+v_{RF} \approx v_{RF}$ — expand $i(v)$ as a Taylor series about the (time-varying) LO operating point $v_{LO}$:

Taylor expansion about the LO operating point $$i(v) = i(v_{LO}) + \left.\frac{di}{dv}\right|_{v_{LO}}\!\! v_{RF} + \frac{1}{2!}\left.\frac{d^2 i}{dv^2}\right|_{v_{LO}}\!\! v_{RF}^2 + \frac{1}{3!}\left.\frac{d^3 i}{dv^3}\right|_{v_{LO}}\!\! v_{RF}^3 + \frac{1}{4!}\left.\frac{d^4 i}{dv^4}\right|_{v_{LO}}\!\! v_{RF}^4 + \cdots$$

Keeping the Taylor terms only up to 1st order in $v_{RF}$:

First-order truncation → time-varying conductance $$i(t) \approx i_{LO} + \bbox[#fff3b0,2px]{g(t)\,v_{RF}(t)} = i_{LO} + i_o(t), \qquad \bbox[#f8c9c0,3px]{g(t) = \left.\frac{di}{dv}\right|_{v_{LO}}}$$

The wanted output term is $i_o(t) = g(t)\,v_{RF}(t)$, where $g(t) = di/dv\big|_{v_{LO}}$ is the differential conductance with respect to the LO. Since the LO drives $v_{LO}$ periodically, $g(t)$ is itself periodic at $\omega_{LO}$ — which is exactly what the Fourier expansion below uses.

For large LO signals, the LO sets the operating point and $v_{RF}$ is the small perturbation. The differential conductance $g(t) = di/dv\big|_{v_{LO}(t)}$ varies periodically with the LO and is expressed as a Fourier cosine series:

Time-varying conductance (Fourier expansion) $$g(t) = g_0 + 2\sum_{n=1}^{\infty} g_n \cos(n\omega_{LO}t), \quad g_n = I_s\,I_n(\alpha V_{LO})$$

Here $\alpha = q/kT$ (diode ideality factor), $I_n$ is the modified Bessel function of order $n$.

為什麼 $g_n$ 變成 Bessel 函數 (Fourier coefficient → $I_n$)Why $g_n$ becomes a Bessel function (Fourier coefficient → $I_n$)なぜ $g_n$ が Bessel 関数になるのか (Fourier coefficient → $I_n$)

① Fourier 係數.① The Fourier coefficient.① Fourier 係数。 $g(t)$ is periodic ($T=2\pi/\omega_{LO}$) and even, so it expands in a cosine series. The standard cosine coefficient is $a_n=\tfrac{2}{T}\int_0^T g\cos n\omega_{LO}t\,dt$ — but because we deliberately wrote the series with a leading "2" ($g_0+2\sum g_n\cos$), we have $2g_n=a_n$ and the 2 cancels:

$$g_n = \frac{1}{T}\int_0^T g(t)\cos(n\omega_{LO}t)\,dt$$

(Same "2" convention that made the IF amplitude come out as a clean $g_1|V_c|$.)

② Diode conductance. An ideal diode obeys $i=I_s(e^{\alpha v}-1)$, so $g=di/dv=\alpha I_s e^{\alpha v}$. The large LO pumps the operating point as $v=V_{LO}\cos\omega_{LO}t$:

$$g(t) = \alpha I_s\,e^{\alpha V_{LO}\cos\omega_{LO}t}$$

③ Change of variable $\theta=\omega_{LO}t$ (so $dt=d\theta/\omega_{LO}$, $t:0\to T \Rightarrow \theta:0\to 2\pi$, and $\tfrac1T=\tfrac{\omega_{LO}}{2\pi}$). The $\omega_{LO}$ cancels:

$$g_n = \frac{\omega_{LO}}{2\pi}\int_0^{2\pi}\!\alpha I_s\,e^{\alpha V_{LO}\cos\theta}\cos n\theta\,\frac{d\theta}{\omega_{LO}} = \frac{\alpha I_s}{2\pi}\int_0^{2\pi}\! e^{\alpha V_{LO}\cos\theta}\cos n\theta\,d\theta$$

④ Recognise the Bessel integral. The integral representation of the modified Bessel function of the first kind is exactly

$$I_n(z)=\frac{1}{2\pi}\int_0^{2\pi}\! e^{z\cos\theta}\cos n\theta\,d\theta \;\Rightarrow\; g_n = \alpha I_s\,I_n(\alpha V_{LO})$$

為什麼重要:Why it matters:なぜ重要か: 二極體的指數型 I–V 被 LO 抽動後,各 LO 諧波的權重天生由 $I_n(\alpha V_{LO})$ 決定。$g_1$ 管轉換增益;LO 越強($\alpha V_{LO}$ 越大)時 $I_0\approx I_1\approx I_2$ → $g_0\approx g_1\approx g_2$,正是 image 終端難與 RF 終端分離的根本原因。 when the diode's exponential I–V is pumped by the LO, the weight of each LO harmonic is inherently set by $I_n(\alpha V_{LO})$. $g_1$ governs the conversion gain; and as the LO grows stronger (larger $\alpha V_{LO}$), $I_0\approx I_1\approx I_2$ → $g_0\approx g_1\approx g_2$, which is the fundamental reason it is so hard to separate the image termination from the RF termination. ダイオードの指数関数的な I–V が LO でポンピングされると、各 LO 高調波の重みは本質的に $I_n(\alpha V_{LO})$ で決まります。$g_1$ は変換利得を支配し、LO が強くなる($\alpha V_{LO}$ が大きくなる)ほど $I_0\approx I_1\approx I_2$ → $g_0\approx g_1\approx g_2$ となります。これがイメージ終端と RF 終端を分離しにくい根本的な理由です。

The output current $i_o(t) = g(t)\cdot v_{RF}(t)$ produces:

Key results — IF / RF / image currents $$i_{IF}(t) = g_1 V_{RF}\cos\omega_{IF}t, \quad i_{RF}(t) = g_0 V_{RF}\cos\omega_{RF}t, \quad i_{IM}(t) = g_2 V_{RF}\cos\omega_{IM}t$$

Truncating $g(t)$ to its first three Fourier terms and multiplying out via product-to-sum shows exactly where those three components come from:

Output current expansion — RF / IF / image terms $$\begin{aligned} i_o(t) = g(t)\,v_{RF}(t) &\approx \big(g_0 + 2g_1\cos\omega_{LO}t + 2g_2\cos 2\omega_{LO}t\big)\,V_{RF}\cos\omega_{RF}t\\[2pt] &= \underbrace{g_0 V_{RF}\cos\omega_{RF}t}_{i_{RF}(t)} + \underbrace{g_1 V_{RF}\big[\cos(\omega_{LO}-\omega_{RF})t + \cos(\omega_{LO}+\omega_{RF})t\big]}_{i_{IF}(t)}\\[4pt] &\quad + \underbrace{g_2 V_{RF}\big[\cos(2\omega_{LO}-\omega_{RF})t + \cos(2\omega_{LO}+\omega_{RF})t\big]}_{i_{IM}(t)} \end{aligned}$$

Each Fourier term spawns a sum/difference pair: $g_0$ → RF feedthrough, $g_1$ → wanted IF at $\omega_{LO}\pm\omega_{RF}$, $g_2$ → image/2nd-harmonic products. The "2" in $2g_n$ cancels the product-to-sum $\tfrac12$, so the amplitudes drop out clean.

大 LO 電壓的漸進展開 (asymptotic Bessel)Asymptotic expansion for large LO voltage (asymptotic Bessel)大きな LO 電圧に対する漸近展開 (asymptotic Bessel)

對於 LO 輸入功率 10 dBm(10 mW)且 $\alpha \approx 1/(28\text{ mV})$,得到 $\alpha V_{LO} \approx 36$(非常大)。For an LO input power of 10 dBm (10 mW) with $\alpha \approx 1/(28\text{ mV})$, we get $\alpha V_{LO} \approx 36$ (very large).LO 入力電力 10 dBm(10 mW)で $\alpha \approx 1/(28\text{ mV})$ のとき、$\alpha V_{LO} \approx 36$(非常に大きい)が得られます。

此時修正 Bessel 函數的漸近形式為:The asymptotic form of the modified Bessel functions is then:このとき変形 Bessel 関数の漸近形は次のようになります:

$$I_n(\alpha V_{LO}) \approx \frac{e^{\alpha V_{LO}}}{\sqrt{2\pi\,\alpha V_{LO}}}$$ $$\Rightarrow g_0 \approx g_1 \approx g_2 \approx \frac{I_s\,e^{\alpha V_{LO}}}{\sqrt{2\pi\,\alpha V_{LO}}}$$

實際意義:所有三個係數(含 image 項 $g_2$)幾乎相等。可用匹配的 RF 和 IM 終端阻抗來分別消除 $i_{RF}$ 的反射和 $i_{IM}$ 的輻射,但在實踐中 image 和 RF 終端通常共用,難以分開。The practical implication: all three coefficients (including the image term $g_2$) are almost equal. Matched RF and IM terminating impedances could be used to eliminate the reflection of $i_{RF}$ and the radiation of $i_{IM}$ separately, but in practice the image and RF terminations are usually shared and hard to separate.実際的な意味:3 つの係数(イメージ項 $g_2$ を含む)はいずれもほぼ等しくなります。整合した RF と IM の終端インピーダンスを用いれば $i_{RF}$ の反射と $i_{IM}$ の放射をそれぞれ除去できますが、実際にはイメージ終端と RF 終端は共用されることが多く、分離は困難です。

4.2.9 Switching Model

The large-signal analysis leads naturally to a switching model: when the LO is large, the nonlinear device (diode/transistor) acts as a switch, alternating on/off at the LO frequency.

$g_P$ 是什麼?What is $g_P$?$g_P$ とは何か? 它是 device 在 It is the device's conductance これは device が 導通(switch ON)時when it is switched ON導通(switch ON)しているとき的 conductance(P = peak / 導通態)。強 LO 強迫一個 idealisation:device 導通時 conductance $\approx g_P$、截止時 $\approx 0$,所以 $g(t)$ 是高度 $g_P$ 的 (P = peak / the on-state). A strong LO forces an idealization: the device's conductance is $\approx g_P$ when conducting and $\approx 0$ when cut off, so $g(t)$ is a の conductance です(P = peak / 導通状態)。強い LO は一つの理想化を強います:device は導通時に conductance が $\approx g_P$、遮断時に $\approx 0$ となるため、$g(t)$ は square wave(slide 36 的 pulse-train 波形),不是平滑的 $g_m$ 曲線。下面那些 Fourier 係數,就是標準 square-wave 係數乘上 $g_P$。of height $g_P$ (the pulse-train waveform of slide 36), not a smooth $g_m$ curve. The Fourier coefficients below are simply the standard square-wave coefficients multiplied by $g_P$.(高さ $g_P$、slide 36 のパルス列波形)であり、滑らかな $g_m$ 曲線ではありません。以下の Fourier 係数は、標準的な方形波の係数に $g_P$ を掛けただけのものです。

Unipolar Switching Mixer

The switch is ON for half the LO cycle (0 to $\pi$) with conductance $g_P$:

$$g(t) = g_P\!\left[\frac{1}{2} + \frac{2}{\pi}\sum_{n=1}^{\infty}\frac{\sin(n\pi/2)}{n}\cos(n\omega_{LO}t)\right]$$
IF output — unipolar switching $$i_{IF}(t) = \frac{g_P}{\pi}\,V_{RF}\cos(\omega_{RF}-\omega_{LO})t$$
那些 $\tfrac{g_P}{\pi}$ / $\tfrac{2g_P}{\pi}$ 係數怎麼來Where those $\tfrac{g_P}{\pi}$ / $\tfrac{2g_P}{\pi}$ coefficients come fromそれらの $\tfrac{g_P}{\pi}$ / $\tfrac{2g_P}{\pi}$ 係数の由来

① Square-wave Fourier. Unipolar switch 半個 cycle 導通,所以 $g(t)$ 是 $0$ 到 $g_P$ 的 square wave。DC 值是 $\tfrac12 g_P$(一半時間導通),fundamental 係數是標準的 $\tfrac{2}{\pi}g_P$(來自 $\sin(n\pi/2)/n$ 在 $n=1$): A unipolar switch conducts for half of each cycle, so $g(t)$ is a square wave from $0$ to $g_P$. Its DC value is $\tfrac12 g_P$ (conducting half the time) and its fundamental coefficient is the standard $\tfrac{2}{\pi}g_P$ (from $\sin(n\pi/2)/n$ at $n=1$): ユニポーラスイッチは各周期の半分だけ導通するため、$g(t)$ は $0$ から $g_P$ までの方形波になります。その DC 値は $\tfrac12 g_P$(半分の時間だけ導通)で、基本波係数は標準的な $\tfrac{2}{\pi}g_P$ です($n=1$ における $\sin(n\pi/2)/n$ より):

$$g(t) = g_P\!\left[\tfrac12 + \tfrac{2}{\pi}\cos\omega_{LO}t - \cdots\right]$$

② 乘上 RF.② Multiply by the RF.② RF を掛ける。 $i_o(t)=g(t)\,v_{RF}(t)$,其中 $v_{RF}=V_{RF}\cos\omega_{RF}t$。取 fundamental conductance 項做 product-to-sum: $i_o(t)=g(t)\,v_{RF}(t)$ with $v_{RF}=V_{RF}\cos\omega_{RF}t$. Taking the fundamental conductance term and applying product-to-sum: $i_o(t)=g(t)\,v_{RF}(t)$、ここで $v_{RF}=V_{RF}\cos\omega_{RF}t$ です。基本波コンダクタンス項を取り、積和公式を適用します:

$$\tfrac{2g_P}{\pi}\cos\omega_{LO}t \cdot V_{RF}\cos\omega_{RF}t = \tfrac{2g_P}{\pi}V_{RF}\cdot\tfrac12\big[\cos(\omega_{RF}-\omega_{LO})t + \cos(\omega_{RF}+\omega_{LO})t\big]$$

保留 difference(IF)項,$\tfrac{2}{\pi}\times\tfrac12 = \tfrac1\pi$:Keeping the difference (IF) term, with $\tfrac{2}{\pi}\times\tfrac12 = \tfrac1\pi$:差(IF)項を残すと、$\tfrac{2}{\pi}\times\tfrac12 = \tfrac1\pi$:

$$i_{IF}(t) = \frac{g_P}{\pi}V_{RF}\cos(\omega_{RF}-\omega_{LO})t \quad\text{(unipolar)}$$

③ Bipolar = ×2. Bipolar switch 擺幅是 $-g_P\!\to\!+g_P$ 而非 $0\!\to\!g_P$,fundamental 係數加倍成 $\tfrac{4g_P}{\pi}$。同樣 $\times\tfrac12$ 得: A bipolar switch swings from $-g_P$ to $+g_P$ rather than $0$ to $g_P$, doubling the fundamental coefficient to $\tfrac{4g_P}{\pi}$. Multiplying by $\tfrac12$ as before: バイポーラスイッチは $0$ から $g_P$ ではなく $-g_P$ から $+g_P$ まで振れるため、基本波係数は 2 倍の $\tfrac{4g_P}{\pi}$ になります。同様に $\tfrac12$ を掛けると:

$$i_{IF}(t) = \frac{2g_P}{\pi}V_{RF}\cos(\omega_{RF}-\omega_{LO})t \quad\text{(bipolar)}$$

即 unipolar IF 電流的兩倍 → twice the IF current of the unipolar case → ユニポーラの場合の IF 電流の 2 倍 → +6 dB conversion gain。到處的 $\pi$ 就是 square-wave fundamental 因子 $\tfrac{4}{\pi}$(或 $\tfrac{2}{\pi}$)被 product-to-sum 砍半。 conversion gain. The $\pi$ appearing everywhere is the square-wave fundamental factor $\tfrac{4}{\pi}$ (or $\tfrac{2}{\pi}$) halved by the product-to-sum step. の変換利得です。至るところに現れる $\pi$ は、方形波の基本波係数 $\tfrac{4}{\pi}$(または $\tfrac{2}{\pi}$)が積和変換で半分になったものです。

Bipolar Switching Mixer

The switch alternates sign each half-cycle (balanced differential pair):

$$g(t) = g_P\sum_{n=1,3,5,...}\frac{4\sin(n\pi/2)}{n\pi}\cos(n\omega_{LO}t)$$
IF output — bipolar switching $$i_{IF}(t) = \frac{2g_P}{\pi}\,V_{RF}\cos(\omega_{RF}-\omega_{LO})t$$
Bipolar vs Unipolar
The bipolar (differential) switching mixer produces twice the IF current of a unipolar mixer, corresponding to a 6 dB improvement in conversion gain. This is why differential transistor pair mixers are preferred.
LO sine and square-wave conductance plus unipolar and bipolar switching mixer schematics
Lecture slide 39 — unipolar (half-cycle) vs bipolar (differential) switching mixer; bipolar gives +6 dB conversion gain.
Slide 39 — text transcript
  • Top plot: LO voltage waveform $v_{LO}(t)$ and the idealized square-wave conductance waveform $g(t)$ used in the switching model of a mixer.
  • Unipolar switching mixer: single switch S(t) controlled by the LO; conductance switches between $g_P$ (on) and $0$ (off) once per LO cycle; $i_o(t) = g(t)\cdot v_{RF}(t)$.
  • Bipolar switching mixer: crossed-switch (differential) configuration; the sign of conductance flips each LO half-cycle, producing a bipolar square wave of $\pm g_P$.

4.2.10 Single-Ended FET (MESFET) Mixer

The dominant nonlinearity is the transconductance $g_m(v_{GS})$. The gate-to-source voltage is driven by the LO, making $g_m(t)$ a time-varying Fourier series:

$$g_m(t) = g_0 + 2g_1\cos\omega_{LO}t + 2g_2\cos 2\omega_{LO}t + \cdots$$

The output current is this time-varying transconductance multiplied by the RF small-signal voltage — and that product is the mixing operation:

Output current — where the mixing product is born $$i_o(t) = g(t)\,\mathrm{Re}\{V_c e^{j\omega_{RF}t}\} = \big(g_0 + 2g_1\cos\omega_{LO}t + 2g_2\cos 2\omega_{LO}t + \cdots\big)\,|V_c|\cos\omega_{RF}t$$
↗ 完整展開(RF / IF / image 三項)見 §4.2.8↗ For the full expansion (the RF / IF / image terms) see §4.2.8↗ 完全な展開(RF / IF / image の 3 項)は §4.2.8 を参照
$g(t)$ 是什麼?時變跨導 (time-varying transconductance) 怎麼來(點擊展開/收合)What is $g(t)$? Where the time-varying transconductance comes from (click to expand/collapse)$g(t)$ とは何か?時変相互コンダクタンス (time-varying transconductance) はどこから来るのか(クリックで展開/折りたたみ)

跨導 $g_m=\partial I_D/\partial V_{GS}$ 是 $I_D$–$V_{GS}$ 曲線的The transconductance $g_m=\partial I_D/\partial V_{GS}$ is the 相互コンダクタンス $g_m=\partial I_D/\partial V_{GS}$ は斜率slope傾き。一般放大器 bias 固定 → $g_m$ 是常數;但 mixer 把 of the $I_D$–$V_{GS}$ curve. In an ordinary amplifier the bias is fixed → $g_m$ is a constant; but a mixer applies a 、すなわち $I_D$–$V_{GS}$ 曲線の勾配です。通常の増幅器ではバイアスが固定なので → $g_m$ は定数ですが、ミキサでは 大訊號 LO 加在閘極large-signal LO at the gate大信号 LO をゲートに加え,讓工作點沿著非線性曲線來回掃。曲線是彎的,斜率就隨 LO 以 $\omega_{LO}$ 的節奏週期變化 → $g_m$ 成為時間函數 $g(t)$:, sweeping the operating point back and forth along the nonlinear curve. The curve is bent, so the slope varies periodically with the LO at a rate $\omega_{LO}$ → $g_m$ becomes a function of time, $g(t)$:、動作点を非線形曲線に沿って行き来させます。曲線が曲がっているため、傾きは LO に合わせて $\omega_{LO}$ の周期で変化します → $g_m$ は時間の関数 $g(t)$ になります:

VGS ID Vth 小 gsmall g小さい gm 大 glarge g大きい gm LO swing
LO 大訊號掃動工作點:曲線斜率(= $g_m$)隨時間變化 → $g(t)$。The large-signal LO sweeps the operating point: the slope of the curve (= $g_m$) varies with time → $g(t)$.大信号 LO が動作点を掃引します:曲線の傾き(= $g_m$)が時間とともに変化 → $g(t)$。

為什麼能寫成 Fourier 級數:Why it can be written as a Fourier series:なぜ Fourier 級数に書けるのか: $g(t)$ 是週期 $2\pi/\omega_{LO}$ 的週期函數,任何週期函數都能展開:$g_0$ = 平均(DC)跨導;$g_1$ = 基波係數, $g(t)$ is periodic with period $2\pi/\omega_{LO}$, and any periodic function can be expanded: $g_0$ = the average (DC) transconductance; $g_1$ = the fundamental coefficient, which $g(t)$ は周期 $2\pi/\omega_{LO}$ の周期関数であり、任意の周期関数は展開できます:$g_0$ = 平均(DC)相互コンダクタンス、$g_1$ = 基本波係数で、これが做出我們要的 IFproduces the IF we want必要な IF を生み出します($G_c\propto g_1^2$);$g_n$ = $n$ 次諧波(產生 $\omega_{RF}\pm n\omega_{LO}$)。($G_c\propto g_1^2$); $g_n$ = the $n$-th harmonic (generating $\omega_{RF}\pm n\omega_{LO}$).($G_c\propto g_1^2$)。$g_n$ = 第 $n$ 高調波($\omega_{RF}\pm n\omega_{LO}$ を生成)。

為什麼相乘就是混波Why multiplying is mixingなぜ掛け算が混合になるのか
$i_o=g(t)\cdot\mathrm{Re}\{V_c e^{j\omega_{RF}t}\}$ — 時變增益 $g(t)$(LO 那一路) $i_o=g(t)\cdot\mathrm{Re}\{V_c e^{j\omega_{RF}t}\}$ — the time-varying gain $g(t)$ (the LO path) $i_o=g(t)\cdot\mathrm{Re}\{V_c e^{j\omega_{RF}t}\}$ — 時変利得 $g(t)$(LO の経路)に乘上multiplies掛け合わされる RF。 the RF. のが RF です。時域相乘 = 頻域捲積,兩頻率相乘生出和/差頻 $\omega_{RF}\pm n\omega_{LO}$Multiplication in time = convolution in frequency, and multiplying two frequencies generates the sum and difference $\omega_{RF}\pm n\omega_{LO}$時間領域の乗算 = 周波数領域の畳み込みであり、2 つの周波数を掛けると和と差の周波数 $\omega_{RF}\pm n\omega_{LO}$ が生じます。Mixer 本質是乘法器:$g(t)$ 提供 LO、$V_c$ 提供 RF。 . A mixer is fundamentally a multiplier: $g(t)$ supplies the LO and $V_c$ the RF. 。ミキサは本質的に乗算器です:$g(t)$ が LO を、$V_c$ が RF を供給します。

Multiplying each Fourier term by $\cos\omega_{RF}t$ (product-to-sum) spreads the output across the LO harmonic grid:

Source termOutput frequencyMeaning
$g_0$$\omega_{RF}$RF feedthrough (leakage)
$2g_1\cos\omega_{LO}t$$\omega_{RF}\pm\omega_{LO}$wanted IF (sum & difference)
$2g_2\cos 2\omega_{LO}t$$\omega_{RF}\pm 2\omega_{LO}$2nd-harmonic mixing
$2g_n\cos n\omega_{LO}t$$\omega_{RF}\pm n\omega_{LO}$$n$th-harmonic mixing

The fundamental term gives the IF cleanly — the "2" in the series cancels the $\tfrac12$ from product-to-sum:

Fundamental term → IF $$2g_1\cos\omega_{LO}t\cdot|V_c|\cos\omega_{RF}t = g_1|V_c|\big[\cos(\omega_{RF}-\omega_{LO})t + \cos(\omega_{RF}+\omega_{LO})t\big]$$
Why the series is written with a "2"
The IF current amplitude comes out as exactly $g_1|V_c|$ — clean, with no stray factor of $\tfrac12$. That is the whole reason $g(t)$ is written as $g_0+2\sum g_n\cos n\omega_{LO}t$: the leading 2 absorbs the product-to-sum $\tfrac12$. This $i_{D,IF}=g_1 V_c$ is what feeds the conversion-gain result below ($G_c\propto g_1^2$).
Single-ended FET mixer schematic, MESFET equivalent circuit, gm vs Vg curve, conductance waveform
Lecture slide 36 — single-ended FET mixer circuit, MESFET small-signal model, $g_m$-vs-$V_{GS}$ curve, and the resulting square-wave conductance waveform.
Slide 36 — text transcript
  • Top: single-ended FET mixer schematic — RF/LO diplexer + matching network → FET gate; RF/LO chokes; LPF on the drain output; +DC bias on the drain, −DC on the gate.
  • Right: MESFET small-signal equivalent — gate resistance $R_i$, gate-source capacitance $C_{gs}$, voltage-controlled current source $g_m v_c$, drain resistance $R_d$, load $Z_L$ (IF).
  • Lower-left: MESFET transconductance vs gate-to-source voltage — soft "S" curve with pinch-off below $-3$ V, rising to ~30 mS at $V_g=0$ V.
  • Lower-right: resulting transconductance / diode-conductance waveform $g(t)$ when the gate is swung by a large LO — a pulse train (essentially a square wave) at the LO rate.

The drain current phasor at IF frequency is $I_{D,IF} = g_1 V_c$, where $V_c$ is the RF voltage across $C_{gs}$:

$$V_c = \underbrace{\frac{V_{RF}}{j\omega_{RF}C_{gs}\big[(R_i+Z_g)+(j\omega_{RF}C_{gs})^{-1}\big]}}_{\text{分壓原型}} = \frac{V_{RF}}{1 + j\omega_{RF}C_{gs}(R_i+Z_g)}$$ $$V_c = \underbrace{\frac{V_{RF}}{j\omega_{RF}C_{gs}\big[(R_i+Z_g)+(j\omega_{RF}C_{gs})^{-1}\big]}}_{\text{the divider prototype}} = \frac{V_{RF}}{1 + j\omega_{RF}C_{gs}(R_i+Z_g)}$$ $$V_c = \underbrace{\frac{V_{RF}}{j\omega_{RF}C_{gs}\big[(R_i+Z_g)+(j\omega_{RF}C_{gs})^{-1}\big]}}_{\text{the divider prototype}} = \frac{V_{RF}}{1 + j\omega_{RF}C_{gs}(R_i+Z_g)}$$
$V_c$ 是閘極迴路的分壓——兩個等號怎麼來(點擊展開)$V_c$ is a divider in the gate loop — where the two equalities come from (click to expand)$V_c$ はゲートループの分圧——2 つの等号はどこから来るのか(クリックで展開)

混波靠 $g_m$ 被閘源電壓調變,但真正調變 $g_m$ 的是Mixing relies on $g_m$ being modulated by the gate-source voltage, but what actually modulates $g_m$ is the $V_c$混合は $g_m$ がゲート・ソース間電圧で変調されることに依存しますが、実際に $g_m$ を変調するのは $V_c$跨在 $C_{gs}$ 上across $C_{gs}$($C_{gs}$ にかかるもの)的 $V_c$,不是端點的 $V_{RF}$。閘極輸入是一個串聯環:$V_{RF}$ 落在 $Z_g$(源阻抗)、$R_i$(本徵閘極電阻)、$C_{gs}$(阻抗 $Z_C=1/j\omega_{RF}C_{gs}$)三者串聯上,$V_c$ 只取 $C_{gs}$ 那一段:, not the $V_{RF}$ at the terminal. The gate input is a series loop: $V_{RF}$ is dropped across $Z_g$ (the source impedance), $R_i$ (the intrinsic gate resistance) and $C_{gs}$ (of impedance $Z_C=1/j\omega_{RF}C_{gs}$) in series, and $V_c$ takes only the $C_{gs}$ portion:であって、端子における $V_{RF}$ ではありません。ゲート入力は直列ループです:$V_{RF}$ は $Z_g$(ソースインピーダンス)、$R_i$(真性ゲート抵抗)、$C_{gs}$(インピーダンス $Z_C=1/j\omega_{RF}C_{gs}$)の直列に分圧され、$V_c$ はそのうち $C_{gs}$ の分だけを受け取ります:

VRF Zg source Z Ri gate R Cgs Vc
閘極輸入串聯環:$V_c$ = $V_{RF}$ 在 $C_{gs}$ 上的分壓。The gate input series loop: $V_c$ = the portion of $V_{RF}$ divided across $C_{gs}$.ゲート入力の直列ループ:$V_c$ = $V_{RF}$ が $C_{gs}$ に分圧された分。

① 分壓原型.① The divider prototype.① 分圧の原型。

$$V_c = V_{RF}\cdot\frac{Z_C}{(R_i+Z_g)+Z_C},\qquad Z_C=\frac{1}{j\omega_{RF}C_{gs}}$$

② 第一等號.② The first equality.② 第一の等号。 分子分母同除 $Z_C$(等於同乘 $1/Z_C=j\omega_{RF}C_{gs}$),把外面的 $V_{RF}$ 留下: Divide numerator and denominator by $Z_C$ (equivalently multiply by $1/Z_C=j\omega_{RF}C_{gs}$), leaving the outer $V_{RF}$: 分子と分母を $Z_C$ で割り($1/Z_C=j\omega_{RF}C_{gs}$ を掛けるのと同じ)、外側の $V_{RF}$ を残します:

$$V_c = \frac{V_{RF}}{\big[(R_i+Z_g)+Z_C\big]/Z_C} = \frac{V_{RF}}{j\omega_{RF}C_{gs}\big[(R_i+Z_g)+(j\omega_{RF}C_{gs})^{-1}\big]}$$

③ 第二等號.③ The second equality.③ 第二の等号。 把 $j\omega_{RF}C_{gs}$ 乘進中括號,後項相消為 1: Multiply $j\omega_{RF}C_{gs}$ into the bracket, and the trailing term cancels to 1: $j\omega_{RF}C_{gs}$ を括弧の中に掛け込むと、後ろの項は相殺されて 1 になります:

$$j\omega_{RF}C_{gs}(R_i+Z_g)+\underbrace{j\omega_{RF}C_{gs}\cdot(j\omega_{RF}C_{gs})^{-1}}_{=\,1}\;\Rightarrow\; V_c=\frac{V_{RF}}{1+j\omega_{RF}C_{gs}(R_i+Z_g)}$$
物理意義 — 為何高頻增益會掉Physical meaning — why the gain falls at high frequency物理的意味 — なぜ高周波で利得が下がるのか
低頻At low frequency低周波($\omega_{RF}C_{gs}\!\to\!0$):分母 $\approx1$ → $V_c\approx V_{RF}$,RF 幾乎全加到 $C_{gs}$。($\omega_{RF}C_{gs}\!\to\!0$): the denominator $\approx1$ → $V_c\approx V_{RF}$, and essentially all the RF appears across $C_{gs}$.($\omega_{RF}C_{gs}\!\to\!0$):分母 $\approx1$ → $V_c\approx V_{RF}$ となり、RF はほぼすべて $C_{gs}$ に加わります。
高頻At high frequency高周波:$C_{gs}$ 趨近短路,$|V_c|\sim1/\omega_{RF}$ 下降 → 控制電壓變小。$I_{D,IF}=g_1V_c$ 跟著掉,功率取平方,這就是 $G_{c,\max}=\dfrac{g_1^2R_d}{4\omega_{RF}^2C_{gs}^2R_i}$ 分母帶 $\omega_{RF}^2C_{gs}^2$ 的物理來源。 : $C_{gs}$ approaches a short and $|V_c|\sim1/\omega_{RF}$ falls → the controlling voltage shrinks. $I_{D,IF}=g_1V_c$ falls with it, and power goes as the square — which is the physical origin of the $\omega_{RF}^2C_{gs}^2$ in the denominator of $G_{c,\max}=\dfrac{g_1^2R_d}{4\omega_{RF}^2C_{gs}^2R_i}$. :$C_{gs}$ は短絡に近づき、$|V_c|\sim1/\omega_{RF}$ が低下します → 制御電圧が小さくなります。$I_{D,IF}=g_1V_c$ もそれに伴って低下し、電力は 2 乗で効くため、これが $G_{c,\max}=\dfrac{g_1^2R_d}{4\omega_{RF}^2C_{gs}^2R_i}$ の分母に $\omega_{RF}^2C_{gs}^2$ が現れる物理的な起源です。

這個 IF 電流 $I_{D,IF}=g_1V_c$ 再灌進汲極輸出網路 $R_d\parallel Z_L$($R_d$ 汲極電阻、$Z_L$ IF 負載),得到汲極 IF 輸出電壓:This IF current $I_{D,IF}=g_1V_c$ is then driven into the drain output network $R_d\parallel Z_L$ ($R_d$ the drain resistance, $Z_L$ the IF load), giving the drain IF output voltage:この IF 電流 $I_{D,IF}=g_1V_c$ をドレイン出力回路網 $R_d\parallel Z_L$($R_d$ はドレイン抵抗、$Z_L$ は IF 負荷)へ流し込むと、ドレインの IF 出力電圧が得られます:

Drain IF voltage(輸出側)Drain IF voltage (the output side)ドレイン IF 電圧(出力側) $$V_D^{IF} = -g_1 V_c\,\frac{R_d Z_L}{R_d+Z_L} = \frac{-g_1 V_{RF}}{1+j\omega_{RF}C_{gs}(R_i+Z_g)}\,\frac{R_d Z_L}{R_d+Z_L}$$
三段相乘 + 負號哪來(點擊展開)The product of three factors, and where the minus sign comes from (click to expand)三段の積と負号の由来(クリックで展開)

輸出並聯阻抗.The output shunt impedance.出力の並列インピーダンス。 電流源 $g_1V_c$ 看到的是 $R_d$ 與 $Z_L$ 並聯:$R_d\parallel Z_L=\dfrac{R_dZ_L}{R_d+Z_L}$,汲極電壓 = 電流 × 此阻抗。 The current source $g_1V_c$ sees $R_d$ in parallel with $Z_L$: $R_d\parallel Z_L=\dfrac{R_dZ_L}{R_d+Z_L}$, and the drain voltage = the current × that impedance. 電流源 $g_1V_c$ から見えるのは $R_d$ と $Z_L$ の並列で$R_d\parallel Z_L=\dfrac{R_dZ_L}{R_d+Z_L}$、ドレイン電圧 = 電流 × このインピーダンスです。

負號.The minus sign.負号。 common-source 為反相組態:汲極電流源 $g_mv_c$ 把電流抽離輸出節點,$V_D$ 相對閘極反相,故帶 $-$。 Common source is an inverting configuration: the drain current source $g_mv_c$ pulls current out of the output node, so $V_D$ is inverted relative to the gate, hence the $-$. ソース接地は反転構成です:ドレイン電流源 $g_mv_c$ が出力ノードから電流を引き抜くため、$V_D$ はゲートに対して反転し、$-$ が付きます。

第二等號.The second equality.第二の等号。 代入閘極分壓 $V_c=\dfrac{V_{RF}}{1+j\omega_{RF}C_{gs}(R_i+Z_g)}$。 Substitute the gate divider $V_c=\dfrac{V_{RF}}{1+j\omega_{RF}C_{gs}(R_i+Z_g)}$. ゲート分圧 $V_c=\dfrac{V_{RF}}{1+j\omega_{RF}C_{gs}(R_i+Z_g)}$ を代入します。

整條鏈:$V_D^{IF}=\underbrace{(-g_1)}_{\text{跨導}}\times\underbrace{V_c/V_{RF}}_{\text{閘極分壓}}\times\underbrace{(R_d\parallel Z_L)}_{\text{輸出網路}}$ — 這正是 $G_c$ 裡的 $V_D^{IF}/V_{RF}$。The whole chain: $V_D^{IF}=\underbrace{(-g_1)}_{\text{transconductance}}\times\underbrace{V_c/V_{RF}}_{\text{gate divider}}\times\underbrace{(R_d\parallel Z_L)}_{\text{output network}}$ — which is precisely the $V_D^{IF}/V_{RF}$ appearing in $G_c$.全体の連鎖:$V_D^{IF}=\underbrace{(-g_1)}_{\text{transconductance}}\times\underbrace{V_c/V_{RF}}_{\text{gate divider}}\times\underbrace{(R_d\parallel Z_L)}_{\text{output network}}$ — これがまさに $G_c$ の中に現れる $V_D^{IF}/V_{RF}$ です。

$G_c$ 是什麼?What is $G_c$?$G_c$ とは?
$G_c$ 是 $G_c$ is the $G_c$ はconversion gain(轉換增益)conversion gain変換利得(conversion gain):衡量 mixer 把訊號從 RF 搬到 IF 時,功率是放大還是衰減,$G_c = P_{IF}/P_{RF}$。 : a measure of whether the mixer amplifies or attenuates the power as it moves the signal from RF to IF, $G_c = P_{IF}/P_{RF}$. :ミキサが信号を RF から IF へ移すときに、電力が増幅されるのか減衰されるのかを表す指標で、$G_c = P_{IF}/P_{RF}$。
  • $G_c > 1$($>0$ dB):有 gain → $G_c > 1$ ($>0$ dB): there is gain → $G_c > 1$($>0$ dB):利得あり → active mixer(transistor,如 Gilbert cell)。(a transistor mixer such as a Gilbert cell).(Gilbert cell などのトランジスタミキサ)。
  • $G_c < 1$($<0$ dB):有 $G_c < 1$ ($<0$ dB): there is $G_c < 1$($<0$ dB):conversion loss,passive diode mixer 典型 $-6\sim-8$ dB。, typically $-6\sim-8$ dB for a passive diode mixer.があり、パッシブダイオードミキサでは典型的に $-6\sim-8$ dB です。
為什麼 $G_c \propto g_1^2$:Why $G_c \propto g_1^2$:なぜ $G_c \propto g_1^2$ なのか: IF 輸出電流由時變 transconductance 的 fundamental Fourier 係數 $g_1$ 產生($i_{IF}\propto g_1 V_{RF}$),電壓轉換比 $\propto g_1$,而功率是電壓平方,所以 $G_c \propto g_1^2$。$g_1$ 越大(LO 把 device 抽動得越有效)→ 增益越高;這也是 bipolar 切換比 unipolar 多 +6 dB(基波係數翻倍 → 平方差 4 倍)的原因。 the IF output current is produced by the fundamental Fourier coefficient $g_1$ of the time-varying transconductance ($i_{IF}\propto g_1 V_{RF}$), so the voltage conversion ratio $\propto g_1$, and since power goes as voltage squared, $G_c \propto g_1^2$. The larger $g_1$ (the more effectively the LO pumps the device) → the higher the gain; and this is also why bipolar switching gains +6 dB over unipolar (the fundamental coefficient doubles → a factor of 4 when squared). IF 出力電流は時変相互コンダクタンスの基本波フーリエ係数 $g_1$ によって生じ($i_{IF}\propto g_1 V_{RF}$)、電圧変換比は $\propto g_1$、電力は電圧の 2 乗なので $G_c \propto g_1^2$ となります。$g_1$ が大きいほど(LO がデバイスをより効果的に駆動するほど)利得は高くなります。これは bipolar スイッチングが unipolar より +6 dB 有利になる(基本波係数が 2 倍 → 2 乗して 4 倍)理由でもあります。

Conversion gain is defined as available IF output power over available RF source power. Writing the RF source available power as $|V_{RF}|^2/8R_g$ and the IF power delivered to the load $Z_L = R_L + jX_L$ as $0.5\,|V_D^{IF}|^2\,R_L/|Z_L|^2$:

Conversion gain — available-power form $$G_c = \frac{P_{IF}^a}{P_{RF}^a} = \frac{0.5\,\mathrm{Re}\!\left\{V_D^{IF}\left(V_D^{IF}/Z_L\right)^{*}\right\}}{0.5\,\left|V_{RF}/(2R_g)\right|^2 R_g} = \bbox[#f8c9c0,4px]{\frac{4R_g R_L}{|Z_L|^2}\left|\frac{V_D^{IF}}{V_{RF}}\right|^2}$$

為什麼分母帶 $2R_g$?Why does the denominator contain $2R_g$?なぜ分母に $2R_g$ が現れるのか? $V_{RF}$ 是訊號源的 $V_{RF}$ is the source's $V_{RF}$ は信号源の開路電動勢 (open-circuit EMF)open-circuit EMF開放起電力(open-circuit EMF),$R_g$ 是訊號源的 and $R_g$ is the source's 、$R_g$ は信号源の內阻 (source resistance)internal (source) resistance内部抵抗(source resistance)。Available power 指源能交出的最大功率,發生在. Available power is the greatest power the source can deliver, which occurs at 。Available power は信号源が取り出せる最大電力で、これは共軛匹配 (conjugate match)Conjugate match共役整合(conjugate match) 時,即負載電阻 $R_L=R_g$。此時迴路是內阻 $R_g$ 與負載 $R_L=R_g$ , i.e. when the load resistance $R_L=R_g$. The loop is then the internal $R_g$ and the load $R_L=R_g$ のとき、すなわち負荷抵抗が $R_L=R_g$ のときに生じます。このときループは内部抵抗 $R_g$ と負荷 $R_L=R_g$ の串聯in series直列,總電阻 $R_g+R_g=2R_g$,所以分壓後迴路電流 $I=V_{RF}/(2R_g)$;代入 $P=\tfrac12|I|^2R_L$ 即得 $0.5\,|V_{RF}/(2R_g)|^2R_g=|V_{RF}|^2/8R_g$。, a total resistance of $R_g+R_g=2R_g$, so after the division the loop current is $I=V_{RF}/(2R_g)$; substituting into $P=\tfrac12|I|^2R_L$ gives $0.5\,|V_{RF}/(2R_g)|^2R_g=|V_{RF}|^2/8R_g$.接続となり、合計抵抗は $R_g+R_g=2R_g$ なので、分圧後のループ電流は $I=V_{RF}/(2R_g)$、$P=\tfrac12|I|^2R_L$ に代入すると $0.5\,|V_{RF}/(2R_g)|^2R_g=|V_{RF}|^2/8R_g$ が得られます。

VRF open-circuit EMF Rg source resistance I RL = Rg matched load loop sees Rg + Rg = 2Rg in series
Available-power source model — 共軛匹配時負載 $R_L=R_g$,迴路串聯總阻 $2R_g$,故 $I=V_{RF}/(2R_g)$。The available-power source model — at conjugate match $R_L=R_g$, the series loop totals $2R_g$, hence $I=V_{RF}/(2R_g)$.Available-power source model — 共役整合時の負荷は $R_L=R_g$、ループの直列合計抵抗は $2R_g$、したがって $I=V_{RF}/(2R_g)$。

The simplification uses $\mathrm{Re}\{1/Z_L\} = R_L/|Z_L|^2$ and the constant $0.5\times 8 = 4$.

逐等號展開 (step-by-step derivation)Step-by-step derivation, equality by equality等号ごとの展開(step-by-step derivation)

① 定義.① The definition.① 定義。 Conversion gain = available IF output power / available RF source power (上標 $a$ = available): Conversion gain = available IF output power / available RF source power (the superscript $a$ = available): Conversion gain = available IF output power / available RF source power(上付き添字 $a$ = available):

$$G_c = \frac{P_{IF}^a}{P_{RF}^a}$$

② 分母 — RF source available power.② The denominator — the RF source's available power.② 分母 — RF source available power。 Open-circuit EMF $V_{RF}$, source resistance $R_g$. At conjugate match the load current is $V_{RF}/(2R_g)$, and average power into a resistor is $\tfrac12|I|^2R$:

$$P_{RF}^a = 0.5\left|\frac{V_{RF}}{2R_g}\right|^2 R_g = 0.5\cdot\frac{|V_{RF}|^2}{4R_g^2}\cdot R_g = \frac{|V_{RF}|^2}{8R_g}$$

③ 分子 — IF power into the load.③ The numerator — the IF power into the load.③ 分子 — 負荷に入る IF 電力。 Drain IF phasor $V_D^{IF}$ drives load current $I = V_D^{IF}/Z_L$ into $Z_L$. 共軛從哪來:Where the conjugate comes from:共役はどこから来るのか: the phasor average-power formula itself carries one, $P=\tfrac12\mathrm{Re}\{V I^{*}\}$ — the $I^{*}$ is what extracts the in-phase (dissipating) part, $\tfrac12|V||I|\cos(\theta_V-\theta_I)$. Only the resistive component does real work, so the conjugate must be there:

$P=\tfrac12\mathrm{Re}\{VI^{*}\}$ 從時域推一次 + 為什麼是 $I^{*}$(點擊展開)Deriving $P=\tfrac12\mathrm{Re}\{VI^{*}\}$ once in the time domain + why it is $I^{*}$ (click to expand)$P=\tfrac12\mathrm{Re}\{VI^{*}\}$ を時間領域から一度導出 + なぜ $I^{*}$ なのか(クリックで展開)

① 時域.① The time domain.① 時間領域。 相量取峰值,寫回時域: Take the phasors as peak values and write them back in the time domain: フェーザを peak 値で取り、時間領域に書き戻します:

$$v(t)=|V|\cos(\omega t+\theta_V),\qquad i(t)=|I|\cos(\omega t+\theta_I)$$

② 瞬時功率.② The instantaneous power.② 瞬時電力。 $p=vi$,用積化和差 $\cos A\cos B=\tfrac12[\cos(A-B)+\cos(A+B)]$: $p=vi$, using the product-to-sum identity $\cos A\cos B=\tfrac12[\cos(A-B)+\cos(A+B)]$: $p=vi$、積和公式 $\cos A\cos B=\tfrac12[\cos(A-B)+\cos(A+B)]$ を用いて:

$$p(t)=\tfrac12|V||I|\big[\underbrace{\cos(\theta_V-\theta_I)}_{\text{常數 (DC)}}+\underbrace{\cos(2\omega t+\theta_V+\theta_I)}_{2\omega\text{ 項}}\big]$$ $$p(t)=\tfrac12|V||I|\big[\underbrace{\cos(\theta_V-\theta_I)}_{\text{constant (DC)}}+\underbrace{\cos(2\omega t+\theta_V+\theta_I)}_{2\omega\text{ term}}\big]$$ $$p(t)=\tfrac12|V||I|\big[\underbrace{\cos(\theta_V-\theta_I)}_{\text{constant (DC)}}+\underbrace{\cos(2\omega t+\theta_V+\theta_I)}_{2\omega\text{ term}}\big]$$

③ 取平均.③ Take the average.③ 平均を取る。 $2\omega$ 那項一個週期積分為 0,只剩常數項: The $2\omega$ term integrates to 0 over one period, leaving only the constant term: $2\omega$ の項は 1 周期の積分が 0 になり、定数項だけが残ります:

$$P=\langle p(t)\rangle=\tfrac12|V||I|\cos(\theta_V-\theta_I)$$

④ 接回相量.④ Back to phasors.④ フェーザに戻す。 $V=|V|e^{j\theta_V},\ I=|I|e^{j\theta_I}$:

$$VI^{*}=|V|e^{j\theta_V}\cdot|I|e^{-j\theta_I}=|V||I|e^{j(\theta_V-\theta_I)}\;\Rightarrow\;\mathrm{Re}\{VI^{*}\}=|V||I|\cos(\theta_V-\theta_I)$$

兩式對照即得 $\boxed{P=\tfrac12\mathrm{Re}\{VI^{*}\}}$。Comparing the two expressions gives $\boxed{P=\tfrac12\mathrm{Re}\{VI^{*}\}}$.両式を比較すると $\boxed{P=\tfrac12\mathrm{Re}\{VI^{*}\}}$ が得られます。

V I |I|cos(Δθ) in-phase part Δθ
相量投影:I 在 V 方向上的分量 $|I|\cos(\Delta\theta)$ 才耗實功率;垂直分量(reactive)平均功率為 0。The phasor projection: only the component of I along V, $|I|\cos(\Delta\theta)$, dissipates real power; the perpendicular (reactive) component averages to 0.フェーザ投影:I の V 方向成分 $|I|\cos(\Delta\theta)$ だけが実電力を消費し、直交成分(リアクティブ)の平均電力は 0。
為什麼是 $I^{*}$ 不是 $I$Why $I^{*}$ and not $I$なぜ $I^{*}$ であって $I$ ではないのか
$VI=|V||I|e^{j(\theta_V+\theta_I)}$ 給的是相位 $VI=|V||I|e^{j(\theta_V+\theta_I)}$ gives the $VI=|V||I|e^{j(\theta_V+\theta_I)}$ が与えるのは位相の相加sum;但功率只取決於 V 與 I 的相位差of the phases, but power depends only on the phase difference between V and Iですが、電力は V と I の位相差だけで決まります。共軛把 $\theta_I$ 變號,$\theta_V+(-\theta_I)=\theta_V-\theta_I$ → 剛好取出相位差。同相($\Delta\theta=0$)→ $\cos=1$ 功率最大;正交($\Delta\theta=90^\circ$,純電抗)→ $\cos=0$,平均功率為 0。 . The conjugate flips the sign of $\theta_I$, so $\theta_V+(-\theta_I)=\theta_V-\theta_I$ → picking out exactly that difference. In phase ($\Delta\theta=0$) → $\cos=1$ and the power is maximal; in quadrature ($\Delta\theta=90^\circ$, purely reactive) → $\cos=0$ and the average power is zero. 。共役は $\theta_I$ の符号を反転させるので、$\theta_V+(-\theta_I)=\theta_V-\theta_I$ → ちょうど位相差だけを取り出します。同相($\Delta\theta=0$)→ $\cos=1$ で電力は最大、直交($\Delta\theta=90^\circ$、純リアクタンス)→ $\cos=0$ で平均電力は 0 です。
$$P_{IF}^a = 0.5\,\mathrm{Re}\!\left\{V_D^{IF}\Big(\frac{V_D^{IF}}{Z_L}\Big)^{\!*}\right\}$$

共軛怎麼消:How the conjugate cancels:共役はどのように打ち消されるか: use $(ab)^{*}=a^{*}b^{*}$ to pull out $V_D^{IF}(V_D^{IF})^{*}=|V_D^{IF}|^2$ (a real modulus-squared, so it leaves the $\mathrm{Re}\{\}$), then use the fact that taking a conjugate never changes the real part, $\mathrm{Re}\{w^{*}\}=\mathrm{Re}\{w\}$, to turn $(1/Z_L)^{*}$ back into $1/Z_L$:

$$P_{IF}^a = 0.5\,|V_D^{IF}|^2\,\mathrm{Re}\!\left\{\Big(\tfrac{1}{Z_L}\Big)^{\!*}\right\} = 0.5\,|V_D^{IF}|^2\,\mathrm{Re}\!\left\{\frac{1}{Z_L}\right\}$$

With $Z_L = R_L + jX_L$ (only the resistive part $R_L$ dissipates):

$$\mathrm{Re}\!\left\{\frac{1}{Z_L}\right\} = \bbox[#f8c9c0,4px]{\mathrm{Re}\!\left\{\frac{R_L - jX_L}{R_L^2 + X_L^2}\right\}} = \bbox[#f8c9c0,4px]{\frac{R_L}{|Z_L|^2}} \;\Rightarrow\; P_{IF}^a = \frac{0.5\,|V_D^{IF}|^2 R_L}{|Z_L|^2}$$

④ 相除.④ Divide.④ 割り算。 Combine ($0.5\times 8 = 4$):

$$G_c = \frac{\dfrac{0.5\,|V_D^{IF}|^2 R_L}{|Z_L|^2}}{\dfrac{|V_{RF}|^2}{8R_g}} = \frac{4R_g R_L}{|Z_L|^2}\left|\frac{V_D^{IF}}{V_{RF}}\right|^2$$

容易卡的兩點:Two easy places to get stuck:つまずきやすい 2 点: (i) 用峰值相量算平均功率一定帶 $\tfrac12$,匹配分壓給出 $V/(2R_g)$ → 平方成 $1/4R_g^2$,湊起來才是 $|V|^2/8R_g$;(ii) (i) computing average power from peak phasors always brings a $\tfrac12$, and the matched divider gives $V/(2R_g)$ → squaring to $1/4R_g^2$, which together make $|V|^2/8R_g$; (ii) (i) peak フェーザから平均電力を求めると必ず $\tfrac12$ が付き、整合分圧は $V/(2R_g)$ を与え → 2 乗して $1/4R_g^2$、合わせてはじめて $|V|^2/8R_g$ になります。(ii) 只有電阻耗功率,所以 $\mathrm{Re}\{1/Z_L\}=R_L/|Z_L|^2$ 把電抗 $X_L$ 丟掉only resistance dissipates power, so $\mathrm{Re}\{1/Z_L\}=R_L/|Z_L|^2$ discards the reactance $X_L$電力を消費するのは抵抗だけなので、$\mathrm{Re}\{1/Z_L\}=R_L/|Z_L|^2$ はリアクタンス $X_L$ を捨てます.

Conjugate 的三個作用(一眼總結)The three roles of the conjugate (at a glance)共役の三つの役割(ひと目でまとめ)
  1. 取平均功率(average power):Taking the average power:平均電力(average power)を取る:$P=\tfrac12\mathrm{Re}\{V I^{*}\}$ 中的 $I^{*}$ 消去載波相位 $e^{j\omega t}$,只留下 $V$–$I$ 的相位差 $\cos(\theta_V-\theta_I)$,取出真正傳遞的實功率。 the $I^{*}$ in $P=\tfrac12\mathrm{Re}\{V I^{*}\}$ cancels the carrier phase $e^{j\omega t}$ and leaves only the $V$–$I$ phase difference $\cos(\theta_V-\theta_I)$, extracting the real power actually delivered.$I^{*}$ の中の $P=\tfrac12\mathrm{Re}\{V I^{*}\}$ が搬送波位相 $e^{j\omega t}$ を打ち消し、$V$–$I$ の位相差 $\cos(\theta_V-\theta_I)$ だけを残して、実際に伝わる実電力を取り出します。
  2. 複數轉實數:Turning complex into real:複素数を実数に変換する:$V_D^{IF}\,(V_D^{IF})^{*}=|V_D^{IF}|^2$,共軛相乘把相量變成可量測的實數模平方,所以增益 $|V_D^{IF}/V_{RF}|^2$ 是純實數,相位資訊在取模時消失。$V_D^{IF}\,(V_D^{IF})^{*}=|V_D^{IF}|^2$ — multiplying by the conjugate turns a phasor into a measurable real squared modulus, so the gain $|V_D^{IF}/V_{RF}|^2$ is purely real, the phase information vanishing when the modulus is taken.$V_D^{IF}\,(V_D^{IF})^{*}=|V_D^{IF}|^2$、共役との積はフェーザを測定可能な実数の絶対値 2 乗に変えるので、利得 $|V_D^{IF}/V_{RF}|^2$ は純実数となり、位相情報は絶対値を取る時点で失われます。
  3. 只算 real power:Counting only real power:実電力だけを数える:$\mathrm{Re}\{\}$ 排除 reactive(虛)功率;若 $Z_L=R_L+jX_L$ 為複數,只有 $\mathrm{Re}\{Z_L\}=R_L$ 進入功率,這正是最終式出現 $R_L/|Z_L|^2$ 而非 $1/Z_L$ 的原因。$\mathrm{Re}\{\}$ excludes reactive (imaginary) power; if $Z_L=R_L+jX_L$ is complex, only $\mathrm{Re}\{Z_L\}=R_L$ enters the power — which is exactly why the final expression contains $R_L/|Z_L|^2$ rather than $1/Z_L$.$\mathrm{Re}\{\}$ はリアクティブ(虚数)電力を除外します。$Z_L=R_L+jX_L$ が複素数の場合、電力に入るのは $\mathrm{Re}\{Z_L\}=R_L$ だけであり、これが最終式に $1/Z_L$ ではなく $R_L/|Z_L|^2$ が現れる理由です。

一句話:In one sentence: 一言でいえば:conjugate → 取模 → 取實部conjugate → take the modulus → take the real part共役 → 絶対値を取る → 実部を取る 就是把「複數相量」翻成「實際功率」的標準三步,$g_m=g_0+2\sum_n g_n\cos n\omega_{LO}t$ 的傅立葉展開則決定了轉換比 $V_D^{IF}/V_{RF}$ 的大小。 is the standard three-step recipe for turning a “complex phasor” into “actual power”, while the Fourier expansion $g_m=g_0+2\sum_n g_n\cos n\omega_{LO}t$ determines the size of the conversion ratio $V_D^{IF}/V_{RF}$. が「複素フェーザ」を「実際の電力」に翻訳する標準の 3 ステップであり、$g_m=g_0+2\sum_n g_n\cos n\omega_{LO}t$ のフーリエ展開が変換比 $V_D^{IF}/V_{RF}$ の大きさを決めます。

How to read $G_c$
$G_c = \dfrac{4R_g R_L}{|Z_L|^2}\left|\dfrac{V_D^{IF}}{V_{RF}}\right|^2$
The voltage ratio $\left|V_D^{IF}/V_{RF}\right|^2$ carries the device action: since $V_D^{IF}\propto g_1 V_{RF}$, conversion gain scales as $g_1^2$ — the fundamental Fourier coefficient of the time-varying transconductance $g(t)$. The prefactor $4R_gR_L/|Z_L|^2$ is purely a source/load impedance-matching factor. Conjugate-matching both ports maximises it and yields the closed form below.

把 $V_D^{IF}/V_{RF}$ 代進 $G_c$ 取模平方($|Z_L|^2$ 相消),展開成Substituting $V_D^{IF}/V_{RF}$ into $G_c$ and taking the squared modulus (the $|Z_L|^2$ cancelling) expands it into $V_D^{IF}/V_{RF}$ を $G_c$ に代入して絶対値の 2 乗を取ると($|Z_L|^2$ が相殺)、輸入網路 × 輸出網路an input network × an output network入力回路網 × 出力回路網兩段:, in two parts:の 2 段に展開されます:

Conversion gain(完整展開)— 各因子可點,連回來源Conversion gain (fully expanded) — each factor is clickable, linking back to its originConversion gain(完全展開)— 各因子はクリック可能で、出典に戻れます $$G_c = \href{#eq-io}{\textcolor{#3a5998}{\left(\frac{2g_1 R_d}{\omega_{RF}C_{gs}}\right)^{2}}}\;\href{#eq-vc}{\textcolor{#2a6e5a}{\frac{R_g}{(R_i+R_g)^2+\big[X_g-(\omega_{RF}C_{gs})^{-1}\big]^2}}}\cdot\href{#eq-vdif}{\textcolor{#b44a2f}{\frac{R_L}{(R_d+R_L)^2+X_L^2}}}$$
■ 前因子 $g_1,R_d$■ Prefactor $g_1,R_d$■ 前置係数 $g_1,R_d$$g(t)$ 混波 / $i_o(t)$$g(t)$ mixing / $i_o(t)$$g(t)$ ミキシング / $i_o(t)$ · ■ 輸入分母■ Input denominator■ 入力分母$V_c$ 閘極分壓$V_c$, the gate divider$V_c$ ゲート分圧 · ■ 輸出分母■ Output denominator■ 出力分母$R_d\!\parallel\!Z_L$ 汲極輸出 $V_D^{IF}$$R_d\!\parallel\!Z_L$, the drain output $V_D^{IF}$$R_d\!\parallel\!Z_L$ ドレイン出力 $V_D^{IF}$
完整堆導:從三個前提一路推到 $G_c$(逐步,點擊收合)The full derivation: from the three premises all the way to $G_c$ (step by step, click to collapse)完全な導出:三つの前提から $G_c$ まで一気に(ステップごと、クリックで折りたたみ)

三個前提.The three premises.三つの前提。

(A) available-power 形式:(A) The available-power form:(A) available-power 形式:

$$G_c = \frac{4R_gR_L}{|Z_L|^2}\left|\frac{V_D^{IF}}{V_{RF}}\right|^2$$

(B) 跨導 $-g_1$ × 閘極分壓 × 輸出網路 $R_d\!\parallel\!Z_L$ 三段相乘得轉移比:(B) The transconductance $-g_1$ × the gate divider × the output network $R_d\!\parallel\!Z_L$, multiplied together to give the transfer ratio:(B) 相互コンダクタンス $-g_1$ × ゲート分圧 × 出力回路網 $R_d\!\parallel\!Z_L$ の三段の積で伝達比が得られます:

$$\frac{V_D^{IF}}{V_{RF}} = \frac{-g_1}{1+j\omega_{RF}C_{gs}(R_i+Z_g)}\cdot\frac{R_dZ_L}{R_d+Z_L}$$

① 轉移比取模平方① Take the squared modulus of the transfer ratio① 伝達比の絶対値 2 乗を取る(實數 $g_1,R_d$ 提出):(factoring out the real $g_1,R_d$):(実数 $g_1,R_d$ をくくり出す):

$$\left|\frac{V_D^{IF}}{V_{RF}}\right|^2 = \frac{g_1^2}{|1+j\omega_{RF}C_{gs}(R_i+Z_g)|^2}\cdot\frac{R_d^2|Z_L|^2}{|R_d+Z_L|^2}$$

② 代進 (A).② Substitute into (A).② (A) に代入する。

$$G_c = \frac{4R_gR_L}{|Z_L|^2}\cdot\frac{g_1^2R_d^2|Z_L|^2}{|1+j\omega_{RF}C_{gs}(R_i+Z_g)|^2\,|R_d+Z_L|^2}$$

③ $|Z_L|^2$ 上下相消.③ $|Z_L|^2$ cancels top and bottom.③ $|Z_L|^2$ が上下で相殺。

$$G_c = \frac{4g_1^2R_d^2\,R_gR_L}{|1+j\omega_{RF}C_{gs}(R_i+Z_g)|^2\,|R_d+Z_L|^2}$$

④ 展開輸入分母④ Expand the input denominator④ 入力分母を展開する(代 $Z_g=R_g+jX_g$,拆實部/虛部):(substituting $Z_g=R_g+jX_g$ and separating real and imaginary parts):($Z_g=R_g+jX_g$ を代入し、実部/虚部に分ける):

$$1+j\omega_{RF}C_{gs}(R_i+R_g+jX_g)=\underbrace{(1-\omega_{RF}C_{gs}X_g)}_{\text{實部}}+j\,\underbrace{\omega_{RF}C_{gs}(R_i+R_g)}_{\text{虛部}}$$ $$\Rightarrow\;|\cdots|^2=(1-\omega_{RF}C_{gs}X_g)^2+\omega_{RF}^2C_{gs}^2(R_i+R_g)^2=\omega_{RF}^2C_{gs}^2\Big[(R_i+R_g)^2+\big(X_g-(\omega_{RF}C_{gs})^{-1}\big)^2\Big]$$ $$1+j\omega_{RF}C_{gs}(R_i+R_g+jX_g)=\underbrace{(1-\omega_{RF}C_{gs}X_g)}_{\text{real part}}+j\,\underbrace{\omega_{RF}C_{gs}(R_i+R_g)}_{\text{imaginary part}}$$ $$\Rightarrow\;|\cdots|^2=(1-\omega_{RF}C_{gs}X_g)^2+\omega_{RF}^2C_{gs}^2(R_i+R_g)^2=\omega_{RF}^2C_{gs}^2\Big[(R_i+R_g)^2+\big(X_g-(\omega_{RF}C_{gs})^{-1}\big)^2\Big]$$ $$1+j\omega_{RF}C_{gs}(R_i+R_g+jX_g)=\underbrace{(1-\omega_{RF}C_{gs}X_g)}_{\text{real part}}+j\,\underbrace{\omega_{RF}C_{gs}(R_i+R_g)}_{\text{imaginary part}}$$ $$\Rightarrow\;|\cdots|^2=(1-\omega_{RF}C_{gs}X_g)^2+\omega_{RF}^2C_{gs}^2(R_i+R_g)^2=\omega_{RF}^2C_{gs}^2\Big[(R_i+R_g)^2+\big(X_g-(\omega_{RF}C_{gs})^{-1}\big)^2\Big]$$

(用恆等式 $(1-\omega CX_g)^2=\omega^2C^2\,(X_g-1/\omega C)^2$ 把 $\omega_{RF}^2C_{gs}^2$ 提出)(using the identity $(1-\omega CX_g)^2=\omega^2C^2\,(X_g-1/\omega C)^2$ to factor out $\omega_{RF}^2C_{gs}^2$)(恒等式 $(1-\omega CX_g)^2=\omega^2C^2\,(X_g-1/\omega C)^2$ を用いて $\omega_{RF}^2C_{gs}^2$ をくくり出す)

⑤ 展開輸出分母⑤ Expand the output denominator⑤ 出力分母を展開する($Z_L=R_L+jX_L$):

$$|R_d+Z_L|^2=(R_d+R_L)^2+X_L^2$$

⑥ 兩分母代回.⑥ Substitute both denominators back.⑥ 両分母を代入し戻す。

$$G_c=\frac{4g_1^2R_d^2\,R_gR_L}{\omega_{RF}^2C_{gs}^2\big[(R_i+R_g)^2+(X_g-(\omega_{RF}C_{gs})^{-1})^2\big]\big[(R_d+R_L)^2+X_L^2\big]}$$

⑦ 重新分組⑦ Regroup⑦ 再グループ化:把 $\dfrac{4g_1^2R_d^2}{\omega_{RF}^2C_{gs}^2}=\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$ 收成前因子,$R_g$、$R_L$ 各配上對應分母 → 得證:: collect $\dfrac{4g_1^2R_d^2}{\omega_{RF}^2C_{gs}^2}=\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$ into a prefactor and pair $R_g$ and $R_L$ with their respective denominators → QED::$\dfrac{4g_1^2R_d^2}{\omega_{RF}^2C_{gs}^2}=\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$ を前置係数にまとめ、$R_g$、$R_L$ をそれぞれ対応する分母と組み合わせる → 証明完了:

$$G_c=\underbrace{\left(\frac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2}_{\text{前因子}}\frac{R_g}{(R_i+R_g)^2+[X_g-(\omega_{RF}C_{gs})^{-1}]^2}\cdot\frac{R_L}{(R_d+R_L)^2+X_L^2}$$ $$G_c=\underbrace{\left(\frac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2}_{\text{prefactor}}\frac{R_g}{(R_i+R_g)^2+[X_g-(\omega_{RF}C_{gs})^{-1}]^2}\cdot\frac{R_L}{(R_d+R_L)^2+X_L^2}$$ $$G_c=\underbrace{\left(\frac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2}_{\text{prefactor}}\frac{R_g}{(R_i+R_g)^2+[X_g-(\omega_{RF}C_{gs})^{-1}]^2}\cdot\frac{R_L}{(R_d+R_L)^2+X_L^2}$$
展開細節 + 共軛匹配如何給出最大值(點擊展開)The expansion in detail + how conjugate matching gives the maximum (click to expand)展開の詳細 + 共役整合がどのように最大値を与えるか(クリックで展開)

輸入分母.The input denominator.入力分母。 以 $Z_g=R_g+jX_g$ 代入,$|1+j\omega_{RF}C_{gs}(R_i+Z_g)|^2=\omega_{RF}^2C_{gs}^2\big[(R_i+R_g)^2+(X_g-(\omega_{RF}C_{gs})^{-1})^2\big]$;提出的 $\omega_{RF}^2C_{gs}^2$ 併進前因子 $\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$。 Substituting $Z_g=R_g+jX_g$: $|1+j\omega_{RF}C_{gs}(R_i+Z_g)|^2=\omega_{RF}^2C_{gs}^2\big[(R_i+R_g)^2+(X_g-(\omega_{RF}C_{gs})^{-1})^2\big]$; the $\omega_{RF}^2C_{gs}^2$ factored out is absorbed into the prefactor $\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$. $Z_g=R_g+jX_g$ を代入すると $|1+j\omega_{RF}C_{gs}(R_i+Z_g)|^2=\omega_{RF}^2C_{gs}^2\big[(R_i+R_g)^2+(X_g-(\omega_{RF}C_{gs})^{-1})^2\big]$、くくり出した $\omega_{RF}^2C_{gs}^2$ は前置係数 $\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$ に組み込まれます。

輸出分母.The output denominator.出力分母。 $|R_d+Z_L|^2=(R_d+R_L)^2+X_L^2$;分子的 $R_d^2|Z_L|^2$ 與 $G_c$ 前因子的 $|Z_L|^2$ 相消。 $|R_d+Z_L|^2=(R_d+R_L)^2+X_L^2$; the $R_d^2|Z_L|^2$ of the numerator cancels the $|Z_L|^2$ of the $G_c$ prefactor. $|R_d+Z_L|^2=(R_d+R_L)^2+X_L^2$、分子の $R_d^2|Z_L|^2$ と $|Z_L|^2$ 前置係数の $G_c$ が相殺します。

前因子 $\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$ 怎麼湊出來.How the prefactor $\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$ is assembled.前置係数 $\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$ はどう組み立てられるか。 把上面散落的常數收在一起 — $4$(來自 $4R_gR_L$ 的 $0.5\times8$)、$g_1^2$(跨導)、$R_d^2$(輸出)、以及從輸入分母提出的 $1/\omega_{RF}^2C_{gs}^2$: Gather the constants scattered above — the $4$ (from the $0.5\times8$ of $4R_gR_L$), the $g_1^2$ (transconductance), the $R_d^2$ (output), and the $1/\omega_{RF}^2C_{gs}^2$ factored out of the input denominator: 上で散らばっていた定数をまとめます — $4$($0.5\times8$ の $4R_gR_L$ 由来)、$g_1^2$(相互コンダクタンス)、$R_d^2$(出力)、および入力分母からくくり出した $1/\omega_{RF}^2C_{gs}^2$:

$$\underbrace{4}_{\substack{\text{avail. power}\\0.5\times8}}\cdot\underbrace{g_1^2}_{\text{跨導}}\cdot\underbrace{R_d^2}_{\text{輸出}}\cdot\underbrace{\frac{1}{\omega_{RF}^2C_{gs}^2}}_{\substack{\text{輸入分母}\\\text{提出}}}=\frac{(2g_1R_d)^2}{(\omega_{RF}C_{gs})^2}=\left(\frac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$$ $$\underbrace{4}_{\substack{\text{avail. power}\\0.5\times8}}\cdot\underbrace{g_1^2}_{\text{transconductance}}\cdot\underbrace{R_d^2}_{\text{output}}\cdot\underbrace{\frac{1}{\omega_{RF}^2C_{gs}^2}}_{\substack{\text{factored from the}\\\text{input denominator}}}=\frac{(2g_1R_d)^2}{(\omega_{RF}C_{gs})^2}=\left(\frac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$$ $$\underbrace{4}_{\substack{\text{avail. power}\\0.5\times8}}\cdot\underbrace{g_1^2}_{\text{transconductance}}\cdot\underbrace{R_d^2}_{\text{output}}\cdot\underbrace{\frac{1}{\omega_{RF}^2C_{gs}^2}}_{\substack{\text{factored from the}\\\text{input denominator}}}=\frac{(2g_1R_d)^2}{(\omega_{RF}C_{gs})^2}=\left(\frac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2$$

所以那個「$2$」其實是 $\sqrt{4}$,根源是 available-power 比值的常數,跟訊號物理無關。So that “$2$” is really $\sqrt{4}$, rooted in the constant of the available-power ratio and nothing to do with the physics of the signal.つまりあの「$2$」は実際には $\sqrt{4}$ であり、その根源は available-power 比の定数であって、信号の物理とは無関係です。

共軛匹配 (conjugate match)Conjugate match共役整合(conjugate match) — 讓兩段分數各自最大: — maximizing each of the two fractions separately: — 二つの分数をそれぞれ最大にします:

  • 輸入:$X_g=(\omega_{RF}C_{gs})^{-1}$ 抵消電抗,再對 $R_g$ 最佳化 → $R_g=R_i$,得 $\dfrac{R_g}{(R_i+R_g)^2}\to\dfrac{1}{4R_i}$。Input: $X_g=(\omega_{RF}C_{gs})^{-1}$ cancels the reactance, then optimizing over $R_g$ → $R_g=R_i$, giving $\dfrac{R_g}{(R_i+R_g)^2}\to\dfrac{1}{4R_i}$.入力:$X_g=(\omega_{RF}C_{gs})^{-1}$ がリアクタンスを打ち消し、さらに $R_g$ について最適化 → $R_g=R_i$、$\dfrac{R_g}{(R_i+R_g)^2}\to\dfrac{1}{4R_i}$ が得られます。
  • 輸出:$X_L=0$、$R_L=R_d$,得 $\dfrac{R_L}{(R_d+R_L)^2}\to\dfrac{1}{4R_d}$。Output: $X_L=0$ and $R_L=R_d$, giving $\dfrac{R_L}{(R_d+R_L)^2}\to\dfrac{1}{4R_d}$.出力:$X_L=0$、$R_L=R_d$ で $\dfrac{R_L}{(R_d+R_L)^2}\to\dfrac{1}{4R_d}$ が得られます。

為什麼是這些值 — 兩個獨立動作:Why these values — two independent actions:なぜこれらの値なのか — 二つの独立した操作:

① 抵消電抗(分母虛部歸零).① Cancel the reactance (zeroing the imaginary part of the denominator).① リアクタンスを打ち消す(分母の虚部をゼロにする)。 輸入分母含 $[X_g-(\omega_{RF}C_{gs})^{-1}]^2$、輸出含 $X_L^2$,都是「加上去只會變大」的平方項。令 $X_g=(\omega_{RF}C_{gs})^{-1}$、$X_L=0$ 把它們歸零,分母最小。物理上 $C_{gs}$ 電抗為 $-1/\omega_{RF}C_{gs}$,源端擺 $+1/\omega_{RF}C_{gs}$(電感性) The input denominator contains $[X_g-(\omega_{RF}C_{gs})^{-1}]^2$ and the output contains $X_L^2$, both squared terms that can only add to the total. Setting $X_g=(\omega_{RF}C_{gs})^{-1}$ and $X_L=0$ zeroes them, minimizing the denominators. Physically, the reactance of $C_{gs}$ is $-1/\omega_{RF}C_{gs}$, and putting $+1/\omega_{RF}C_{gs}$ (inductive) at the source end 入力分母は $[X_g-(\omega_{RF}C_{gs})^{-1}]^2$ を、出力は $X_L^2$ を含み、いずれも「加われば大きくなるだけ」の 2 乗項です。$X_g=(\omega_{RF}C_{gs})^{-1}$、$X_L=0$ とすればこれらはゼロになり、分母は最小になります。物理的には $C_{gs}$ のリアクタンスは $-1/\omega_{RF}C_{gs}$ であり、信号源側に $+1/\omega_{RF}C_{gs}$(誘導性)を置くと諧振抵消cancels it in resonance共振で打ち消され,RF 電壓才全落在閘極。, so that the whole RF voltage lands on the gate.、RF 電圧がすべてゲートに掛かります。

② 電阻匹配(對 $R$ 微分取極值).② Resistance matching (differentiate with respect to $R$ and find the extremum).② 抵抗整合($R$ で微分して極値を求める)。 電抗歸零後輸入剩 $\dfrac{R_g}{(R_i+R_g)^2}$,$\dfrac{d}{dR_g}=0$: With the reactance zeroed, the input reduces to $\dfrac{R_g}{(R_i+R_g)^2}$, and $\dfrac{d}{dR_g}=0$ gives: リアクタンスをゼロにすると入力は $\dfrac{R_g}{(R_i+R_g)^2}$ だけとなり、$\dfrac{d}{dR_g}=0$:

$$\frac{(R_i+R_g)^2-R_g\cdot2(R_i+R_g)}{(R_i+R_g)^4}=0\;\Rightarrow\;(R_i+R_g)-2R_g=0\;\Rightarrow\;R_g=R_i$$

which is the これは最大功率傳輸定理maximum power transfer theorem最大電力伝送の定理:源電阻 = 負載(閘極)電阻;輸出同理 $R_L=R_d$。四條件合起來就是輸入/輸出兩端的共軛匹配(電阻相等、電抗抵消)。: the source resistance = the load (gate) resistance; and likewise $R_L=R_d$ at the output. The four conditions together are conjugate matching at both the input and the output (equal resistances, cancelled reactances).です:信号源抵抗 = 負荷(ゲート)抵抗であり、出力も同様に $R_L=R_d$ です。四つの条件を合わせると、入力/出力両端での共役整合(抵抗が等しく、リアクタンスが打ち消される)になります。

代回:$\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2\cdot\dfrac{1}{4R_i}\cdot\dfrac{1}{4R_d}=\dfrac{g_1^2R_d}{4\omega_{RF}^2C_{gs}^2R_i}$。Substituting back: $\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2\cdot\dfrac{1}{4R_i}\cdot\dfrac{1}{4R_d}=\dfrac{g_1^2R_d}{4\omega_{RF}^2C_{gs}^2R_i}$.代入し戻すと:$\left(\dfrac{2g_1R_d}{\omega_{RF}C_{gs}}\right)^2\cdot\dfrac{1}{4R_i}\cdot\dfrac{1}{4R_d}=\dfrac{g_1^2R_d}{4\omega_{RF}^2C_{gs}^2R_i}$。

分母帶 $\omega_{RF}^2C_{gs}^2$ → The $\omega_{RF}^2C_{gs}^2$ in the denominator means 分母に $\omega_{RF}^2C_{gs}^2$ がある → 頻率越高 / $C_{gs}$ 越大,增益越小the higher the frequency and the larger $C_{gs}$, the lower the gain周波数が高いほど / $C_{gs}$ が大きいほど、利得は小さくなります(junction 電容把 RF 旁路掉)。(the junction capacitance shunts the RF away).(接合容量が RF を分流してしまうため)。

Maximum conversion gain (conjugate match at both RF and IF ports):

Max conversion gain — MESFET mixer $$G_{c,\max} = \frac{g_1^2 R_d}{4\omega_{RF}^2 C_{gs}^2 R_i}$$

4.2.11 Balanced Transistor Mixers

Single-Balanced Transistor Mixer (Differential Pair)

A differential transistor pair acts as a bipolar switch driven by the LO. A third transistor in common-source/emitter converts the RF voltage to current for the differential pair. Baluns on LO and IF ports provide balanced-to-unbalanced transitions.

IF output — single-balanced transistor mixer $$i_{IF}(t) = \frac{2g_P}{\pi}\,V_{RF}\cos\omega_{IF}t = \frac{2g_m}{\pi}\,V_{RF}\cos\omega_{IF}t$$
Active vs Passive Balun
Passive baluns (transformers, hybrids) offer better input match and isolation. Active baluns can be integrated into ICs but degrade noise figure and linearity. For IC integration, active baluns are usually acceptable in non-base-station applications.
H8 SBM vs DBM: single-balanced transistor mixer; differential pair switched by LO S(t), common-source RF transconductor g_p=g_m
Lecture slide 41 — single-balanced transistor mixer: differential pair switched by LO + common-source RF transconductor + LO/IF baluns.
Slide 41 — text transcript
  • A single-balanced transistor mixer uses two transistors in a differential configuration. The pair operates as an alternating switch, with the LO turning the two transistors on and off on alternate cycles. The switching waveform is similar to an ideal bipolar switching mixer.
  • A third transistor in common-source / common-emitter configuration operates in saturation to convert RF voltage at gate/base into RF drain/collector current, which is then pumped into the differential pair via the LO.
  • The balun (balanced-to-unbalanced) networks on the LO and IF ports provide a transition between a two-wire differential signal and a single-wire single-ended signal relative to ground.
  • Both passive and active baluns may be used. Passive baluns offer better input match and isolation, but active baluns are required for IC integration.
  • IF current: $i_{IF}(t) = \dfrac{2g_p V_{RF}}{\pi}\cos(\omega_{RF}-\omega_{LO})t = \dfrac{2g_m V_{RF}}{\pi}\cos\omega_{IF}t$, with $g_p = g_m$.
  • (Note "H8: SBM vs DBM" — homework prompt to compare single- vs double-balanced topologies.)

Double-Balanced Transistor Mixer — Gilbert Cell

Two single-balanced mixers share a common RF transconductor stage. Proposed by Gilbert (1967). Requires baluns on all three ports (RF, LO, IF).

  • High conversion gain
  • Low LO power requirement
  • Good isolation between all ports
  • Good linearity (even-order cancellation)
  • Monolithic integration — dominant in wireless front-end ICs
H8 SBM vs DBM: Gilbert cell double-balanced transistor mixer; LO and RF baluns highlighted (both differential signals)
Lecture slide 42 — Gilbert cell: two single-balanced mixers sharing a common RF transconductor; baluns on RF, LO, IF.
Slide 42 — text transcript
  • The double-balanced transistor mixer, originally conceived by Gilbert in 1967, consists of two single-balanced mixers. Baluns are required on all RF, LO and IF ports.
  • Gilbert-cell mixers are dominant in wireless front-end ICs because of: (a) high conversion gain, (b) low LO power, (c) good isolation, (d) good linearity, (e) monolithic integration capability.

Summary (slide 42 bullets):

  • Single-ended mixers rely on external filters (diplexer at input, LPF at output) to achieve port-to-port isolation.
  • Single-balanced mixers are commonly used to achieve better input match and isolation than single-ended mixers.
  • Double-balanced mixers are commonly used to achieve better conversion gain/loss and linearity than single-balanced mixers.

📝 HW8:Single-Balanced (SBM) vs Double-Balanced (DBM) Transistor Mixer📝 HW8: Single-Balanced (SBM) vs Double-Balanced (DBM) Transistor Mixer📝 HW8:単平衡(SBM)vs 二重平衡(DBM)トランジスタミキサ

作業要求:The assignment:課題の要求:比較 Compare the topology, signal balance, balun requirements and performance of the 以下では單平衡晶體混頻器(SBM,differential transistor mixer)single-balanced transistor mixer (SBM, a differential transistor mixer)単平衡トランジスタミキサ(SBM、differential transistor mixer) and 雙平衡 Gilbert cell 混頻器(DBM)double-balanced Gilbert cell mixer (DBM)二重平衡 Gilbert cell ミキサ(DBM)的拓樸、訊號平衡性、baluns 需求與性能差異。下面整理兩張投影片的內容、電路圖、IF 輸出推導與比較表。 . Below is a summary of the two slides, the schematics, the IF-output derivation and a comparison table. のトポロジー、信号の平衡性、バラン要求、性能の違いを比較します。以下に 2 枚のスライドの内容、回路図、IF 出力の導出と比較表をまとめます。

(A) Single-Balanced Transistor Mixer(差動晶體混頻器)(A) Single-balanced transistor mixer (a differential transistor mixer)(A) 単平衡トランジスタミキサ(差動トランジスタミキサ)

單平衡晶體混頻器使用A single-balanced transistor mixer uses 単平衡トランジスタミキサは兩顆電晶體組成差動對(differential pair)two transistors forming a differential pair2 個のトランジスタで構成した差動対(differential pair)。這對差動對當作. The differential pair acts as an を使用します。この差動対は交替開關(alternating switch)alternating switch交互スイッチ(alternating switch):LO 在每個半週期把兩顆電晶體輪流打開/關閉,產生的切換波形 $S(t)$ 與理想 bipolar switching mixer 相同。另一顆: on each half-cycle the LO turns the two transistors on and off in turn, producing a switching waveform $S(t)$ identical to that of an ideal bipolar switching mixer. A separate として働きます:LO が半周期ごとに 2 個のトランジスタを交互にオン/オフし、生じるスイッチング波形 $S(t)$ は理想的な bipolar switching mixer と同じです。別の共源/共射(common-source / common-emitter)電晶體工作在飽和區common-source / common-emitter transistor operating in saturation飽和領域で動作するソース接地/エミッタ接地(common-source / common-emitter)トランジスタ,把加在 gate/base 的 RF converts the RF applied to its gate/base が、gate/base に加えられた RF 電壓Voltage電圧訊號轉成 drain/collector 的 RF into an RF current at the drain/collector 信号を drain/collector の RF 電流Current電流,再被 LO 的開關動作「抽運(pumped)」。, which is then “pumped” by the switching action of the LO.に変換し、それが LO のスイッチング動作によって「ポンピング(pumped)」されます。

Balun(balanced-to-unbalanced)Balun (balanced-to-unbalanced)バラン(balanced-to-unbalanced)網路裝在 LO 與 IF 埠上,負責在「對地平衡(差動)的兩線訊號」與「對地不平衡(單端)的單線訊號」之間轉換。被動與主動 balun 都可使用: networks are fitted at the LO and IF ports, converting between “a two-wire signal balanced about ground (differential)” and “a single-wire signal unbalanced with respect to ground (single-ended)”. Both passive and active baluns can be used: 回路網は LO と IF ポートに設けられ、「接地に対して平衡な(差動)2 線信号」と「接地に対して不平衡な(シングルエンド)1 線信号」との間を変換します。受動バランも能動バランも使用できます:被動 baluna passive balun受動バラン 提供較佳的輸入匹配與隔離,但 gives a better input match and isolation but は入力整合とアイソレーションが優れますが、無法整合進 ICcannot be integrated on an ICIC に集積できません; 主動 balunan active balun能動バラン 可整合但會犧牲 NF 與線性度。 can be integrated but sacrifices NF and linearity. は集積できますが、NF と直線性を犠牲にします。

Single-balanced transistor mixer — IF balun, LO balun differential switch pair S(t), common-source RF transconductor g_p=g_m
單平衡晶體混頻器:差動對 M1/M2 由 LO 切換(開關函數 $S(t)$),共源 M3 把 RF 電壓轉成電流。LO、IF 兩埠用 balun;RF 為單端輸入。A single-balanced transistor mixer: the differential pair M1/M2 is switched by the LO (the switching function $S(t)$), while the common-source M3 converts the RF voltage into a current. Baluns are used at the LO and IF ports; the RF input is single-ended.単平衡トランジスタミキサ:差動対 M1/M2 は LO で切り替えられ(スイッチング関数 $S(t)$)、ソース接地の M3 が RF 電圧を電流に変換。LO と IF の 2 ポートにバランを使用、RF はシングルエンド入力。
IF 輸出(單平衡晶體混頻器)The IF output (single-balanced transistor mixer)IF 出力(単平衡トランジスタミキサ) $$i_{IF}(t) = \frac{2g_p V_{RF}}{\pi}\cos(\omega_{RF}-\omega_{LO})\,t = \frac{2g_m V_{RF}}{\pi}\cos\omega_{IF}\,t \qquad (g_p = g_m)$$

係數 $2/\pi$ 來自方波開關函數 $S(t)$ 基頻的 Fourier 係數($4/\pi$ 的一半,因為只取一個 sideband)。M3 的跨導 $g_m$ 設定 RF→電流的轉換增益。The coefficient $2/\pi$ comes from the fundamental Fourier coefficient of the square-wave switching function $S(t)$ (half of $4/\pi$, since only one sideband is taken). M3's transconductance $g_m$ sets the RF→current conversion gain.係数 $2/\pi$ は方形波スイッチング関数 $S(t)$ の基本波フーリエ係数に由来します($4/\pi$ の半分、側波帯を 1 つしか取らないため)。M3 の相互コンダクタンス $g_m$ が RF→電流の変換利得を決めます。

(B) Double-Balanced (Gilbert Cell) Mixer

雙平衡晶體混頻器由 The double-balanced transistor mixer was proposed by 二重平衡トランジスタミキサは Gilbert 於 1967 年Gilbert in 1967Gilbert が 1967 年に提出,結構上是; structurally it is 提案したもので、構造的には兩個單平衡混頻器疊起來two single-balanced mixers stacked together2 つの単平衡ミキサを積み重ねたもの.です。RF、LO、IF 三個埠都需要 balunall three ports — RF, LO and IF — need balunsRF、LO、IF の 3 ポートすべてにバランが必要(三者皆為差動訊號)。(all three being differential signals).です(3 つとも差動信号のため)。

Gilbert cell 是The Gilbert cell is the Gilbert cell は無線前端 IC 的主流mainstay of wireless front-end ICs無線フロントエンド IC の主流,因為具備:(a) 高轉換增益、(b) 低 LO 功率需求、(c) 良好的埠間隔離、(d) 良好的線性度、(e) 可單晶整合(monolithic integration)。, because it offers: (a) high conversion gain, (b) a low LO power requirement, (c) good port-to-port isolation, (d) good linearity, and (e) monolithic integration.です。理由は (a) 高い変換利得、(b) 低い LO 電力要求、(c) 良好なポート間アイソレーション、(d) 良好な直線性、(e) モノリシック集積が可能、という点にあります。

Double-balanced Gilbert-cell mixer — IF/LO/RF baluns, all differential signals
Gilbert cell:上方 switching quad(M1–M4,兩個交叉耦合差動對)由 LO 差動驅動;下方 M5/M6 為 RF 差動跨導級。RF、LO、IF 三埠皆需 balun。= 兩個單平衡混頻器疊加。The Gilbert cell: the upper switching quad (M1–M4, two cross-coupled differential pairs) is driven differentially by the LO; the lower M5/M6 form the RF differential transconductance stage. All three ports — RF, LO and IF — need baluns. = two single-balanced mixers stacked.Gilbert cell:上側のスイッチングクワッド(M1–M4、2 組の交差結合差動対)は LO で差動駆動され、下側の M5/M6 が RF 差動相互コンダクタンス段。RF、LO、IF の 3 ポートすべてにバランが必要。= 単平衡ミキサ 2 つの積み重ね。

(C) SBM vs DBM 比較(C) SBM vs DBM compared(C) SBM と DBM の比較

項目Item項目 Single-Balanced (SBM) Double-Balanced (Gilbert / DBM)
結構Structure構造 1 個差動對開關 + 1 個共源跨導級1 differential-pair switch + 1 common-source transconductance stage差動対スイッチ 1 個 + ソース接地相互コンダクタンス段 1 個 2 個單平衡混頻器(switching quad + 差動跨導對)2 single-balanced mixers (a switching quad + a differential transconductance pair)単平衡ミキサ 2 個(switching quad + 差動相互コンダクタンス対)
Balun 需求Balun requirementバラン要求 LO、IF(RF 單端)LO and IF (RF single-ended)LO と IF(RF はシングルエンド) RF、LO、IF 三者全要all three: RF, LO and IFRF、LO、IF の 3 つすべて
RF 訊號RF signal'sRF 信号 單端(single-ended)single-endedシングルエンド(single-ended) 差動(differential)differential差動(differential)
LO 抑制LO suppressionLO 抑制 LO feed-through 部分抵消(僅一側平衡)LO feedthrough partly cancelled (balanced on one side only)LO フィードスルーは部分的に相殺(片側のみ平衡) LO 與 RF feed-through 都抵消(雙側平衡)✓both LO and RF feedthrough cancelled (balanced on both sides) ✓LO と RF のフィードスルーがともに相殺(両側平衡)✓
偶次諧波Even harmonics偶数次高調波 僅抑制 LO 偶次only even LO harmonics suppressedLO の偶数次のみ抑制 RF 與 LO 偶次都抑制 ✓both even RF and even LO harmonics suppressed ✓RF と LO の偶数次をともに抑制 ✓
埠間隔離Port-to-port isolationポート間アイソレーション 中等moderate中程度 優(best)✓best ✓最良(best)✓
線性度 / IP3Linearity / IP3直線性 / IP3 中等moderate中程度 較高 ✓higher ✓より高い ✓
IC 整合IC integrationIC 集積 可,但 RF 單端較不利共模抑制possible, but a single-ended RF is less favourable for common-mode rejection可能だが、RF がシングルエンドのため同相除去には不利 最適合單晶整合(無線前端主流)✓best suited to monolithic integration (the mainstay of wireless front ends) ✓モノリシック集積に最適(無線フロントエンドの主流)✓
面積 / 功耗Area / power面積 / 消費電力 較小、較省smaller and more frugal小さく、省電力 較大(多一倍電晶體 + 多一個 balun)larger (twice the transistors plus one more balun)大きい(トランジスタ 2 倍 + バラン 1 個追加)
一句話總結In one sentence一言でまとめると
SBM=一個差動對開關+一個共源 RF 跨導級,只在 LO/IF 用 balun,RF 是單端 → 結構簡單、省面積,但只有「單側平衡」。= one differential-pair switch plus one common-source RF transconductance stage, with baluns only at LO/IF and a single-ended RF → simple and area-efficient, but balanced on one side only.=差動対スイッチ 1 個+ソース接地 RF 相互コンダクタンス段 1 個、バランは LO/IF のみ、RF はシングルエンド → 構造が単純で面積効率が良い一方、平衡は「片側のみ」です。
DBM(Gilbert cell)DBM (the Gilbert cell)DBM(Gilbert cell)=把兩個 SBM 疊起來,三埠全差動全 balun → LO 與 RF feed-through 雙雙抵消、線性度與隔離最佳,是無線前端 IC 的標準選擇,代價是面積與功耗加倍。 = two SBMs stacked, with all three ports differential and all with baluns → both LO and RF feedthrough cancel, giving the best linearity and isolation. It is the standard choice for wireless front-end ICs, at the cost of double the area and power. = SBM を 2 つ積み重ねたもので、3 ポートすべてが差動でバラン付き → LO と RF のフィードスルーがともに相殺され、直線性とアイソレーションは最良です。無線フロントエンド IC の標準的な選択肢であり、代償は面積と消費電力が倍になることです。
為什麼 Gilbert cell 是「double-balanced」?Why is the Gilbert cell “double-balanced”?なぜ Gilbert cell は「double-balanced」なのか?
第一重平衡The first balance第一の平衡來自 LO 開關 quad 的差動驅動(抵消 LO feed-through); comes from the differential drive of the LO switching quad (cancelling LO feedthrough); は LO スイッチングクワッドの差動駆動によるもの(LO フィードスルーを相殺);第二重平衡the second balance第二の平衡來自 RF 跨導級的差動輸入(抵消 RF feed-through)。兩重平衡疊加 → RF 與 LO 都不會洩漏到 IF,只有想要的 $\omega_{RF}\pm\omega_{LO}$ 乘積項留下來。SBM 只有 LO 那一重平衡,所以 RF feed-through 仍會出現在輸出。 comes from the differential input of the RF transconductance stage (cancelling RF feedthrough). With both balances superimposed → neither RF nor LO leaks into the IF, and only the wanted $\omega_{RF}\pm\omega_{LO}$ product terms survive. An SBM has only the LO balance, so RF feedthrough still appears at its output. は RF 相互コンダクタンス段の差動入力によるもの(RF フィードスルーを相殺)。二重の平衡が重なると → RF も LO も IF に漏れず、必要な $\omega_{RF}\pm\omega_{LO}$ の積の項だけが残ります。SBM は LO の平衡しか持たないため、RF フィードスルーは依然として出力に現れます。

Summary — Mixer Topology Comparison

Topology Match (RF) LO-RF Isolation LO AM Noise Even Harmonics IP3 LO Power
Single-ended diode Poor Poor No No Low Low
Single-balanced (90°) Good ✓ Moderate Yes ✓ LO only Medium Medium
Single-balanced (180°) Moderate Good ✓ Yes ✓ LO only Medium Medium
Double-balanced (ring) Moderate Good ✓ Yes ✓ RF & LO ✓ High ✓ High
Image reject diode Good Good Yes ✓ Yes Medium Medium
Gilbert cell Good Good ✓ Yes ✓ RF & LO ✓ Medium Low ✓
Single-ended

Rely on external diplexer and LPF for isolation. Simplest but worst all-round performance.

Single-balanced

Better match and isolation than single-ended. 90° hybrid → match; 180° hybrid → LO-RF isolation. Both suppress LO AM noise.

Double-balanced

Best linearity and even-harmonic rejection. Highest LO power demand. Dominant in base-station diode applications.

Gilbert cell

IC-integrable double-balanced transistor mixer. Low LO drive. Dominant in wireless front-end ICs (phones, WiFi).

Practice Example 2 — Image Frequency Calculation

Problem: A receiver uses $f_{LO} = 2.1$ GHz and $f_{IF} = 200$ MHz. (a) What is the desired RF frequency? (b) What is the image frequency? (c) What type of filter should be placed before the mixer to suppress the image?

Click to reveal solution

(a) $f_{RF} = f_{LO} + f_{IF} = 2.1 + 0.2 = 2.3$ GHz (using upper sideband selection)

(b) $f_{IM} = f_{LO} - f_{IF} = 2.1 - 0.2 = 1.9$ GHz

Note: $f_{RF} - f_{IM} = 2f_{IF} = 400$ MHz — the two are separated by twice the IF.

(c) A bandpass filter (image-reject filter) centered at 2.3 GHz with a stopband at 1.9 GHz. Since the separation is 400 MHz, the filter must have a steep roll-off. A higher IF frequency would increase the separation and ease the filter requirement.

Practice Example 3 — DSB vs SSB Noise Figure

Problem: A diode mixer has a measured DSB noise figure of $NF_{DSB} = 5.5$ dB and conversion loss $L_C = 6$ dB. (a) What is the SSB noise figure? (b) If this mixer follows an LNA with $G=12$ dB and $NF=2$ dB, find the cascaded $NF$ using the SSB noise figure.

Click to reveal solution

(a) $NF_{SSB} = NF_{DSB} + 3 = 5.5 + 3 = 8.5$ dB, or equivalently $F_{SSB} = 2F_{DSB} = 2 \times 10^{0.55} \approx 7.08$.

(b) Cascaded noise figure (Friis): $$F_{LNA} = 10^{0.2} \approx 1.585, \quad F_{mix} = 10^{0.85} \approx 7.079, \quad G_{LNA} = 10^{1.2} \approx 15.85$$ $$F_{casc} = 1.585 + \frac{7.079-1}{15.85} = 1.585 + 0.384 = 1.969$$ $$NF_{casc} = 10\log_{10}(1.969) \approx 2.94\;\text{dB}$$

Appendix — dB Cheat Sheet(必背 dB 換算表)Appendix — dB Cheat SheetAppendix — dB Cheat Sheet(必修の dB 換算表)

Power ratios use $10\log_{10}$; voltage / field / current ratios use $20\log_{10}$. Memorize the bold rows — every other value is built from them.

Power ratio   $\mathrm{dB} = 10\log_{10}(P_2/P_1)$

dB Linear (power) Notes
$-30$$0.001$ (1/1000)
$-20$$0.01$ (1/100)
$-10$$0.1$ (1/10)
$-6$$\approx 0.25$ (1/4)
$-3$$\approx 0.5$ (1/2)"half-power" point
$-1$$\approx 0.794$
$0$$1$unity
$1$$\approx 1.259$$10^{0.1}$
$3$$\approx 2$"double-power" rule
$6$$\approx 4$$2\times 2$
$10$$10$decade
$13$$\approx 20$$10+3$
$20$$100$$10\times 10$
$30$$1000$
$40$$10^4$
$60$$10^6$

Voltage / field ratio   $\mathrm{dB} = 20\log_{10}(V_2/V_1)$

dB Linear (voltage) Notes
$-20$$0.1$
$-6$$\approx 0.5$ (1/2)
$-3$$\approx 0.707$$1/\sqrt 2$
$0$$1$unity
$3$$\approx 1.414$$\sqrt 2$
$6$$\approx 2$"double-voltage" rule
$10$$\approx 3.162$$\sqrt{10}$
$20$$10$decade in voltage
$40$$100$
$60$$1000$

Absolute power (dBm, dBW)

Value Linear power Notes
$-90$ dBm$10^{-12}$ W (1 pW)typical RX sensitivity floor
$-30$ dBm$10^{-6}$ W (1 µW)
$0$ dBm$1$ mWreference
$10$ dBm$10$ mWtypical mixer LO drive
$30$ dBm$1$ W= 0 dBW
$0$ dBW$1$ W
Mental-arithmetic recipe

Any integer dB value can be reconstructed from three building blocks:

e.g. 23 dB (power) $= 20 + 3 = 100 \times 2 = 200$.   $-7$ dB $= -10 + 3 = 0.1 \times 2 = 0.2$.

Appendix — Bessel Functions & FM/PM Sidebands

Bessel functions $J_n(x)$ are the cylindrical analog of $\sin/\cos$ — they appear in any RF problem with a sinusoidally pumped phase: the angle-modulated (FM/PM) spectrum, and the large-signal mixer driven by a sinusoidal LO. The bridge between the two is the Jacobi–Anger expansion.

Why a mixer chapter cares about Bessel functions

When a nonlinear device's phase is driven by a strong sinusoid (the LO), $e^{j\beta\sin\omega t}$ expands into a comb of harmonics whose weights are $J_n(\beta)$. The same math describes the sidebands of an FM signal of modulation index $\beta$. So the harmonic content of a sine-pumped mixer and the spectrum of an FM carrier are governed by the identical Bessel comb.

Jacobi–Anger expansion (the key identity)

Sine-driven phase → Bessel comb $$e^{\,j\beta\sin\omega t} = \sum_{n=-\infty}^{\infty} J_n(\beta)\,e^{\,jn\omega t} \;\;\Longrightarrow\;\; \cos(\omega_c t + \beta\sin\omega_m t) = \sum_{n=-\infty}^{\infty} J_n(\beta)\,\cos\big((\omega_c + n\omega_m)t\big)$$

So an FM/PM carrier of modulation index $\beta = \Delta f / f_m$ (FM) or $\beta = $ peak phase deviation (PM) breaks into a carrier line plus sideband pairs spaced by $f_m$, with amplitude $J_n(\beta)$ on the $n$-th pair:

FM/PM line spectrum $$v(t) = A\Big[\,\underbrace{J_0(\beta)\cos\omega_c t}_{\text{carrier}} + \sum_{n=1}^{\infty} J_n(\beta)\big\{\cos(\omega_c + n\omega_m)t + (-1)^n\cos(\omega_c - n\omega_m)t\big\}\Big]$$

$J_0$–$J_3$ at a glance

2.4055.520 8.65411.79 024 681012 1.00.50-0.5 argument  x  (= modulation index β for FM) J₀(x) J₁(x) J₂(x) J₃(x)
$J_0$ starts at 1 (carrier), higher orders start at 0 and "turn on" only once $\beta \gtrsim n$. Dashed red lines mark the zeros of $J_0$ — the carrier-null deviations.

Sideband amplitude table $J_n(\beta)$

$\beta$ $J_0$ (carrier) $J_1$ $J_2$ $J_3$ $J_4$ $J_5$
$0.2$ (NBFM)$0.990$$0.100$$0.005$
$0.5$$0.938$$0.242$$0.031$$0.003$
$1$$0.765$$0.440$$0.115$$0.020$$0.002$
$2$$0.224$$0.577$$0.353$$0.129$$0.034$$0.007$
$2.405$$0.000$ ⚑$0.519$$0.432$$0.199$$0.065$$0.016$
$5$$-0.178$$-0.328$$0.047$$0.365$$0.391$$0.261$
$5.520$$0.000$ ⚑$-0.340$$-0.123$$0.251$$0.396$$0.323$

⚑ = carrier null ($J_0=0$). Note the spectrum stays bounded: $\displaystyle\sum_{n=-\infty}^{\infty} J_n^2(\beta) = 1$ — every row's squared entries (carrier + both sidebands) sum to 1, so FM/PM redistributes power without changing total power.

Carrier nulls — deviation calibration

Because $J_0(\beta)=0$ at fixed values of $\beta$, sweeping the deviation until the carrier disappears on a spectrum analyzer gives an absolute measurement of $\beta=\Delta f/f_m$ — independent of analyzer amplitude calibration. The first few nulls:

Zeros of $J_0$ (carrier-null $\beta$ values) $$\beta = 2.405,\;\; 5.520,\;\; 8.654,\;\; 11.792,\;\; 14.931,\;\ldots$$
Example — Bessel-null deviation check
Given: An FM source modulated at $f_m = 1\text{ kHz}$. The deviation is increased from zero; the carrier line first vanishes, reappears, and vanishes a second time on the analyzer.
Find: the peak deviation $\Delta f$ at the second null.
Solution: The 2nd zero of $J_0$ is $\beta = 5.520$. With $\beta = \Delta f/f_m$: $$\Delta f = \beta\,f_m = 5.520 \times 1\text{ kHz} = 5.52\text{ kHz}.$$ The first null ($\beta=2.405$) would have been $\Delta f = 2.405$ kHz — so each successive null is a precise, calibration-free deviation checkpoint.

Bandwidth — Carson's rule

Although the Bessel comb is formally infinite, $J_n(\beta)\to 0$ rapidly once $n>\beta$ (visible in the plot: each $J_n$ stays flat near zero until $x\approx n$). Keeping sidebands down to ~1% of the carrier gives the engineering bandwidth:

Carson's rule $$B \approx 2(\beta+1)f_m = 2(\Delta f + f_m)$$
Two regimes

Useful $J_n$ identities

Cheat sheet

Practice Problems

Spec-design oriented · Easy / Medium / Hard · Answers included

Difficulty Easy Direct application of formula / definition Medium Multi-step, circuit reasoning required Hard Full spec-design from requirements Design Open-ended spec → component values

Topic 1Mixer Fundamentals

Frequency conversion, conversion loss, port definitions (RF/LO/IF).

1.1 Easy
Identify IF output frequencies
Given RF and LO, find both mixer output products.

A mixer receives an RF signal at $f_{RF} = 2.45\,\text{GHz}$ and a LO at $f_{LO} = 2.00\,\text{GHz}$.

(a) Find both first-order mixing products $f_{IF}$.

(b) Which one is the upper sideband (USB) and which is the lower sideband (LSB)?

(c) If the IF filter passes only frequencies below 1 GHz, which product is selected?

Answer

已知/目標:Given / required:既知/目標:$f_{RF} = 2.45\,\text{GHz}$、$f_{LO} = 2.00\,\text{GHz}$。求兩個 first-order mixing products $f_{IF}$,判別 USB/LSB,並決定通帶 < 1 GHz 的 IF filter 會選到哪一個。$f_{RF} = 2.45\,\text{GHz}$ and $f_{LO} = 2.00\,\text{GHz}$. Find the two first-order mixing products $f_{IF}$, identify the USB/LSB, and determine which one an IF filter with a passband below 1 GHz will select.$f_{RF} = 2.45\,\text{GHz}$、$f_{LO} = 2.00\,\text{GHz}$。2 つの一次混合積 $f_{IF}$ を求め、USB/LSB を判別し、通過帯域が < 1 GHz の IF filter がどちらを選ぶかを決定します。

公式與來源:Formulas and sources:公式と出典:mixer 本質上是把 RF 與 LO 相乘,由積化和差恆等式 $\cos A\cos B = \tfrac{1}{2}[\cos(A-B)+\cos(A+B)]$,一階輸出只有兩條譜線: A mixer is essentially multiplying the RF by the LO, and by the product-to-sum identity $\cos A\cos B = \tfrac{1}{2}[\cos(A-B)+\cos(A+B)]$ the first-order output has only two spectral lines:ミキサは本質的に RF と LO を掛け合わせるものであり、積和公式 $\cos A\cos B = \tfrac{1}{2}[\cos(A-B)+\cos(A+B)]$ により、一次出力にはスペクトル線が 2 本しかありません:

$$f_{IF} = |f_{RF} \pm f_{LO}|$$

逐步代入:Substituting step by step:順に代入:

(a) 和頻(sum product):$f_{RF} + f_{LO} = 2.45 + 2.00 = 4.45\,\text{GHz}$;差頻(difference product):$|f_{RF} - f_{LO}| = 2.45 - 2.00 = 0.45\,\text{GHz} = 450\,\text{MHz}$。(a) The sum product: $f_{RF} + f_{LO} = 2.45 + 2.00 = 4.45\,\text{GHz}$; the difference product: $|f_{RF} - f_{LO}| = 2.45 - 2.00 = 0.45\,\text{GHz} = 450\,\text{MHz}$.(a) 和の積(sum product):$f_{RF} + f_{LO} = 2.45 + 2.00 = 4.45\,\text{GHz}$;差の積(difference product):$|f_{RF} - f_{LO}| = 2.45 - 2.00 = 0.45\,\text{GHz} = 450\,\text{MHz}$。

USB: 2.45 + 2.00 = 4.45 GHz LSB: 2.45 − 2.00 = 0.45 GHz

(b) 依輸出頻譜的相對位置命名:(b) Named by their relative position in the output spectrum: (b) 出力スペクトル内の相対位置で命名します:USB(upper sideband)= 和頻 (upper sideband) = the sum product (upper sideband)= 和の積 4.45 GHz; LSB(lower sideband)= 差頻 (lower sideband) = the difference product (lower sideband)= 差の積 0.45 GHz.

(c) IF filter 通帶只到 1 GHz:$4.45\,\text{GHz} > 1\,\text{GHz}$ 被擋掉;$450\,\text{MHz} < 1\,\text{GHz}$ 通過 → 選到 (c) The IF filter passband extends only to 1 GHz: $4.45\,\text{GHz} > 1\,\text{GHz}$ is blocked; $450\,\text{MHz} < 1\,\text{GHz}$ passes → selecting the (c) IF filter の通過帯域は 1 GHz までなので:$4.45\,\text{GHz} > 1\,\text{GHz}$ は阻止され、$450\,\text{MHz} < 1\,\text{GHz}$ は通過 → 選ばれるのは LSB at 450 MHz.

物理詮釋:Physical interpretation:物理的な解釈:mixer 一定「同時」產生 sum 與 difference 兩個 product,down-converter 是靠 IF filter 把不要的那一條濾掉,而不是 mixer 自己只產生一條。決定要哪一條譜線(以及 filter 截止頻率)就是接收機 frequency plan 的第一步。 A mixer always produces the sum and difference products *together*; a downconverter relies on the IF filter to remove the unwanted one, rather than the mixer producing only one. Deciding which line you want (and where to put the filter cutoff) is the first step of a receiver's frequency plan.ミキサは必ず和と差の積を「同時に」生成します。ダウンコンバータは、ミキサ自身が片方だけを生成するのではなく、IF filter で不要な方を除去することに頼っています。どのスペクトル線が欲しいか(およびフィルタの遮断周波数をどこに置くか)を決めることが、受信機の frequency plan の第一歩です。

1.2 Easy
Conversion loss calculation
$L_C = P_{RF,avail}/P_{IF,avail}$

A passive diode mixer has a conversion loss of $L_C = 7\,\text{dB}$.

(a) If the available RF input power is $P_{RF} = -20\,\text{dBm}$, find $P_{IF}$.

(b) A gain block with $G = 15\,\text{dB}$ and $NF = 3\,\text{dB}$ is placed after the mixer. Find the cascade gain from RF to IF amplifier output.

Answer

已知/目標:Given / required:既知/目標:passive diode mixer 的 conversion loss $L_C = 7\,\text{dB}$,available RF input power $P_{RF} = -20\,\text{dBm}$;mixer 之後接 gain block($G = 15\,\text{dB}$、NF = 3 dB)。求 (a) $P_{IF}$,(b) RF 到 IF amplifier 輸出端的 cascade gain。 A passive diode mixer has a conversion loss $L_C = 7\,\text{dB}$ with an available RF input power $P_{RF} = -20\,\text{dBm}$; a gain block ($G = 15\,\text{dB}$, NF = 3 dB) follows the mixer. Find (a) $P_{IF}$ and (b) the cascade gain from RF to the IF amplifier output.passive diode mixer の変換損失は $L_C = 7\,\text{dB}$、available RF input power は $P_{RF} = -20\,\text{dBm}$;ミキサの後段に gain block($G = 15\,\text{dB}$、NF = 3 dB)が続きます。(a) $P_{IF}$、(b) RF から IF amplifier 出力端までの cascade gain を求めます。

公式與來源:Formulas and sources:公式と出典:conversion loss 定義為 available power 的比值 $L_C = P_{RF,avail}/P_{IF,avail}$,取 dB 後乘除變加減: Conversion loss is defined as the ratio of available powers, $L_C = P_{RF,avail}/P_{IF,avail}$, so in dB multiplication and division become addition and subtraction:変換損失は available power の比 $L_C = P_{RF,avail}/P_{IF,avail}$ として定義され、dB を取ると乗除が加減になります:

$$P_{IF}\,[\text{dBm}] = P_{RF}\,[\text{dBm}] - L_C\,[\text{dB}], \qquad G_{cascade}\,[\text{dB}] = -L_C + G$$

逐步代入:Substituting step by step:順に代入:

(a) $P_{IF} = -20 - 7 = \mathbf{-27\,dBm}$。(a) $P_{IF} = -20 - 7 = \mathbf{-27\,dBm}$.(a) $P_{IF} = -20 - 7 = \mathbf{-27\,dBm}$。

用 linear 單位驗證:$-20\,\text{dBm} = 10^{-20/10}\,\text{mW} = 0.01\,\text{mW} = 10\,\mu\text{W}$;除以 $L_C$ 的 linear 值 $10^{7/10} = 5.01$ 得 $10/5.01 \approx 2.0\,\mu\text{W}$,而 $10\log_{10}(0.002\,\text{mW}) = -27\,\text{dBm}$ ✓Check in linear units: $-20\,\text{dBm} = 10^{-20/10}\,\text{mW} = 0.01\,\text{mW} = 10\,\mu\text{W}$; dividing by the linear value of $L_C$, $10^{7/10} = 5.01$, gives $10/5.01 \approx 2.0\,\mu\text{W}$, and $10\log_{10}(0.002\,\text{mW}) = -27\,\text{dBm}$ ✓linear 単位で検証:$-20\,\text{dBm} = 10^{-20/10}\,\text{mW} = 0.01\,\text{mW} = 10\,\mu\text{W}$;$L_C$ の linear 値 $10^{7/10} = 5.01$ で割ると $10/5.01 \approx 2.0\,\mu\text{W}$、そして $10\log_{10}(0.002\,\text{mW}) = -27\,\text{dBm}$ ✓

(b) cascade gain $= -L_C + G = -7 + 15 = \mathbf{+8\,dB}$;IF amplifier 輸出功率 $= P_{RF} + G_{cascade} = -20 + 8 = \mathbf{-12\,dBm}$。(b) The cascade gain $= -L_C + G = -7 + 15 = \mathbf{+8\,dB}$; the IF amplifier output power $= P_{RF} + G_{cascade} = -20 + 8 = \mathbf{-12\,dBm}$.(b) cascade gain $= -L_C + G = -7 + 15 = \mathbf{+8\,dB}$;IF amplifier の出力電力 $= P_{RF} + G_{cascade} = -20 + 8 = \mathbf{-12\,dBm}$。

P_IF = −27 dBm;RF→IF amp 輸出 cascade gain = +8 dB → 輸出功率 −12 dBmP_IF = −27 dBm; the RF→IF-amp cascade gain = +8 dB → an output power of −12 dBmP_IF = −27 dBm;RF→IF amp の cascade gain = +8 dB → 出力電力 −12 dBm

物理詮釋:Physical interpretation:物理的な解釈:passive mixer 必然是損耗元件($L_C > 0$),所以實際接收鏈幾乎都在 mixer 後緊接 IF gain block 把訊號電平拉回來。題目給的 NF = 3 dB 在純功率計算用不到,但它會進 Friis 雜訊串級(Topic 7)— mixer 的高 NF 之所以可容忍,正是因為它前面通常還有 LNA 的增益壓制。 A passive mixer is necessarily a lossy element ($L_C > 0$), which is why almost every real receive chain puts an IF gain block immediately after the mixer to pull the signal level back up. The NF = 3 dB given in the problem is not needed for the pure power calculation, but it enters the Friis noise cascade (Topic 7) — and the reason a mixer's high NF is tolerable is precisely that there is usually LNA gain ahead of it to suppress its contribution.passive mixer は必然的に損失素子($L_C > 0$)であるため、実際の受信チェーンではほとんどの場合、ミキサの直後に IF gain block を置いて信号レベルを引き戻します。問題で与えられた NF = 3 dB は純粋な電力計算には使いませんが、Friis 雑音縦続(Topic 7)では効いてきます — ミキサの高い雑音指数が許容できるのは、その前段に通常 LNA の利得があって寄与を抑えているからです。

1.3 Medium
Image frequency and rejection requirement
Image sits at $f_{IM} = f_{LO} - f_{IF}$ for low-side LO injection.

A superheterodyne receiver uses low-side LO injection: $f_{RF} = 900\,\text{MHz}$, $f_{LO} = 800\,\text{MHz}$, so $f_{IF} = 100\,\text{MHz}$.

(a) Find the image frequency $f_{IM}$.

(b) A blocker at $f_{IM}$ is 60 dB stronger than the desired signal. How much image rejection must the RF preselector filter provide to make the image equal to the desired signal at the mixer input?

(c) What is the bandwidth separation between $f_{RF}$ and $f_{IM}$? Does a simple LC bandpass filter with $Q=50$ at 900 MHz provide enough rejection?

Answer

已知/目標:Given / required:既知/目標:low-side LO injection:$f_{RF} = 900\,\text{MHz}$、$f_{LO} = 800\,\text{MHz}$、$f_{IF} = 100\,\text{MHz}$。求 image frequency、所需 image rejection(image blocker 比訊號強 60 dB),並檢驗 $Q = 50$ 的 LC BPF 夠不夠。 Low-side LO injection: $f_{RF} = 900\,\text{MHz}$, $f_{LO} = 800\,\text{MHz}$, $f_{IF} = 100\,\text{MHz}$. Find the image frequency and the image rejection required (the image blocker is 60 dB stronger than the signal), and check whether a $Q = 50$ LC BPF is sufficient.low-side LO injection:$f_{RF} = 900\,\text{MHz}$、$f_{LO} = 800\,\text{MHz}$、$f_{IF} = 100\,\text{MHz}$。イメージ周波数と必要なイメージ除去量(イメージブロッカーは信号より 60 dB 強い)を求め、$Q = 50$ の LC BPF で十分かを検証します。

公式與來源:Formulas and sources:公式と出典:image 是「與 $f_{RF}$ 鏡像對稱於 $f_{LO}$、會混到同一個 $f_{IF}$」的頻率。low-side injection($f_{LO} < f_{RF}$)時: The image is the frequency “mirror-symmetric to $f_{RF}$ about $f_{LO}$, which mixes down to the same $f_{IF}$”. For low-side injection ($f_{LO} < f_{RF}$):イメージとは「$f_{LO}$ を中心に $f_{RF}$ と鏡像対称で、同じ $f_{IF}$ に変換されてしまう」周波数です。low-side injection($f_{LO} < f_{RF}$)の場合:

$$f_{IM} = f_{LO} - f_{IF} = f_{RF} - 2f_{IF}$$

因為 $|f_{IM} - f_{LO}| = f_{IF}$,mixer 無法分辨 $f_{RF}$ 與 $f_{IM}$ — 兩者都落到 $f_{IF}$。Since $|f_{IM} - f_{LO}| = f_{IF}$, the mixer cannot tell $f_{RF}$ from $f_{IM}$ — both land on $f_{IF}$.$|f_{IM} - f_{LO}| = f_{IF}$ であるため、ミキサは $f_{RF}$ と $f_{IM}$ を区別できません — どちらも $f_{IF}$ に落ちます。

逐步代入:Substituting step by step:順に代入:

(a) $f_{IM} = 800 - 100 = \mathbf{700\,MHz}$(驗證:$|700 - 800| = 100\,\text{MHz} = f_{IF}$ ✓)。(a) $f_{IM} = 800 - 100 = \mathbf{700\,MHz}$ (check: $|700 - 800| = 100\,\text{MHz} = f_{IF}$ ✓).(a) $f_{IM} = 800 - 100 = \mathbf{700\,MHz}$(検証:$|700 - 800| = 100\,\text{MHz} = f_{IF}$ ✓)。

(b) blocker 在 mixer 輸入要被壓到與訊號等高:所需 rejection $=$ 強度差 $= \mathbf{60\,dB}$(實務上還要再加 margin,例如 +10 dB 使 image 低於訊號 10 dB)。(b) The blocker must be pushed down to the level of the signal at the mixer input: the rejection required $=$ the difference in level $= \mathbf{60\,dB}$ (in practice with additional margin, say +10 dB to put the image 10 dB below the signal).(b) ブロッカーはミキサ入力で信号と同じレベルまで抑える必要があります:必要な rejection $=$ レベル差 $= \mathbf{60\,dB}$(実務ではさらにマージン、例えば +10 dB を加えてイメージを信号より 10 dB 低くします)。

(c) 頻率間隔 $= |f_{RF} - f_{IM}| = 900 - 700 = 200\,\text{MHz}$。BPF 的 3 dB 頻寬由 $Q$ 決定:(c) The frequency separation $= |f_{RF} - f_{IM}| = 900 - 700 = 200\,\text{MHz}$. The BPF's 3 dB bandwidth is set by $Q$:(c) 周波数間隔 $= |f_{RF} - f_{IM}| = 900 - 700 = 200\,\text{MHz}$。BPF の 3 dB 帯域幅は $Q$ で決まります:

$$BW_{3dB} = \frac{f_0}{Q} = \frac{900}{50} = 18\,\text{MHz}$$

歸一化偏移:$\dfrac{2\Delta f}{BW_{3dB}} = \dfrac{2\times 200}{18} \approx 22.2$(即偏移約為半頻寬 9 MHz 的 22 倍)。2nd-order(雙極點)BPF 的遠端滾降約 40 dB/decade:The normalized offset: $\dfrac{2\Delta f}{BW_{3dB}} = \dfrac{2\times 200}{18} \approx 22.2$ (i.e. an offset about 22 times the 9 MHz half-bandwidth). A 2nd-order (two-pole) BPF rolls off far from centre at about 40 dB/decade:正規化オフセット:$\dfrac{2\Delta f}{BW_{3dB}} = \dfrac{2\times 200}{18} \approx 22.2$(すなわちオフセットは半帯域幅 9 MHz の約 22 倍)。2nd-order(2 極)BPF の遠方でのロールオフは約 40 dB/decade です:

$$\text{rejection} \approx 40\log_{10}\!\left(\frac{2\Delta f}{BW_{3dB}}\right) = 40\log_{10}(22.2) = 40\times 1.35 \approx \mathbf{54\,dB}$$

f_IM = 700 MHz;需 60 dB image rejection;Q=50 的 2 階 LC BPF 只給 ≈54 dB → 差 ≈6 dB,建議提高階數或用 SAW/cavity filterf_IM = 700 MHz; 60 dB of image rejection is needed; a 2nd-order LC BPF with Q=50 gives only ≈54 dB → about 6 dB short, so raise the order or use a SAW/cavity filterf_IM = 700 MHz;必要なイメージ除去は 60 dB;Q=50 の 2 次 LC BPF では ≈54 dB しか得られず → ≈6 dB 不足、次数を上げるか SAW/cavity filter の使用を推奨

物理詮釋:Physical interpretation:物理的な解釈:image 與 RF 的間隔恆為 $2f_{IF}$ — IF 選得越低,image 越貼近訊號、preselector 越難做。這就是 superheterodyne 設計中「高 IF 利於 image rejection、低 IF 利於 channel filter」的經典取捨(也是 image-reject mixer 存在的理由,見 Topic 6)。 The separation between the image and the RF is always $2f_{IF}$ — the lower the IF, the closer the image sits to the signal and the harder the preselector becomes. This is the classic superheterodyne trade-off that “a high IF favours image rejection while a low IF favours the channel filter” (and the reason image-reject mixers exist, see Topic 6).イメージと RF の間隔は常に $2f_{IF}$ です — IF を低く選ぶほどイメージは信号に近づき、プリセレクタは作りにくくなります。これがスーパーヘテロダイン設計における「高い IF はイメージ除去に有利、低い IF はチャネルフィルタに有利」という古典的なトレードオフであり、イメージ除去ミキサが存在する理由でもあります(Topic 6 参照)。

1.4 Design
Spec a down-conversion stage for a 5G NR sub-6 GHz receiver
Translate system requirements to mixer specs: $L_C$, isolation, IP3.

System requirements for a 5G NR n78 (3.5 GHz) receiver front-end:

System Spec System NF ≤ 6 dB (referred to antenna)
System IIP3 ≥ −15 dBm (input-referred)
LNA gain = 20 dB, LNA NF = 2 dB, LNA IIP3 = −5 dBm
IF amplifier NF = 4 dB, IF amplifier gain = 20 dB
RF frequency: 3.5 GHz, IF: 200 MHz, LO: 3.3 GHz

(a) Using Friis formula, what is the maximum allowable mixer SSB NF to meet the system NF spec?

(b) Using the cascade IIP3 formula, what is the minimum mixer IIP3?

(c) Choose between a single-balanced 90° or double-balanced mixer and justify in 2 sentences.

Answer

已知/目標:Given / required:既知/目標:system NF ≤ 6 dB、system IIP3 ≥ −15 dBm;LNA:G = 20 dB / NF = 2 dB / IIP3 = −5 dBm;IF amp:NF = 4 dB / G = 20 dB。求 (a) mixer 的 SSB NF 上限、(b) mixer 的 IIP3 下限、(c) 拓樸選擇。 A system NF ≤ 6 dB and a system IIP3 ≥ −15 dBm; the LNA has G = 20 dB / NF = 2 dB / IIP3 = −5 dBm; the IF amp has NF = 4 dB / G = 20 dB. Find (a) the upper limit on the mixer's SSB NF, (b) the lower limit on the mixer's IIP3, and (c) the topology to choose.system NF ≤ 6 dB、system IIP3 ≥ −15 dBm;LNA:G = 20 dB / NF = 2 dB / IIP3 = −5 dBm;IF amp:NF = 4 dB / G = 20 dB。(a) ミキサの SSB NF の上限、(b) ミキサの IIP3 の下限、(c) トポロジーの選択を求めます。

(a) Mixer NF budget — 公式與來源(Friis cascade formula):(a) The mixer NF budget — formula and source (the Friis cascade formula):(a) ミキサ NF バジェット — 公式と出典(Friis cascade formula):

$$NF_{sys} = NF_{LNA} + \dfrac{NF_{mix}-1}{G_{LNA}} + \dfrac{NF_{IFA}-1}{G_{LNA}\cdot G_{mix}} \quad (\text{全部 linear 值})$$$$NF_{sys} = NF_{LNA} + \dfrac{NF_{mix}-1}{G_{LNA}} + \dfrac{NF_{IFA}-1}{G_{LNA}\cdot G_{mix}} \quad (\text{all in linear values})$$$$NF_{sys} = NF_{LNA} + \dfrac{NF_{mix}-1}{G_{LNA}} + \dfrac{NF_{IFA}-1}{G_{LNA}\cdot G_{mix}} \quad (\text{all in linear values})$$

逐步代入Substituting step by step順に代入 — 先把 dB 轉 linear: — first convert dB to linear: — まず dB を linear に変換:

  • $NF_{sys,lin} \leq 10^{6/10} = 3.981$
  • $NF_{LNA,lin} = 10^{2/10} = 1.585$,$G_{LNA,lin} = 10^{20/10} = 100$$NF_{LNA,lin} = 10^{2/10} = 1.585$ and $G_{LNA,lin} = 10^{20/10} = 100$$NF_{LNA,lin} = 10^{2/10} = 1.585$、$G_{LNA,lin} = 10^{20/10} = 100$
  • $NF_{IFA,lin} = 10^{4/10} = 2.512$
  • mixer conversion loss 假設 $\approx 8\,\text{dB}$(典型 passive mixer)→ $G_{mix} = 10^{-8/10} = 0.158$assume the mixer conversion loss $\approx 8\,\text{dB}$ (typical of a passive mixer) → $G_{mix} = 10^{-8/10} = 0.158$ミキサの変換損失を $\approx 8\,\text{dB}$ と仮定(パッシブミキサの典型値)→ $G_{mix} = 10^{-8/10} = 0.158$

第三項(IF amp 經兩級增益回推):$\dfrac{NF_{IFA}-1}{G_{LNA}G_{mix}} = \dfrac{2.512-1}{100\times 0.158} = \dfrac{1.512}{15.8} \approx 0.096$the third term (the IF amp referred back through two stages of gain): $\dfrac{NF_{IFA}-1}{G_{LNA}G_{mix}} = \dfrac{2.512-1}{100\times 0.158} = \dfrac{1.512}{15.8} \approx 0.096$第 3 項(IF amp を 2 段の利得で入力換算):$\dfrac{NF_{IFA}-1}{G_{LNA}G_{mix}} = \dfrac{2.512-1}{100\times 0.158} = \dfrac{1.512}{15.8} \approx 0.096$

解 mixer 項:$\dfrac{NF_{mix}-1}{100} \leq 3.981 - 1.585 - 0.096 = 2.30$Solving for the mixer term: $\dfrac{NF_{mix}-1}{100} \leq 3.981 - 1.585 - 0.096 = 2.30$ミキサの項について解く:$\dfrac{NF_{mix}-1}{100} \leq 3.981 - 1.585 - 0.096 = 2.30$

$$\Rightarrow NF_{mix,lin} \leq 1 + 100\times 2.30 = 231$$

NF_mix ≤ 10·log10(231) ≈ 23.6 dB SSB (passive diode mixers at 8–12 dB easily meet this)

(b) Mixer IIP3 budget — 公式與來源(IIP3 cascade 公式,兩級近似,全 linear mW):(b) The mixer IIP3 budget — formula and source (the IIP3 cascade formula, two-stage approximation, all in linear mW):(b) ミキサ IIP3 バジェット — 公式と出典(IIP3 cascade 公式、2 段近似、すべて linear mW):

$$\dfrac{1}{IIP3_{sys}} \approx \dfrac{1}{IIP3_{LNA}} + \dfrac{G_{LNA}}{IIP3_{mix}}$$

逐步代入Substituting step by step順に代入 — dBm 轉 mW:$IIP3_{LNA} = 10^{-5/10} = 0.316\,\text{mW}$、$IIP3_{sys} = 10^{-15/10} = 0.0316\,\text{mW}$、$G_{LNA} = 100$: — converting dBm to mW: $IIP3_{LNA} = 10^{-5/10} = 0.316\,\text{mW}$, $IIP3_{sys} = 10^{-15/10} = 0.0316\,\text{mW}$ and $G_{LNA} = 100$: — dBm を mW に変換:$IIP3_{LNA} = 10^{-5/10} = 0.316\,\text{mW}$、$IIP3_{sys} = 10^{-15/10} = 0.0316\,\text{mW}$、$G_{LNA} = 100$:

$$\dfrac{1}{0.0316} \leq \dfrac{1}{0.316} + \dfrac{100}{IIP3_{mix}} \;\Rightarrow\; 31.6 \leq 3.16 + \dfrac{100}{IIP3_{mix}}$$

$$\dfrac{100}{IIP3_{mix}} \geq 31.6 - 3.16 = 28.4 \;\Rightarrow\; IIP3_{mix} \geq \dfrac{100}{28.4} = 3.52\,\text{mW} = 10\log_{10}(3.52) = +5.5\,\text{dBm}$$

IIP3_mix ≥ +5.5 dBm (input-referred to mixer RF port)

(c) Topology choice:(c) Topology choice:(c) トポロジーの選択: choose the 選ぶべきは double-balanced ring mixer。理由:(1) 它同時抑制 even LO harmonics 與 even RF harmonics,在頻譜擁擠的 5G sub-6 GHz 環境提供必要的 spectral purity;(2) LO-to-RF isolation(~40 dB)防止 LO 洩漏經 LNA 反向輻射到天線。. The reasons: (1) it suppresses both even LO harmonics and even RF harmonics, providing the spectral purity needed in the crowded 5G sub-6 GHz environment; (2) its LO-to-RF isolation (~40 dB) stops the LO leaking back through the LNA and radiating from the antenna.です。理由:(1) even LO harmonics と even RF harmonics を同時に抑制し、スペクトルが混雑する 5G sub-6 GHz 環境で必要な spectral purity を提供すること、(2) LO-to-RF アイソレーション(~40 dB)が、LO の漏れが LNA を逆流してアンテナから放射されるのを防ぐこと。

物理詮釋:Physical interpretation:物理的解釈:注意兩個 budget 的「方向」相反 — LNA 增益越大,mixer 的 NF 要求越鬆(除以 $G_{LNA}$)、IIP3 要求卻越緊(乘上 $G_{LNA}$)。NF 上限 23.6 dB 很容易達標,但 IIP3 ≥ +5.5 dBm 需要夠大的 LO drive(level-13 以上的 diode ring 或 active mixer),這就是「線性度由後級主宰、雜訊由前級主宰」的典型表現。 Note that the two budgets pull in opposite directions — the greater the LNA gain, the looser the mixer's NF requirement (divided by $G_{LNA}$) but the tighter its IIP3 requirement (multiplied by $G_{LNA}$). An NF ceiling of 23.6 dB is easily met, but an IIP3 ≥ +5.5 dBm needs a substantial LO drive (a level-13 or higher diode ring, or an active mixer) — a textbook illustration of “linearity is dominated by the later stages, noise by the earlier ones”.二つのバジェットの「方向」が逆であることに注意してください — LNA 利得が大きいほどミキサの NF 要求は緩くなり($G_{LNA}$ で割る)、逆に IIP3 要求は厳しくなります($G_{LNA}$ を掛ける)。NF 上限 23.6 dB は容易に達成できますが、IIP3 ≥ +5.5 dBm には十分大きな LO ドライブ(level-13 以上のダイオードリング、またはアクティブミキサ)が必要で、これは「直線性は後段が支配し、雑音は前段が支配する」の典型的な現れです。

Topic 2Intermodulation & Spurious Products

IM3, IIP3, 1 dB compression, harmonic mixing, spur table.

2.1 Easy
IM3 level from IIP3
$P_{IM3} = 3P_{in} - 2\cdot IIP3$ (all in dBm)

A mixer has $IIP3 = +5\,\text{dBm}$ (input-referred). Two equal-power RF tones at $-10\,\text{dBm}$ each enter the RF port.

(a) Find the output IM3 power relative to the fundamental IF output (conversion loss = 7 dB).

(b) What input power would cause the IM3 to equal the fundamental (intercept check)?

Answer

已知/目標:Given / required:既知/目標:mixer 的 $IIP3 = +5\,\text{dBm}$(input-referred)、兩個等功率 RF tone 各 $P_{in} = -10\,\text{dBm}$、conversion loss $L_C = 7\,\text{dB}$。求 (a) 輸出端 IM3 相對 fundamental 的電平、(b) IM3 追上 fundamental 的輸入功率。 A mixer has $IIP3 = +5\,\text{dBm}$ (input-referred), two equal-power RF tones each at $P_{in} = -10\,\text{dBm}$, and a conversion loss $L_C = 7\,\text{dB}$. Find (a) the IM3 level relative to the fundamental at the output and (b) the input power at which IM3 catches up with the fundamental.ミキサの $IIP3 = +5\,\text{dBm}$(input-referred)、等電力の RF トーン 2 波が各 $P_{in} = -10\,\text{dBm}$、変換損失 $L_C = 7\,\text{dB}$。(a) 出力端における基本波に対する IM3 のレベル、(b) IM3 が基本波に追いつく入力電力を求めます。

公式與來源:Formulas and sources:公式と出典:IM3 以 3:1 斜率成長、fundamental 以 1:1 成長,兩線外插交於 intercept point。由幾何關係(全部 dBm/dB): IM3 grows with a 3:1 slope and the fundamental with a 1:1 slope, and the two extrapolated lines cross at the intercept point. From the geometry (everything in dBm/dB):IM3 は 3:1 の傾きで、基本波は 1:1 の傾きで増加し、両者の外挿直線はインターセプト点で交わります。幾何関係より(すべて dBm/dB):

$$P_{IM3,in} = 3P_{in} - 2\cdot IIP3, \qquad \Delta_{IM3} = P_{fund} - P_{IM3} = 2(IIP3 - P_{in})$$

逐步代入:Substituting step by step:順に代入:

(a) input-referred IM3:$P_{IM3,in} = 3\times(-10) - 2\times(+5) = -30 - 10 = -40\,\text{dBm}$。(a) Input-referred IM3: $P_{IM3,in} = 3\times(-10) - 2\times(+5) = -30 - 10 = -40\,\text{dBm}$.(a) input-referred IM3:$P_{IM3,in} = 3\times(-10) - 2\times(+5) = -30 - 10 = -40\,\text{dBm}$。

兩者同樣經過 $L_C = 7\,\text{dB}$ 轉到 IF:fundamental IF $= -10 - 7 = -17\,\text{dBm}$;IM3 at IF $= -40 - 7 = -47\,\text{dBm}$。相對電平 $= -17-(-47) = 30\,\text{dB}$(conversion loss 對兩者相同,所以相對值不變)。Both pass through the same $L_C = 7\,\text{dB}$ on the way to IF: the fundamental at IF $= -10 - 7 = -17\,\text{dBm}$ and the IM3 at IF $= -40 - 7 = -47\,\text{dBm}$. The relative level $= -17-(-47) = 30\,\text{dB}$ (the conversion loss being the same for both, the relative value is unchanged).両者は同じ $L_C = 7\,\text{dB}$ を経て IF に変換されます:IF での基本波 $= -10 - 7 = -17\,\text{dBm}$、IF での IM3 $= -40 - 7 = -47\,\text{dBm}$。相対レベルは $= -17-(-47) = 30\,\text{dB}$(変換損失は両者に共通なので相対値は変わりません)。

交叉驗證:$\Delta_{IM3} = 2(IIP3 - P_{in}) = 2\times(5-(-10)) = 2\times 15 = 30\,\text{dB}$ ✓Cross-check: $\Delta_{IM3} = 2(IIP3 - P_{in}) = 2\times(5-(-10)) = 2\times 15 = 30\,\text{dB}$ ✓交差検証:$\Delta_{IM3} = 2(IIP3 - P_{in}) = 2\times(5-(-10)) = 2\times 15 = 30\,\text{dB}$ ✓

IM3 is 30 dB below fundamental (ΔIM3 = 2(IIP3 − Pin) = 2×15 = 30 dB)

(b) 由 intercept 的定義,IM3 線與 fundamental 線交點就在 $P_{in} = IIP3$,故 $P_{in} = \mathbf{+5\,dBm}$(此時 $\Delta_{IM3} = 2(5-5) = 0$ ✓)。(b) By the definition of the intercept, the IM3 line meets the fundamental line at $P_{in} = IIP3$, so $P_{in} = \mathbf{+5\,dBm}$ (at which $\Delta_{IM3} = 2(5-5) = 0$ ✓).(b) インターセプトの定義より、IM3 の直線と基本波の直線の交点はちょうど $P_{in} = IIP3$ にあり、したがって $P_{in} = \mathbf{+5\,dBm}$(このとき $\Delta_{IM3} = 2(5-5) = 0$ ✓)。

物理詮釋:Physical interpretation:物理的解釈:記住兩條速算規則 — 輸入每升 1 dB,IM3 相對電平惡化 2 dB;輸入離 IIP3 每差 1 dB,IM3 就低於主訊號 2 dB。實際上 mixer 在遠未到 IIP3 前就已壓縮($P_{1dB} \approx IIP3 - 10\,\text{dB}$),所以 (b) 的交點是外插出來的虛擬點,不能真的打到 +5 dBm 量它。 Remember two quick rules — every 1 dB rise in input worsens the relative IM3 level by 2 dB; and for every 1 dB the input sits below IIP3, the IM3 is 2 dB below the wanted signal. In reality a mixer compresses long before it reaches IIP3 ($P_{1dB} \approx IIP3 - 10\,\text{dB}$), so the crossing point in (b) is an extrapolated fiction and you cannot actually drive it to +5 dBm to measure it.二つの速算則を覚えてください — 入力が 1 dB 上がるごとに IM3 の相対レベルは 2 dB 悪化します。入力が IIP3 より 1 dB 低いごとに、IM3 は主信号より 2 dB 低くなります。実際にはミキサは IIP3 に達するはるか手前で圧縮するため($P_{1dB} \approx IIP3 - 10\,\text{dB}$)、(b) の交点は外挿による仮想的な点であり、実際に +5 dBm まで加えて測定することはできません。

2.2 Medium
Spur identification — which harmonics fall in band?
Spur frequency: $f_{spur} = mf_{LO} \pm nf_{RF}$

A mixer is used with $f_{LO} = 1000\,\text{MHz}$ and $f_{RF} = 1100\,\text{MHz}$, targeting $f_{IF} = 100\,\text{MHz}$.

Consider harmonics up to order 3 ($m, n \leq 3$). List all spurs $|mf_{LO} - nf_{RF}|$ that fall within the IF passband $80\text{–}120\,\text{MHz}$.

Tip: build a 3×3 table for $(m,n)$ from (1,1) to (3,3) and evaluate $|m\times 1000 - n\times 1100|$.
Answer

已知/目標:Given / required:既知/目標:$f_{LO} = 1000\,\text{MHz}$、$f_{RF} = 1100\,\text{MHz}$、目標 $f_{IF} = 100\,\text{MHz}$,IF passband 80–120 MHz。掃描 $m, n \leq 3$ 的所有 spur,找出落入通帶者。$f_{LO} = 1000\,\text{MHz}$, $f_{RF} = 1100\,\text{MHz}$, a target $f_{IF} = 100\,\text{MHz}$ and an IF passband of 80–120 MHz. Sweep all spurs with $m, n \leq 3$ and find those falling in the passband.$f_{LO} = 1000\,\text{MHz}$、$f_{RF} = 1100\,\text{MHz}$、目標 $f_{IF} = 100\,\text{MHz}$、IF 通過帯域 80–120 MHz。$m, n \leq 3$ のすべてのスプリアスを掃引し、通過帯域に落ちるものを求めます。

公式與來源:Formulas and sources:公式と出典:非線性元件會產生所有諧波組合(spur chart 的依據): A nonlinear element produces every combination of harmonics (the basis of the spur chart):非線形素子はあらゆる高調波の組み合わせを生成します(スプリアスチャートの根拠):

$$f_{spur} = |m f_{LO} - n f_{RF}|, \quad m, n = 1, 2, 3$$

判斷準則:$80 \leq f_{spur} \leq 120\,\text{MHz}$ 即為 in-band spur。The criterion: $80 \leq f_{spur} \leq 120\,\text{MHz}$ makes it an in-band spur.判定基準:$80 \leq f_{spur} \leq 120\,\text{MHz}$ であれば帯域内スプリアスです。

逐步代入Substituting step by step順に代入 — 完整 3×3 掃描: — the complete 3×3 sweep: — 完全な 3×3 掃引:

(m, n)|m·1000 − n·1100| MHzIn band?
(1,1)|1000−1100|=100✓ 100 MHz(desired)✓ 100 MHz (desired)✓ 100 MHz(希望波)
(1,2)|1000−2200|=1200
(1,3)|1000−3300|=2300
(2,1)|2000−1100|=900
(2,2)|2000−2200|=200
(2,3)|2000−3300|=1300
(3,1)|3000−1100|=1900
(3,2)|3000−2200|=800
(3,3)|3000−3300|=300

結果:The result:結果:除了 desired 的 (1,1) 之外,最近的 spur 是 (2,2) 的 200 MHz,離通帶上緣 120 MHz 還有 80 MHz 的餘裕。 Apart from the desired (1,1), the nearest spur is the 200 MHz of (2,2), which leaves 80 MHz of clearance above the 120 MHz upper passband edge.希望波の (1,1) を除くと、最も近いスプリアスは (2,2) の 200 MHz で、通過帯域上端の 120 MHz からまだ 80 MHz の余裕があります。

Only the (1,1) product falls in band — good mixer selection.

延伸檢查:A further check:拡張チェック:若改 $f_{RF} = 1033\,\text{MHz}$($f_{IF} = 33\,\text{MHz}$,假設通帶也跟著搬到 33 MHz 附近):(3,2) 給 $|3000-2066|=934$、(2,3) 給 $|2000-3099|=1099$、(3,3) 給 $|3000-3099|=99$ — (3,3) spur 與 3 倍 IF 的關係變得逼近,顯示 RF/IF 比例的選擇直接決定 spur clearance。 If $f_{RF}$ were changed to $1033\,\text{MHz}$ ($f_{IF} = 33\,\text{MHz}$, assuming the passband moves to around 33 MHz too): (3,2) gives $|3000-2066|=934$, (2,3) gives $|2000-3099|=1099$ and (3,3) gives $|3000-3099|=99$ — the (3,3) spur comes uncomfortably close to 3× the IF, showing that the choice of RF/IF ratio directly determines the spur clearance.$f_{RF}$ を $1033\,\text{MHz}$ に変えた場合($f_{IF} = 33\,\text{MHz}$、通過帯域も 33 MHz 付近に移るとします):(3,2) は $|3000-2066|=934$、(2,3) は $|2000-3099|=1099$、(3,3) は $|3000-3099|=99$ — (3,3) のスプリアスが IF の 3 倍に不都合なほど近づき、RF/IF 比の選択がスプリアスの余裕を直接決めることを示しています。

物理詮釋:Physical interpretation:物理的解釈:spur 落點只由 frequency plan 決定(跟 mixer 品質無關);mixer 拓樸(balanced 與否)只能決定 spur 的「強度」。所以正確順序是:先用 spur table 把高階產物排開,再用 balanced topology 把剩下壓低 — 這正是 9.2 題的主題。 Where the spurs land is determined solely by the frequency plan (independently of the quality of the mixer); the mixer topology (balanced or not) only determines the *strength* of the spurs. So the correct order is: first use a spur table to move the high-order products out of the way, then use a balanced topology to push down what remains — which is exactly the subject of problem 9.2.スプリアスの落ちる位置は周波数プランのみで決まり(ミキサの品質とは無関係)、ミキサのトポロジー(平衡か否か)はスプリアスの「強度」だけを決めます。したがって正しい順序は:まずスプリアス表で高次の積を外し、次に平衡トポロジーで残りを押し下げる — これがまさに 9.2 題のテーマです。

2.3 Design
Specify IF filter to meet IM-free dynamic range
IMFDR3 $= \frac{2}{3}(IIP3 - NF - kTB)$

A direct-conversion receiver (ZIF) at 2.4 GHz has the following post-mixer IF path:

Given Mixer: L_C = 8 dB, SSB NF = 10 dB, IIP3 = +5 dBm
IF bandwidth: 20 MHz (802.11g channel)
Temperature: 290 K

(a) Calculate the mixer noise floor (in dBm) in a 20 MHz IF bandwidth.

(b) Calculate the IM-free dynamic range (IMFDR3) of this mixer.

(c) A strong adjacent-channel blocker appears at $\Delta f = 40\,\text{MHz}$ from the desired channel at $-30\,\text{dBm}$. The IM3 product falls in-channel. Find the IM3 level and determine if it violates a −90 dBm in-channel noise floor requirement.

Answer

已知/目標:Given / required:既知/目標:ZIF receiver,mixer 規格 $L_C = 8\,\text{dB}$、$NF_{SSB} = 10\,\text{dB}$、$IIP3 = +5\,\text{dBm}$;IF 頻寬 $B = 20\,\text{MHz}$、$T = 290\,\text{K}$。求 (a) noise floor、(b) IMFDR3、(c) −30 dBm blocker 的 IM3 是否違反 −90 dBm 要求。 A ZIF receiver with a mixer specified as $L_C = 8\,\text{dB}$, $NF_{SSB} = 10\,\text{dB}$ and $IIP3 = +5\,\text{dBm}$; the IF bandwidth is $B = 20\,\text{MHz}$ at $T = 290\,\text{K}$. Find (a) the noise floor, (b) the IMFDR3, and (c) whether the IM3 from a −30 dBm blocker violates the −90 dBm requirement.ZIF 受信機、ミキサ仕様は $L_C = 8\,\text{dB}$、$NF_{SSB} = 10\,\text{dB}$、$IIP3 = +5\,\text{dBm}$;IF 帯域幅 $B = 20\,\text{MHz}$、$T = 290\,\text{K}$。(a) 雑音フロア、(b) IMFDR3、(c) −30 dBm ブロッカーの IM3 が −90 dBm 要求に違反するかどうかを求めます。

公式與來源:Formulas and sources:公式と出典:熱雜訊底(thermal noise floor)與 IM-free dynamic range 定義: The thermal noise floor and the definition of IM-free dynamic range:熱雑音フロア(thermal noise floor)と IM-free dynamic range の定義:

$$P_{noise} = \underbrace{-174\,\text{dBm/Hz}}_{kT,\,290\,\text{K}} + 10\log_{10}B + NF, \qquad IMFDR3 = \frac{2}{3}\left(IIP3 - P_{noise}\right)$$

係數 $2/3$ 來自:兩音輸入升到「IM3 恰好觸及 noise floor」時,可用範圍是 $(IIP3-P_{noise})$ 的 2/3(IM3 斜率 3 的幾何結果)。The factor $2/3$ arises because, as the two-tone input rises to the point where the IM3 just touches the noise floor, the usable range is 2/3 of $(IIP3-P_{noise})$ (a geometric consequence of the 3:1 IM3 slope).係数 $2/3$ は、2 トーン入力を「IM3 がちょうど雑音フロアに接する」点まで上げたとき、使用可能な範囲が $(IIP3-P_{noise})$ の 2/3 になることに由来します(IM3 の傾きが 3 であることの幾何学的帰結)。

逐步代入:Substituting step by step:順に代入:

(a) 頻寬項:$10\log_{10}(20\times 10^6) = 10\times 7.30 = 73\,\text{dB}$(a) The bandwidth term: $10\log_{10}(20\times 10^6) = 10\times 7.30 = 73\,\text{dB}$(a) 帯域幅項:$10\log_{10}(20\times 10^6) = 10\times 7.30 = 73\,\text{dB}$

$kTB = -174 + 73 = -101\,\text{dBm}$;noise floor $= kTB + NF_{mix} = -101 + 10 = \mathbf{-91\,dBm}$$kTB = -174 + 73 = -101\,\text{dBm}$; the noise floor $= kTB + NF_{mix} = -101 + 10 = \mathbf{-91\,dBm}$$kTB = -174 + 73 = -101\,\text{dBm}$;雑音フロア $= kTB + NF_{mix} = -101 + 10 = \mathbf{-91\,dBm}$

(b) $IMFDR3 = \dfrac{2}{3}\left(+5 - (-91)\right) = \dfrac{2}{3}\times 96 = \mathbf{64\,dB}$

(c) 兩個 blocker 等效兩音測試,IM3 落回 in-channel:(c) The two blockers are equivalent to a two-tone test, with the IM3 falling back in-channel:(c) 2 つのブロッカーは 2 トーン試験と等価で、IM3 はチャネル内に落ち戻ります:

$$P_{IM3} = 3P_{blocker} - 2\cdot IIP3 = 3\times(-30) - 2\times(+5) = -90 - 10 = -100\,\text{dBm}$$

IM3 = −100 dBm, noise floor = −91 dBm → IM3 is 9 dB below the noise floor → just acceptable.

敏感度檢查:A sensitivity check:感度チェック:若 blocker 升到 −25 dBm(只差 5 dB):$P_{IM3} = 3(-25) - 10 = -85\,\text{dBm} > -90\,\text{dBm}$ → 違規。blocker 每升 1 dB,IM3 升 3 dB — 所以 IF low-pass filter 必須在 IM3 生成之前(mixer 前或 mixer 內)就把 $\Delta f = 40\,\text{MHz}$ 的 blocker 壓低至少 5 dB。 If the blocker rose to −25 dBm (only 5 dB more): $P_{IM3} = 3(-25) - 10 = -85\,\text{dBm} > -90\,\text{dBm}$ → a violation. Every 1 dB rise in the blocker raises the IM3 by 3 dB — so the IF low-pass filter must attenuate the blocker at $\Delta f = 40\,\text{MHz}$ by at least 5 dB *before* the IM3 is generated (ahead of, or inside, the mixer).ブロッカーが −25 dBm まで上がると(わずか 5 dB 差):$P_{IM3} = 3(-25) - 10 = -85\,\text{dBm} > -90\,\text{dBm}$ → 違反です。ブロッカーが 1 dB 上がるごとに IM3 は 3 dB 上がります — したがって IF 低域通過フィルタは、IM3 が生成される前(ミキサの前またはミキサ内)に $\Delta f = 40\,\text{MHz}$ のブロッカーを少なくとも 5 dB 減衰させる必要があります。

物理詮釋:Physical interpretation:物理的解釈:ZIF 架構沒有 RF channel filter,blocker 全額打進 mixer,所以 mixer 的 IIP3 直接決定抗鄰道能力。IMFDR3 = 64 dB 是這顆 mixer「雜訊底與失真底之間」可用的訊號窗口;設計餘裕要從提高 IIP3(LO drive)或縮小 $B$(降 noise floor)兩端去買。 A ZIF architecture has no RF channel filter, so the blocker hits the mixer at full strength and the mixer's IIP3 directly determines the adjacent-channel rejection. IMFDR3 = 64 dB is this mixer's usable signal window “between the noise floor and the distortion floor”; design margin has to be bought either by raising IIP3 (more LO drive) or by narrowing $B$ (lowering the noise floor).ZIF アーキテクチャには RF チャネルフィルタがないため、ブロッカーはそのままの大きさでミキサに入り、ミキサの IIP3 が隣接チャネル抑圧能力を直接決めます。IMFDR3 = 64 dB はこのミキサの「雑音フロアと歪みフロアの間」で使用できる信号窓です。設計余裕は IIP3 を上げる(LO ドライブを増やす)か、$B$ を狭める(雑音フロアを下げる)かのどちらかで買う必要があります。

Topic 3Single-Balanced 90° Mixer

Branch-line hybrid, LO AM noise cancellation, 3 dB split with 90° phase.

3.1 Easy
Signal distribution through a 90° hybrid
Use the S-matrix: $[S]_{BL} = \frac{-1}{\sqrt{2}}\begin{bmatrix}0&j&1&0\\j&0&0&1\\1&0&0&j\\0&1&j&0\end{bmatrix}$

RF signal $V_{RF}\angle 0°$ enters port 1 of a lossless 90° branch-line hybrid.

(a) What are the amplitudes and phases at ports 2 and 3? Port 4?

(b) Two diode mixers are connected to ports 2 and 3. LO enters port 4. Write the IF outputs from diode 1 (port 2) and diode 2 (port 3) separately, assuming each diode multiplies the signals reaching it.

(c) Show that adding the two IF outputs cancels the LO AM noise component.

Answer

已知/目標:Given / required:既知/目標:lossless branch-line hybrid 的 $[S]_{BL}$(見提示)、RF 從 port 1 進、LO(含 AM noise)從 port 4 進、兩顆 diode 接在 port 2/3。求各埠的振幅相位、兩路 IF 輸出,並證明 LO AM noise 在 IF combiner 相消。 The ideal $[S]_{BL}$ of a lossless branch-line hybrid (see the hint), with the RF entering port 1, the LO (including AM noise) entering port 4, and the two diodes at ports 2/3. Find the amplitude and phase at each port and the two IF outputs, and prove that the LO AM noise cancels at the IF combiner.無損失の branch-line hybrid の $[S]_{BL}$(ヒント参照)、RF は port 1 から入り、LO(AM ノイズを含む)は port 4 から入り、2 個のダイオードは port 2/3 に接続されます。各ポートの振幅と位相、2 系統の IF 出力を求め、LO の AM ノイズが IF 合成器で相殺されることを証明します。

公式與來源:Formulas and sources:公式と出典:$b_i = S_{ij} a_j$(S-parameter 定義);diode mixer 的 difference-frequency 產物相位為 $\phi_{IF} = \phi_{RF} - \phi_{LO}$(由 $\cos(\omega_{RF}t+\phi_{RF})\cos(\omega_{LO}t+\phi_{LO})$ 積化和差的差頻項直接讀出)。$b_i = S_{ij} a_j$ (the S-parameter definition); the phase of a diode mixer's difference-frequency product is $\phi_{IF} = \phi_{RF} - \phi_{LO}$ (read directly off the difference term of the product-to-sum expansion of $\cos(\omega_{RF}t+\phi_{RF})\cos(\omega_{LO}t+\phi_{LO})$).$b_i = S_{ij} a_j$(S-parameter の定義);ダイオードミキサの差周波数積の位相は $\phi_{IF} = \phi_{RF} - \phi_{LO}$ です($\cos(\omega_{RF}t+\phi_{RF})\cos(\omega_{LO}t+\phi_{LO})$ の積和公式における差周波数項から直接読み取れます)。

(a) 逐步代入(a) Substituting step by step(a) 順に代入 — port 1 激發 $a_1 = V_{RF}$,讀 $[S]_{BL}$ 第一行(column 1): — exciting port 1 with $a_1 = V_{RF}$ and reading column 1 of $[S]_{BL}$: — port 1 を $a_1 = V_{RF}$ で励振し、$[S]_{BL}$ の第 1 列(column 1)を読みます:

Port 2:$S_{21}V_{RF} = \frac{-1}{\sqrt{2}}(j)V_{RF} = \frac{-j}{\sqrt{2}}V_{RF}$(−3 dB, Port 2: $S_{21}V_{RF} = \frac{-1}{\sqrt{2}}(j)V_{RF} = \frac{-j}{\sqrt{2}}V_{RF}$ (−3 dB, Port 2:$S_{21}V_{RF} = \frac{-1}{\sqrt{2}}(j)V_{RF} = \frac{-j}{\sqrt{2}}V_{RF}$(−3 dB, −90°)
Port 3:$S_{31}V_{RF} = \frac{-1}{\sqrt{2}}V_{RF}$(−3 dB, Port 3: $S_{31}V_{RF} = \frac{-1}{\sqrt{2}}V_{RF}$ (−3 dB, Port 3:$S_{31}V_{RF} = \frac{-1}{\sqrt{2}}V_{RF}$(−3 dB, −180°)
Port 4:$S_{41}V_{RF} = 0$( Port 4: $S_{41}V_{RF} = 0$ ( Port 4:$S_{41}V_{RF} = 0$(isolated — RF 不會漏進 LO 埠) — the RF does not leak into the LO port) — RF は LO ポートへ漏れません)

(b) LO 從 port 4 進($a_4 = V_{LO}[1+m(t)]$,$m(t)$ 為 AM noise)。由 $[S]_{BL}$ 第 4 行:$S_{24} = \frac{-1}{\sqrt{2}}$(−180°)、$S_{34} = \frac{-j}{\sqrt{2}}$(−90°)。 The LO enters at port 4 ($a_4 = V_{LO}[1+m(t)]$, with $m(t)$ the AM noise). From row 4 of $[S]_{BL}$: $S_{24} = \frac{-1}{\sqrt{2}}$ (−180°) and $S_{34} = \frac{-j}{\sqrt{2}}$ (−90°). LO は port 4 から入ります($a_4 = V_{LO}[1+m(t)]$、$m(t)$ は AM ノイズ)。$[S]_{BL}$ の第 4 行より:$S_{24} = \frac{-1}{\sqrt{2}}$(−180°)、$S_{34} = \frac{-j}{\sqrt{2}}$(−90°)。(更正:原版把 LO 在 port 2/3 的相位寫反;結論不變。)(Correction: the original version had the LO phases at ports 2/3 the wrong way round; the conclusion is unchanged.)(訂正:元の版では port 2/3 における LO の位相が逆に書かれていました。結論は変わりません。)整理各 diode 收到的訊號: Collecting what each diode receives:各ダイオードが受け取る信号を整理します:

RF 相位RF phaseRF 位相LO 相位LO phaseLO 位相IF 相位 $\phi_{RF}-\phi_{LO}$IF phase $\phi_{RF}-\phi_{LO}$IF 位相 $\phi_{RF}-\phi_{LO}$
Diode 1 (port 2)−90°−180°+90°
Diode 2 (port 3)−180°−90°−90°

所以兩路 IF(振幅各 $\propto \tfrac{1}{2}V_{RF}V_{LO}[1+m]$):So the two IF paths (each of amplitude $\propto \tfrac{1}{2}V_{RF}V_{LO}[1+m]$):したがって 2 系統の IF(振幅は各 $\propto \tfrac{1}{2}V_{RF}V_{LO}[1+m]$):

$$i_1 \propto \tfrac{1}{2}V_{RF}V_{LO}[1+m]\cos(\omega_{IF}t + 90°), \qquad i_2 \propto \tfrac{1}{2}V_{RF}V_{LO}[1+m]\cos(\omega_{IF}t - 90°)$$

兩路 IF 訊號The two IF signals are 2 系統の IF 信号は反相(相差 180°)in anti-phase (180° apart)逆相(位相差 180°)→ 在減法 combiner($i_1 - i_2$,或其中一顆 diode 反接)中相加。→ so they add in a subtracting combiner ($i_1 - i_2$, or with one of the diodes reversed).→ 減算合成器($i_1 - i_2$、またはダイオードの一方を逆接続)では加算されます。

(c) AM noise 相消:(c) AM-noise cancellation:(c) AM ノイズの相殺:LO 的 AM noise 不是經由 mixing 產生,而是經 diode 的 The LO's AM noise is not produced by mixing but appears directly in the IF band through the diode's LO の AM ノイズはミキシングによって生じるのではなく、ダイオードの包絡檢波(envelope detection)envelope detection包絡線検波(envelope detection)直接出現在 IF 頻帶:$i_{n} \propto V_{LO}\,m(t)$。這個檢波輸出只跟 LO 的「振幅」有關、不繼承 carrier 相位 — 而兩顆 diode 收到的 LO 振幅相同(同樣 −3 dB split),所以兩路的 AM noise 電流: $i_{n} \propto V_{LO}\,m(t)$. This detected output depends only on the LO's *amplitude* and does not inherit the carrier phase — and both diodes receive the same LO amplitude (the same −3 dB split), so the AM-noise currents on the two paths are により直接 IF 帯域に現れます:$i_{n} \propto V_{LO}\,m(t)$。この検波出力は LO の「振幅」だけに依存し、搬送波の位相を引き継ぎません — そして 2 個のダイオードが受け取る LO の振幅は同じ(同じ −3 dB 分配)なので、2 系統の AM ノイズ電流は同相(common mode)in phase (common mode)同相(common mode)。在減法 combiner:. In the subtracting combiner:。減算合成器では:

$$i_{IF} = i_1 - i_2 = \underbrace{2\times\text{desired IF}}_{\text{反相 → 相減後相加}} + \underbrace{(i_{n} - i_{n})}_{=\,0}$$$$i_{IF} = i_1 - i_2 = \underbrace{2\times\text{desired IF}}_{\text{anti-phase → adds on subtraction}} + \underbrace{(i_{n} - i_{n})}_{=\,0}$$$$i_{IF} = i_1 - i_2 = \underbrace{2\times\text{desired IF}}_{\text{anti-phase → adds on subtraction}} + \underbrace{(i_{n} - i_{n})}_{=\,0}$$

LO AM noise cancellation: paths are symmetric → common-mode rejection at IF combiner.

物理詮釋:Physical interpretation:物理的解釈:balanced mixer 的核心就是讓「要的訊號」與「不要的訊號」在 combiner 處呈現不同的 odd/even 對稱性:desired IF 帶著 ±90° 的路徑相位差(differential),AM noise 不帶相位(common)— 一個減法就同時做到「訊號相加、噪聲相消」。抑制量受兩顆 diode 匹配度限制,實務約 20–30 dB。 The essence of a balanced mixer is to give the wanted and the unwanted signals different odd/even symmetries at the combiner: the desired IF carries a ±90° path phase difference (differential) while the AM noise carries none (common mode) — so a single subtraction simultaneously adds the signal and cancels the noise. The depth of suppression is limited by how well the two diodes are matched, typically 20–30 dB in practice.平衡ミキサの核心は、「必要な信号」と「不要な信号」を合成器において異なる奇偶対称性で現れさせることです:希望 IF は ±90° の経路位相差を持ち(differential)、AM ノイズは位相差を持ちません(common) — 一度の減算で「信号は加算、雑音は相殺」を同時に実現します。抑圧量は 2 個のダイオードの整合度で制限され、実務では約 20–30 dB です。

3.2 Medium
Branch-line coupler dimensional design at 2.4 GHz
Line impedances: shunt arms $Z_0/\sqrt{2}$, series arms $Z_0$; all λ/4.

Design a microstrip branch-line coupler for $f_0 = 2.4\,\text{GHz}$, $Z_0 = 50\,\Omega$, on Rogers RO4003C ($\varepsilon_r = 3.55$, $h = 0.508\,\text{mm}$, $\tan\delta = 0.0027$).

(a) State the required line impedances for the four arms.

(b) For a $50\,\Omega$ microstrip on RO4003C, the effective dielectric constant $\varepsilon_{eff} \approx 2.64$ and $W \approx 1.13\,\text{mm}$. For a $35.4\,\Omega$ ($= 50/\sqrt{2}$) line, $\varepsilon_{eff} \approx 2.85$ and $W \approx 2.08\,\text{mm}$. Find the physical length of each arm.

(c) Estimate the total footprint area (assume the coupler is a square ring of $\lambda/4$ arms).

Answer

已知/目標:Given / required:既知/目標:$f_0 = 2.4\,\text{GHz}$、$Z_0 = 50\,\Omega$,基板 RO4003C;$50\,\Omega$ 線 $\varepsilon_{eff} \approx 2.64$、$35.4\,\Omega$ 線 $\varepsilon_{eff} \approx 2.85$。求四臂阻抗、各臂實體長度、總面積。$f_0 = 2.4\,\text{GHz}$ and $Z_0 = 50\,\Omega$ on RO4003C substrate; the $50\,\Omega$ line has $\varepsilon_{eff} \approx 2.64$ and the $35.4\,\Omega$ line $\varepsilon_{eff} \approx 2.85$. Find the four arm impedances, the physical length of each arm, and the total area.$f_0 = 2.4\,\text{GHz}$、$Z_0 = 50\,\Omega$、基板は RO4003C;$50\,\Omega$ 線は $\varepsilon_{eff} \approx 2.64$、$35.4\,\Omega$ 線は $\varepsilon_{eff} \approx 2.85$。四つのアームのインピーダンス、各アームの実体長、総面積を求めます。

公式與來源:Formulas and sources:公式と出典:branch-line coupler 的標準設計(由 even-odd mode 匹配條件導出):series 臂 $Z_0$、shunt 臂 $Z_0/\sqrt{2}$,全部 $\lambda_g/4$。microstrip 的 guided wavelength: The standard branch-line coupler design (derived from the even-odd mode matching conditions): series arms of $Z_0$, shunt arms of $Z_0/\sqrt{2}$, all $\lambda_g/4$ long. The guided wavelength in microstrip:branch-line coupler の標準設計(even-odd mode の整合条件から導出):series アーム $Z_0$、shunt アーム $Z_0/\sqrt{2}$、すべて $\lambda_g/4$。microstrip の guided wavelength:

$$\lambda_g = \frac{c}{f_0\sqrt{\varepsilon_{eff}}}, \qquad L_{arm} = \frac{\lambda_g}{4}$$

注意 $\varepsilon_{eff}$ 與線寬有關:低阻抗線較寬 → 場更集中於介質 → $\varepsilon_{eff}$ 較大 → 同頻率下波長略短。Note that $\varepsilon_{eff}$ depends on line width: a lower-impedance line is wider → the field is more concentrated in the dielectric → $\varepsilon_{eff}$ is larger → the wavelength is slightly shorter at the same frequency.$\varepsilon_{eff}$ が線幅に依存する点に注意してください:低インピーダンス線は幅が広い → 電界が誘電体により集中する → $\varepsilon_{eff}$ が大きくなる → 同じ周波数で波長がやや短くなります。

逐步代入:Substituting step by step:順に代入:

(a) series 臂(上/下):$Z_0 = 50\,\Omega$;shunt 臂(左/右):$Z_0/\sqrt{2} = 50/1.414 = 35.4\,\Omega$。(a) The series arms (top/bottom): $Z_0 = 50\,\Omega$; the shunt arms (left/right): $Z_0/\sqrt{2} = 50/1.414 = 35.4\,\Omega$.(a) series アーム(上/下):$Z_0 = 50\,\Omega$;shunt アーム(左/右):$Z_0/\sqrt{2} = 50/1.414 = 35.4\,\Omega$。

(b) 50 Ω 臂($c = 3\times 10^{11}\,\text{mm/s}$、$f_0 = 2.4\times 10^9\,\text{Hz}$):(b) The 50 Ω arms (with $c = 3\times 10^{11}\,\text{mm/s}$ and $f_0 = 2.4\times 10^9\,\text{Hz}$):(b) 50 Ω アーム($c = 3\times 10^{11}\,\text{mm/s}$、$f_0 = 2.4\times 10^9\,\text{Hz}$):

$$\lambda_{g,50} = \dfrac{3\times10^{11}}{2.4\times10^9\times\sqrt{2.64}} = \dfrac{125\,\text{mm}}{1.625} = 76.9\,\text{mm} \;\Rightarrow\; L_{50} = \frac{76.9}{4} = \mathbf{19.2\,mm}\;(W = 1.13\,\text{mm})$$

35.4 Ω 臂:The 35.4 Ω arms:35.4 Ω アーム:

$$\lambda_{g,35.4} = \dfrac{3\times10^{11}}{2.4\times10^9\times\sqrt{2.85}} = \dfrac{125\,\text{mm}}{1.688} = 74.0\,\text{mm} \;\Rightarrow\; L_{35.4} = \frac{74.0}{4} = \mathbf{18.5\,mm}\;(W = 2.08\,\text{mm})$$

(c) coupler 為一個約 $19.2\,\text{mm} \times 18.5\,\text{mm}$ 的方環(series 臂長 × shunt 臂長)。(c) The coupler is a square ring of about $19.2\,\text{mm} \times 18.5\,\text{mm}$ (series arm length × shunt arm length).(c) coupler は約 $19.2\,\text{mm} \times 18.5\,\text{mm}$ の正方形リングです(series アーム長 × shunt アーム長)。

Footprint ≈ 19.2 × 18.5 = 355 mm² ≈ 3.6 cm²

物理詮釋:Physical interpretation:物理的解釈:3.6 cm² 在 2.4 GHz 的 PCB 上已算可觀 — 分佈式 hybrid 的面積 $\propto \lambda_g^2 \propto 1/f_0^2$,這就是為什麼低頻(sub-GHz)寧可用 lumped LC 或 transformer hybrid,而毫米波段 branch-line 反而變得迷你。兩臂長度略有差(19.2 vs 18.5 mm)提醒:不同阻抗的線必須各自用自己的 $\varepsilon_{eff}$ 算長度,不能共用。 3.6 cm² is already substantial on a PCB at 2.4 GHz — the area of a distributed hybrid goes as $\lambda_g^2 \propto 1/f_0^2$, which is why sub-GHz designs prefer lumped LC or transformer hybrids, while at millimetre-wave a branch-line becomes tiny. The slight difference in arm lengths (19.2 vs 18.5 mm) is a reminder that lines of different impedance must each use their own $\varepsilon_{eff}$; you cannot share one value.3.6 cm² は 2.4 GHz の PCB 上では既にかなりの大きさです — 分布定数ハイブリッドの面積は $\lambda_g^2 \propto 1/f_0^2$ なので、低周波(sub-GHz)では lumped LC やトランス型ハイブリッドが好まれ、ミリ波帯では branch-line は逆に極めて小さくなります。両アームの長さがわずかに異なる(19.2 対 18.5 mm)ことは、インピーダンスの異なる線はそれぞれ自分の $\varepsilon_{eff}$ を用いて長さを計算しなければならず、共用できないことを示しています。

3.3 Design
Design a 90° single-balanced mixer for Wi-Fi 5 GHz

Spec for a single-balanced 90° diode mixer:

Requirements RF: 5.18–5.32 GHz (UNII-2 Wi-Fi band)
LO: 5.08–5.22 GHz (fixed 100 MHz IF)
IF: 100 MHz ± 10 MHz
Conversion loss ≤ 9 dB
LO-to-RF isolation ≥ 25 dB
LO AM noise rejection ≥ 20 dB
Substrate: FR4 (ε_r=4.4, h=1.6 mm) (for size estimate only)

(a) What bandwidth must the branch-line coupler cover? Express as a fractional bandwidth.

(b) Sketch the block diagram and label the four ports (RF in, LO in, 2× diode mixers, IF combiner).

(c) Why is the 90° hybrid preferable to the 180° hybrid for LO AM noise rejection in this topology? (1–2 sentences)

(d) Estimate the coupler physical size on FR4 at $f_0 = 5.25\,\text{GHz}$, assuming $\varepsilon_{eff,50\Omega} \approx 3.2$.

Answer

已知/目標:Given / required:既知/目標:RF 5.18–5.32 GHz、LO 5.08–5.22 GHz、IF 100 MHz、$L_C \leq 9\,\text{dB}$、LO-RF isolation ≥ 25 dB、AM noise rejection ≥ 20 dB,FR4 基板($\varepsilon_{eff,50\Omega} \approx 3.2$)。求 coupler 頻寬需求、方塊圖、90° vs 180° 的理由、實體尺寸。 RF 5.18–5.32 GHz, LO 5.08–5.22 GHz, IF 100 MHz, $L_C \leq 9\,\text{dB}$, LO-RF isolation ≥ 25 dB, AM noise rejection ≥ 20 dB, on FR4 substrate ($\varepsilon_{eff,50\Omega} \approx 3.2$). Find the coupler bandwidth requirement, the block diagram, the reasoning for 90° vs 180°, and the physical dimensions.RF 5.18–5.32 GHz、LO 5.08–5.22 GHz、IF 100 MHz、$L_C \leq 9\,\text{dB}$、LO-RF アイソレーション ≥ 25 dB、AM ノイズ除去 ≥ 20 dB、FR4 基板($\varepsilon_{eff,50\Omega} \approx 3.2$)。coupler の帯域幅要求、ブロック図、90° と 180° の選択理由、実体寸法を求めます。

(a) 公式與來源(a) Formula and source(a) 公式と出典 — fractional bandwidth 定義 $FBW = BW/f_0$。 — the definition of fractional bandwidth, $FBW = BW/f_0$. — fractional bandwidth の定義 $FBW = BW/f_0$。逐步代入:Substituting step by step:順に代入:hybrid 必須同時涵蓋 RF(5.18–5.32)與 LO(5.08–5.22)→ 總涵蓋範圍 5.08–5.32 GHz,但以 RF band 計: The hybrid must cover both the RF (5.18–5.32) and the LO (5.08–5.22) → a total span of 5.08–5.32 GHz, but reckoned against the RF band:hybrid は RF(5.18–5.32)と LO(5.08–5.22)の両方を同時にカバーする必要があります → 総カバー範囲は 5.08–5.32 GHz ですが、RF band を基準に計算すると:

$$BW = 5.32 - 5.18 = 0.14\,\text{GHz}, \quad f_0 = \frac{5.18+5.32}{2} = 5.25\,\text{GHz}, \quad FBW = \frac{140}{5250} \approx \mathbf{2.7\%}$$

branch-line coupler 典型可用頻寬約 ±10–15%(amplitude/phase balance 劣化前),2.7% 遠在舒適區內 ✓(即使算入 LO 下緣 5.08 GHz,總 FBW 也只有 $240/5200 \approx 4.6\%$。)A branch-line coupler's typical usable bandwidth is about ±10–15% (before the amplitude/phase balance degrades), so 2.7% is comfortably inside it ✓ (even counting the LO's lower edge at 5.08 GHz, the total FBW is only $240/5200 \approx 4.6\%$.)branch-line coupler の典型的な使用可能帯域幅は約 ±10–15%(振幅/位相バランスが劣化する前)なので、2.7% は十分に余裕のある範囲です ✓(LO 下端の 5.08 GHz を含めても、総 FBW は $240/5200 \approx 4.6\%$ にすぎません。)

(b) Block diagram(文字版): Block diagram (in words): ブロック図(文章版):
   RF → Port 1 of 90° Hybrid
   LO → Port 4 of 90° Hybrid(port 1↔4 isolated → 天然提供 LO-RF isolation)    LO → Port 4 of the 90° hybrid (ports 1↔4 are isolated → providing LO-RF isolation naturally)    LO → 90° Hybrid の Port 4(port 1↔4 は isolated → LO-RF アイソレーションを自然に提供)
   Port 2 → Diode D1 → IF1
   Port 3 → Diode D2 → IF2
   IF1, IF2 → 180° power combiner (or balun) → IF out(100 MHz low-pass 取出差頻)    IF1, IF2 → a 180° power combiner (or balun) → IF out (a 100 MHz low-pass extracts the difference frequency)    IF1, IF2 → 180° power combiner (or balun) → IF out(100 MHz low-pass で差周波数を取り出す)

(c) 90° hybrid 讓兩顆 diode 收到的 LO 振幅相同、AM noise 經 envelope detection 後在兩路 The 90° hybrid gives both diodes the same LO amplitude, so after envelope detection the AM noise is 90° hybrid は 2 個のダイオードに同じ LO 振幅を与えるので、AM ノイズは envelope detection を経た後、2 系統で同相(common mode)in phase (common mode)同相(common mode),IF 減法 combiner 直接相消(見 3.1 題推導);desired IF 則帶 ±90° 路徑相位差呈 differential 而相加。spec 要求的 ≥ 20 dB AM noise rejection 由兩 diode 的匹配度保證(典型可達 20–30 dB)。 on the two paths and the subtracting IF combiner cancels it outright (see the derivation of problem 3.1); the desired IF, carrying a ±90° path phase difference, is differential and adds. The ≥ 20 dB of AM noise rejection required by the spec is guaranteed by how well the two diodes are matched (20–30 dB is typically achievable).となり、IF 減算合成器がそのまま相殺します(3.1 題の導出を参照);希望 IF のほうは ±90° の経路位相差を持つため differential となり加算されます。spec が要求する ≥ 20 dB の AM ノイズ除去は 2 個のダイオードの整合度によって保証されます(典型的に 20–30 dB が得られます)。

(d) 逐步代入(d) Substituting step by step(d) 順に代入 — FR4 上 $f_0 = 5.25\,\text{GHz}$: — on FR4 at $f_0 = 5.25\,\text{GHz}$: — FR4 上で $f_0 = 5.25\,\text{GHz}$:

$$\lambda_{g} = \dfrac{c}{f_0\sqrt{\varepsilon_{eff}}} = \dfrac{3\times10^{11}\,\text{mm/s}}{5.25\times10^9\times\sqrt{3.2}} = \dfrac{57.1\,\text{mm}}{1.789} = 31.9\,\text{mm}$$

$$L_{arm} = \lambda_g/4 = 31.9/4 \approx 8.0\,\text{mm} \;\Rightarrow\; \text{footprint} \approx 8\times 8 = 64\,\text{mm}^2$$

Size: ~8 mm × 8 mm on FR4 at 5.25 GHz

物理詮釋:Physical interpretation:物理的解釈:這個設計把三個 spec 一次收掉:isolation 來自 hybrid 的 port 隔離、AM rejection 來自 balanced 對稱性、$L_C \leq 9\,\text{dB}$ 則靠單顆 diode 的 ~6–8 dB 加上 hybrid 損耗 0.5–1 dB(FR4 在 5 GHz 的 $\tan\delta$ 損耗偏高,留 margin)。注意 FR4 只用來估尺寸 — 量產 5 GHz 設計多半改用低損耗基板。 This design settles all three specs at once: the isolation comes from the hybrid's port isolation, the AM rejection from the balanced symmetry, and the $L_C \leq 9\,\text{dB}$ from a single diode's ~6–8 dB plus 0.5–1 dB of hybrid loss (FR4's $\tan\delta$ loss is rather high at 5 GHz, so leave margin). Note that FR4 is used here only to estimate dimensions — a production 5 GHz design would usually move to a low-loss substrate.この設計は三つの spec を一度に満たします:アイソレーションは hybrid のポート隔離から、AM rejection は平衡対称性から、$L_C \leq 9\,\text{dB}$ は単一ダイオードの ~6–8 dB に hybrid 損失 0.5–1 dB を加えたものから得られます(FR4 の 5 GHz における $\tan\delta$ 損失は高めなので margin を残します)。FR4 はここでは寸法の見積もりに使っているだけである点に注意してください — 量産の 5 GHz 設計では通常、低損失基板に変更します。

Topic 4Single-Balanced 180° Mixer (Rat-Race)

Rat-race ring hybrid, Σ/Δ ports, even-odd mode analysis, S-matrix.

4.1 Easy
Rat-race S-matrix element lookup

The ideal rat-race (port numbering: 1=Σ, 2=through, 3=through, 4=Δ) S-matrix is:

$$[S] = \frac{-j}{\sqrt{2}}\begin{bmatrix}0&1&1&0\\1&0&0&-1\\1&0&0&1\\0&-1&1&0\end{bmatrix}$$

(a) RF enters port 1 (Σ). What are the amplitudes and phases at ports 2, 3, 4?

(b) RF enters port 4 (Δ). What are the amplitudes and phases at ports 2, 3? Is port 1 isolated?

(c) Which port pair provides 180° phase difference for Δ excitation?

Answer

已知/目標:Given / required:既知/目標:rat-race 的理想 $[S]$(題目給定,係數 $\frac{-j}{\sqrt{2}}$)。分別以 port 1(Σ)與 port 4(Δ)激發,讀出各埠輸出的振幅與相位。 The ideal $[S]$ of a rat-race (given in the problem, with a coefficient of $\frac{-j}{\sqrt{2}}$). Excite port 1 (Σ) and port 4 (Δ) in turn, and read off the amplitude and phase at each output port.rat-race の理想的な $[S]$(問題で与えられ、係数は $\frac{-j}{\sqrt{2}}$)。port 1(Σ)と port 4(Δ)で順に励振し、各ポートの出力の振幅と位相を読み取ります。

公式與來源:Formulas and sources:公式と出典:S-parameter 定義 $b_i = \sum_j S_{ij}a_j$;單埠激發時 $b_i = S_{ij}a_j$,直接「查矩陣第 $j$ 行」。振幅 $\left|\frac{-j}{\sqrt{2}}\right| = \frac{1}{\sqrt{2}}$ → $20\log_{10}\frac{1}{\sqrt{2}} = -3\,\text{dB}$;相位 $\angle(-j) = -90°$、$\angle(+j) = +90°$。 The S-parameter definition $b_i = \sum_j S_{ij}a_j$; for a single-port excitation $b_i = S_{ij}a_j$, so you simply “read column $j$ of the matrix”. The amplitude $\left|\frac{-j}{\sqrt{2}}\right| = \frac{1}{\sqrt{2}}$ → $20\log_{10}\frac{1}{\sqrt{2}} = -3\,\text{dB}$; the phases are $\angle(-j) = -90°$ and $\angle(+j) = +90°$.S-parameter の定義 $b_i = \sum_j S_{ij}a_j$;単一ポート励振では $b_i = S_{ij}a_j$ なので、そのまま「行列の第 $j$ 列を読む」だけです。振幅 $\left|\frac{-j}{\sqrt{2}}\right| = \frac{1}{\sqrt{2}}$ → $20\log_{10}\frac{1}{\sqrt{2}} = -3\,\text{dB}$;位相は $\angle(-j) = -90°$、$\angle(+j) = +90°$。

(a) 逐步代入(a) Substituting step by step(a) 順に代入 — port 1(Σ)激發 $a_1$,讀第 1 行(matrix 元素依序 $0, 1, 1, 0$): — exciting port 1 (Σ) with $a_1$ and reading row 1 (matrix elements $0, 1, 1, 0$ in order): — port 1(Σ)を $a_1$ で励振し、第 1 行を読みます(行列要素は順に $0, 1, 1, 0$):

$b_2 = S_{21}a_1 = \frac{-j}{\sqrt{2}}\cdot 1\cdot a_1$ → −3 dB, −90°

$b_3 = S_{31}a_1 = \frac{-j}{\sqrt{2}}\cdot 1\cdot a_1$ → −3 dB, −90°(與 port 2 同相)(in phase with port 2)(port 2 と同相)

$b_4 = S_{41}a_1 = \frac{-j}{\sqrt{2}}\cdot 0 = 0$ → isolated

Σ port 對 ports 2、3 做等幅同相分配(in-phase split)。The Σ port splits equally and in phase to ports 2 and 3 (an in-phase split).Σ port は ports 2、3 へ等振幅・同相で分配します(in-phase split)。

(b) port 4(Δ)激發 $a_4$,讀第 4 行(元素 $0, -1, 1, 0$): Exciting port 4 (Δ) with $a_4$ and reading row 4 (elements $0, -1, 1, 0$): port 4(Δ)を $a_4$ で励振し、第 4 行を読みます(要素 $0, -1, 1, 0$):

$b_2 = S_{24}a_4 = \frac{-j}{\sqrt{2}}\cdot(-1)\cdot a_4 = \frac{+j}{\sqrt{2}}a_4$ → −3 dB, +90°

$b_3 = S_{34}a_4 = \frac{-j}{\sqrt{2}}\cdot(+1)\cdot a_4 = \frac{-j}{\sqrt{2}}a_4$ → −3 dB, −90°

$b_1 = S_{14}a_4 = 0$ → port 1(Σ)對 Δ 隔離port 1 (Σ) is isolated from Δport 1(Σ)は Δ に対してアイソレート

(c) Δ 激發下 port 2 為 +90°、port 3 為 −90° → 相位差 $+90° - (-90°) = \mathbf{180°}$(anti-phase pair)。 Under Δ excitation port 2 is at +90° and port 3 at −90° → a phase difference of $+90° - (-90°) = \mathbf{180°}$ (an anti-phase pair). Δ 励振では port 2 が +90°、port 3 が −90° → 位相差 $+90° - (-90°) = \mathbf{180°}$(anti-phase pair)。

Σ→ equal phase split; Δ→ 180° anti-phase split; each port is isolated from its "partner" (Σ↔Δ).

物理詮釋:Physical interpretation:物理的解釈:Σ/Δ 的行為來自環上兩條路徑的長度差:到兩輸出埠等長(各 $\lambda/4$)→ 同相;經 $3\lambda/4$ 長臂的那條多走 $\lambda/2$ → 反相。這對 0°/180° 輸出正是 180° single-balanced mixer 要的:LO 從 Δ 進讓兩顆 diode 反相切換、RF 從 Σ 進同相饋入(見 4.4 題),且 Σ↔Δ 隔離天然給出 LO-RF isolation。 The Σ/Δ behaviour comes from the difference in path length around the ring: the paths to the two output ports are equal (each $\lambda/4$) → in phase; while the path through the $3\lambda/4$ long arm travels an extra $\lambda/2$ → in anti-phase. This pair of 0°/180° outputs is exactly what a 180° single-balanced mixer needs: the LO entering at Δ switches the two diodes in anti-phase and the RF entering at Σ feeds them in phase (see problem 4.4), while the natural Σ↔Δ isolation provides LO-RF isolation.Σ/Δ の振る舞いはリング上の 2 経路の長さの差から生じます:2 つの出力ポートまでは等長(各 $\lambda/4$)→ 同相、$3\lambda/4$ の長いアームを通る経路は $\lambda/2$ 余分に進む → 逆相。この 0°/180° 出力の組はまさに 180° 単平衡ミキサが必要とするものです:Δ から入る LO が 2 個のダイオードを逆相で切り替え、Σ から入る RF が同相で給電され(4.4 題参照)、さらに Σ↔Δ の隔離が自然に LO-RF アイソレーションを与えます。

4.2 Medium
Rat-race design at 10 GHz — find ring dimensions

Design a rat-race coupler for $f_0 = 10\,\text{GHz}$, $Z_0 = 50\,\Omega$ on RT/Duroid 5880 ($\varepsilon_r = 2.2$, $h = 0.254\,\text{mm}$).

For this substrate and impedance, use: $Z_R = \sqrt{2}\times 50 = 70.7\,\Omega$, $\varepsilon_{eff,70.7} \approx 1.84$, $W_{70.7} \approx 0.28\,\text{mm}$.

(a) Find the guided wavelength at 10 GHz.

(b) Find the physical lengths of the three $\lambda/4$ arms and the one $3\lambda/4$ arm.

(c) What is the total ring circumference in mm?

Answer

已知/目標:Given / required:既知/目標:$f_0 = 10\,\text{GHz}$、$Z_0 = 50\,\Omega$,RT/Duroid 5880;環阻抗 $Z_R = \sqrt{2}\times 50 = 70.7\,\Omega$,此線寬下 $\varepsilon_{eff} \approx 1.84$、$W \approx 0.28\,\text{mm}$。求 $\lambda_g$、三段 $\lambda/4$ 臂與一段 $3\lambda/4$ 臂的實體長度、環總周長。$f_0 = 10\,\text{GHz}$ and $Z_0 = 50\,\Omega$ on RT/Duroid 5880; the ring impedance is $Z_R = \sqrt{2}\times 50 = 70.7\,\Omega$, and at that line width $\varepsilon_{eff} \approx 1.84$ with $W \approx 0.28\,\text{mm}$. Find $\lambda_g$, the physical lengths of the three $\lambda/4$ arms and the one $3\lambda/4$ arm, and the total ring circumference.$f_0 = 10\,\text{GHz}$、$Z_0 = 50\,\Omega$、RT/Duroid 5880;リングのインピーダンスは $Z_R = \sqrt{2}\times 50 = 70.7\,\Omega$、この線幅では $\varepsilon_{eff} \approx 1.84$、$W \approx 0.28\,\text{mm}$。$\lambda_g$、3 本の $\lambda/4$ アームと 1 本の $3\lambda/4$ アームの実体長、リングの総周長を求めます。

公式與來源:Formulas and sources:公式と出典:rat-race 的標準結構 — 環特性阻抗 $Z_R = \sqrt{2}Z_0$(使每個 tee junction 對外匹配)、三段 $\lambda_g/4$ 短臂加一段 $3\lambda_g/4$ 長臂。guided wavelength: The standard rat-race structure — a ring characteristic impedance $Z_R = \sqrt{2}Z_0$ (making each tee junction matched to the outside), three $\lambda_g/4$ short arms plus one $3\lambda_g/4$ long arm. The guided wavelength:rat-race の標準構造 — リングの特性インピーダンス $Z_R = \sqrt{2}Z_0$(各 tee junction を外部に対して整合させる)、3 本の $\lambda_g/4$ 短アームと 1 本の $3\lambda_g/4$ 長アーム。guided wavelength:

$$\lambda_g = \frac{c}{f_0\sqrt{\varepsilon_{eff}}}$$

逐步代入:Substituting step by step:順に代入:

(a) $\lambda_g = \dfrac{3\times10^{11}\,\text{mm/s}}{10^{10}\,\text{Hz}\times\sqrt{1.84}} = \dfrac{30\,\text{mm}}{1.356} = \mathbf{22.1\,mm}$

(對照自由空間 $\lambda_0 = 30\,\text{mm}$:介質讓波長縮短 $\sqrt{1.84} = 1.36$ 倍。)(Compare free space, $\lambda_0 = 30\,\text{mm}$: the dielectric shortens the wavelength by a factor of $\sqrt{1.84} = 1.36$.)(自由空間 $\lambda_0 = 30\,\text{mm}$ との対照:誘電体により波長は $\sqrt{1.84} = 1.36$ 倍短くなります。)

(b) $L_{\lambda/4} = \dfrac{22.1}{4} = \mathbf{5.53\,mm}$(共三段);$L_{3\lambda/4} = 3\times 5.53 = \mathbf{16.6\,mm}$(一段)。(b) $L_{\lambda/4} = \dfrac{22.1}{4} = \mathbf{5.53\,mm}$ (three of them); $L_{3\lambda/4} = 3\times 5.53 = \mathbf{16.6\,mm}$ (one).(b) $L_{\lambda/4} = \dfrac{22.1}{4} = \mathbf{5.53\,mm}$(計 3 本);$L_{3\lambda/4} = 3\times 5.53 = \mathbf{16.6\,mm}$(1 本)。

(c) 總周長 $= 3\times L_{\lambda/4} + L_{3\lambda/4} = 3\times 5.53 + 16.6 = 16.6 + 16.6 = \mathbf{33.2\,mm}$。(c) The total circumference $= 3\times L_{\lambda/4} + L_{3\lambda/4} = 3\times 5.53 + 16.6 = 16.6 + 16.6 = \mathbf{33.2\,mm}$.(c) 総周長 $= 3\times L_{\lambda/4} + L_{3\lambda/4} = 3\times 5.53 + 16.6 = 16.6 + 16.6 = \mathbf{33.2\,mm}$。

Ring circumference = 3×(λ/4) + 1×(3λ/4) = 6λ/4 = 3λ_g/2 = 1.5 × 22.1 = 33.2 mm ✓

物理詮釋:Physical interpretation:物理的解釈:rat-race 周長固定是 $1.5\lambda_g$(直徑 $\approx 1.5\lambda_g/\pi \approx 10.6\,\text{mm}$),比同頻的 branch-line(周長 $\approx \lambda_g$)大一半 — 換來的是 Σ/Δ(0°/180°)功能與較佳的輸出隔離。低 $\varepsilon_r$ 的 5880(2.2)在 X-band 常用,因為損耗低、色散小;代價是電路略大。 A rat-race's circumference is always $1.5\lambda_g$ (a diameter of $\approx 1.5\lambda_g/\pi \approx 10.6\,\text{mm}$), half again as large as a branch-line at the same frequency (circumference $\approx \lambda_g$) — bought in exchange for the Σ/Δ (0°/180°) function and better output isolation. The low-$\varepsilon_r$ 5880 (2.2) is common at X-band because it is low-loss and low-dispersion; the price is a slightly larger circuit.rat-race の周長は常に $1.5\lambda_g$(直径 $\approx 1.5\lambda_g/\pi \approx 10.6\,\text{mm}$)で、同じ周波数の branch-line(周長 $\approx \lambda_g$)より 1.5 倍大きくなります — その代わりに Σ/Δ(0°/180°)機能と優れた出力アイソレーションが得られます。低 $\varepsilon_r$ の 5880(2.2)は損失が小さく分散も小さいため X-band でよく使われ、代償は回路がやや大きくなることです。

4.3 Medium
Even-odd mode analysis: why are ports 1 and 4 isolated?

For a rat-race excited at port 1 (Σ), the even-odd mode decomposition places:

  • Even mode: +½ at port 1, +½ at port 4
  • Odd mode: +½ at port 1, −½ at port 4

(a) In even mode, a magnetic (open) wall bisects the ring. For the stub seen from port 4: it sees a λ/4 short-circuit stub in parallel with a 3λ/4 short-circuit stub. Evaluate both input impedances. What is the parallel combination?

(b) Repeat for odd mode (electric/short wall). What impedances do the stubs present?

(c) Using $b_4 = \frac{1}{2}(\Gamma_e - \Gamma_o)$, show that port 4 is isolated.

Answer

已知/目標:Given / required:既知/目標:port 1(Σ)激發,even/odd 分解為 $(+\tfrac{1}{2}, +\tfrac{1}{2})$ 與 $(+\tfrac{1}{2}, -\tfrac{1}{2})$ 加在 ports 1、4。求兩個 mode 下 port 4 看到的 stub 阻抗與 $\Gamma_e$、$\Gamma_o$,再用 $b_4 = \tfrac{1}{2}(\Gamma_e - \Gamma_o)$ 證明隔離。 Exciting port 1 (Σ), decomposed into even/odd excitations of $(+\tfrac{1}{2}, +\tfrac{1}{2})$ and $(+\tfrac{1}{2}, -\tfrac{1}{2})$ applied at ports 1 and 4. Find the stub impedances seen from port 4 in each mode along with $\Gamma_e$ and $\Gamma_o$, then prove the isolation using $b_4 = \tfrac{1}{2}(\Gamma_e - \Gamma_o)$.port 1(Σ)励振を、ports 1、4 に加える $(+\tfrac{1}{2}, +\tfrac{1}{2})$ と $(+\tfrac{1}{2}, -\tfrac{1}{2})$ の even/odd 励振に分解します。両モードで port 4 から見える stub のインピーダンスと $\Gamma_e$、$\Gamma_o$ を求め、さらに $b_4 = \tfrac{1}{2}(\Gamma_e - \Gamma_o)$ を用いてアイソレーションを証明します。

更正:Correction:訂正:對稱面是把每段 arm 從 the plane of symmetry cuts each arm at its 対称面は各アームを中點midpoint中点切開,所以 port 4 看到的是「半臂」stub:$\lambda/8$($\lambda/4$ 臂的一半)與 $3\lambda/8$($3\lambda/4$ 臂的一半),而不是整段 $\lambda/4$、$3\lambda/4$。終端條件由 wall 決定:magnetic wall(even)= 切口開路、electric wall(odd)= 切口短路。原版答案把 open/short stub 公式對調且長度未除半,以下按正確幾何計算(與本講義 even-odd 推導章節一致),結論不變。, so what port 4 sees are “half-arm” stubs: $\lambda/8$ (half of a $\lambda/4$ arm) and $3\lambda/8$ (half of the $3\lambda/4$ arm), not whole $\lambda/4$ and $3\lambda/4$ sections. The termination is set by the wall: a magnetic wall (even) = an open at the cut, an electric wall (odd) = a short. The original answer had the open/short stub formulas interchanged and did not halve the lengths; the calculation below uses the correct geometry (consistent with the even-odd derivation section of these notes), and the conclusion is unchanged.で切るので、port 4 から見えるのは「半アーム」の stub です:$\lambda/8$($\lambda/4$ アームの半分)と $3\lambda/8$($3\lambda/4$ アームの半分)であって、$\lambda/4$、$3\lambda/4$ の全長ではありません。終端条件は wall で決まります:magnetic wall(even)= 切り口が開放、electric wall(odd)= 短絡。元の解答は open/short stub の公式が入れ替わっており、長さも半分にしていませんでした。以下は正しい幾何に基づく計算(本講義の even-odd 導出の節と一致)で、結論は変わりません。

公式與來源:Formulas and sources:公式と出典:傳輸線 stub 輸入阻抗(環阻抗 $Z_R = \sqrt{2}Z_0$): The input impedance of a transmission-line stub (ring impedance $Z_R = \sqrt{2}Z_0$):伝送線路 stub の入力インピーダンス(リングのインピーダンス $Z_R = \sqrt{2}Z_0$):

$$Z_{in}^{oc} = -jZ_R\cot(\beta\ell)\;(\text{開路端}), \qquad Z_{in}^{sc} = +jZ_R\tan(\beta\ell)\;(\text{短路端}), \qquad \beta\ell = \frac{2\pi}{\lambda}\ell$$$$Z_{in}^{oc} = -jZ_R\cot(\beta\ell)\;(\text{open-circuited end}), \qquad Z_{in}^{sc} = +jZ_R\tan(\beta\ell)\;(\text{short-circuited end}), \qquad \beta\ell = \frac{2\pi}{\lambda}\ell$$$$Z_{in}^{oc} = -jZ_R\cot(\beta\ell)\;(\text{open-circuited end}), \qquad Z_{in}^{sc} = +jZ_R\tan(\beta\ell)\;(\text{short-circuited end}), \qquad \beta\ell = \frac{2\pi}{\lambda}\ell$$

(a) Even mode — magnetic(open)wall → 兩條 open-circuit stub,逐步代入: — a magnetic (open) wall → two open-circuit stubs; substituting step by step: — magnetic(open)wall → 2 本の open-circuit stub、順に代入:

$\lambda/8$ stub:$\beta\ell = 45°$ → $Z = -jZ_R\cot 45° = -jZ_R$(容性)the $\lambda/8$ stub: $\beta\ell = 45°$ → $Z = -jZ_R\cot 45° = -jZ_R$ (capacitive)$\lambda/8$ stub:$\beta\ell = 45°$ → $Z = -jZ_R\cot 45° = -jZ_R$(容量性)

$3\lambda/8$ stub:$\beta\ell = 135°$ → $\cot 135° = -1$ → $Z = +jZ_R$(感性)the $3\lambda/8$ stub: $\beta\ell = 135°$ → $\cot 135° = -1$ → $Z = +jZ_R$ (inductive)$3\lambda/8$ stub:$\beta\ell = 135°$ → $\cot 135° = -1$ → $Z = +jZ_R$(誘導性)

並聯:$(-jZ_R)\parallel(+jZ_R) = \dfrac{(-jZ_R)(+jZ_R)}{-jZ_R + jZ_R} = \dfrac{Z_R^2}{0} \to \boxed{\infty}$(parallel resonance,開路)→ $\Gamma_e = +1$In parallel: $(-jZ_R)\parallel(+jZ_R) = \dfrac{(-jZ_R)(+jZ_R)}{-jZ_R + jZ_R} = \dfrac{Z_R^2}{0} \to \boxed{\infty}$ (a parallel resonance, an open circuit) → $\Gamma_e = +1$並列:$(-jZ_R)\parallel(+jZ_R) = \dfrac{(-jZ_R)(+jZ_R)}{-jZ_R + jZ_R} = \dfrac{Z_R^2}{0} \to \boxed{\infty}$(parallel resonance、開放)→ $\Gamma_e = +1$

(b) Odd mode — electric(short)wall → 兩條 short-circuit stub: — an electric (short) wall → two short-circuit stubs: — electric(short)wall → 2 本の short-circuit stub:

$\lambda/8$ stub:$Z = +jZ_R\tan 45° = +jZ_R$;$3\lambda/8$ stub:$\tan 135° = -1$ → $Z = -jZ_R$the $\lambda/8$ stub: $Z = +jZ_R\tan 45° = +jZ_R$; the $3\lambda/8$ stub: $\tan 135° = -1$ → $Z = -jZ_R$$\lambda/8$ stub:$Z = +jZ_R\tan 45° = +jZ_R$;$3\lambda/8$ stub:$\tan 135° = -1$ → $Z = -jZ_R$

並聯:$(+jZ_R)\parallel(-jZ_R) \to \boxed{\infty}$ → 同樣開路 → $\Gamma_o = +1$In parallel: $(+jZ_R)\parallel(-jZ_R) \to \boxed{\infty}$ → again an open circuit → $\Gamma_o = +1$並列:$(+jZ_R)\parallel(-jZ_R) \to \boxed{\infty}$ → 同様に開放 → $\Gamma_o = +1$

(c) 結果:(c) The result:(c) 結果:

$$b_4 = \tfrac{1}{2}(\Gamma_e - \Gamma_o) = \tfrac{1}{2}(+1 - (+1)) = \mathbf{0} \;\Rightarrow\; S_{41} = 0,\;\text{port 4 isolated ✓}$$

Equal reflection in both modes (Γe = Γo = +1) → zero net wave out of port 4 → Σ and Δ are isolated.

物理詮釋:Physical interpretation:物理的解釈:隔離的本質是兩條 stub 的電納 $\pm jY_R$ 恰好互消 — 而這成立的條件正是兩臂長度差 $\lambda/4 - 3\lambda/4 \to$ 半臂差 $\lambda/4$,對應整環兩路徑差 $\lambda/2 = 180°$ 的干涉相消圖像。也因為 $\cot/\tan$ 只在 $f_0$ 恰好互為相反數,rat-race 的 isolation 是 The essence of the isolation is that the susceptances of the two stubs, $\pm jY_R$, cancel exactly — and the condition for that is precisely the arm-length difference $\lambda/4 - 3\lambda/4 \to$ a half-arm difference of $\lambda/4$, corresponding to the picture of destructive interference between two paths differing by $\lambda/2 = 180°$ around the whole ring. And because $\cot$ and $\tan$ are exact negatives of each other only at $f_0$, a rat-race's isolation is a アイソレーションの本質は 2 本の stub のサセプタンス $\pm jY_R$ がちょうど互いに打ち消し合うことです — そしてそれが成り立つ条件がまさに両アームの長さの差 $\lambda/4 - 3\lambda/4 \to$、すなわち半アームの差 $\lambda/4$ であり、環全体で 2 経路の差が $\lambda/2 = 180°$ であることによる干渉相殺の描像に対応します。また $\cot$ と $\tan$ が $f_0$ でのみ厳密に互いの符号を反転したものとなるため、rat-race のアイソレーションは窄頻narrowband狭帯域特性,離開中心頻率會迅速劣化(典型 20–30 dB 隔離頻寬約 ±10%)。 property that degrades rapidly away from the centre frequency (a typical 20–30 dB isolation bandwidth is about ±10%).特性であり、中心周波数から離れると急速に劣化します(典型的な 20–30 dB のアイソレーション帯域幅は約 ±10%)。

4.4 Design
Full 180° SB mixer design for cellular band (1.9 GHz)
Requirements RF: 1900 MHz, LO: 1800 MHz, IF: 100 MHz
Z₀ = 50 Ω, substrate: RO4003C (ε_r=3.55, h=0.508 mm)
For Z_R=70.7 Ω on RO4003C: ε_eff≈2.98, W≈0.62 mm
Conversion loss ≤ 8 dB, LO-to-RF isolation ≥ 30 dB

(a) Design the rat-race ring: compute $\lambda_g$, $L_{\lambda/4}$, $L_{3\lambda/4}$.

(b) Sketch the full mixer block diagram: rat-race ports labeled (1=Σ, 4=Δ), RF/LO port assignment, two diodes, IF summing network.

(c) Which port (Σ or Δ) should receive the RF signal? Which receives LO? Justify using the S-matrix phase relationship to maximize IF output from both diodes in-phase.

(d) Compared to a 90° single-balanced mixer, what is the key advantage of this 180° topology for this cellular application?

Answer

已知/目標:Given / required:既知/目標:RF 1900 MHz、LO 1800 MHz、IF 100 MHz;$Z_0 = 50\,\Omega$、RO4003C,$Z_R = 70.7\,\Omega$ 線的 $\varepsilon_{eff} \approx 2.98$、$W \approx 0.62\,\text{mm}$。求 ring 尺寸、方塊圖、RF/LO 埠指派與 90° 版的比較。 RF 1900 MHz, LO 1800 MHz, IF 100 MHz; $Z_0 = 50\,\Omega$ on RO4003C, with the $Z_R = 70.7\,\Omega$ line having $\varepsilon_{eff} \approx 2.98$ and $W \approx 0.62\,\text{mm}$. Find the ring dimensions, the block diagram, the RF/LO port assignment, and compare with the 90° version.RF 1900 MHz、LO 1800 MHz、IF 100 MHz;$Z_0 = 50\,\Omega$、RO4003C、$Z_R = 70.7\,\Omega$ 線の $\varepsilon_{eff} \approx 2.98$、$W \approx 0.62\,\text{mm}$。ring の寸法、ブロック図、RF/LO のポート指定、および 90° 版との比較を求めます。

(a) 公式與來源:(a) Formula and source:(a) 公式と出典:rat-race 設計頻率取 RF/LO 都要通過的中心 $f_0 \approx 1.9\,\text{GHz}$(RF 與 LO 只差 5%,一個環即可同時服務);$\lambda_g = c/(f_0\sqrt{\varepsilon_{eff}})$。 The rat-race design frequency is taken as the centre through which both RF and LO must pass, $f_0 \approx 1.9\,\text{GHz}$ (RF and LO differ by only 5%, so a single ring can serve both); $\lambda_g = c/(f_0\sqrt{\varepsilon_{eff}})$.rat-race の設計周波数は RF と LO の両方が通る中心 $f_0 \approx 1.9\,\text{GHz}$ とします(RF と LO の差はわずか 5% なので、一つの環で両方に対応できます);$\lambda_g = c/(f_0\sqrt{\varepsilon_{eff}})$。逐步代入:Substituting step by step:順に代入:

$$\lambda_g = \dfrac{3\times10^{11}\,\text{mm/s}}{1.9\times10^9\times\sqrt{2.98}} = \dfrac{157.9\,\text{mm}}{1.726} = 91.5\,\text{mm}$$

$$L_{\lambda/4} = \frac{91.5}{4} = 22.9\,\text{mm}, \qquad L_{3\lambda/4} = 3\times 22.9 = 68.6\,\text{mm}$$

L(λ/4) = 22.9 mm, L(3λ/4) = 68.6 mm, W = 0.62 mm(環周長 1.5λ_g ≈ 137 mm)L(λ/4) = 22.9 mm, L(3λ/4) = 68.6 mm, W = 0.62 mm (a ring circumference of 1.5λ_g ≈ 137 mm)L(λ/4) = 22.9 mm, L(3λ/4) = 68.6 mm, W = 0.62 mm(環の周長 1.5λ_g ≈ 137 mm)

(b) Block diagram:RF → Port 1(Σ);LO → Port 4(Δ);Port 2、Port 3 各接一顆 diode(其中一顆反接);兩 diode 輸出在 IF 節點匯合(或經 IF transformer combiner),後接 100 MHz low-pass。Σ↔Δ 的隔離直接給出 LO-to-RF isolation ≥ 30 dB 的基礎。 Block diagram: RF → Port 1 (Σ); LO → Port 4 (Δ); ports 2 and 3 each carry a diode (one of them reversed); the two diode outputs meet at the IF node (or through an IF transformer combiner), followed by a 100 MHz low-pass. The Σ↔Δ isolation directly provides the basis for LO-to-RF isolation ≥ 30 dB. ブロック図:RF → Port 1(Σ);LO → Port 4(Δ);Port 2、Port 3 にそれぞれダイオードを 1 個接続(うち 1 個は逆接続);2 つのダイオード出力は IF ノードで合流し(または IF transformer combiner を経由)、後段に 100 MHz low-pass を接続します。Σ↔Δ の隔離が LO-to-RF アイソレーション ≥ 30 dB の基礎を直接与えます。

(c) 埠指派與相位推導(c) Port assignment and the phase derivation(c) ポート指定と位相の導出(用 4.1 題的 $[S]$):(using the $[S]$ of problem 4.1): (4.1 題の $[S]$ を使用):RF → Σ (port 1) and LO → Δ (port 4):

  • 從 Σ:ports 2、3 都拿到 $-j/\sqrt{2}$(同相)→ RF 同相到達兩顆 diodeFrom Σ: ports 2 and 3 both receive $-j/\sqrt{2}$ (in phase) → the RF reaches both diodes in phaseΣ から:ports 2、3 はいずれも $-j/\sqrt{2}$ を受け取ります(同相)→ RF は両方のダイオードに同相で到達します
  • 從 Δ:port 2 拿 $+j/\sqrt{2}$、port 3 拿 $-j/\sqrt{2}$(反相)→ LO 以 180° 相位差驅動兩顆 diode(反相切換)From Δ: port 2 receives $+j/\sqrt{2}$ and port 3 $-j/\sqrt{2}$ (in anti-phase) → the LO drives the two diodes 180° apart (switching in anti-phase)Δ から:port 2 は $+j/\sqrt{2}$、port 3 は $-j/\sqrt{2}$ を受け取ります(逆相)→ LO は 180° の位相差で 2 個のダイオードを駆動します(逆相スイッチング)
  • IF 相位 $= \phi_{RF} - \phi_{LO}$:diode 1 得 $(-90°)-(+90°) = -180°$、diode 2 得 $(-90°)-(-90°) = 0°$ — 兩路 IF 反相,把其中一顆 diode 反接(或用 IF transformer 取差)後即同相相加The IF phase $= \phi_{RF} - \phi_{LO}$: diode 1 gets $(-90°)-(+90°) = -180°$ and diode 2 gets $(-90°)-(-90°) = 0°$ — the two IF paths are in anti-phase, so reversing one diode (or taking the difference with an IF transformer) makes them add in phaseIF 位相 $= \phi_{RF} - \phi_{LO}$:diode 1 は $(-90°)-(+90°) = -180°$、diode 2 は $(-90°)-(-90°) = 0°$ — 2 系統の IF は逆相なので、一方のダイオードを逆接続する(または IF transformer で差をとる)と同相で加算されます

反向指派(LO→Σ、RF→Δ)電路上也可運作,但 LO AM noise 與 even-harmonic 抑制特性較差,慣例上 LO 走 Δ。The reverse assignment (LO→Σ, RF→Δ) would also work electrically, but the LO AM noise and even-harmonic suppression are worse, so by convention the LO goes to Δ.逆の指定(LO→Σ、RF→Δ)でも電気的には動作しますが、LO の AM ノイズと偶数次高調波の抑圧特性が劣るため、慣例として LO は Δ に入れます。

(d) 與 90° 版相比,180° 拓樸的關鍵優勢是 Compared with the 90° version, the key advantage of the 180° topology is 90° 版と比べると、180° トポロジーの重要な利点は even LO harmonic rejection:LO 反相切換使 $2f_{LO} = 3.6\,\text{GHz}$、$4f_{LO}$ 等偶次諧波產物在 IF 端呈 common-mode 而相消 — 在擁擠的 cellular 頻譜中,3.6 GHz 恰好落在其他通訊頻段附近,這個抑制直接減少互擾。(90° 版的強項則是 RF/LO port 的寬頻 match 與 AM noise rejection。): switching the LO in anti-phase makes the even-harmonic products at $2f_{LO} = 3.6\,\text{GHz}$, $4f_{LO}$ and so on common-mode at the IF, where they cancel — and in the crowded cellular spectrum 3.6 GHz falls close to other communication bands, so this suppression directly reduces mutual interference. (The strengths of the 90° version, by contrast, are the broadband match at the RF/LO ports and its AM noise rejection.):LO を逆相で切り替えることにより、$2f_{LO} = 3.6\,\text{GHz}$、$4f_{LO}$ などの偶数次高調波の産物が IF 端で common-mode となって相殺されます — 混雑した cellular スペクトルでは 3.6 GHz がちょうど他の通信帯域の近くに落ちるため、この抑圧は相互干渉を直接減らします。(一方、90° 版の強みは RF/LO port の広帯域 match と AM ノイズ除去です。)

物理詮釋:Physical interpretation:物理的解釈:1.9 GHz 的 rat-race 周長已達 137 mm — 比 5 GHz 設計大得多,PCB 上常改用 lumped-element 等效(LC π-network 模擬 λ/4 線)或多層折疊來縮小。拓樸選擇從來不只看電氣規格,尺寸/成本也是 spec 的一部分。 A 1.9 GHz rat-race already has a circumference of 137 mm — far larger than a 5 GHz design, so PCBs often switch to a lumped-element equivalent (an LC π-network emulating the λ/4 line) or fold it across several layers to shrink it. Choosing a topology is never purely about electrical specifications; size and cost are part of the spec too.1.9 GHz の rat-race は周長が既に 137 mm に達し — 5 GHz 設計よりはるかに大きいため、PCB では lumped-element 等価(LC π-network で λ/4 線を模擬)や多層折り返しによる小型化がよく用いられます。トポロジーの選択は電気的仕様だけで決まるものではなく、寸法やコストも spec の一部です。

Topic 5Double-Balanced Ring Mixer

4-diode ring, 180° hybrid + balun, harmonic suppression, waveform analysis.

5.1 Easy
Diode switching and IF polarity

In a double-balanced ring mixer, the 4 diodes (D1–D4) are arranged in a ring. With LO positive half-cycle: D1 and D2 conduct; with LO negative half-cycle: D3 and D4 conduct.

(a) For LO positive ($+V_{LO}$): write the expression for $i_{IF}$ in terms of $v_{RF}$ and a sign $(\pm 1)$.

(b) For LO negative ($-V_{LO}$): write $i_{IF}$.

(c) Express $i_{IF}(t)$ using the signum function $\text{sgn}(v_{LO})$ and show it equals a multiplication by a square wave with fundamental $f_{LO}$.

Answer

已知/目標:Given / required:既知/目標:4 顆 diode 的 ring mixer:LO 正半週 D1、D2 導通,負半週 D3、D4 導通。求 $i_{IF}$ 在兩個半週的表達式,並用 $\text{sgn}(v_{LO})$ 證明它等效於「乘上一個 $f_{LO}$ 的方波」。 A 4-diode ring mixer: D1 and D2 conduct on the positive LO half-cycle and D3 and D4 on the negative. Find the expression for $i_{IF}$ in each half-cycle, and use $\text{sgn}(v_{LO})$ to prove it is equivalent to “multiplying by a square wave at $f_{LO}$”.4 個のダイオードの ring mixer:LO の正の半周期では D1、D2 が導通し、負の半周期では D3、D4 が導通します。両方の半周期における $i_{IF}$ の表式を求め、$\text{sgn}(v_{LO})$ を用いて「$f_{LO}$ の方形波を掛けること」と等価であることを証明します。

公式與來源:Formulas and sources:公式と出典:大訊號 LO 把 diode 當開關用(switching mixer model):哪一對 diode 導通,決定 RF 訊號以「正接」還是「反接」的極性接到 IF 端。 A large-signal LO uses the diodes as switches (the switching-mixer model): which pair conducts determines whether the RF signal is connected to the IF end with “normal” or “reversed” polarity.大信号の LO はダイオードをスイッチとして使います(switching mixer model):どちらの対のダイオードが導通するかによって、RF 信号が「正接続」と「逆接続」のどちらの極性で IF 端に接続されるかが決まります。

逐步代入:Substituting step by step:順に代入:

(a) LO+(D1、D2 on):RF 經 D1/D2 以正極性接到 IF → $i_{IF} = +K\,v_{RF}(t)$($K$ 為導通路徑的轉移常數)。(a) LO+ (D1, D2 on): the RF reaches the IF through D1/D2 with positive polarity → $i_{IF} = +K\,v_{RF}(t)$ ($K$ being the transfer constant of the conducting path).(a) LO+(D1、D2 on):RF は D1/D2 を経て正極性で IF に接続されます → $i_{IF} = +K\,v_{RF}(t)$($K$ は導通経路の伝達定数)。

(b) LO−(D3、D4 on):導通對換到 ring 的另一側,等效把 RF 的接線「翻面」→ $i_{IF} = -K\,v_{RF}(t)$。(b) LO− (D3, D4 on): conduction moves to the other side of the ring, which is equivalent to “flipping over” the RF connection → $i_{IF} = -K\,v_{RF}(t)$.(b) LO−(D3、D4 on):導通が ring の反対側に移り、RF の配線を「裏返す」のと等価です → $i_{IF} = -K\,v_{RF}(t)$。

(c) 兩個半週合併,極性恰好跟著 LO 的符號走:(c) Combining the two half-cycles, the polarity follows the sign of the LO exactly:(c) 二つの半周期を合わせると、極性はちょうど LO の符号に従います:

$$i_{IF}(t) = K\,v_{RF}(t)\cdot \text{sgn}(v_{LO}(t))$$

$\text{sgn}(v_{LO})$ 是 ±1 方波,其 Fourier series 只含$\text{sgn}(v_{LO})$ is a ±1 square wave, and its Fourier series contains only $\text{sgn}(v_{LO})$ は ±1 の方形波であり、その Fourier series に含まれるのは奇次odd奇数次諧波: harmonics:高調波:

$$\text{sgn}(v_{LO}) = \sum_{n\,\text{odd}} \frac{4}{n\pi}\sin(n\omega_{LO}t) = \frac{4}{\pi}\sin\omega_{LO}t + \frac{4}{3\pi}\sin 3\omega_{LO}t + \frac{4}{5\pi}\sin 5\omega_{LO}t + \cdots$$

乘上 $v_{RF} = V_{RF}\cos\omega_{RF}t$ 後,輸出只在 $|nf_{LO} \pm f_{RF}|$($n$ 奇數)有譜線:$n=1$ 給 desired IF,$n=3$ 給 $\tfrac{1}{3}$ 振幅(−9.5 dB)的高階產物——而 $2f_{LO}$、$4f_{LO}$ 相關的產物完全不存在。Multiplying by $v_{RF} = V_{RF}\cos\omega_{RF}t$, the output has spectral lines only at $|nf_{LO} \pm f_{RF}|$ ($n$ odd): $n=1$ gives the desired IF and $n=3$ a higher-order product at $\tfrac{1}{3}$ the amplitude (−9.5 dB) — while products related to $2f_{LO}$ and $4f_{LO}$ do not exist at all.$v_{RF} = V_{RF}\cos\omega_{RF}t$ を掛けると、出力は $|nf_{LO} \pm f_{RF}|$($n$ は奇数)にのみスペクトル線を持ちます:$n=1$ が所望の IF を与え、$n=3$ は $\tfrac{1}{3}$ の振幅(−9.5 dB)の高次成分を与えます——一方 $2f_{LO}$、$4f_{LO}$ に関連する成分は全く存在しません。

Even LO harmonics (2f_LO, 4f_LO, …) are suppressed by the balanced ring topology.

物理詮釋:Physical interpretation:物理的解釈:偶次諧波消失不是濾波濾掉的,而是 The even harmonics do not vanish because a filter removed them but as a consequence of 偶数次高調波が消えるのはフィルタで除去されたからではなく、對稱性symmetry対称性的結果:±1 方波是奇函數(半波對稱 $s(t+T/2) = -s(t)$),其偶次 Fourier 係數恆為零。實務上抑制深度由 4 顆 diode 與 balun 的匹配度決定(典型 30–50 dB)——一旦 diode 不匹配,「被對稱性藏起來」的偶次產物就會漏出來。: a ±1 square wave is an odd function (half-wave symmetric, $s(t+T/2) = -s(t)$), and its even Fourier coefficients are identically zero. In practice the depth of suppression is set by how well the 4 diodes and the balun are matched (typically 30–50 dB) — as soon as the diodes are mismatched, the even-order products that symmetry was hiding leak out.の結果です:±1 の方形波は奇関数(半波対称 $s(t+T/2) = -s(t)$)であり、その偶数次フーリエ係数は恒等的にゼロです。実務では抑圧の深さは 4 個のダイオードとバランの整合度で決まります(典型的に 30–50 dB)——ダイオードが不整合になった途端、対称性が隠していた偶数次成分が漏れ出します。

5.2 Medium
Derive the IF Fourier component at f_RF − f_LO

Given:

  • $v_{RF}(t) = V_{RF}\cos(\omega_{RF}t)$
  • $\text{sgn}(v_{LO}) = \dfrac{4}{\pi}\sum_{k=0}^{\infty}\dfrac{(-1)^k}{2k+1}\cos((2k+1)\omega_{LO}t)$

(a) Write the full product $v_{RF}\cdot\text{sgn}(v_{LO})$ and identify the $k=0$ term.

(b) Extract only the IF component at $f_{IF} = f_{RF} - f_{LO}$ and find its amplitude relative to $V_{RF}$.

(c) What is the single-sideband conversion loss (in dB) due solely to the Fourier coefficient?

Answer

已知/目標:Given / required:既知/目標:$v_{RF} = V_{RF}\cos(\omega_{RF}t)$,switching function $\text{sgn}(v_{LO}) = \frac{4}{\pi}\sum_k \frac{(-1)^k}{2k+1}\cos((2k+1)\omega_{LO}t)$。求乘積中 $f_{IF} = f_{RF}-f_{LO}$ 分量的振幅,與純粹由 Fourier 係數造成的 SSB conversion loss。$v_{RF} = V_{RF}\cos(\omega_{RF}t)$ with a switching function $\text{sgn}(v_{LO}) = \frac{4}{\pi}\sum_k \frac{(-1)^k}{2k+1}\cos((2k+1)\omega_{LO}t)$. Find the amplitude of the $f_{IF} = f_{RF}-f_{LO}$ component of the product, and the SSB conversion loss arising purely from the Fourier coefficients.$v_{RF} = V_{RF}\cos(\omega_{RF}t)$、スイッチング関数は $\text{sgn}(v_{LO}) = \frac{4}{\pi}\sum_k \frac{(-1)^k}{2k+1}\cos((2k+1)\omega_{LO}t)$。積のうち $f_{IF} = f_{RF}-f_{LO}$ 成分の振幅と、純粋にフーリエ係数から生じる SSB 変換損失を求めます。

公式與來源:Formulas and sources:公式と出典:switching mixer 模型 $i_{IF} = v_{RF}\cdot\text{sgn}(v_{LO})$(5.1 題),配合積化和差 $\cos A\cos B = \tfrac{1}{2}[\cos(A-B)+\cos(A+B)]$;conversion loss 定義為輸入可用功率對單一 sideband 輸出功率之比。 The switching-mixer model $i_{IF} = v_{RF}\cdot\text{sgn}(v_{LO})$ (problem 5.1), together with the product-to-sum identity $\cos A\cos B = \tfrac{1}{2}[\cos(A-B)+\cos(A+B)]$; conversion loss is defined as the ratio of the available input power to the output power in a single sideband.スイッチングミキサモデル $i_{IF} = v_{RF}\cdot\text{sgn}(v_{LO})$(問題 5.1)と、積和公式 $\cos A\cos B = \tfrac{1}{2}[\cos(A-B)+\cos(A+B)]$ を用います;変換損失は入力有能電力と単一側波帯の出力電力の比として定義されます。

逐步代入:Substituting step by step:逐次代入:

(a) 完整乘積,取 $k=0$(基波)項:(a) The full product, taking the $k=0$ (fundamental) term:(a) 完全な積、$k=0$(基本波)の項を取ります:

$$i_{IF} = V_{RF}\cos(\omega_{RF}t)\cdot\frac{4}{\pi}\cos(\omega_{LO}t) + \underbrace{\text{3rd, 5th,}\ldots}_{\text{higher order}}$$

積化和差展開 $k=0$ 項:Expanding the $k=0$ term by product-to-sum:$k=0$ の項を積和公式で展開します:

$$\frac{4}{\pi}V_{RF}\cos(\omega_{RF}t)\cos(\omega_{LO}t) = \frac{2}{\pi}V_{RF}\bigl[\cos((\omega_{RF}-\omega_{LO})t) + \cos((\omega_{RF}+\omega_{LO})t)\bigr]$$

(b) 取出 $\omega_{RF}-\omega_{LO}$(difference)分量:振幅 $= \dfrac{2}{\pi}V_{RF} \approx 0.637\,V_{RF}$ — 注意 $\frac{4}{\pi}$ 的能量被「平分」到 sum 與 difference 兩個 sideband,各拿 $\frac{2}{\pi}$。(b) Extracting the $\omega_{RF}-\omega_{LO}$ (difference) component: the amplitude $= \dfrac{2}{\pi}V_{RF} \approx 0.637\,V_{RF}$ — note that the energy of the $\frac{4}{\pi}$ is “split evenly” between the sum and difference sidebands, each taking $\frac{2}{\pi}$.(b) $\omega_{RF}-\omega_{LO}$(差)成分を取り出します:振幅は $= \dfrac{2}{\pi}V_{RF} \approx 0.637\,V_{RF}$ — $\frac{4}{\pi}$ のエネルギーが和と差の二つの側波帯に「均等分配」され、それぞれ $\frac{2}{\pi}$ を取ることに注意してください。

(c) 電壓比的平方 = 功率比:(c) The square of the voltage ratio = the power ratio:(c) 電圧比の二乗 = 電力比:

$$\frac{P_{IF}}{P_{RF}} = \left(\frac{2}{\pi}\right)^2 = \frac{4}{\pi^2} \approx 0.405 \;\Rightarrow\; L_C = -10\log_{10}(0.405) = \mathbf{3.92\,dB}$$

L_C(theoretical) = −10·log(4/π²) = 10·log(π²/4) ≈ 3.92 dB

物理詮釋:Physical interpretation:物理的解釈:3.9 dB 是理想 switching mixer 的物理下限,由兩個因素組成:方波基波只佔 $4/\pi$(≈ 0.9 dB 的「切換損」)、再加上能量平分給兩個 sideband 只取其一(3 dB 的 SSB 損)。實際 diode ring 還要加 junction resistance $r_j$、balun 損與 mismatch(見 5.3 題),所以商品 mixer 的 $L_C$ 典型落在 6–8 dB — 量測值若遠低於 3.9 dB 必有量測錯誤。 3.9 dB is the physical lower bound for an ideal switching mixer, made up of two parts: the square wave's fundamental accounts for only $4/\pi$ (a ≈ 0.9 dB “switching loss”), plus the energy being split evenly between two sidebands of which only one is taken (the 3 dB SSB loss). A real diode ring adds junction resistance $r_j$, balun loss and mismatch (see problem 5.3), so a commercial mixer's $L_C$ typically lands at 6–8 dB — a measured value far below 3.9 dB must be a measurement error.3.9 dB は理想的なスイッチングミキサの物理的な下限であり、二つの要因から成ります:方形波の基本波が $4/\pi$ しか占めないこと(≈ 0.9 dB の「スイッチング損失」)、さらにエネルギーが二つの側波帯に均等分配されそのうち一方だけを取ること(3 dB の SSB 損失)。実際のダイオードリングではさらに接合抵抗 $r_j$、バラン損失と不整合が加わるため(問題 5.3 参照)、市販ミキサの $L_C$ は典型的に 6–8 dB になります — 測定値が 3.9 dB を大きく下回れば必ず測定の誤りです。

5.3 Design
Specify a double-balanced mixer for a 2 GHz WCDMA receiver
System Requirements RF: 1920–1980 MHz (WCDMA Band 1 UL)
LO: 1870–1930 MHz (low-side injection)
IF: 50 MHz ±10 MHz
Required: conversion loss ≤ 8 dB, IIP3 ≥ +10 dBm, LO-RF isolation ≥ 40 dB
LO drive level: +7 dBm available

(a) Why is a double-balanced ring mixer preferred over a single-balanced mixer for this application (LO leakage and harmonic suppression)?

(b) The ring uses 4 matched Schottky diodes with junction resistance $r_j = 4\,\Omega$ each. Estimate the contribution of $r_j$ to conversion loss using $L_{r_j} \approx 2r_j/Z_0$ (linear, add to 3.92 dB theoretical).

(c) Specify the rat-race hybrid center frequency and bandwidth. The balun must handle IF down to 40 MHz — is a standard center-tap transformer feasible? What is the low-frequency cutoff constraint?

Answer

已知/目標:Given / required:既知/目標:RF 1920–1980 MHz、LO 1870–1930 MHz(low-side)、IF 50 MHz ± 10 MHz;spec:$L_C \leq 8\,\text{dB}$、IIP3 ≥ +10 dBm、LO-RF isolation ≥ 40 dB、LO drive +7 dBm;4 顆 matched Schottky diode,$r_j = 4\,\Omega$。求拓樸理由、$r_j$ 對 $L_C$ 的貢獻、hybrid/balun 規格。 RF 1920–1980 MHz, LO 1870–1930 MHz (low-side), IF 50 MHz ± 10 MHz; the specs: $L_C \leq 8\,\text{dB}$, IIP3 ≥ +10 dBm, LO-RF isolation ≥ 40 dB, LO drive +7 dBm; 4 matched Schottky diodes with $r_j = 4\,\Omega$. Give the reasoning for the topology, $r_j$'s contribution to $L_C$, and the hybrid/balun specifications.RF 1920–1980 MHz、LO 1870–1930 MHz(ロー側)、IF 50 MHz ± 10 MHz;仕様:$L_C \leq 8\,\text{dB}$、IIP3 ≥ +10 dBm、LO-RF アイソレーション ≥ 40 dB、LO ドライブ +7 dBm;整合した 4 個のショットキーダイオード、$r_j = 4\,\Omega$。トポロジーの理由、$r_j$ の $L_C$ への寄与、ハイブリッド/バランの仕様を求めます。

(a) 拓樸理由:(a) The reasoning for the topology:(a) トポロジーの理由:double-balanced ring 勝過 single-balanced 之處:(1) where a double-balanced ring beats single-balanced: (1) 二重平衡リングが単平衡に勝る点:(1) LO-to-RF isolation 多 ~20 dB~20 dB more LO-to-RF isolationLO-RF アイソレーションが ~20 dB 大きい — LO 在 RF balun 的兩臂等幅反相、在 RF 埠相消(single-balanced 只靠 hybrid 一層隔離 ~25 dB,ring 可達 40–50 dB,正好滿足 ≥ 40 dB spec);(2) — the LO is equal and opposite in the two arms of the RF balun and cancels at the RF port (single-balanced relies on the hybrid's single layer of isolation for ~25 dB, whereas a ring reaches 40–50 dB, exactly meeting the ≥ 40 dB spec); (2) — LO は RF バランの両アームで等振幅逆相となり、RF ポートで打ち消されます(単平衡はハイブリッド一層のアイソレーション ~25 dB に頼るのみですが、リングは 40–50 dB に達し、≥ 40 dB の仕様をちょうど満たします);(2) even harmonics $2f_{LO}, 4f_{LO}$ 被抑制the even harmonics $2f_{LO}$ and $4f_{LO}$ are suppressed偶数次高調波 $2f_{LO}$ と $4f_{LO}$ が抑圧される — WCDMA 的 $2f_{LO} \approx 3.8\,\text{GHz}$ 鄰近其他通訊頻段,不抑制會經天線輻射造成干擾。 — WCDMA's $2f_{LO} \approx 3.8\,\text{GHz}$ is adjacent to other communication bands, and without suppression it would radiate from the antenna and cause interference. — WCDMA の $2f_{LO} \approx 3.8\,\text{GHz}$ は他の通信帯域に隣接しており、抑圧しなければアンテナから放射されて干渉を引き起こします。

(b) 公式與來源:(b) Formula and source:(b) 公式と出典:導通時兩顆 diode 串在訊號路徑上,$2r_j$ 與 $Z_0$ 形成分壓;以電壓比估損耗(rule of thumb): when conducting, two diodes sit in series in the signal path and $2r_j$ forms a divider with $Z_0$; estimating the loss from the voltage ratio (a rule of thumb):導通時には 2 個のダイオードが信号経路に直列に入り、$2r_j$ と $Z_0$ が分圧を形成します;電圧比から損失を見積もります(経験則):

$$L_{r_j} \approx 20\log_{10}\!\left(1+\frac{2r_j}{Z_0}\right)$$

逐步代入:Substituting step by step:逐次代入:$L_{r_j} = 20\log_{10}\!\left(1+\dfrac{2\times 4}{50}\right) = 20\log_{10}(1.16) = 20\times 0.0645 = \mathbf{1.3\,dB}$

加上理想切換損(5.2 題):$L_C \approx 3.92 + 1.3 = \mathbf{5.2\,dB}$。再加 balun 損與 mismatch 的 ~2–3 dB → 總計 ~7–8 dB,Adding the ideal switching loss (problem 5.2): $L_C \approx 3.92 + 1.3 = \mathbf{5.2\,dB}$. Adding a further ~2–3 dB of balun loss and mismatch → about 7–8 dB in total, which 理想的なスイッチング損失(問題 5.2)を加えると:$L_C \approx 3.92 + 1.3 = \mathbf{5.2\,dB}$。さらにバラン損失と不整合の ~2–3 dB を加えると → 合計で約 7–8 dB となり、滿足 ≤ 8 dB specmeets the ≤ 8 dB spec≤ 8 dB の仕様を満たします(無多少餘裕,balun 品質是關鍵)。(with little margin, so the quality of the balun is critical).(余裕はほとんどないため、バランの品質が鍵となります)。

(c) Hybrid 與 balun 規格:(c) Hybrid and balun specifications:(c) ハイブリッドとバランの仕様:

RF 側 rat-race:中心 $f_0 = \dfrac{1920+1980}{2} = 1950\,\text{MHz}$;需求頻寬 60 MHz → fractional BW $= 60/1950 = 3.1\%$,遠小於 rat-race 的 ±10% 可用範圍 ✓The RF-side rat-race: centre $f_0 = \dfrac{1920+1980}{2} = 1950\,\text{MHz}$; the bandwidth required is 60 MHz → a fractional BW $= 60/1950 = 3.1\%$, far inside the rat-race's usable ±10% ✓RF 側のラットレース:中心 $f_0 = \dfrac{1920+1980}{2} = 1950\,\text{MHz}$;必要な帯域幅は 60 MHz → 比帯域 $= 60/1950 = 3.1\%$ で、ラットレースの使用可能な ±10% の範囲に十分収まります ✓

IF balun 低頻截止:center-tap transformer 的 magnetizing inductance $L_m$ 與負載構成 high-pass,$f_{low} = \dfrac{R_{load}}{2\pi L_m}$。要 $f_{low} \leq 40\,\text{MHz}$、$R = 50\,\Omega$:The IF balun's low-frequency cutoff: the centre-tapped transformer's magnetizing inductance $L_m$ forms a high-pass with the load, $f_{low} = \dfrac{R_{load}}{2\pi L_m}$. Requiring $f_{low} \leq 40\,\text{MHz}$ with $R = 50\,\Omega$:IF バランの低域遮断:センタータップトランスの励磁インダクタンス $L_m$ が負荷とともに高域通過を形成し、$f_{low} = \dfrac{R_{load}}{2\pi L_m}$。$f_{low} \leq 40\,\text{MHz}$、$R = 50\,\Omega$ を要求すると:

$$L_m \geq \frac{50}{2\pi\times 40\times 10^6} = \frac{50}{2.51\times 10^8} = 199\times 10^{-9}\,\text{H} \approx 200\,\text{nH}$$

Use a ferrite-core center-tap balun with L_m ≥ 200 nH to cover IF down to 40 MHz.

物理詮釋:Physical interpretation:物理的解釈:wound ferrite transformer 容易做到數百 nH~數 µH,可行;air-core 或 microstrip balun 在 40 MHz 的感抗不足會把低端 IF 衰減掉。另注意 LO drive 只有 +7 dBm(level-7):由 $IIP3 \approx P_{LO} + 4\sim 8\,\text{dB}$ 的經驗律,+10 dBm 的 IIP3 spec 已在邊緣 — 若量產要留 margin,應考慮 level-10/13 drive(對照 9.4 題的取捨)。 A wound ferrite transformer easily reaches hundreds of nH to a few µH, so this is feasible; an air-core or microstrip balun would not have enough reactance at 40 MHz and would attenuate the low end of the IF. Note also that the LO drive is only +7 dBm (level-7): by the rule of thumb $IIP3 \approx P_{LO} + 4\sim 8\,\text{dB}$, an IIP3 spec of +10 dBm is already marginal — if production margin is wanted, consider a level-10/13 drive (compare the trade-offs of problem 9.4).巻線フェライトトランスは数百 nH~数 µH を容易に実現できるので実現可能です;空芯やマイクロストリップのバランでは 40 MHz でのリアクタンスが不足し、IF の低域を減衰させてしまいます。また LO ドライブが +7 dBm(level-7)しかないことにも注意してください:$IIP3 \approx P_{LO} + 4\sim 8\,\text{dB}$ の経験則によれば、+10 dBm の IIP3 仕様はすでに限界的です — 量産で余裕を残したい場合は level-10/13 のドライブを検討すべきです(問題 9.4 のトレードオフと対照)。

Topic 6Image-Reject Mixer

Two 90° hybrids, USB/LSB separation, image rejection ratio.

6.1 Easy
Identify USB and LSB for given RF/LO

An image-reject mixer receives: $f_{LO} = 5.8\,\text{GHz}$, two RF input signals at $f_{RF1} = 5.9\,\text{GHz}$ and $f_{RF2} = 5.7\,\text{GHz}$.

(a) Find the IF frequency for each RF input.

(b) Which is the USB (upper sideband) and which is the LSB (lower sideband)?

(c) The image-reject mixer routes USB to output port A and LSB to output port B. At which port does each signal appear?

Answer

已知/目標:Given / required:既知/目標:$f_{LO} = 5.8\,\text{GHz}$,兩個 RF 輸入 $f_{RF1} = 5.9\,\text{GHz}$、$f_{RF2} = 5.7\,\text{GHz}$。求各自的 IF、判別 USB/LSB、以及在 image-reject mixer 的哪個輸出埠出現。$f_{LO} = 5.8\,\text{GHz}$ with two RF inputs, $f_{RF1} = 5.9\,\text{GHz}$ and $f_{RF2} = 5.7\,\text{GHz}$. Find their IFs, identify the USB/LSB, and say at which output port of an image-reject mixer each appears.$f_{LO} = 5.8\,\text{GHz}$、二つの RF 入力 $f_{RF1} = 5.9\,\text{GHz}$、$f_{RF2} = 5.7\,\text{GHz}$。それぞれの IF を求め、USB/LSB を判別し、イメージ除去ミキサのどの出力ポートに現れるかを述べます。

公式與來源:Formulas and sources:公式と出典:down-conversion 後 $f_{IF} = |f_{RF} - f_{LO}|$;sideband 以「RF 相對 LO 的位置」定義 — 在 LO 上方為 USB、下方為 LSB(兩者互為對方的 image)。 After downconversion $f_{IF} = |f_{RF} - f_{LO}|$; the sideband is defined by “where the RF sits relative to the LO” — above the LO is the USB and below it the LSB (each being the other's image).ダウンコンバージョン後は $f_{IF} = |f_{RF} - f_{LO}|$;側波帯は「RF が LO に対してどこにあるか」で定義され — LO の上側が USB、下側が LSB です(両者は互いのイメージです)。

逐步代入:Substituting step by step:逐次代入:

(a) $f_{IF1} = |5.9 - 5.8| = 0.1\,\text{GHz} = 100\,\text{MHz}$;$f_{IF2} = |5.7 - 5.8| = 0.1\,\text{GHz} = 100\,\text{MHz}$。(a) $f_{IF1} = |5.9 - 5.8| = 0.1\,\text{GHz} = 100\,\text{MHz}$; $f_{IF2} = |5.7 - 5.8| = 0.1\,\text{GHz} = 100\,\text{MHz}$.(a) $f_{IF1} = |5.9 - 5.8| = 0.1\,\text{GHz} = 100\,\text{MHz}$;$f_{IF2} = |5.7 - 5.8| = 0.1\,\text{GHz} = 100\,\text{MHz}$。

兩個輸入落到Both inputs land on 二つの入力が落ちるのは完全相同的 IFexactly the same IF全く同じ IF — 一般 mixer + IF filter 對此無能為力,這正是 image problem 本身,也是 image-reject mixer 要解的問題。 — an ordinary mixer plus IF filter can do nothing about it, which is the image problem itself and precisely what an image-reject mixer sets out to solve. — 通常のミキサと IF フィルタではどうにもできず、これがイメージ問題そのものであり、まさにイメージ除去ミキサが解こうとする問題です。

(b) USB:$f_{RF1} = 5.9\,\text{GHz}$($= f_{LO} + f_{IF}$,在 LO 上方);LSB:$f_{RF2} = 5.7\,\text{GHz}$($= f_{LO} - f_{IF}$,在 LO 下方,即 image)。(b) USB: $f_{RF1} = 5.9\,\text{GHz}$ ($= f_{LO} + f_{IF}$, above the LO); LSB: $f_{RF2} = 5.7\,\text{GHz}$ ($= f_{LO} - f_{IF}$, below the LO — the image).(b) USB:$f_{RF1} = 5.9\,\text{GHz}$($= f_{LO} + f_{IF}$、LO の上側);LSB:$f_{RF2} = 5.7\,\text{GHz}$($= f_{LO} - f_{IF}$、LO の下側、すなわちイメージ)。

(c) 依題目的埠定義:USB → (c) With the port definitions of the problem: USB → (c) 問題のポート定義に従うと:USB → Port A($f_{RF1}$);LSB/image → ($f_{RF1}$); LSB/image → ($f_{RF1}$);LSB/イメージ → Port B($f_{RF2}$)。分離的原理:USB 與 LSB 經 I/Q 兩路混頻後的 IF 相位差符號相反($\mp 90°$),輸出 90° hybrid 對一個 sideband 建設性疊加、對另一個破壞性相消(詳見 6.2 題推導)。($f_{RF2}$). The separation principle: after mixing through the I/Q paths, the USB and LSB have IF phase differences of opposite sign ($\mp 90°$), and the output 90° hybrid adds one sideband constructively while cancelling the other destructively (see the derivation of problem 6.2).($f_{RF2}$)。分離の原理:USB と LSB は I/Q 二経路で混合された後の IF 位相差の符号が逆であり($\mp 90°$)、出力の 90° ハイブリッドが一方の側波帯を建設的に加算し、他方を破壊的に打ち消します(問題 6.2 の導出を参照)。

Image-reject mixer separates same-IF-frequency signals by sideband — no additional RF preselector needed.

物理詮釋:Physical interpretation:物理的解釈:「同一個 IF 頻率」在頻域上已無法區分,image-reject mixer 利用的是頻域看不到的“The same IF frequency” cannot be distinguished in the frequency domain at all; what an image-reject mixer exploits is the 「同じ IF 周波数」は周波数領域ではもはや区別できず、イメージ除去ミキサが利用するのは周波数領域では見えない相位資訊phase information位相情報:上下 sideband 混頻時相位旋轉方向相反。實際可分離的深度(IRR)受 I/Q 振幅/相位平衡限制(6.3 題),典型 25–40 dB,所以它是「補濾波器之不足」,極強的 image 仍需 preselector 分擔。 invisible there: the upper and lower sidebands rotate in phase in opposite directions when mixed. The achievable separation (the IRR) is limited by the I/Q amplitude/phase balance (problem 6.3), typically 25–40 dB, so it “makes up for what the filter cannot do”, and a very strong image still needs a preselector to share the load.です:上側と下側の側波帯は混合時に位相の回転方向が逆になります。実際に分離できる深さ(IRR)は I/Q の振幅/位相平衡に制限され(問題 6.3)、典型的に 25–40 dB なので、これは「フィルタの不足を補う」ものであり、非常に強いイメージにはやはりプリセレクタの分担が必要です。

6.2 Medium
Phase analysis — how the 90° IF hybrid separates sidebands

In an image-reject mixer using two 90° hybrids:

  • Input RF hybrid splits RF into arms I and Q with 0°/−90°.
  • LO is split into arms I and Q with 0°/−90°.
  • Mixer I: multiplies RF_I × LO_I; Mixer Q: multiplies RF_Q × LO_Q.
  • IF hybrid recombines with 0°/−90° at output.

(a) For USB input ($\omega_{RF} = \omega_{LO} + \omega_{IF}$): trace the phase of the IF signal through each arm and show it sums at output port A.

(b) For LSB input ($\omega_{RF} = \omega_{LO} - \omega_{IF}$): show the signals cancel at port A and add at port B.

Answer

已知/目標:Given / required:既知/目標:image-reject (Hartley) mixer:RF 經 90° hybrid 分成 I(0°)/Q(−90°)兩路,各自與 LO 混頻,IF 再經 90° hybrid 合成。證明 USB 在 port A 相加、LSB 在 port A 相消(轉而出現在 port B)。 An image-reject (Hartley) mixer: the RF is split by a 90° hybrid into I (0°) and Q (−90°) paths, each mixed with the LO, and the IFs recombined through a 90° hybrid. Prove that the USB adds at port A while the LSB cancels there (appearing at port B instead).イメージ除去(Hartley)ミキサ:RF は 90° ハイブリッドで I(0°)/Q(−90°)の二経路に分けられ、それぞれ LO と混合され、IF は再び 90° ハイブリッドで合成されます。USB が port A で加算され、LSB が port A で打ち消される(代わりに port B に現れる)ことを証明します。

架構釐清:A clarification of the architecture:アーキテクチャの整理:若 RF 與 LO if RF と LO の兩者都both両方とも做 0°/−90° quadrature split,相位差會在混頻時互相抵消($\phi_{RF}-\phi_{LO}$ 兩路相同)而無法分離 sideband。標準 Hartley 架構是 the RF and the LO were split into 0°/−90° quadrature, the phase differences would cancel during mixing ($\phi_{RF}-\phi_{LO}$ being the same on both paths) and no sideband separation would occur. The standard Hartley architecture has 0°/−90° の直交分配を行うと、位相差は混合時に互いに打ち消し合い($\phi_{RF}-\phi_{LO}$ が両経路で同じ)、側波帯を分離できません。標準的な Hartley アーキテクチャは只有一路 quadraturequadrature on one side only片側のみ直交(RF 或 LO 擇一),另一路同相分配 — 以下取 RF quadrature、LO 同相($LO_I = LO_Q = \cos\omega_{LO}t$)推導,結論與原答案相同。(either the RF or the LO), the other being split in phase — the derivation below takes RF quadrature with an in-phase LO ($LO_I = LO_Q = \cos\omega_{LO}t$), reaching the same conclusion as the original answer.(RF か LO のいずれか)で、もう一方は同相で分配されます — 以下では RF を直交、LO を同相($LO_I = LO_Q = \cos\omega_{LO}t$)として導出し、結論は元の解答と同じになります。

公式與來源:Formulas and sources:公式と出典:difference product 的 IF 相位(積化和差直接讀出):down-convert 時若 $\omega_{RF} > \omega_{LO}$(USB),$\phi_{IF} = \phi_{RF} - \phi_{LO}$;若 $\omega_{RF} < \omega_{LO}$(LSB),差頻項是 $\cos[(\omega_{LO}-\omega_{RF})t + \phi_{LO} - \phi_{RF}]$,相位 The IF phase of the difference product (read straight off the product-to-sum expansion): on downconversion, if $\omega_{RF} > \omega_{LO}$ (USB) then $\phi_{IF} = \phi_{RF} - \phi_{LO}$; if $\omega_{RF} < \omega_{LO}$ (LSB) the difference term is $\cos[(\omega_{LO}-\omega_{RF})t + \phi_{LO} - \phi_{RF}]$ and the phase 差成分の IF 位相(積和公式から直接読み取れます):ダウンコンバージョン時に $\omega_{RF} > \omega_{LO}$(USB)なら $\phi_{IF} = \phi_{RF} - \phi_{LO}$;$\omega_{RF} < \omega_{LO}$(LSB)なら差周波の項は $\cos[(\omega_{LO}-\omega_{RF})t + \phi_{LO} - \phi_{RF}]$ となり、位相の符號反轉changes sign符号が反転します:$\phi_{IF} = \phi_{LO} - \phi_{RF}$。這個符號差就是整個架構的槓桿點。: $\phi_{IF} = \phi_{LO} - \phi_{RF}$. That sign difference is the lever the whole architecture works on.:$\phi_{IF} = \phi_{LO} - \phi_{RF}$。この符号の差こそが、アーキテクチャ全体を動かす梃子となります。

(a) USB($\omega_{RF} = \omega_{LO}+\omega_{IF}$)逐步代入:(a) USB ($\omega_{RF} = \omega_{LO}+\omega_{IF}$), substituting step by step:(a) USB($\omega_{RF} = \omega_{LO}+\omega_{IF}$)を逐次代入:

Arm I:$RF_I = \cos(\omega_{RF}t)$(0°)× $LO_I$ → $IF_I = \tfrac{1}{2}\cos(\omega_{IF}t)$,相位 Arm I: $RF_I = \cos(\omega_{RF}t)$ (0°) × $LO_I$ → $IF_I = \tfrac{1}{2}\cos(\omega_{IF}t)$, phase アーム I:$RF_I = \cos(\omega_{RF}t)$(0°)× $LO_I$ → $IF_I = \tfrac{1}{2}\cos(\omega_{IF}t)$、位相

Arm Q:$RF_Q = \cos(\omega_{RF}t - 90°)$ × $LO_Q$ → $IF_Q = \tfrac{1}{2}\cos(\omega_{IF}t - 90°)$,相位 Arm Q: $RF_Q = \cos(\omega_{RF}t - 90°)$ × $LO_Q$ → $IF_Q = \tfrac{1}{2}\cos(\omega_{IF}t - 90°)$, phase アーム Q:$RF_Q = \cos(\omega_{RF}t - 90°)$ × $LO_Q$ → $IF_Q = \tfrac{1}{2}\cos(\omega_{IF}t - 90°)$、位相 −90°

IF hybrid 在 port A 對 I 路再加 −90°、Q 路直通後相加:At port A the IF hybrid adds a further −90° to path I and passes Q straight through, then sums:IF ハイブリッドは port A で I 経路にさらに −90° を加え、Q 経路はそのまま通してから加算します:

$$\text{Port A} = \tfrac{1}{2}\cos(\omega_{IF}t - 90°) + \tfrac{1}{2}\cos(\omega_{IF}t - 90°) = \cos(\omega_{IF}t - 90°) \;\rightarrow\; \textbf{constructive:USB 出現在 A}$$$$\text{Port A} = \tfrac{1}{2}\cos(\omega_{IF}t - 90°) + \tfrac{1}{2}\cos(\omega_{IF}t - 90°) = \cos(\omega_{IF}t - 90°) \;\rightarrow\; \textbf{constructive: the USB appears at A}$$$$\text{Port A} = \tfrac{1}{2}\cos(\omega_{IF}t - 90°) + \tfrac{1}{2}\cos(\omega_{IF}t - 90°) = \cos(\omega_{IF}t - 90°) \;\rightarrow\; \textbf{constructive: the USB appears at A}$$

(port B 則是 Q 路加 −90°、I 路直通:$\tfrac{1}{2}\cos(\omega_{IF}t) + \tfrac{1}{2}\cos(\omega_{IF}t - 180°) = 0$ — USB 不漏到 B ✓)(At port B, path Q gets the −90° and path I passes straight through: $\tfrac{1}{2}\cos(\omega_{IF}t) + \tfrac{1}{2}\cos(\omega_{IF}t - 180°) = 0$ — no USB leaks to B ✓)(port B では Q 経路に −90° を加え、I 経路はそのまま通します:$\tfrac{1}{2}\cos(\omega_{IF}t) + \tfrac{1}{2}\cos(\omega_{IF}t - 180°) = 0$ — USB は B に漏れません ✓)

(b) LSB($\omega_{RF} = \omega_{LO}-\omega_{IF}$):(b) LSB ($\omega_{RF} = \omega_{LO}-\omega_{IF}$):(b) LSB($\omega_{RF} = \omega_{LO}-\omega_{IF}$):相位符號反轉 $\phi_{IF} = \phi_{LO} - \phi_{RF}$: With the phase sign reversed, $\phi_{IF} = \phi_{LO} - \phi_{RF}$:位相の符号が反転し、$\phi_{IF} = \phi_{LO} - \phi_{RF}$:

Arm I:$IF_I = \tfrac{1}{2}\cos(\omega_{IF}t)$,相位 0°(與 USB 相同)Arm I: $IF_I = \tfrac{1}{2}\cos(\omega_{IF}t)$, phase 0° (the same as the USB)アーム I:$IF_I = \tfrac{1}{2}\cos(\omega_{IF}t)$、位相 0°(USB と同じ)

Arm Q:$\phi_{IF} = 0 - (-90°) = \mathbf{+90°}$ → $IF_Q = \tfrac{1}{2}\cos(\omega_{IF}t + 90°)$(與 USB 的 −90° 反號!)Arm Q: $\phi_{IF} = 0 - (-90°) = \mathbf{+90°}$ → $IF_Q = \tfrac{1}{2}\cos(\omega_{IF}t + 90°)$ (the opposite sign to the USB's −90°!)アーム Q:$\phi_{IF} = 0 - (-90°) = \mathbf{+90°}$ → $IF_Q = \tfrac{1}{2}\cos(\omega_{IF}t + 90°)$(USB の −90° とは逆符号!)

$$\text{Port A} = \underbrace{\tfrac{1}{2}\cos(\omega_{IF}t - 90°)}_{I\,\text{路}+(-90°)} + \underbrace{\tfrac{1}{2}\cos(\omega_{IF}t + 90°)}_{Q\,\text{路直通}} = 0 \;\rightarrow\; \textbf{cancel}$$$$\text{Port A} = \underbrace{\tfrac{1}{2}\cos(\omega_{IF}t - 90°)}_{I\,\text{path}+(-90°)} + \underbrace{\tfrac{1}{2}\cos(\omega_{IF}t + 90°)}_{Q\,\text{path, straight through}} = 0 \;\rightarrow\; \textbf{cancel}$$$$\text{Port A} = \underbrace{\tfrac{1}{2}\cos(\omega_{IF}t - 90°)}_{I\,\text{path}+(-90°)} + \underbrace{\tfrac{1}{2}\cos(\omega_{IF}t + 90°)}_{Q\,\text{path, straight through}} = 0 \;\rightarrow\; \textbf{cancel}$$

$$\text{Port B} = \tfrac{1}{2}\cos(\omega_{IF}t) + \tfrac{1}{2}\cos(\omega_{IF}t + 90° - 90°) = \cos(\omega_{IF}t) \;\rightarrow\; \textbf{constructive:LSB 出現在 B}$$$$\text{Port B} = \tfrac{1}{2}\cos(\omega_{IF}t) + \tfrac{1}{2}\cos(\omega_{IF}t + 90° - 90°) = \cos(\omega_{IF}t) \;\rightarrow\; \textbf{constructive: the LSB appears at B}$$$$\text{Port B} = \tfrac{1}{2}\cos(\omega_{IF}t) + \tfrac{1}{2}\cos(\omega_{IF}t + 90° - 90°) = \cos(\omega_{IF}t) \;\rightarrow\; \textbf{constructive: the LSB appears at B}$$

The 90° phase relationship of USB vs LSB mixing products causes constructive/destructive interference at the IF output ports.

物理詮釋:Physical interpretation:物理的解釈:USB 與 LSB 在 Q 路拿到的 IF 相位是 $\mp 90°$(鏡像),輸出 hybrid 固定再補一個 −90°,恰好把其中一個補成同相(相加)、另一個補成反相(相消)。等效地說:I/Q 訊號分辨的是「正負頻率」(complex envelope 的旋轉方向),這是單一實數訊號做不到的 — 也因此任何 I/Q 不平衡都會讓 image 漏過來(6.3 題)。 The IF phases the USB and LSB acquire on the Q path are $\mp 90°$ (mirror images), and the output hybrid always adds one more −90°, which happens to bring one of them into phase (adding) and the other into anti-phase (cancelling). Equivalently: I/Q signals distinguish “positive from negative frequency” (the direction of rotation of the complex envelope), which a single real signal cannot do — and that is also why any I/Q imbalance lets the image leak through (problem 6.3).USB と LSB が Q 経路で得る IF 位相は $\mp 90°$(鏡像)であり、出力ハイブリッドは常にさらに −90° を加えるため、ちょうど一方を同相(加算)に、他方を逆相(相殺)にします。等価的に言えば:I/Q 信号が区別するのは「正負の周波数」(複素包絡線の回転方向)であり、これは単一の実信号にはできません — だからこそ I/Q の不平衡があるとイメージが漏れてくるのです(問題 6.3)。

6.3 Design
Image rejection ratio vs amplitude/phase imbalance

The image rejection ratio (IRR) for an image-reject mixer with amplitude imbalance $\varepsilon$ (fractional) and phase imbalance $\Delta\phi$ (radians) is:

$$IRR = \frac{1 + 2\varepsilon\cos\Delta\phi + \varepsilon^2}{1 - 2\varepsilon\cos\Delta\phi + \varepsilon^2}$$

(a) For perfect balance ($\varepsilon = 0$), what is IRR (in dB)?

(b) A spec requires IRR ≥ 40 dB. Using the small-error approximation $IRR \approx 1/((\varepsilon^2+\Delta\phi^2))$ (power), find the combined imbalance budget.

(c) If the two hybrid chips have amplitude matching guaranteed to ±0.5 dB, what is the maximum phase imbalance allowed to still meet 40 dB IRR?

(d) Is 40 dB IRR achievable with a single RF preselector filter in a typical cellular design? What does the image-reject mixer add?

Answer

已知/目標:Given / required:既知/目標:IRR 公式(題目給定,$\varepsilon$ 為振幅不平衡、$\Delta\phi$ 為相位不平衡)。求 (a) 完美平衡時的 IRR、(b) 40 dB IRR 的不平衡 budget、(c) ±0.5 dB 振幅匹配下允許的相位誤差、(d) 與 preselector 的分工。 The IRR formula (given in the problem, with $\varepsilon$ the amplitude imbalance and $\Delta\phi$ the phase imbalance). Find (a) the IRR under perfect balance, (b) the imbalance budget for 40 dB IRR, (c) the phase error permitted with ±0.5 dB amplitude matching, and (d) how the work is divided with the preselector.IRR の公式(問題で与えられ、$\varepsilon$ は振幅不平衡、$\Delta\phi$ は位相不平衡)。(a) 完全平衡時の IRR、(b) 40 dB IRR のための不平衡バジェット、(c) ±0.5 dB の振幅整合で許される位相誤差、(d) プリセレクタとの分担を求めます。

(a) 注意題目公式中的 $\varepsilon$ 是 Note that the $\varepsilon$ in the problem's formula is an 問題の公式中の $\varepsilon$ は振幅比amplitude ratio振幅比(perfect balance 對應 $\varepsilon = 1$):代入 $\varepsilon = 1$、$\Delta\phi = 0$,分母 $= 1 - 2 + 1 = 0$ → $IRR \to \infty$(無限大 dB)。若改用「振幅誤差」定義(perfect balance 為誤差 $= 0$,即小誤差近似的寫法),同樣得 image 殘量為零 → $IRR = \infty$。實際上限永遠由殘餘不平衡決定。(perfect balance corresponding to $\varepsilon = 1$): substituting $\varepsilon = 1$ and $\Delta\phi = 0$ makes the denominator $= 1 - 2 + 1 = 0$ → $IRR \to \infty$ (infinite in dB). Using the “amplitude error” definition instead (perfect balance being an error of $0$, as in the small-error form), the residual image is likewise zero → $IRR = \infty$. In reality the ceiling is always set by the residual imbalance.であることに注意してください(完全平衡は $\varepsilon = 1$ に対応):$\varepsilon = 1$、$\Delta\phi = 0$ を代入すると分母は $= 1 - 2 + 1 = 0$ → $IRR \to \infty$(dB では無限大)。「振幅誤差」の定義(完全平衡は誤差 $0$、すなわち小誤差近似の書き方)を用いても、同様にイメージ残留はゼロ → $IRR = \infty$ となります。実際の上限は常に残留する不平衡で決まります。

(b) 公式與來源:(b) Formula and source:(b) 公式と出典:小誤差展開後 image 殘量功率 $\propto \varepsilon^2 + \Delta\phi^2$(兩種誤差以 RSS 疊加),題目給的近似: expanding for small errors, the residual image power $\propto \varepsilon^2 + \Delta\phi^2$ (the two errors combining in RSS), giving the approximation stated in the problem:小誤差で展開するとイメージ残留電力は $\propto \varepsilon^2 + \Delta\phi^2$(二つの誤差は RSS で合成)となり、問題で与えられた近似は:

$$IRR \approx \frac{1}{\varepsilon^2 + \Delta\phi^2}$$

逐步代入:Substituting step by step:逐次代入:$IRR \geq 40\,\text{dB} \Rightarrow IRR_{lin} = 10^{40/10} = 10^4$:$IRR \geq 40\,\text{dB} \Rightarrow IRR_{lin} = 10^{40/10} = 10^4$:$IRR \geq 40\,\text{dB} \Rightarrow IRR_{lin} = 10^{40/10} = 10^4$:

$$\varepsilon^2 + \Delta\phi^2 \leq 10^{-4} \;\Rightarrow\; \sqrt{\varepsilon^2+\Delta\phi^2} \leq 0.01$$

平均分配(RSS split):$\varepsilon = \Delta\phi \leq 0.01/\sqrt{2} \approx 0.007$,即振幅誤差 $20\log_{10}(1.007) \approx 0.06\,\text{dB}$、相位誤差 $0.007\,\text{rad} \approx 0.4°$。(若用更精確的 $IRR \approx 4/(\varepsilon^2+\Delta\phi^2)$,budget 放寬 2 倍:各 ≈ 0.12 dB / 0.8°。)Splitting it evenly (an RSS split): $\varepsilon = \Delta\phi \leq 0.01/\sqrt{2} \approx 0.007$, i.e. an amplitude error of $20\log_{10}(1.007) \approx 0.06\,\text{dB}$ and a phase error of $0.007\,\text{rad} \approx 0.4°$. (With the more precise $IRR \approx 4/(\varepsilon^2+\Delta\phi^2)$ the budget relaxes by a factor of 2: about 0.12 dB / 0.8° each.)均等に配分すると(RSS 分割):$\varepsilon = \Delta\phi \leq 0.01/\sqrt{2} \approx 0.007$、すなわち振幅誤差 $20\log_{10}(1.007) \approx 0.06\,\text{dB}$、位相誤差 $0.007\,\text{rad} \approx 0.4°$。(より正確な $IRR \approx 4/(\varepsilon^2+\Delta\phi^2)$ を用いると、バジェットは 2 倍に緩和され、それぞれ約 0.12 dB / 0.8° となります。)

(c) 逐步代入(c) Substituting step by step(c) 逐次代入 — dB 轉 linear:振幅比 $A = 10^{0.5/20} = 10^{0.025} = 1.0593$ → 振幅誤差 $\varepsilon = A - 1 = 0.0593$。 — converting dB to linear: the amplitude ratio $A = 10^{0.5/20} = 10^{0.025} = 1.0593$ → an amplitude error $\varepsilon = A - 1 = 0.0593$. — dB をリニアに変換します:振幅比 $A = 10^{0.5/20} = 10^{0.025} = 1.0593$ → 振幅誤差 $\varepsilon = A - 1 = 0.0593$。

$\varepsilon^2 = (0.0593)^2 = 3.5\times 10^{-3}$,$\varepsilon^2 = (0.0593)^2 = 3.5\times 10^{-3}$, which $\varepsilon^2 = (0.0593)^2 = 3.5\times 10^{-3}$ となり、單獨就超過on its own already exceedsこれだけで既に超えているのは 40 dB 所需的 $10^{-4}$(即使含因子 4 的版本 $4\times10^{-4}$ 也超過 ~9 倍)→ the $10^{-4}$ required for 40 dB (even the version with the factor of 4, $4\times10^{-4}$, is exceeded by ~9×) → 40 dB に必要な $10^{-4}$ です(因子 4 を含む版 $4\times10^{-4}$ でさえ ~9 倍超えます)→ 允許的相位誤差為零仍不達標,±0.5 dB 下 40 dB IRR 不可能even with zero phase error it cannot be met, so 40 dB IRR is impossible at ±0.5 dB位相誤差がゼロでも満たせず、±0.5 dB では 40 dB の IRR は不可能。振幅誤差單獨造成的上限:. The ceiling imposed by the amplitude error alone:。振幅誤差だけで決まる上限は:

$$IRR \approx \frac{4}{\varepsilon^2} = \frac{4}{3.5\times 10^{-3}} = 1137 \;\Rightarrow\; 10\log_{10}(1137) \approx \mathbf{31\,dB}$$

(更正:原版誤用 $1/(4\varepsilon^2)$ 算出 18 dB;正確的小誤差公式是 $4/\varepsilon^2$,以精確公式驗證 $\left(\frac{1+1.0593}{1-1.0593}\right)^2 = 1206 \to 30.8\,\text{dB}$ ✓)(Correction: the original version wrongly used $1/(4\varepsilon^2)$ and obtained 18 dB; the correct small-error formula is $4/\varepsilon^2$, verified against the exact formula $\left(\frac{1+1.0593}{1-1.0593}\right)^2 = 1206 \to 30.8\,\text{dB}$ ✓)(訂正:元の版では誤って $1/(4\varepsilon^2)$ を用いて 18 dB としていました;正しい小誤差公式は $4/\varepsilon^2$ であり、厳密式 $\left(\frac{1+1.0593}{1-1.0593}\right)^2 = 1206 \to 30.8\,\text{dB}$ で検証済みです ✓)

±0.5 dB amplitude imbalance alone caps IRR at ≈ 31 dB(原版 18 dB 為誤算)。40 dB IRR requires ≈ ±0.1 dB amplitude match and < 1° phase balance.±0.5 dB amplitude imbalance alone caps IRR at ≈ 31 dB (the original 18 dB was a miscalculation). 40 dB IRR requires ≈ ±0.1 dB amplitude match and < 1° phase balance.±0.5 dB の振幅不平衡だけで IRR は ≈ 31 dB に制限(元の 18 dB は誤算)。40 dB の IRR には ≈ ±0.1 dB の振幅整合と < 1° の位相平衡が必要

(d) 40 dB 的 RF preselector 在 5.9 GHz 要窄帶 SAW/BAW filter(手機常見)。image-reject mixer 的價值是用電子方式取得 image rejection、放鬆 preselector 規格 — 在 low-IF / direct-conversion 接收機中尤其關鍵,因為 image 與訊號只差 $2f_{IF}$(低 IF → 間隔極窄),任何實體濾波器都分不開。實務常用「preselector 30 dB + IRM 30 dB」分層達成 60 dB 總抑制。 A 40 dB RF preselector at 5.9 GHz means a narrowband SAW/BAW filter (common in handsets). The value of an image-reject mixer is that it obtains image rejection electronically and so relaxes the preselector specification — which is especially critical in low-IF and direct-conversion receivers, where the image is only $2f_{IF}$ from the signal (a low IF → an extremely narrow separation) and no physical filter can separate them. In practice a layered “30 dB preselector + 30 dB IRM” is commonly used to reach 60 dB of total suppression. 5.9 GHz で 40 dB の RF プリセレクタとなると、狭帯域の SAW/BAW フィルタが必要です(携帯端末では一般的)。イメージ除去ミキサの価値は、電子的にイメージ除去を得てプリセレクタの仕様を緩和できる点にあります — これは low-IF やダイレクトコンバージョン受信機で特に重要で、イメージと信号の差はわずか $2f_{IF}$(低い IF → 間隔が極めて狭い)であり、いかなる物理的なフィルタでも分離できないからです。実務では「プリセレクタ 30 dB + IRM 30 dB」と階層化して合計 60 dB の抑圧を達成することがよくあります。

物理詮釋:Physical interpretation:物理的解釈:IRR 對振幅誤差極其敏感:0.5 dB(人耳都聽不出的誤差)就把 IRR 鎖死在 ~31 dB。所以高 IRR 設計不靠元件先天匹配,而靠 The IRR is extraordinarily sensitive to amplitude error: 0.5 dB (an error the ear could not even detect) pins the IRR at ~31 dB. So high-IRR designs do not rely on inherent component matching but on IRR は振幅誤差に極めて敏感です:0.5 dB(人の耳では聞き分けられないほどの誤差)で IRR は ~31 dB に固定されてしまいます。そのため高 IRR 設計は素子固有の整合には頼らず、I/Q calibration(數位估測並補償 $\varepsilon, \Delta\phi$),校正後 50–60 dB 才有可能。(digitally estimating and compensating $\varepsilon$ and $\Delta\phi$), after which 50–60 dB becomes possible.($\varepsilon$ と $\Delta\phi$ をディジタルに推定して補償)、校正後には 50–60 dB が可能になります。

Topic 7Noise Figure Budget

SSB vs DSB NF, Friis cascade, receiver sensitivity.

7.1 Easy
SSB vs DSB noise figure conversion

(a) A mixer datasheet states $NF_{DSB} = 6\,\text{dB}$. What is the SSB noise figure?

(b) Why does SSB NF differ from DSB NF for a mixer? (1–2 sentences, physical reasoning.)

(c) A system noise figure spec is always given as SSB. You measure a mixer DSB NF of 7 dB. What SSB NF do you enter into the Friis cascade?

Answer

已知/目標:Given / required:既知/目標:(a) datasheet 給 $NF_{DSB} = 6\,\text{dB}$,求 $NF_{SSB}$;(b) 解釋兩者差異的物理原因;(c) 量得 $NF_{DSB} = 7\,\text{dB}$,求進 Friis cascade 的值。(a) The datasheet gives $NF_{DSB} = 6\,\text{dB}$; find $NF_{SSB}$. (b) Explain the physical reason for the difference. (c) A measurement gives $NF_{DSB} = 7\,\text{dB}$; find the value to use in the Friis cascade.(a) データシートは $NF_{DSB} = 6\,\text{dB}$ を与えており、$NF_{SSB}$ を求めます;(b) 両者の差の物理的な理由を説明します;(c) 測定で $NF_{DSB} = 7\,\text{dB}$ が得られたとき、Friis 縦続計算に用いる値を求めます。

公式與來源:Formulas and sources:公式と出典:mixer 同時把 signal band 與 image band 的雜訊都搬到 IF。若訊號只佔一個 sideband(SSB,實際接收的情況),雜訊卻來自兩個 band → 有效 NF 比 DSB 量測高 3 dB(假設兩 sideband 的 conversion gain 相同): A mixer brings the noise of both the signal band and the image band down to the IF. If the signal occupies only one sideband (SSB, as in real reception) while the noise comes from both bands → the effective NF is 3 dB higher than the DSB measurement (assuming equal conversion gain for the two sidebands):ミキサは信号帯域とイメージ帯域の雑音を同時に IF へ移します。信号が一方の側波帯だけを占め(SSB、実際の受信の場合)、雑音は両方の帯域から来る場合 → 実効 NF は DSB 測定より 3 dB 高くなります(両側波帯の変換利得が等しいと仮定):

$$NF_{SSB} = NF_{DSB} + 3\,\text{dB}$$

逐步代入:Substituting step by step:逐次代入:

(a) $NF_{SSB} = 6 + 3 = \mathbf{9\,dB}$。(a) $NF_{SSB} = 6 + 3 = \mathbf{9\,dB}$.(a) $NF_{SSB} = 6 + 3 = \mathbf{9\,dB}$。

(b) DSB 量測時,訊號同時放在 signal 與 image 兩個 sideband,訊號功率與雜訊功率「成對」出現;SSB 工作時訊號只在一個 sideband,但 image band 的雜訊照樣被折下來疊在 IF 上 — 雜訊翻倍(+3 dB)而訊號不變 → SNR 退化 3 dB,反映為 NF 高 3 dB。(b) In a DSB measurement the signal is placed in both the signal and image sidebands, so signal power and noise power appear “in pairs”; in SSB operation the signal is in only one sideband, yet the image band's noise is still folded down onto the IF — the noise doubles (+3 dB) while the signal does not → the SNR degrades by 3 dB, which shows up as a 3 dB higher NF.(b) DSB 測定では信号を信号側とイメージ側の両方の側波帯に置くため、信号電力と雑音電力が「対で」現れます;SSB 動作では信号は一方の側波帯のみですが、イメージ帯域の雑音は同じように折り返されて IF に重なります — 雑音は 2 倍(+3 dB)になるのに信号は変わらない → SNR が 3 dB 劣化し、NF が 3 dB 高いという形で現れます。

(c) $NF_{SSB} = 7 + 3 = \mathbf{10\,dB}$ — Friis cascade 要用這個值。(c) $NF_{SSB} = 7 + 3 = \mathbf{10\,dB}$ — this is the value to use in the Friis cascade.(c) $NF_{SSB} = 7 + 3 = \mathbf{10\,dB}$ — Friis 縦続計算にはこの値を使います。

NF_SSB = NF_DSB + 3 dB;系統雜訊預算一律用 SSB 值NF_SSB = NF_DSB + 3 dB; always use the SSB value in a system noise budgetNF_SSB = NF_DSB + 3 dB;システム雑音バジェットでは常に SSB 値を使用

物理詮釋:Physical interpretation:物理的解釈:datasheet 偏好標 DSB(數字好看 3 dB),但 superheterodyne 接收的訊號是 SSB 的 — 拿 DSB 值直接進 Friis 會把系統 NF 低估 3 dB,這是新手最常見的雜訊預算錯誤。例外:direct-conversion(ZIF)接收機的「image」就是訊號自己的另一半 sideband,此時 DSB NF 才是正確的工作定義。若 mixer 前有 image-reject filter 把 image band 雜訊濾掉,有效 NF 介於兩者之間。 Datasheets prefer to quote DSB (a figure 3 dB more flattering), but a superheterodyne receives an SSB signal — feeding the DSB value straight into Friis underestimates the system NF by 3 dB, the commonest noise-budget mistake beginners make. The exception: in a direct-conversion (ZIF) receiver the “image” is the signal's own other sideband, and there the DSB NF is the correct working definition. If an image-reject filter ahead of the mixer removes the image-band noise, the effective NF lies between the two.データシートは DSB を記載したがりますが(数字が 3 dB 良く見えます)、スーパーヘテロダインが受信する信号は SSB です — DSB 値をそのまま Friis に入れるとシステム NF を 3 dB 過小評価してしまい、これは初心者が最もよく犯す雑音バジェットの誤りです。例外:ダイレクトコンバージョン(ZIF)受信機の「イメージ」は信号自身のもう半分の側波帯であり、この場合は DSB の NF こそが正しい実用上の定義です。ミキサの前にイメージ除去フィルタがあってイメージ帯域の雑音を除去する場合、実効 NF は両者の中間になります。

7.2 Medium
Receiver sensitivity calculation

A receiver chain: LNA (G=18 dB, NF=1.5 dB) → Mixer (L_C=7 dB, SSB NF=10 dB) → IF Amp (G=20 dB, NF=4 dB).

(a) Find the cascaded system NF using Friis formula.

(b) For a required SNR of 10 dB and bandwidth B = 200 kHz (GSM), find the minimum detectable signal (sensitivity) at the antenna input.

(c) If sensitivity must be −110 dBm, does this chain meet the spec?

Answer

已知/目標:Given / required:既知/目標:鏈路 LNA(G = 18 dB, NF = 1.5 dB)→ Mixer($L_C$ = 7 dB, $NF_{SSB}$ = 10 dB)→ IF Amp(G = 20 dB, NF = 4 dB);GSM 通道 $B = 200\,\text{kHz}$、需求 SNR = 10 dB。求 (a) 系統 NF、(b) sensitivity、(c) 是否達 −110 dBm spec。 A chain of LNA (G = 18 dB, NF = 1.5 dB) → Mixer ($L_C$ = 7 dB, $NF_{SSB}$ = 10 dB) → IF Amp (G = 20 dB, NF = 4 dB); a GSM channel of $B = 200\,\text{kHz}$ requiring SNR = 10 dB. Find (a) the system NF, (b) the sensitivity, and (c) whether it meets the −110 dBm spec.LNA(G = 18 dB, NF = 1.5 dB)→ ミキサ($L_C$ = 7 dB, $NF_{SSB}$ = 10 dB)→ IF アンプ(G = 20 dB, NF = 4 dB)の縦続;GSM チャネル $B = 200\,\text{kHz}$、要求 SNR = 10 dB。(a) システム NF、(b) 感度、(c) −110 dBm の仕様を満たすかを求めます。

公式與來源:Formulas and sources:公式と出典:Friis cascade formula(linear 值;mixer 的「增益」是 $1/L_C$)與 sensitivity 公式: The Friis cascade formula (in linear values; the mixer's “gain” being $1/L_C$) and the sensitivity formula:Friis の縦続公式(リニア値;ミキサの「利得」は $1/L_C$)と感度の公式:

$$F_{sys} = F_1 + \frac{F_2 - 1}{G_1} + \frac{F_3 - 1}{G_1 G_2}, \qquad P_{sens} = kTB + NF_{sys} + SNR_{req}$$

(a) 逐步代入(a) Substituting step by step(a) 逐次代入 — dB 全部轉 linear: — converting everything from dB to linear: — すべて dB からリニアに変換します:

  • $F_1 = 10^{1.5/10} = 1.413$,$G_1 = 10^{18/10} = 63.1$$F_1 = 10^{1.5/10} = 1.413$ and $G_1 = 10^{18/10} = 63.1$$F_1 = 10^{1.5/10} = 1.413$、$G_1 = 10^{18/10} = 63.1$
  • $F_{mix} = 10^{10/10} = 10$,$G_{mix} = 10^{-7/10} = 0.2$(conversion loss 取負號進指數)$F_{mix} = 10^{10/10} = 10$ and $G_{mix} = 10^{-7/10} = 0.2$ (the conversion loss entering the exponent with a minus sign)$F_{mix} = 10^{10/10} = 10$、$G_{mix} = 10^{-7/10} = 0.2$(変換損失は負号を付けて指数に入ります)
  • $F_{IFA} = 10^{4/10} = 2.512$

$$F_{sys} = 1.413 + \dfrac{10-1}{63.1} + \dfrac{2.512-1}{63.1\times 0.2} = 1.413 + 0.143 + 0.120 = 1.676$$

NF_sys = 10·log(1.676) = 2.24 dB

(b) 逐步代入(b) Substituting step by step(b) 逐次代入 — 先算熱雜訊底: — first the thermal noise floor: — まず熱雑音フロアを計算します:

$$kTB = -174\,\text{dBm/Hz} + 10\log_{10}(200\times 10^3) = -174 + 53 = -121\,\text{dBm}$$

$$P_{sens} = -121 + 2.24 + 10 = \mathbf{-108.8\,dBm}$$

(c) 需求 ≤ −110 dBm,但 $-108.8 > -110$ → The requirement is ≤ −110 dBm, but $-108.8 > -110$ → 要求は ≤ −110 dBm ですが、$-108.8 > -110$ → 差 1.2 dB,不達標1.2 dB short, a fail1.2 dB 足りず、不合格.

物理詮釋:Physical interpretation:物理的解釈:分解貢獻:LNA 自己佔 1.5 dB,mixer 經 LNA 增益壓制後仍貢獻 0.143(折合約 0.4 dB)、IF amp 貢獻 0.120(約 0.3 dB)— mixer 的 10 dB NF 被 18 dB 增益除到只剩零頭,驗證「第一級主宰雜訊」。修正方向:要過 −110 dBm 需 $NF_{sys} \leq 1.0\,\text{dB}$($F_{sys} \leq 1.26$)— 即使 LNA NF 降到 1.0 dB 且增益拉到 22 dB,$F_{sys} = 1.259 + 0.057 + 0.048 = 1.364 \to 1.35\,\text{dB}$ 仍差 0.35 dB;實際要 LNA NF ≈ 0.7 dB、G ≥ 22 dB 才穩過,顯示 sensitivity spec 幾乎完全壓在 LNA 身上。 Breaking down the contributions: the LNA itself accounts for 1.5 dB; the mixer, suppressed by the LNA gain, still contributes 0.143 (about 0.4 dB) and the IF amp 0.120 (about 0.3 dB) — the mixer's 10 dB NF is divided by 18 dB of gain down to a trifle, confirming that “the first stage dominates the noise”. The way forward: to pass −110 dBm requires $NF_{sys} \leq 1.0\,\text{dB}$ ($F_{sys} \leq 1.26$) — and even lowering the LNA NF to 1.0 dB and raising its gain to 22 dB gives $F_{sys} = 1.259 + 0.057 + 0.048 = 1.364 \to 1.35\,\text{dB}$, still 0.35 dB short; in practice an LNA NF ≈ 0.7 dB with G ≥ 22 dB is needed to pass comfortably, showing that the sensitivity spec rests almost entirely on the LNA.寄与を分解すると:LNA 自身が 1.5 dB を占め、ミキサは LNA の利得で抑えられてもなお 0.143(約 0.4 dB)、IF アンプは 0.120(約 0.3 dB)を寄与します — ミキサの 10 dB の NF は 18 dB の利得で割られてわずかしか残らず、「初段が雑音を支配する」ことを裏付けます。改善の方向:−110 dBm を満たすには $NF_{sys} \leq 1.0\,\text{dB}$($F_{sys} \leq 1.26$)が必要です — LNA の NF を 1.0 dB に下げ利得を 22 dB に上げても $F_{sys} = 1.259 + 0.057 + 0.048 = 1.364 \to 1.35\,\text{dB}$ で、まだ 0.35 dB 足りません;実際には LNA の NF ≈ 0.7 dB、G ≥ 22 dB が必要で、感度仕様がほぼ完全に LNA に懸かっていることが分かります。

7.3 Design
Design a receive chain to meet NF and IIP3 simultaneously
Requirements System NF ≤ 4 dB, System IIP3 ≥ −20 dBm
Frequency: 900 MHz
Available components: LNA (G=15dB, NF=1.5dB, IIP3=−5dBm), Mixer (L=7dB, NF_SSB=10dB, IIP3=+5dBm), IF Amp (G=20dB, NF=4dB, IIP3=+15dBm)

(a) Verify NF with this chain.

(b) Verify IIP3 with this chain using $\dfrac{1}{IIP3_{sys}} = \dfrac{1}{IIP3_1} + \dfrac{G_1}{IIP3_2} + \dfrac{G_1 G_2}{IIP3_3}$.

(c) If either spec fails, propose one change (add/change one component) to fix it.

Answer

已知/目標:Given / required:既知/目標:spec:NF ≤ 4 dB、IIP3 ≥ −20 dBm。鏈路:LNA(G = 15 dB, NF = 1.5 dB, IIP3 = −5 dBm)→ Mixer(L = 7 dB, $NF_{SSB}$ = 10 dB, IIP3 = +5 dBm)→ IF Amp(G = 20 dB, NF = 4 dB, IIP3 = +15 dBm)。同時驗證兩項 spec。 The specs: NF ≤ 4 dB and IIP3 ≥ −20 dBm. The chain: LNA (G = 15 dB, NF = 1.5 dB, IIP3 = −5 dBm) → Mixer (L = 7 dB, $NF_{SSB}$ = 10 dB, IIP3 = +5 dBm) → IF Amp (G = 20 dB, NF = 4 dB, IIP3 = +15 dBm). Verify both specs.仕様:NF ≤ 4 dB、IIP3 ≥ −20 dBm。縦続:LNA(G = 15 dB, NF = 1.5 dB, IIP3 = −5 dBm)→ ミキサ(L = 7 dB, $NF_{SSB}$ = 10 dB, IIP3 = +5 dBm)→ IF アンプ(G = 20 dB, NF = 4 dB, IIP3 = +15 dBm)。両方の仕様を同時に検証します。

(a) NF check — Friis(linear):(a) NF check — Friis (linear):(a) NF の確認 — Friis(リニア):$F_1 = 10^{0.15} = 1.413$、$G_1 = 10^{1.5} = 31.6$、$F_2 = 10$、$G_2 = 10^{-0.7} = 0.2$、$F_3 = 2.512$:$F_1 = 10^{0.15} = 1.413$, $G_1 = 10^{1.5} = 31.6$, $F_2 = 10$, $G_2 = 10^{-0.7} = 0.2$, $F_3 = 2.512$:$F_1 = 10^{0.15} = 1.413$、$G_1 = 10^{1.5} = 31.6$、$F_2 = 10$、$G_2 = 10^{-0.7} = 0.2$、$F_3 = 2.512$:

$$F_{sys} = 1.413 + \frac{10-1}{31.6} + \frac{2.512-1}{31.6\times 0.2} = 1.413 + 0.285 + 0.239 = 1.937$$

$$NF_{sys} = 10\log_{10}(1.937) = \mathbf{2.87\,dB} \;\checkmark\;(\leq 4\,\text{dB})$$

(b) IIP3 check — cascade 公式(全部 mW,逐項換算):(b) IIP3 check — the cascade formula (all in mW, converted term by term):(b) IIP3 の確認 — 縦続公式(すべて mW、項ごとに換算):

  • $IIP3_{LNA} = 10^{-5/10} = 0.316\,\text{mW}$;$G_{LNA} = 31.6$$IIP3_{LNA} = 10^{-5/10} = 0.316\,\text{mW}$; $G_{LNA} = 31.6$$IIP3_{LNA} = 10^{-5/10} = 0.316\,\text{mW}$;$G_{LNA} = 31.6$
  • $IIP3_{mix} = 10^{5/10} = 3.16\,\text{mW}$;$G_{mix} = 0.2$$IIP3_{mix} = 10^{5/10} = 3.16\,\text{mW}$; $G_{mix} = 0.2$$IIP3_{mix} = 10^{5/10} = 3.16\,\text{mW}$;$G_{mix} = 0.2$
  • $IIP3_{IFA} = 10^{15/10} = 31.6\,\text{mW}$

$$\dfrac{1}{IIP3_{sys}} = \dfrac{1}{0.316} + \dfrac{31.6}{3.16} + \dfrac{31.6\times 0.2}{31.6} = 3.16 + 10 + 0.2 = 13.36\,\text{mW}^{-1}$$

$$IIP3_{sys} = \frac{1}{13.36} = 0.0749\,\text{mW} = 10\log_{10}(0.0749) = -11.25\,\text{dBm}$$

IIP3_sys = −11.25 dBm ≥ −20 dBm → ✓ PASSES(注意 dBm 比較方向:−11.25 比 −20 「大」);NF = 2.87 dB ≤ 4 dB → ✓ PASSESIIP3_sys = −11.25 dBm ≥ −20 dBm → ✓ PASSES (mind the direction of the dBm comparison: −11.25 is “larger” than −20); NF = 2.87 dB ≤ 4 dB → ✓ PASSESIIP3_sys = −11.25 dBm ≥ −20 dBm → ✓ PASSES(dBm の比較方向に注意:−11.25 は −20 より「大きい」);NF = 2.87 dB ≤ 4 dB → ✓ PASSES

三項中 mixer 的貢獻(10 mW⁻¹)佔 75% — 它是 IIP3 瓶頸;IF amp 雖在最後一級,但其貢獻被 mixer 的損耗壓回($G_1G_2 = 6.3$),只佔 1.5%。Of the three terms the mixer's contribution (10 mW⁻¹) accounts for 75% — it is the IIP3 bottleneck; the IF amp, though last in the chain, has its contribution pushed back down by the mixer's loss ($G_1G_2 = 6.3$) and accounts for only 1.5%.三つの項のうちミキサの寄与(10 mW⁻¹)が 75% を占め — これが IIP3 のボトルネックです;IF アンプは最終段にありますが、その寄与はミキサの損失によって押し戻され($G_1G_2 = 6.3$)、わずか 1.5% にとどまります。

(c) 兩項 spec 都過, Both specs pass, so 両方の仕様を満たすため、不需更動no change is needed変更は不要です。延伸:若 spec 收緊,換 IIP3 = +12 dBm(15.85 mW)的 mixer 後第二項從 10 降到 $31.6/15.85 = 2.0$ → 總和 $5.36\,\text{mW}^{-1}$ → $IIP3_{sys} = 0.187\,\text{mW} = -7.3\,\text{dBm}$。再注意一個極限:即使 mixer 無限好,$1/IIP3_{sys} \geq 3.16 + 0.2 = 3.36$ → $IIP3_{sys} \leq -5.3\,\text{dBm}$ — . An extension: if the spec tightened, switching to a mixer with IIP3 = +12 dBm (15.85 mW) would drop the second term from 10 to $31.6/15.85 = 2.0$ → a total of $5.36\,\text{mW}^{-1}$ → $IIP3_{sys} = 0.187\,\text{mW} = -7.3\,\text{dBm}$. Note one further limit: even with an infinitely good mixer, $1/IIP3_{sys} \geq 3.16 + 0.2 = 3.36$ → $IIP3_{sys} \leq -5.3\,\text{dBm}$ — 。発展:仕様が厳しくなった場合、IIP3 = +12 dBm(15.85 mW)のミキサに替えると第二項は 10 から $31.6/15.85 = 2.0$ に下がり → 合計 $5.36\,\text{mW}^{-1}$ → $IIP3_{sys} = 0.187\,\text{mW} = -7.3\,\text{dBm}$ となります。さらにもう一つの限界に注意してください:ミキサが無限に良くても $1/IIP3_{sys} \geq 3.16 + 0.2 = 3.36$ → $IIP3_{sys} \leq -5.3\,\text{dBm}$ — LNA 的 IIP3 與增益設定了天花板the LNA's IIP3 and gain set the ceilingLNA の IIP3 と利得が上限を決めます,若 spec 要 ≥ −5 dBm,光換 mixer 永遠不夠,得換 LNA 或降其增益。, so if the spec demanded ≥ −5 dBm, changing the mixer alone would never suffice and the LNA would have to be changed or its gain reduced.。仕様が ≥ −5 dBm を要求するなら、ミキサだけを替えても決して足りず、LNA を替えるか、その利得を下げる必要があります。

物理詮釋:Physical interpretation:物理的解釈:NF 與 IIP3 的 budget 拉力方向相反:增益大 → 後級雜訊被壓低(NF 好)但後級提早飽和(IIP3 差)。本鏈用 15 dB 的中等 LNA 增益取得平衡,兩項 spec 同時過 — 這正是 receiver line-up 設計的核心習題。 The NF and IIP3 budgets pull in opposite directions: more gain → the later stages' noise is suppressed (better NF) but they saturate sooner (worse IIP3). This chain strikes the balance with a moderate 15 dB of LNA gain and passes both specs at once — the essential exercise of receiver line-up design.NF と IIP3 のバジェットは逆方向に引き合います:利得を大きくする → 後段の雑音は抑えられます(NF は良い)が、後段が早く飽和します(IIP3 は悪い)。この縦続は 15 dB という中庸な LNA 利得で釣り合いを取り、両方の仕様を同時に満たしています — これこそ受信機ラインアップ設計の要となる練習問題です。

Topic 8Topology Selection

Compare single-ended, single-balanced, double-balanced, image-reject. Given an application, choose and justify.

8.1 Easy
Match topology to application

For each application, choose the best mixer topology from: (A) single-ended, (B) single-balanced 90°, (C) single-balanced 180°, (D) double-balanced, (E) image-reject.

  1. Narrowband ISM 433 MHz sensor, cost is critical, LO leakage is acceptable, bandwidth < 1 MHz.
  2. Cellular phone receiver, LO-to-RF isolation > 40 dB required, even harmonic suppression needed.
  3. Low-IF Bluetooth receiver (IF = 1 MHz), image is at $f_{LO} - 1\,\text{MHz}$, image separation only 2 MHz.
  4. Laboratory test equipment up/converter, maximum dynamic range, LO drive is plentiful.
  5. A system requiring rejection of AM noise on LO, and LO is single-ended (no differential).
Answer

已知/目標:Given / required:既知/目標:五個應用情境,從 (A) single-ended、(B) single-balanced 90°、(C) single-balanced 180°、(D) double-balanced、(E) image-reject 中各選一個最適拓樸。 Five application scenarios; for each, choose the most suitable topology from (A) single-ended, (B) single-balanced 90°, (C) single-balanced 180°, (D) double-balanced and (E) image-reject.五つの応用シナリオについて、(A) シングルエンド、(B) 単平衡 90°、(C) 単平衡 180°、(D) 二重平衡、(E) イメージ除去 の中から最適なトポロジーをそれぞれ一つ選びます。

判斷原則(公式與來源):The guiding principle (formula and source):判断の原則(公式と出典):拓樸選擇 = 把應用的「最痛規格」對到各拓樸的強項 — diode 數量/成本:A < B,C < D,E;LO-RF isolation:A(~15 dB)< B,C(~25 dB)< D(~40–50 dB);harmonic 抑制:A 無、B/C 抑 even-LO、D 抑 even-LO 與 even-RF;近距 image 分離:唯有 E。 choosing a topology = matching the application's most painful specification to each topology's strength — diode count/cost: A < B,C < D,E; LO-RF isolation: A (~15 dB) < B,C (~25 dB) < D (~40–50 dB); harmonic suppression: A none, B/C suppress even-LO, D suppresses both even-LO and even-RF; separating a nearby image: only E.トポロジーの選択 = 応用で最も苦しい仕様を各トポロジーの強みに対応させること — ダイオード数/コスト:A < B,C < D,E;LO-RF アイソレーション:A(~15 dB)< B,C(~25 dB)< D(~40–50 dB);高調波抑圧:A はなし、B/C は偶数次 LO を抑圧、D は偶数次 LO と偶数次 RF を抑圧;近接したイメージの分離:E のみ。

逐題代入:Working through each:各シナリオへの適用:

  1. (A) Single-ended — 痛點是成本。433 MHz、頻寬 < 1 MHz、LO leakage 可接受 → 一顆 diode + 簡單 LC post-filter 即可;balanced 結構的 hybrid/balun 在這裡是浪費。 — the pain point is cost. At 433 MHz with a bandwidth < 1 MHz and tolerable LO leakage → a single diode plus a simple LC post-filter will do; a balanced structure's hybrid/balun would be wasted here. — 痛点はコストです。433 MHz、帯域幅 < 1 MHz、LO 漏れは許容できる → ダイオード 1 個と簡単な LC ポストフィルタで十分です;平衡構造のハイブリッド/バランはここでは無駄になります。
  2. (D) Double-balanced — 痛點是 isolation > 40 dB「加上」even harmonic 抑制:single-balanced 兩者只各拿一半(isolation ~25 dB),只有 ring 同時滿足兩條。 — the pain point is isolation > 40 dB *plus* even-harmonic suppression: single-balanced gives only half of each (isolation ~25 dB), and only a ring meets both. — 痛点はアイソレーション > 40 dB「かつ」偶数次高調波の抑圧です:単平衡ではどちらも半分ずつしか得られず(アイソレーション ~25 dB)、リングだけが両方を同時に満たします。
  3. (E) Image-reject — image 距離訊號僅 $2f_{IF} = 2\,\text{MHz}$,2.4 GHz 級的 RF preselector 不可能有 0.08% 的選擇性 → 必須用相位法電子分離(low-IF 接收機的標準解)。 — the image is only $2f_{IF} = 2\,\text{MHz}$ from the signal, and no 2.4 GHz RF preselector can have 0.08% selectivity → electronic separation by the phase method is essential (the standard solution for a low-IF receiver). — イメージは信号からわずか $2f_{IF} = 2\,\text{MHz}$ しか離れておらず、2.4 GHz 帯の RF プリセレクタに 0.08% の選択度は不可能です → 位相法による電子的な分離が必須です(low-IF 受信機の標準的な解)。
  4. (D) Double-balanced — 實驗室儀器要最大 dynamic range:ring 的 IIP3 最高(LO drive 充足可上 level-13/17)、spur 最少 → SFDR 最大。 — laboratory instruments need the greatest dynamic range: a ring has the highest IIP3 (with ample LO drive it can go to level-13/17) and the fewest spurs → the greatest SFDR. — 実験室用計測器には最大のダイナミックレンジが必要です:リングは IIP3 が最も高く(LO ドライブが十分なら level-13/17 まで上げられます)、スプリアスが最も少ない → SFDR が最大になります。
  5. (B) Single-balanced 90° — 需求是 LO AM noise rejection 且 LO 為 single-ended:90° hybrid 的對稱路徑讓 AM noise 在 IF combiner 呈 common-mode 相消(3.1 題),且 hybrid 的輸入埠本來就吃 single-ended 訊號,不需差動 LO 或 balun。 — the requirement is LO AM noise rejection with a single-ended LO: the symmetric paths of a 90° hybrid make the AM noise common-mode at the IF combiner where it cancels (problem 3.1), and the hybrid's input port takes a single-ended signal natively, needing no differential LO or balun. — 要求は LO の AM 雑音除去で、しかも LO はシングルエンドです:90° ハイブリッドの対称な経路は AM 雑音を IF 合成器で同相成分にして打ち消し(問題 3.1)、またハイブリッドの入力ポートはもともとシングルエンド信号を受け付けるため、差動 LO やバランを必要としません。
選型口訣:成本→single-ended;AM noise→90° SB;isolation+harmonics→DB ring;近距 image→image-rejectA selection mnemonic: cost→single-ended; AM noise→90° SB; isolation+harmonics→DB ring; a nearby image→image-reject選定の覚え方:コスト→シングルエンド;AM 雑音→90° SB;アイソレーション+高調波→DB リング;近接イメージ→イメージ除去

物理詮釋:Physical interpretation:物理的解釈:每往 balanced 階梯上爬一級,就是「用更多 diode 與更大 LO 功率,買一種對稱性抵消」— 所以選型的問題永遠是:這個應用最貴的雜質是哪一種?而不是「哪個拓樸最好」。 Every step up the balanced ladder is “spending more diodes and more LO power to buy one more symmetry cancellation” — so the selection question is always: which impurity is most expensive in *this* application? Never “which topology is best”.平衡構成の階段を一段上がるごとに、「より多くのダイオードとより大きな LO 電力を払って、対称性による打ち消しを一つ買う」ことになります — したがって選定の問いは常に:この応用で最も高くつく不要成分はどれか?であって、「どのトポロジーが最良か」ではありません。

8.2 Medium
Comparison table — suppressed harmonics

Complete the harmonic suppression table. For each topology, state which LO harmonics are suppressed at the IF output (ideally).

TopologySuppressed LO harmonicsReason
Single-ended??
Single-balanced (90° or 180°)??
Double-balanced ring??
Answer

已知/目標:Given / required:既知/目標:比較三種拓樸在 IF 輸出端(理想匹配下)抑制了哪些 LO harmonics,並給出對稱性理由。 Compare which LO harmonics the three topologies suppress at the IF output (under ideal matching), and give the symmetry reason.三つのトポロジーが IF 出力端で(理想的な整合のもとで)どの LO 高調波を抑圧するかを比較し、対称性による理由を示します。

公式與來源:Formulas and sources:公式と出典:balanced mixer 的抑制都來自同一機制 — 把某類產物安排成 combiner 處的 All the suppression in a balanced mixer comes from the same mechanism — arranging one class of product to be 平衡ミキサの抑圧はすべて同じ機構から生じます — ある種類の成分を合成器において common-mode(同相 → 相減消失),desired 產物則是 differential(反相 → 相減相加)。double-balanced 的 switching function $\text{sgn}(v_{LO})$ 是奇函數,只含奇次 LO 諧波(5.1 題);RF 側 balun 同理只讓奇次 RF 諧波產物通過。at the combiner (in phase → vanishing on subtraction) while the desired products are differential (in anti-phase → adding on subtraction). A double-balanced mixer's switching function $\text{sgn}(v_{LO})$ is an odd function containing only odd LO harmonics (problem 5.1); the RF-side balun likewise passes only odd RF harmonic products.になるように配置し(同相 → 減算で消滅)、所望の成分は差動にします(逆相 → 減算で加算)。二重平衡ミキサのスイッチング関数 $\text{sgn}(v_{LO})$ は奇関数で、奇数次の LO 高調波のみを含みます(問題 5.1);RF 側のバランも同様に、奇数次 RF 高調波の成分のみを通します。

TopologySuppressed LO harmonicsReason
Single-endedNone — all $nf_{LO}$ present單一非線性元件,沒有任何對稱路徑可供抵消;所有 $mf_{LO} \pm nf_{RF}$ 全額輸出,只能靠濾波器收拾A single nonlinear element with no symmetric path available for cancellation; every $mf_{LO} \pm nf_{RF}$ comes out in full and only a filter can clear it up単一の非線形素子で、打ち消しに使える対称な経路がありません。すべての $mf_{LO} \pm nf_{RF}$ が全量そのまま出力され、フィルタでしか処理できません
Single-balancedEven harmonics: $2f_{LO}$, $4f_{LO}$, …LO 以 180° 反相驅動兩顆 diode:奇次 LO 項在兩路反相(differential)→ combiner 相減後保留;偶次 LO 項 $\cos(2n\omega_{LO}t)$ 對 180° 平移不變、兩路同相(common-mode)→ 相減歸零The LO drives the two diodes 180° out of phase: the odd LO terms are in anti-phase on the two paths (differential) → they survive the subtracting combiner; the even LO terms $\cos(2n\omega_{LO}t)$ are invariant under a 180° shift and are in phase on both paths (common-mode) → they vanish on subtractionLO が 2 個のダイオードを 180° 逆相で駆動します:奇数次 LO 項は両経路で逆相(differential)→ 合成器で減算した後も残ります;偶数次 LO 項 $\cos(2n\omega_{LO}t)$ は 180° シフトに対して不変で両経路が同相(common-mode)→ 減算によりゼロになります
Double-balanced ringEven LO harmonics AND even RF harmonics(即所有 even-$m$ 或 even-$n$ 的 $mf_{LO} \pm nf_{RF}$ 產物)Even LO harmonics AND even RF harmonics (i.e. every $mf_{LO} \pm nf_{RF}$ product with even $m$ or even $n$)偶数次 LO 高調波と偶数次 RF 高調波(すなわち偶数の $m$ または偶数の $n$ を持つすべての $mf_{LO} \pm nf_{RF}$ 産物)4-diode ring 對 LO 與 RF 各有一層 balance:switching function 為奇函數(只含奇 $m$)、RF balun 對偶 $n$ 產物同樣 common-mode 抵消 → 只剩 odd × odd 產物A 4-diode ring has one layer of balance for the LO and one for the RF: the switching function is an odd function (containing only odd $m$), and the RF balun likewise makes the even-$n$ products common-mode and cancels them → only odd × odd products remain4-diode ring は LO と RF にそれぞれ 1 層ずつの平衡を持ちます:スイッチング関数は奇関数(奇数の $m$ のみを含む)であり、RF バランも同様に偶数 $n$ の産物を同相(common-mode)にして打ち消します → odd × odd の産物のみが残ります
每加一層 balance,就消去一邊(LO 或 RF)的 even-order 產物:SE 全留 → SB 消 even-LO → DB 消 even-LO 與 even-RF(僅剩 odd-m × odd-n)。Each extra layer of balance eliminates the even-order products on one side (LO or RF): SE keeps everything → SB kills even-LO → DB kills both even-LO and even-RF (leaving only odd-m × odd-n).平衡を 1 層追加するごとに、片側(LO または RF)の偶数次産物が消えます:SE はすべて残る → SB は even-LO を消す → DB は even-LO と even-RF の両方を消す(odd-m × odd-n のみが残る)。

物理詮釋:Physical interpretation:物理的解釈:這張表也是 spur chart 的「強度欄」:頻率規劃(2.2 題)決定 spur 落點,拓樸決定哪些 spur 先天被壓 30–60 dB。例如 half-IF spur 是 (2,2) 產物 — even×even — 在 DB ring 被雙重抑制,這正是 9.2、9.4 題中 frequency plan 與拓樸要一起設計的原因。理想抑制是無限大,實際由 diode/balun 匹配度限制在 ~30–60 dB。 This table is also the “strength column” of a spur chart: the frequency plan (problem 2.2) decides where the spurs land, and the topology decides which of them are inherently 30–60 dB down. The half-IF spur, for instance, is a (2,2) product — even×even — doubly suppressed in a DB ring, which is exactly why problems 9.2 and 9.4 require the frequency plan and the topology to be designed together. Ideal suppression is infinite; in practice diode/balun matching limits it to ~30–60 dB.この表は spur chart の「強度欄」でもあります:周波数計画(問題 2.2)が spur の落ちる位置を決め、トポロジーがどの spur が本質的に 30–60 dB 抑えられるかを決めます。例えば half-IF spur は (2,2) 産物 — even×even — であり、DB ring では二重に抑圧されます。これがまさに問題 9.2、9.4 で frequency plan とトポロジーを一緒に設計しなければならない理由です。理想的な抑圧は無限大ですが、実際にはダイオード/バランの整合度によって ~30–60 dB に制限されます。

8.3 Design
Choose and fully specify a mixer for a direct-broadcast satellite (DBS) receiver
System Context DBS receiver block (outdoor unit), Ku-band: RF 10.7–11.7 GHz
LO: 10.6 GHz (fixed), IF: 100–1100 MHz (L-band output)
LNA gain: 55 dB, NF: 1 dB (integrated in LNB block)
Conversion loss budget: ≤ 10 dB
Image band: 9.6–10.6 GHz (just below LO)
Image power: −100 to −80 dBm typical from satellite interference
Desired signal: −110 dBm at mixer input (after 55 dB LNA gain)

(a) Calculate how much image rejection is needed to prevent the −80 dBm image from exceeding the desired signal level.

(b) Choose between a double-balanced mixer and an image-reject mixer for this Ku-band application. Justify in 3 points.

(c) With a double-balanced mixer: what additional component provides the required image rejection, and what must its selectivity be?

Answer

已知/目標:Given / required:既知/目標:Ku-band LNB:RF 10.7–11.7 GHz、LO 10.6 GHz(fixed)、IF 100–1100 MHz;image band 9.6–10.6 GHz、image 干擾最強 −80 dBm;desired 在 mixer 輸入 −110 dBm(LNA 後)。求 (a) 所需 image rejection、(b) 拓樸選擇、(c) image filter 規格。 A Ku-band LNB: RF 10.7–11.7 GHz, LO 10.6 GHz (fixed), IF 100–1100 MHz; the image band is 9.6–10.6 GHz with the strongest image interference at −80 dBm; the desired signal at the mixer input is −110 dBm (after the LNA). Find (a) the image rejection required, (b) the topology choice, and (c) the image-filter specification.Ku-band LNB:RF 10.7–11.7 GHz、LO 10.6 GHz(固定)、IF 100–1100 MHz;イメージ帯域は 9.6–10.6 GHz で、最も強いイメージ干渉は −80 dBm;ミキサ入力での希望信号は −110 dBm(LNA 後)。(a) 必要なイメージ除去量、(b) トポロジーの選択、(c) イメージフィルタの仕様を求めます。

(a) 公式與來源:(a) Formula and source:(a) 公式と出典:要把 image 壓到不超過 desired 電平,所需 rejection = 兩者的 dB 差: to push the image no higher than the desired level, the rejection required = the difference between them in dB:イメージを希望信号レベル以下に抑えるには、必要な除去量 = 両者の dB 差:

$$\text{Rejection} \geq P_{image} - P_{desired} = -80 - (-110) = \mathbf{30\,dB\;minimum}$$

(這只是「打平」;要 image 產物低於 desired 10 dB,就再加 10 dB margin → 40 dB 目標較穩。)(That merely brings them level; to put the image product 10 dB below the desired signal, add 10 dB of margin → a 40 dB target is safer.)(これは単に同レベルにするだけです;イメージ産物を希望信号より 10 dB 低くするには、さらに 10 dB のマージンを加えます → 40 dB を目標にするほうが安全です。)

(b) 拓樸選擇:double-balanced mixer + 專用 image BPF(b) The topology choice: a double-balanced mixer plus a dedicated image BPF(b) トポロジーの選択:二重平衡ミキサ + 専用イメージ BPF(而非 image-reject mixer),三點理由:(rather than an image-reject mixer), for three reasons:(イメージ除去ミキサではなく)、理由は次の 3 点:

  1. 頻寬 vs 平衡度:Bandwidth vs balance:帯域幅 vs 平衡度:RF band 寬 1 GHz(fractional BW ≈ 9%),image-reject mixer 要在整個頻帶維持 I/Q 振幅/相位平衡 — 由 6.3 題,IRR 30 dB 已需 ±0.5 dB / ~2° 的全頻帶平衡,Ku-band 寬頻下實際只能做到 IRR < 25 dB,連 30 dB 門檻都不穩。 the RF band is 1 GHz wide (a fractional BW ≈ 9%), and an image-reject mixer would have to hold its I/Q amplitude/phase balance across the whole band — by problem 6.3, even 30 dB of IRR needs ±0.5 dB / ~2° of balance across the band, and over a wide Ku-band the achievable IRR is really < 25 dB, not even reliably clearing the 30 dB threshold.RF 帯域は 1 GHz 幅(比帯域 ≈ 9%)であり、イメージ除去ミキサは全帯域にわたって I/Q の振幅/位相平衡を保たなければなりません — 問題 6.3 より、IRR 30 dB でも全帯域で ±0.5 dB / ~2° の平衡が必要であり、広帯域の Ku-band で実際に達成できる IRR は < 25 dB にとどまり、30 dB のしきい値すら安定して越えられません。
  2. 頻率間隔有利於濾波:The frequency separation favours filtering:周波数間隔が濾波に有利:image band 上緣(10.6 GHz)距 RF band 下緣(10.7 GHz)雖只有 100 MHz,但主要 image 能量在更低處,且 Ku-band 的 cavity / coupled-line filter 百分比頻寬選擇性好,30–35 dB 的近端 rejection 可實作。 although the top of the image band (10.6 GHz) is only 100 MHz below the bottom of the RF band (10.7 GHz), the bulk of the image energy is further down, and Ku-band cavity / coupled-line filters have good percentage-bandwidth selectivity, so 30–35 dB of close-in rejection is realizable.イメージ帯域の上端(10.6 GHz)は RF 帯域の下端(10.7 GHz)からわずか 100 MHz しか離れていませんが、イメージエネルギーの大半はさらに低い側にあり、また Ku-band の cavity / coupled-line フィルタは比帯域あたりの選択度が良いため、30–35 dB の近接除去は実現可能です。
  3. 線性度:Linearity:直線性:55 dB LNA 增益直接打進 mixer,IIP3 壓力大 — double-balanced 的 IIP3 與 harmonic 抑制遠優於 IRM 中兩顆需精確匹配的小 mixer。 55 dB of LNA gain drives straight into the mixer, putting heavy pressure on IIP3 — a double-balanced mixer's IIP3 and harmonic suppression are far better than those of the two precisely matched small mixers inside an IRM.55 dB の LNA 利得がそのままミキサに入るため、IIP3 への負担が大きくなります — 二重平衡ミキサの IIP3 と高調波抑圧は、IRM 内部で精密な整合を要する 2 個の小型ミキサよりはるかに優れています。

(c) Image filter 規格(c) Image-filter specification(c) イメージフィルタの仕様(microstrip 或 waveguide BPF):(a microstrip or waveguide BPF):(マイクロストリップまたは導波管 BPF):

  • 中心:$f_c = (10.7+11.7)/2 = 11.2\,\text{GHz}$;passband 10.7–11.7 GHz(插損 ≤ 1 dB 以免吃掉 NF)Centre: $f_c = (10.7+11.7)/2 = 11.2\,\text{GHz}$; passband 10.7–11.7 GHz (insertion loss ≤ 1 dB so as not to eat into the NF)中心:$f_c = (10.7+11.7)/2 = 11.2\,\text{GHz}$;通過帯域 10.7–11.7 GHz(NF を損なわないよう挿入損失 ≤ 1 dB)
  • stopband:≤ 10.6 GHz 處 ≥ 35 dB(30 dB 需求 + 5 dB guard)Stopband: ≥ 35 dB at and below 10.6 GHz (the 30 dB requirement + 5 dB of guard)阻止帯域:10.6 GHz 以下で ≥ 35 dB(30 dB の要求 + 5 dB のガード)
  • 選擇性需求:從 10.7(通)到 10.6(−35 dB)只有 0.9% 的過渡帶 → 需 5-pole coupled-resonator 設計(LNB 業界標準作法)Selectivity required: from 10.7 (pass) to 10.6 (−35 dB) is a transition band of only 0.9% → a 5-pole coupled-resonator design (standard practice in the LNB industry)必要な選択度:10.7(通過)から 10.6(−35 dB)まで遷移帯域はわずか 0.9% → 5-pole coupled-resonator 設計が必要(LNB 業界の標準的な手法)
DBS LNB design: 5-pole Ku-band BPF for image rejection + double-balanced HEMT mixer for conversion.

物理詮釋:Physical interpretation:物理的解釈:這題示範拓樸選擇的負面教材思維:image-reject mixer 不是「更高級就更好」— 它的 IRR 是 This problem illustrates the cautionary side of topology selection: an image-reject mixer is not “better because it is more sophisticated” — its IRR is a この問題はトポロジー選択の反面教師的な考え方を示します:イメージ除去ミキサは「より高度だから優れている」わけではありません — その IRR は頻寬的函數function of bandwidth帯域幅の関数,寬頻系統會把它的優勢磨光。當 image 與訊號間隔「夠遠可以濾」時,濾波器 + 簡單拓樸永遠是更便宜更穩的解;IRM 留給「近到濾不掉」的場合(8.1 題第 3 例)。, and a wideband system grinds its advantage away. When the image and the signal are “far enough apart to be filtered”, a filter plus a simple topology is always the cheaper and more robust solution; IRMs are reserved for cases “too close to filter” (the third example of problem 8.1).であり、広帯域システムはその優位性を削り取ってしまいます。イメージと信号が「濾波できるほど十分に離れている」場合、フィルタ + 単純なトポロジーが常により安価で堅牢な解です;IRM は「近すぎて濾波できない」場合(問題 8.1 の 3 番目の例)のために取っておきます。

Topic 9Cross-Section Synthesis

Each problem deliberately chains material from several topics — frequency planning, image rejection, NF/IIP3 budgets, spur charts, balanced topologies, and LO spectral purity — the way a real design review does.

9.1 Hard
5.8 GHz receiver — frequency plan, image budget, NF, sensitivity
Synthesis of Topic 1 (frequency plan), Topic 6 (image rejection), Topic 7 (Friis & sensitivity).
System 5.8 GHz ISM-band OFDM receiver: $f_{RF} = 5.8\,\text{GHz}$, $f_{IF} = 200\,\text{MHz}$, low-side LO injection
Preselector: 30 dB attenuation at the image frequency
Worst-case image-band blocker: −50 dBm at the antenna; weakest desired signal: −80 dBm; image products must end up ≥ 12 dB below the desired signal
Chain: LNA (G = 16 dB, NF = 2 dB) → image-reject mixer ($L_C$ = 6.5 dB, $NF_{SSB}$ = 7 dB) → IF amp (G = 25 dB, NF = 3.5 dB)
Channel bandwidth B = 20 MHz, required SNR = 18 dB, sensitivity spec ≤ −80 dBm

(a) Find $f_{LO}$ and the image frequency. Where does the image sit relative to the RX band?

(b) How much total image rejection is required, and how much must the image-reject mixer contribute beyond the preselector? Adding 8 dB of margin, translate the mixer IRR requirement into an amplitude/phase imbalance budget $(\varepsilon,\ \Delta\phi)$ using $IRR \approx 1/(\varepsilon^2+\Delta\phi^2)$.

(c) Compute the cascaded system NF (Friis).

(d) Compute the sensitivity and the margin against the −80 dBm spec. Which stage dominates the noise budget?

Answer

已知/目標:Given / required:既知/目標:$f_{RF} = 5.8\,\text{GHz}$、$f_{IF} = 200\,\text{MHz}$、low-side LO;preselector 在 image 處 30 dB;image blocker −50 dBm、最弱訊號 −80 dBm、image 產物需低於訊號 ≥ 12 dB;鏈路 LNA(16 dB / 2 dB)→ IRM($L_C$ 6.5 dB / $NF_{SSB}$ 7 dB)→ IF amp(25 dB / 3.5 dB);$B = 20\,\text{MHz}$、SNR 18 dB、sensitivity ≤ −80 dBm。$f_{RF} = 5.8\,\text{GHz}$, $f_{IF} = 200\,\text{MHz}$, low-side LO; the preselector gives 30 dB at the image; the image blocker is −50 dBm, the weakest signal −80 dBm, and the image product must be ≥ 12 dB below the signal; the chain is LNA (16 dB / 2 dB) → IRM ($L_C$ 6.5 dB / $NF_{SSB}$ 7 dB) → IF amp (25 dB / 3.5 dB); $B = 20\,\text{MHz}$, SNR 18 dB, sensitivity ≤ −80 dBm.$f_{RF} = 5.8\,\text{GHz}$、$f_{IF} = 200\,\text{MHz}$、low-side LO;プリセレクタはイメージ点で 30 dB;イメージブロッカーは −50 dBm、最弱信号は −80 dBm、イメージ産物は信号より ≥ 12 dB 低い必要があります;チェーンは LNA(16 dB / 2 dB)→ IRM($L_C$ 6.5 dB / $NF_{SSB}$ 7 dB)→ IF 増幅器(25 dB / 3.5 dB);$B = 20\,\text{MHz}$、SNR 18 dB、感度 ≤ −80 dBm。

(a) 公式:(a) Formula:(a) 公式:low-side injection:$f_{LO} = f_{RF} - f_{IF}$、$f_{im} = f_{LO} - f_{IF} = f_{RF} - 2f_{IF}$。 low-side injection: $f_{LO} = f_{RF} - f_{IF}$ and $f_{im} = f_{LO} - f_{IF} = f_{RF} - 2f_{IF}$.low-side injection:$f_{LO} = f_{RF} - f_{IF}$、$f_{im} = f_{LO} - f_{IF} = f_{RF} - 2f_{IF}$。代入:Substituting:代入:

$$f_{LO} = 5.8 - 0.2 = \mathbf{5.6\,GHz}, \qquad f_{im} = 5.6 - 0.2 = \mathbf{5.4\,GHz}$$

image 在 RX band 下方 $2f_{IF} = 400\,\text{MHz}$ 處,與訊號分居 LO 兩側。The image sits $2f_{IF} = 400\,\text{MHz}$ below the RX band, on the opposite side of the LO from the signal.イメージは RX 帯域の $2f_{IF} = 400\,\text{MHz}$ 下方にあり、信号とは LO を挟んで反対側に位置します。

(b) 逐步代入:(b) Substituting step by step:(b) 段階的に代入:image 產物的等效輸入電平需 ≤ $-80 - 12 = -92\,\text{dBm}$,blocker 是 −50 dBm: The image product's equivalent input level must be ≤ $-80 - 12 = -92\,\text{dBm}$, and the blocker is −50 dBm:イメージ産物の等価入力レベルは ≤ $-80 - 12 = -92\,\text{dBm}$ である必要があり、ブロッカーは −50 dBm です:

Required total image rejection = −50 − (−92) = 42 dB

preselector 出 30 dB → mixer 至少補 $42 - 30 = 12\,\text{dB}$;加 8 dB margin → 設計目標 $IRR = 20\,\text{dB}$。轉成 imbalance budget($IRR \approx 1/(\varepsilon^2+\Delta\phi^2)$):The preselector supplies 30 dB → the mixer must make up at least $42 - 30 = 12\,\text{dB}$; adding 8 dB of margin → a design target of $IRR = 20\,\text{dB}$. Converting that into an imbalance budget ($IRR \approx 1/(\varepsilon^2+\Delta\phi^2)$):プリセレクタが 30 dB を提供 → ミキサは少なくとも $42 - 30 = 12\,\text{dB}$ を補う必要があります;8 dB のマージンを加えて → 設計目標は $IRR = 20\,\text{dB}$。これを imbalance budget($IRR \approx 1/(\varepsilon^2+\Delta\phi^2)$)に換算すると:

$$\varepsilon^2 + \Delta\phi^2 \leq 10^{-20/10} = 10^{-2}$$

RSS 平均分配:$\varepsilon = \Delta\phi = \sqrt{0.01/2} = 0.071$ → 相位平衡 $0.071\,\text{rad} \approx 4.0°$、振幅平衡 $20\log_{10}(1.071) \approx 0.6\,\text{dB}$ — 很寬鬆的目標,不需校正即可達成(對照 6.3 題:40 dB IRR 需 ≈1° 與 ±0.1 dB)。Splitting it evenly in RSS: $\varepsilon = \Delta\phi = \sqrt{0.01/2} = 0.071$ → a phase balance of $0.071\,\text{rad} \approx 4.0°$ and an amplitude balance of $20\log_{10}(1.071) \approx 0.6\,\text{dB}$ — a very relaxed target, achievable without any calibration (compare problem 6.3: 40 dB IRR needs ≈1° and ±0.1 dB).RSS で均等に分配:$\varepsilon = \Delta\phi = \sqrt{0.01/2} = 0.071$ → 位相平衡 $0.071\,\text{rad} \approx 4.0°$、振幅平衡 $20\log_{10}(1.071) \approx 0.6\,\text{dB}$ — 非常に緩い目標であり、校正なしでも達成できます(問題 6.3 と比較:40 dB の IRR には ≈1° と ±0.1 dB が必要)。

(c) Friis — dB 轉 linear:(c) Friis — converting dB to linear:(c) Friis — dB を linear に換算:$F_1 = 10^{0.2} = 1.585$、$G_1 = 10^{1.6} = 39.8$;$F_2 = 10^{0.7} = 5.01$、$G_2 = 10^{-0.65} = 0.224$;$F_3 = 10^{0.35} = 2.24$:$F_1 = 10^{0.2} = 1.585$, $G_1 = 10^{1.6} = 39.8$; $F_2 = 10^{0.7} = 5.01$, $G_2 = 10^{-0.65} = 0.224$; $F_3 = 10^{0.35} = 2.24$:$F_1 = 10^{0.2} = 1.585$、$G_1 = 10^{1.6} = 39.8$;$F_2 = 10^{0.7} = 5.01$、$G_2 = 10^{-0.65} = 0.224$;$F_3 = 10^{0.35} = 2.24$:

$$F_{sys} = 1.585 + \dfrac{5.01-1}{39.8} + \dfrac{2.24-1}{39.8\times 0.224} = 1.585 + 0.101 + 0.139 = 1.825$$

NF_sys = 10·log(1.825) = 2.6 dB

(d) 代入 sensitivity 公式:(d) Substituting into the sensitivity formula:(d) 感度の式に代入:

$$kTB = -174 + 10\log_{10}(20\times 10^6) = -174 + 73 = -101\,\text{dBm}$$

$$P_{sens} = kTB + NF_{sys} + SNR_{req} = -101 + 2.6 + 18 = \mathbf{-80.4\,dBm}$$

過 −80 dBm spec,但只有 It passes the −80 dBm spec, but with only −80 dBm の spec は満たしますが、わずか 0.4 dB margin。雜訊由 LNA 主宰:$F_1 = 1.585$ 佔 $F_{sys} = 1.825$ 的 ≈ 87% — 要買 margin 先改 LNA:NF = 1.2 dB / G = 18 dB 時 $F_{sys} \approx 1.318 + 0.064 + 0.088 \approx 1.47 \to NF_{sys} \approx 1.7\,\text{dB}$,多出 ≈ 1 dB margin。. The noise is dominated by the LNA: $F_1 = 1.585$ accounts for ≈ 87% of $F_{sys} = 1.825$ — so buying margin means changing the LNA first: with NF = 1.2 dB and G = 18 dB, $F_{sys} \approx 1.318 + 0.064 + 0.088 \approx 1.47 \to NF_{sys} \approx 1.7\,\text{dB}$, gaining ≈ 1 dB more margin.。雑音は LNA が支配します:$F_1 = 1.585$ が $F_{sys} = 1.825$ の ≈ 87% を占めます — マージンを買うにはまず LNA を変更します:NF = 1.2 dB / G = 18 dB のとき $F_{sys} \approx 1.318 + 0.064 + 0.088 \approx 1.47 \to NF_{sys} \approx 1.7\,\text{dB}$ となり、≈ 1 dB のマージンが増えます。

物理詮釋:Physical interpretation:物理的解釈:本題演示完整的 spec flow-down:frequency plan 決定 image 位置 → image budget 拆給 preselector 與 IRM(濾波器吃大頭、IRM 補尾巴,所以 IRM 只需 20 dB 的寬鬆平衡)→ Friis 驗證雜訊 → sensitivity 收斂。0.4 dB margin 太薄是常見的審查紅旗:溫度、老化、批次變異輕易吃掉 1–2 dB,設計審查會要求至少 2–3 dB。 This problem demonstrates a complete spec flow-down: the frequency plan fixes where the image lands → the image budget is divided between the preselector and the IRM (the filter takes the lion's share and the IRM mops up, so the IRM needs only a relaxed 20 dB of balance) → Friis verifies the noise → the sensitivity converges. A margin as thin as 0.4 dB is a common review red flag: temperature, ageing and lot-to-lot variation easily eat 1–2 dB, and a design review would demand at least 2–3 dB.この問題は完全な spec flow-down を示します:frequency plan がイメージの落ちる位置を決める → イメージ予算をプリセレクタと IRM に分配する(フィルタが大半を担い IRM が残りを処理するので、IRM には 20 dB という緩い平衡しか要りません)→ Friis で雑音を検証する → 感度が収束する。0.4 dB という薄いマージンは審査でよくある赤信号です:温度、経年変化、ロット間ばらつきが 1–2 dB を簡単に食ってしまうため、設計審査では少なくとも 2–3 dB が要求されます。

9.2 Hard
Half-IF spur, topology suppression, and LO leakage — one frequency plan, three checks
Synthesis of Topic 2 (spur chart), Topics 4–5 (balanced suppression), Topic 8 (selection).
Frequency plan Downconverter: $f_{RF} = 900\,\text{MHz}$, $f_{LO} = 800\,\text{MHz}$ (low-side), $f_{IF} = 100\,\text{MHz}$, IF passband 90–110 MHz
Mixer-first front end: the mixer RF port sees the antenna through a preselector (10 dB attenuation at out-of-band frequencies, including $f_{LO}$)
LO drive +7 dBm; regulatory LO emission limit at the antenna: −30 dBm
Candidates: single-ended (LO–RF isolation ≈ 15 dB, no spur suppression) · single-balanced (≈ 25 dB, even-$n$ LO spurs suppressed ≈ 30 dB) · double-balanced ring (≈ 45 dB, even-$m$/even-$n$ spurs suppressed ≈ 65 dB)

(a) Which interferer frequencies $f_x$ between 700 and 900 MHz can reach the IF passband through spurious products $|m f_x - n f_{LO}|$ with $m, n \leq 3$? Identify the $(m,n)$ pair of each.

(b) For each candidate topology, state which of those spur paths are suppressed and why.

(c) A −30 dBm interferer sits at 850 MHz and the preselector attenuates it by 10 dB. With the double-balanced ring's 65 dB (2,2) rejection, what is its input-referred level at IF, and what is the margin below a −85 dBm desired signal?

(d) Check the LO emission limit for all three topologies. Which topology survives all three checks?

Answer

已知/目標:Given / required:既知/目標:$f_{LO} = 800\,\text{MHz}$、IF passband 90–110 MHz;mixer-first 前端,preselector 帶外 10 dB;LO drive +7 dBm、天線端 LO 洩漏限制 −30 dBm。求 700–900 MHz 內的 spur 路徑、三種拓樸的抑制情況、(2,2) spur 電平、洩漏合規性。$f_{LO} = 800\,\text{MHz}$ with an IF passband of 90–110 MHz; a mixer-first front end with 10 dB of out-of-band preselector rejection; an LO drive of +7 dBm and an antenna LO-leakage limit of −30 dBm. Find the spur paths within 700–900 MHz, the suppression offered by the three topologies, the (2,2) spur level, and whether the leakage complies.$f_{LO} = 800\,\text{MHz}$、IF 通過帯域 90–110 MHz;mixer-first フロントエンドで、プリセレクタの帯域外除去は 10 dB;LO ドライブ +7 dBm、アンテナ端の LO 漏れ制限は −30 dBm。700–900 MHz 内の spur 経路、3 種類のトポロジーによる抑圧、(2,2) spur のレベル、漏れが規格に適合するかを求めます。

(a) 公式:(a) Formula:(a) 公式:spur 進入 IF 的條件 $|m f_x - n\cdot 800| \in [90, 110]\,\text{MHz}$。 The condition for a spur to enter the IF is $|m f_x - n\cdot 800| \in [90, 110]\,\text{MHz}$.spur が IF に入る条件は $|m f_x - n\cdot 800| \in [90, 110]\,\text{MHz}$。逐一解出 $f_x$(限 700–900 MHz):Solving for $f_x$ one case at a time (restricted to 700–900 MHz):$f_x$ を場合ごとに解きます(700–900 MHz に限定):

  • $(m,n) = (1,1)$:$800 - f_x = 100 \Rightarrow f_x = \mathbf{700\,MHz}$ → 這就是 $(m,n) = (1,1)$: $800 - f_x = 100 \Rightarrow f_x = \mathbf{700\,MHz}$ → this is the $(m,n) = (1,1)$:$800 - f_x = 100 \Rightarrow f_x = \mathbf{700\,MHz}$ → これが image.
  • $(m,n) = (2,2)$:$2f_x - 1600 = 100 \Rightarrow f_x = \mathbf{850\,MHz}$ → $(m,n) = (2,2)$: $2f_x - 1600 = 100 \Rightarrow f_x = \mathbf{850\,MHz}$ → $(m,n) = (2,2)$:$2f_x - 1600 = 100 \Rightarrow f_x = \mathbf{850\,MHz}$ → half-IF spur,恰在 $(f_{RF}+f_{LO})/2$。, exactly at $(f_{RF}+f_{LO})/2$.、ちょうど $(f_{RF}+f_{LO})/2$ にあります。
  • $(m,n) = (3,3)$:$3f_x - 2400 = 100 \Rightarrow f_x = \mathbf{833.3\,MHz}$ → half-IF 的三階類比。$(m,n) = (3,3)$: $3f_x - 2400 = 100 \Rightarrow f_x = \mathbf{833.3\,MHz}$ → the third-order analogue of the half-IF.$(m,n) = (3,3)$:$3f_x - 2400 = 100 \Rightarrow f_x = \mathbf{833.3\,MHz}$ → half-IF の三次版に相当します。

(其餘 $m,n \leq 3$ 組合解出的 $f_x$ 都落在 700–900 MHz 之外,例如 (2,1) 給 445–455 MHz、(1,2) 給 1490–1510 MHz。)(Every other combination with $m,n \leq 3$ solves to an $f_x$ outside 700–900 MHz; for example (2,1) gives 445–455 MHz and (1,2) gives 1490–1510 MHz.)(その他の $m,n \leq 3$ の組み合わせから解かれる $f_x$ はすべて 700–900 MHz の外に落ちます。例えば (2,1) では 445–455 MHz、(1,2) では 1490–1510 MHz となります。)

(b) 對照 8.2 題的拓樸抑制表:(b) Against the topology suppression table of problem 8.2:(b) 問題 8.2 のトポロジー抑圧表と照合:

  • Single-ended:Single-ended:Single-ended:無任何抵消 — 三條 spur 路徑全部以原生轉換效率存活。 no cancellation of any kind — all three spur paths survive at their native conversion efficiency.打ち消しは一切なし — 3 本の spur 経路がすべて本来の変換効率のまま残ります。
  • Single-balanced:Single-balanced:Single-balanced:抑制 even- suppresses even-even-LO-harmonic 產物 → (2,2) half-IF spur 壓 ≈ 30 dB;(3,3)(odd×odd)與 (1,1) image 存活。-harmonic products → the (2,2) half-IF spur is pushed down ≈ 30 dB; (3,3) (odd×odd) and the (1,1) image survive.-harmonic 産物を抑圧します → (2,2) half-IF spur は ≈ 30 dB 押し下げられます;(3,3)(odd×odd)と (1,1) イメージは残ります。
  • Double-balanced ring:Double-balanced ring:Double-balanced ring:只有 odd-$m$ × odd-$n$ 產物能轉換 → (2,2) 被 RF、LO 兩層 balance 同時抑制(≈ 65 dB)。(3,3) 是 odd×odd,balance 救不了 — 只能靠 preselector 裙邊與元件線性度壓制。(1,1) image 更要注意:它就是 desired 的同階產物、只是頻率不同 — only odd-$m$ × odd-$n$ products can convert → (2,2) is suppressed by both the RF and LO layers of balance at once (≈ 65 dB). (3,3) is odd×odd and balance cannot save it — only the preselector skirt and device linearity can push it down. The (1,1) image deserves particular care: it is the same order of product as the desired signal, differing only in frequency — odd-$m$ × odd-$n$ の産物のみが変換できます → (2,2) は RF と LO の両方の平衡層によって同時に抑圧されます(≈ 65 dB)。(3,3) は odd×odd であり平衡では救えません — プリセレクタのスカートと素子の直線性でしか押し下げられません。(1,1) イメージにはとくに注意が必要です:希望信号と同じ次数の産物であり、違うのは周波数だけです — 任何Anyいかなる balance 都不抑制它,必須靠 preselector(此處距離 200 MHz → 容易濾)或 image-reject 架構(Topic 6)。 no amount of balance suppresses it, and it must be handled by the preselector (200 MHz away here → easy to filter) or an image-reject architecture (Topic 6). balance でもこれを抑圧できず、プリセレクタ(ここでは 200 MHz 離れている → 濾波は容易)またはイメージ除去アーキテクチャ(Topic 6)で対処するしかありません。

(c) 逐步代入:(c) Substituting step by step:(c) 段階的に代入:850 MHz 干擾 −30 dBm,經 preselector −10 dB、再經 ring 的 (2,2) 抑制 −65 dB: The 850 MHz interferer at −30 dBm, after −10 dB from the preselector and a further −65 dB of (2,2) suppression from the ring:850 MHz の干渉波は −30 dBm、プリセレクタによる −10 dB と、リングの (2,2) 抑圧によるさらに −65 dB を経て:

$$P_{spur,in} = -30 - 10 - 65 = \mathbf{-105\,dBm}$$

Margin below the −85 dBm desired signal = −85 − (−105) = 20 dB ✓

(d) LO 洩漏檢查(d) LO leakage check(d) LO 漏れの検査 — 天線端電平 $= P_{LO} - \text{isolation} - 10\,\text{dB(preselector @ 800 MHz)}$,逐一代入: — the level at the antenna $= P_{LO} - \text{isolation} - 10\,\text{dB (the preselector at 800 MHz)}$, case by case: — アンテナ端のレベル $= P_{LO} - \text{isolation} - 10\,\text{dB (the preselector at 800 MHz)}$ を、場合ごとに代入します:

  • Single-ended:$7 - 15 - 10 = -18\,\text{dBm}$ → Single-ended: $7 - 15 - 10 = -18\,\text{dBm}$ → Single-ended:$7 - 15 - 10 = -18\,\text{dBm}$ → fails(超限 12 dB)。(12 dB over the limit).(制限を 12 dB 超過)。
  • Single-balanced:$7 - 25 - 10 = -28\,\text{dBm}$ → Single-balanced: $7 - 25 - 10 = -28\,\text{dBm}$ → Single-balanced:$7 - 25 - 10 = -28\,\text{dBm}$ → fails by 2 dB(−28 > −30)。(−28 > −30).(−28 > −30)。
  • Double-balanced:$7 - 45 - 10 = -48\,\text{dBm}$ → 過關,18 dB margin ✓Double-balanced: $7 - 45 - 10 = -48\,\text{dBm}$ → passes, with 18 dB of margin ✓Double-balanced:$7 - 45 - 10 = -48\,\text{dBm}$ → 合格、18 dB のマージン ✓
Only the double-balanced ring passes spur suppression AND emission simultaneously; the image and the (3,3) spur are then closed out by the preselector.

物理詮釋:Physical interpretation:物理的解釈:三個檢查各考一種機制:image 考「頻率規劃 + 濾波」、half-IF 考「拓樸對稱性」、LO 洩漏考「isolation 預算」。mixer-first 架構(無 LNA 反向隔離可借)把所有壓力堆在 mixer 的 isolation 上,因此 DB ring 幾乎是唯一解 — 這也解釋了為何商用 mixer-first SDR 前端清一色用高 isolation 的 ring 或 switch mixer。 Each of the three checks tests a different mechanism: the image tests “frequency planning + filtering”, the half-IF tests “topological symmetry”, and the LO leakage tests the “isolation budget”. A mixer-first architecture (with no LNA reverse isolation to borrow) piles all the pressure onto the mixer's isolation, so a DB ring is virtually the only answer — which is also why commercial mixer-first SDR front ends universally use high-isolation ring or switching mixers.3 つの検査はそれぞれ異なるメカニズムを試します:イメージは「周波数計画 + 濾波」、half-IF は「トポロジーの対称性」、LO 漏れは「アイソレーション予算」を問います。mixer-first アーキテクチャ(借りられる LNA の逆方向アイソレーションがない)はすべての負担をミキサのアイソレーションに集中させるため、DB ring がほぼ唯一の答えになります — これは商用の mixer-first SDR フロントエンドが一様に高アイソレーションのリングミキサやスイッチングミキサを使う理由でもあります。

9.3 Medium
LO spectral purity — FM sidebands ride through the mixer
Bessel appendix + Topic 1 + dB cheat sheet: $\beta = \Delta f/f_m$, sideband ≈ $\beta/2$; mixing convolves the LO spectrum onto the IF.
Setup LO: 800 MHz, +7 dBm, with residual FM (e.g. supply ripple on the VCO): deviation $\Delta f = 2\,\text{kHz}$ at rate $f_m = 100\,\text{kHz}$
Downconverted IF carrier measured at −7 dBm
Spec: any spur at ±100 kHz from the IF carrier ≤ −50 dBc

(a) Compute the modulation index $\beta$ and the first FM sideband level in dBc (narrowband FM).

(b) Explain why those LO sidebands reappear at ±100 kHz around the IF carrier at the same dBc level — and why no IF filter can remove them.

(c) Does the spec pass? If not, find the maximum tolerable deviation $\Delta f$.

(d) Express the sideband's absolute power in dBm and nW (dB cheat-sheet drill). Bonus: at what $\beta$ would the carrier itself vanish?

Answer

已知/目標:Given / required:既知/目標:LO 800 MHz / +7 dBm,殘餘 FM:$\Delta f = 2\,\text{kHz}$、$f_m = 100\,\text{kHz}$;IF carrier 量得 −7 dBm;spec:IF carrier ±100 kHz 處 spur ≤ −50 dBc。求 sideband 電平、解釋它如何穿過 mixer、合規檢查與允許的 $\Delta f$。 An LO at 800 MHz / +7 dBm with residual FM of $\Delta f = 2\,\text{kHz}$ at $f_m = 100\,\text{kHz}$; the IF carrier measures −7 dBm; the spec is spurs ≤ −50 dBc at ±100 kHz from the IF carrier. Find the sideband level, explain how it passes through the mixer, check compliance, and find the permissible $\Delta f$.LO は 800 MHz / +7 dBm、残留 FM は $\Delta f = 2\,\text{kHz}$、$f_m = 100\,\text{kHz}$;IF キャリアの測定値は −7 dBm;spec は IF キャリアから ±100 kHz で spur ≤ −50 dBc。側波帯のレベル、それがミキサをどのように通り抜けるかの説明、適合性の確認、および許容される $\Delta f$ を求めます。

(a) 公式與來源:(a) Formula and source:(a) 公式と出典:FM 調變指數與 narrowband FM($\beta \ll 1$)的一階 Bessel 近似: The FM modulation index and the first-order Bessel approximation for narrowband FM ($\beta \ll 1$):FM 変調指数と、narrowband FM($\beta \ll 1$)に対する一次ベッセル近似:

$$\beta = \frac{\Delta f}{f_m}, \qquad \frac{J_1(\beta)}{J_0(\beta)} \approx \frac{\beta}{2}$$

逐步代入:Substituting step by step:段階的に代入:$\beta = \dfrac{2\,\text{kHz}}{100\,\text{kHz}} = 0.02 \ll 1$ ✓ → 第一對 sideband 相對 carrier 的振幅比 $\approx \beta/2 = 0.01$:$\beta = \dfrac{2\,\text{kHz}}{100\,\text{kHz}} = 0.02 \ll 1$ ✓ → the amplitude of the first sideband pair relative to the carrier is $\approx \beta/2 = 0.01$:$\beta = \dfrac{2\,\text{kHz}}{100\,\text{kHz}} = 0.02 \ll 1$ ✓ → 第 1 側波帯対のキャリアに対する振幅比は $\approx \beta/2 = 0.01$:

Sideband level = 20·log10(0.01) = −40 dBc at ±100 kHz

(b) 時域相乘 = Multiplication in time = 時間領域での乗算 = 頻域卷積(convolution)convolution in frequency周波数領域での畳み込み(convolution)。LO 頻譜 = carrier + ±100 kHz 的 −40 dBc sidebands;卷積時每一條 LO 譜線都把 RF 各自搬移一份 → IF carrier 帶著一模一樣的 −40 dBc / ±100 kHz sidebands。它們是 angle-modulation sidebands,. The LO spectrum = a carrier plus −40 dBc sidebands at ±100 kHz; on convolution every LO spectral line translates its own copy of the RF → so the IF carrier carries exactly the same −40 dBc / ±100 kHz sidebands. They are angle-modulation sidebands that 。LO スペクトル = キャリア + ±100 kHz の −40 dBc 側波帯;畳み込みでは LO の各スペクトル線がそれぞれ RF を 1 部ずつ移動させます → IF キャリアはまったく同じ −40 dBc / ±100 kHz の側波帯を伴います。これらは角度変調の側波帯であり、跟著 carrier 一起移動move with the carrierキャリアとともに移動し、永遠貼在通道內或鄰道 — 以 carrier 為中心的 IF filter 無法分離。唯一解法是清理 LO 本身(VCO 供電濾波、PLL 頻寬設計)。 and stay permanently in-channel or in the adjacent channel — an IF filter centred on the carrier cannot separate them. The only remedy is to clean up the LO itself (filtering the VCO supply, designing the PLL bandwidth).、常にチャネル内または隣接チャネルに貼り付いたままです — キャリアを中心とする IF フィルタでは分離できません。唯一の解決策は LO 自体をきれいにすることです(VCO 電源の濾波、PLL 帯域幅の設計)。

(c) 合規檢查:(c) Compliance check:(c) 適合性の確認:−40 dBc > −50 dBc → fail 10 dB。反推允許值:. Working back to the permitted value:。許容値を逆算すると:

$$\frac{\beta}{2} \leq 10^{-50/20} = 3.16\times 10^{-3} \;\Rightarrow\; \beta \leq 6.3\times 10^{-3}$$

Δf ≤ β·f_m = 0.0063 × 100 kHz ≈ 630 Hz (deviation must drop ≈ 3.2×, i.e. −10 dB in sideband level)

(d) dB 換算練習:(d) A dB conversion exercise:(d) dB 換算の練習:sideband 絕對功率 $= P_{IF} + (\text{dBc}) = -7 - 40 = -47\,\text{dBm}$。心算:$-47 = -50 + 3$ → $10^{-5}\,\text{mW}\times 2 = 2\times 10^{-5}\,\text{mW} = \mathbf{20\,nW}$。 The absolute sideband power $= P_{IF} + (\text{dBc}) = -7 - 40 = -47\,\text{dBm}$. Mentally: $-47 = -50 + 3$ → $10^{-5}\,\text{mW}\times 2 = 2\times 10^{-5}\,\text{mW} = \mathbf{20\,nW}$.側波帯の絶対電力 $= P_{IF} + (\text{dBc}) = -7 - 40 = -47\,\text{dBm}$。暗算では:$-47 = -50 + 3$ → $10^{-5}\,\text{mW}\times 2 = 2\times 10^{-5}\,\text{mW} = \mathbf{20\,nW}$。

Bonus:carrier 消失於 $J_0(\beta) = 0$ 的第一個根 $\beta = 2.405$ → $\Delta f = 2.405\times 100\,\text{kHz} = 240.5\,\text{kHz}$ — 經典的 Bonus: the carrier vanishes at the first root of $J_0(\beta) = 0$, $\beta = 2.405$ → $\Delta f = 2.405\times 100\,\text{kHz} = 240.5\,\text{kHz}$ — the classic Bonus:キャリアは $J_0(\beta) = 0$ の最初の根 $\beta = 2.405$ → $\Delta f = 2.405\times 100\,\text{kHz} = 240.5\,\text{kHz}$ で消失します — 古典的な Bessel-null 法,用頻譜儀看 carrier 歸零來校正 FM deviation。 method, calibrating FM deviation by watching the carrier null on a spectrum analyser. 法で、スペクトラムアナライザでキャリアのヌルを観察して FM の周波数偏移を校正します。

物理詮釋:Physical interpretation:物理的解釈:這題是「LO 品質 = IF 品質」的核心教訓:mixer 把 LO 的全部頻譜雜質(FM/PM sidebands、phase noise)原封不動「蓋章」到每一個轉換後的訊號上,且以 dBc 計完全不打折。接收機的 blocker 場景更糟 — 強鄰道訊號與 LO sidebands 卷積會把雜訊倒進通道(reciprocal mixing),所以 LO 純度規格通常由系統最嚴的 blocking 情境反推出來。 This problem carries the central lesson that “LO quality = IF quality”: a mixer stamps every spectral impurity of the LO (FM/PM sidebands, phase noise) onto every converted signal completely unaltered, with no discount whatsoever in dBc. The blocker scenario in a receiver is worse still — convolving a strong adjacent-channel signal with the LO sidebands pours noise into the channel (reciprocal mixing), so the LO purity specification is usually derived backwards from the system's most severe blocking scenario.この問題は「LO の品質 = IF の品質」という中心的な教訓を伝えます:ミキサは LO のあらゆるスペクトル不純物(FM/PM 側波帯、位相雑音)を、変換後のすべての信号にそのまま「押印」し、dBc で見て一切割り引かれません。受信機のブロッカー場面はさらに悪く — 強い隣接チャネル信号と LO 側波帯の畳み込みがチャネル内へ雑音を注ぎ込みます(reciprocal mixing)。そのため LO の純度仕様は通常、システムで最も厳しいブロッキング場面から逆算して決められます。

9.4 Design
PCS downconverter — LO level, linearity, leakage, and spur plan
Synthesis of Topics 1, 2, 5, 7, 8. Diode-ring rules of thumb: $P_{1dB} \approx P_{LO} - 6\,\text{dB}$, $IIP3 \approx P_{1dB} + 10\,\text{dB}$.
Spec RF 1.93–1.99 GHz (PCS RX), $f_{IF} = 190\,\text{MHz}$, high-side LO
Diode double-balanced ring, LO–RF isolation 40 dB; LNA precedes the mixer with 20 dB reverse isolation; antenna emission limit −30 dBm
Two −20 dBm in-band blockers at the mixer input; weakest desired signal at the mixer input −75 dBm; IM3 must stay ≥ 10 dB below it
Available LO drive: level-7 (+7 dBm) or level-13 (+13 dBm)
Preselector: ≥ 25 dB attenuation outside 1.93–1.99 GHz

(a) Give the LO band and image band. Is an image-reject mixer needed here?

(b) For level-7 drive: estimate $P_{1dB}$ and IIP3, compute the two-tone IM3 level, and check the spec.

(c) Repeat for level-13, then check the LO leakage reaching the antenna.

(d) Locate the half-IF (2,2) interferer band and argue why this frequency plan is safe.

Answer

已知/目標:Given / required:既知/目標:RF 1.93–1.99 GHz、$f_{IF} = 190\,\text{MHz}$、high-side LO;DB ring(LO–RF iso 40 dB)、LNA 反向隔離 20 dB、天線洩漏限制 −30 dBm;兩個 −20 dBm blocker、最弱訊號 −75 dBm、IM3 需低 ≥ 10 dB;LO drive 可選 +7 或 +13 dBm;preselector 帶外 ≥ 25 dB。 RF 1.93–1.99 GHz, $f_{IF} = 190\,\text{MHz}$, high-side LO; a DB ring (LO–RF isolation 40 dB), 20 dB of LNA reverse isolation, and an antenna leakage limit of −30 dBm; two −20 dBm blockers, a weakest signal of −75 dBm, and IM3 required to be ≥ 10 dB below it; an LO drive of either +7 or +13 dBm; a preselector giving ≥ 25 dB out of band.RF 1.93–1.99 GHz、$f_{IF} = 190\,\text{MHz}$、high-side LO;DB ring(LO–RF アイソレーション 40 dB)、LNA の逆方向アイソレーション 20 dB、アンテナ漏れ制限 −30 dBm;−20 dBm のブロッカー 2 波、最弱信号 −75 dBm、IM3 はそれより ≥ 10 dB 低いことが必要;LO ドライブは +7 または +13 dBm から選択;プリセレクタは帯域外 ≥ 25 dB。

(a) 公式:(a) Formula:(a) 公式:high-side:$f_{LO} = f_{RF} + f_{IF}$、$f_{im} = f_{LO} + f_{IF} = f_{RF} + 2f_{IF}$。 High-side: $f_{LO} = f_{RF} + f_{IF}$ and $f_{im} = f_{LO} + f_{IF} = f_{RF} + 2f_{IF}$.high-side:$f_{LO} = f_{RF} + f_{IF}$、$f_{im} = f_{LO} + f_{IF} = f_{RF} + 2f_{IF}$。代入:Substituting:代入:

$$f_{LO} = (1.93\text{–}1.99) + 0.19 = \mathbf{2.12\text{–}2.18\,GHz}, \qquad f_{im} = f_{RF} + 0.38 = \mathbf{2.31\text{–}2.37\,GHz}$$

image band 在 RX 上緣之外足足 320 MHz → preselector 的 ≥ 25 dB stopband 直接涵蓋(若 image band 內有強發射源再加 pole)。The image band lies a full 320 MHz above the top of the RX band → the preselector's ≥ 25 dB stopband covers it directly (add a pole if there is a strong transmitter within the image band).イメージ帯域は RX 帯域の上端よりさらに 320 MHz 上にあります → プリセレクタの ≥ 25 dB 阻止帯域が直接カバーします(イメージ帯域内に強い送信源があれば極を 1 つ追加します)。不需 image-reject mixerNo image-reject mixer is neededイメージ除去ミキサは不要 — 那個架構的複雜度只有在「image 近到濾不掉」(low-IF,Topic 6)才划算。 — that architecture's complexity only pays off when the image is “too close to filter” (low IF, Topic 6). — あのアーキテクチャの複雑さは「イメージが近すぎて濾波できない」(low-IF、Topic 6)場合にのみ見合います。

(b) Level-7 — 公式與來源(diode-ring 經驗律):(b) Level-7 — formula and source (the diode-ring rule of thumb):(b) Level-7 — 公式と出典(diode-ring の経験則):$P_{1dB} \approx P_{LO} - 6\,\text{dB}$、$IIP3 \approx P_{1dB} + 10\,\text{dB}$。$P_{1dB} \approx P_{LO} - 6\,\text{dB}$ and $IIP3 \approx P_{1dB} + 10\,\text{dB}$.$P_{1dB} \approx P_{LO} - 6\,\text{dB}$、$IIP3 \approx P_{1dB} + 10\,\text{dB}$。逐步代入:Substituting step by step:段階的に代入:

$$P_{1dB} \approx 7 - 6 = +1\,\text{dBm}, \qquad IIP3 \approx 1 + 10 = +11\,\text{dBm}$$

兩音 IM3(input-referred):$P_{IM3} = 3P_{in} - 2\,IIP3 = 3\times(-20) - 2\times 11 = -60 - 22 = \mathbf{-82\,dBm}$Two-tone IM3 (input-referred): $P_{IM3} = 3P_{in} - 2\,IIP3 = 3\times(-20) - 2\times 11 = -60 - 22 = \mathbf{-82\,dBm}$2 音 IM3(input-referred):$P_{IM3} = 3P_{in} - 2\,IIP3 = 3\times(-20) - 2\times 11 = -60 - 22 = \mathbf{-82\,dBm}$

需求:$\leq -75 - 10 = -85\,\text{dBm}$ → $-82 > -85$ → The requirement: $\leq -75 - 10 = -85\,\text{dBm}$ → $-82 > -85$ → 要求:$\leq -75 - 10 = -85\,\text{dBm}$ → $-82 > -85$ → fails by 3 dB.

(c) Level-13:(c) Level-13:(c) Level-13:$P_{1dB} \approx 13 - 6 = +7\,\text{dBm}$、$IIP3 \approx +17\,\text{dBm}$:$P_{1dB} \approx 13 - 6 = +7\,\text{dBm}$ and $IIP3 \approx +17\,\text{dBm}$:$P_{1dB} \approx 13 - 6 = +7\,\text{dBm}$、$IIP3 \approx +17\,\text{dBm}$:

$$P_{IM3} = 3\times(-20) - 2\times 17 = -60 - 34 = \mathbf{-94\,dBm} \;\rightarrow\; \text{比 −85 dBm 低 9 dB,passes ✓}$$$$P_{IM3} = 3\times(-20) - 2\times 17 = -60 - 34 = \mathbf{-94\,dBm} \;\rightarrow\; \text{9 dB below −85 dBm, passes ✓}$$$$P_{IM3} = 3\times(-20) - 2\times 17 = -60 - 34 = \mathbf{-94\,dBm} \;\rightarrow\; \text{9 dB below −85 dBm, passes ✓}$$

LO 洩漏逐級遞減:mixer RF port $= +13 - 40 = -27\,\text{dBm}$ → 經 LNA 反向隔離 20 dB → $-47\,\text{dBm}$ → preselector 對 2.12–2.18 GHz 再衰減 ≥ 25 dB → 天線端 $\approx -72\,\text{dBm}$,遠低於 −30 dBm 限制 ✓The LO leakage falls stage by stage: at the mixer RF port $= +13 - 40 = -27\,\text{dBm}$ → through 20 dB of LNA reverse isolation → $-47\,\text{dBm}$ → the preselector attenuates 2.12–2.18 GHz by a further ≥ 25 dB → about $-72\,\text{dBm}$ at the antenna, far below the −30 dBm limit ✓LO 漏れは段ごとに減衰します:ミキサ RF ポートで $= +13 - 40 = -27\,\text{dBm}$ → LNA の逆方向アイソレーション 20 dB を経て → $-47\,\text{dBm}$ → プリセレクタが 2.12–2.18 GHz をさらに ≥ 25 dB 減衰させ → アンテナ端で約 $-72\,\text{dBm}$ となり、−30 dBm の制限をはるかに下回ります ✓

Choose the level-13 ring: +6 dB of LO buys 6 dB of IIP3 (3 dB of IM3 margin per dB of blocker), and the higher leakage is absorbed by isolation + reverse isolation + preselector.

(d) Half-IF (2,2) 干擾帶:(d) The half-IF (2,2) interference band:(d) Half-IF (2,2) 干渉帯域:位於 $\dfrac{f_{RF}+f_{LO}}{2} = f_{RF} + \dfrac{f_{IF}}{2} = f_{RF} + 95\,\text{MHz}$ → It lies at $\dfrac{f_{RF}+f_{LO}}{2} = f_{RF} + \dfrac{f_{IF}}{2} = f_{RF} + 95\,\text{MHz}$ → $\dfrac{f_{RF}+f_{LO}}{2} = f_{RF} + \dfrac{f_{IF}}{2} = f_{RF} + 95\,\text{MHz}$ に位置します → 2.025–2.085 GHz,在 RX passband 之外。保護是雙重的:preselector ≥ 25 dB , outside the RX passband. The protection is twofold: ≥ 25 dB from the preselector 、RX 通過帯域の外です。保護は二重で、プリセレクタの ≥ 25 dB 加上plusに加えて DB ring 對 even-order (2,2) 的 ≈ 60–65 dB 抑制。high-side plan 把 image 與 half-IF band 都推到 RX band「同一側」(上方)→ 一道濾波器裙邊同時罩住兩者。 the DB ring's ≈ 60–65 dB suppression of the even-order (2,2). The high-side plan pushes both the image and the half-IF band to the *same* side of the RX band (above it) → so a single filter skirt covers them both. DB ring による偶数次 (2,2) の ≈ 60–65 dB の抑圧があります。high-side plan はイメージと half-IF 帯域の両方を RX 帯域の「同じ側」(上方)に押しやります → 1 枚のフィルタスカートで両方を同時に覆えます。

物理詮釋:Physical interpretation:物理的解釈:LO drive 的選擇是典型的「拿功率買線性度」:+6 dB LO → IIP3 +6 dB → IM3 改善 12 dB(從差 3 dB 變賺 9 dB),代價是洩漏 +6 dB — 但洩漏走的是「isolation 鏈」(40+20+25 dB 層層遞減),而 IM3 沒有任何下游補救。結論也展示了 frequency plan、拓樸、濾波器三者是一個耦合決策,不能分開各自最佳化。 The choice of LO drive is a textbook case of “spending power to buy linearity”: +6 dB of LO → +6 dB of IIP3 → a 12 dB improvement in IM3 (from 3 dB short to 9 dB in hand), at the cost of +6 dB of leakage — but the leakage travels down an “isolation chain” (40+20+25 dB of successive attenuation) whereas the IM3 has no downstream remedy at all. The conclusion also shows that the frequency plan, the topology and the filters are one coupled decision and cannot be optimized separately.LO ドライブの選択は「電力を使って直線性を買う」典型例です:LO +6 dB → IIP3 +6 dB → IM3 が 12 dB 改善(3 dB 不足から 9 dB の余裕へ)、代償は漏れ +6 dB — しかし漏れは「アイソレーションの鎖」(40+20+25 dB と層ごとに減衰)を通るのに対し、IM3 には下流での救済手段がまったくありません。この結論は、frequency plan、トポロジー、フィルタの三者が 1 つの結合した決定であり、個別に最適化できないことも示しています。

Appendix — Interactive: Double-Balanced Ring Mixer

雙平衡二極體環式混頻器 — 互動式動畫解說Double-Balanced Diode Ring Mixer — Interactive Animated Walkthrough二重平衡ダイオードリングミキサ — インタラクティブ・アニメーション解説
slide 29: how 4 diodes switch & balun gives mixing & IF output

點擊 Phase A / Phase B,或按 ▶ 自動播放,觀察四顆二極體在 LO 正負半週的切換Click Phase A / Phase B, or press ▶ to play, and watch the four diodes commutate over the positive and negative half-cycles of the LOPhase A / Phase B をクリックするか、▶ を押して自動再生し、4 個のダイオードが LO の正負半周期で切り替わる様子を観察してください

LO phase 角度:LO phase angle:LO 位相角度:
RF input Δ 180° hybrid · rat-race λ/4 (−90°) λ/4 (−90°) λ/4 (−90°) 3λ/4 (−270°) Δ 2 3 Σ −90° +90° Σ Z₀ +v_RF −v_RF T (RF+) B (RF−) L R D1 D2 D3 D4 IF (+ relative to GND below) IF amp reference IF GND (= chassis GND) ↑ LO 源負端 = 同一個 chassis GND↑ the negative terminal of the LO source = the same chassis GND↑ LO 源の負端 = 同じ chassis GND balun LO source (±sign shown above) + + +RF

LO 半週相位LO half-cycle phaseLO 半周期の位相

Phase A · LO > 0

導通的二極體Conducting diodes導通しているダイオード

D1, D2

切斷的二極體Cut-off diodes遮断されているダイオード

D3, D4

當前 IF 極性Current IF polarity現在の IF 極性

+RF

同步波形圖(時間軸跟著 LO 走)Synchronised waveform plot (the time axis follows the LO)同期波形図(時間軸は LO に追従)

t LO RF IF (chopped) LO: +0.00 RF: +0.00 IF: +0.00

↔ 點擊或拖曳波形圖上任意位置,↔ Click or drag anywhere on the waveform plot and the ↔ 波形図上の任意の位置をクリックまたはドラッグすると、紅色游標red cursor赤いカーソル會跟著移動,旁邊小框即時顯示該時刻的 LO / RF / IF 數值 follows; the small box beside it shows the LO / RF / IF values at that instant in real timeが追従して動き、隣の小さな枠がその瞬間の LO / RF / IF の値をリアルタイムで表示します

它是怎麼運作的?How does it work?どのように動作するのか?

核心觀念:四顆二極體 = 一個被 LO 控制的「斬波開關」The core idea: four diodes = a “chopping switch” controlled by the LO核心となる考え方:4 個のダイオード = LO に制御される「チョッピングスイッチ」

把 4 顆二極體想像成Think of the 4 diodes as 4 個のダイオードを一組由 LO 切換的開關a set of switches commutated by the LOLO によって切り替えられる一組のスイッチ。在 LO 的. On だと考えてください。LO の每半週each half-cycle半周期ごと,剛好有兩顆二極體導通、另外兩顆切斷: of the LO, exactly two diodes conduct while the other two are cut off:に、ちょうど 2 個のダイオードが導通し、他の 2 個は遮断されます:

結果:IF 看到的是 The result: what the IF sees is 結果:IF が見るのは RF × sgn(LO) — 一個被 LO 方波斬波過的 RF,這就是混頻! — an RF chopped by the LO square wave, and that is mixing! — LO の方形波でチョッピングされた RF であり、これがまさに混合(ミキシング)です!

① 180° hybrid 把 RF 分成 +RF 與 −RF① The 180° hybrid splits the RF into +RF and −RF① 180° ハイブリッドが RF を +RF と −RF に分ける

RF 進入 Δ 埠(Σ 埠以 $Z_0$ 終結),hybrid 在兩個輸出產生The RF enters the Δ port (the Σ port is terminated in $Z_0$), and the hybrid produces RF は Δ ポートに入り(Σ ポートは $Z_0$ で終端)、ハイブリッドは 2 つの出力に反相 (anti-phase)anti-phase逆相 (anti-phase) 的 RF: RF at its two outputs: の RF を生成します:

$$v_{T}(t) = +\tfrac{1}{\sqrt{2}}\,V_{RF}\cos\omega_{RF}t, \qquad v_{B}(t) = -\tfrac{1}{\sqrt{2}}\,V_{RF}\cos\omega_{RF}t$$

分別接到二極體環的which connect respectively to the diode ring's 接続先はそれぞれダイオードリングの上頂點 Ttop vertex T上頂点 T and 下頂點 Bbottom vertex B下頂点 B.

整顆 180° hybrid 的The 180° ハイブリッド全体の散射矩陣scattering matrix散乱行列(埠號照 ①②③④,Δ = port 4,Σ = port 1,輸出 port 2 接 T、port 3 接 B): of the complete 180° hybrid (port numbering as ①②③④, with Δ = port 4, Σ = port 1, output port 2 to T and port 3 to B):(ポート番号は ①②③④、Δ = port 4、Σ = port 1、出力 port 2 が T、port 3 が B):

$$[S]_{180°} = \frac{-j}{\sqrt{2}}\begin{bmatrix} 0 & \cssId{s12r}{1} & \cssId{s13r}{1} & 0 \\ \cssId{s21r}{1} & 0 & 0 & \cssId{s24r}{-1} \\ \cssId{s31r}{1} & 0 & 0 & \cssId{s34r}{1} \\ 0 & \cssId{s42r}{-1} & \cssId{s43r}{1} & 0 \end{bmatrix}$$
[S] 矩陣元素解讀(port 1 = Σ, 2 = T (上 diode), 3 = B (下 diode), 4 = Δ = RF in)Reading the [S]-matrix elements (port 1 = Σ, 2 = T (upper diode), 3 = B (lower diode), 4 = Δ = RF in)[S] 行列要素の読み方(port 1 = Σ, 2 = T(上側 diode), 3 = B(下側 diode), 4 = Δ = RF in)

讀法:How to read it:読み方: $S_{ij}=$ 從 port $j$ 入射 → 在 port $i$ 出射的複數電壓比。所有非零元素的大小都是 $1/\sqrt{2}$(3 dB 等分),加上總體 $-j$ 前綴;矩陣裡的 $\pm 1$ 決定 $0°$ 或 $180°$ 的額外相位。 $S_{ij}=$ the complex voltage ratio for a wave incident at port $j$ and emerging at port $i$. Every non-zero element has magnitude $1/\sqrt{2}$ (an equal 3 dB split) plus an overall $-j$ prefix; the $\pm 1$ in the matrix sets the extra $0°$ or $180°$ of phase. $S_{ij}=$ は port $j$ から入射 → port $i$ から出射する複素電圧比です。非ゼロ要素の大きさはすべて $1/\sqrt{2}$(3 dB の等分割)で、全体に $-j$ の前置係数が付きます;行列中の $\pm 1$ が $0°$ または $180°$ の追加位相を決めます。

元素Element要素 Value 路徑($j\to i$)Path ($j\to i$)経路($j\to i$) 物理意義Physical meaning物理的意味
$S_{ii}$ $0$ 自己 → 自己self → self自分 → 自分 四個埠都All four ports are 4 つのポートはすべて完美匹配perfectly matched完全に整合
$S_{14}=S_{41}$ $0$ Σ ↔ Δ Σ 與 Δ 兩埠The Σ and Δ ports Σ と Δ の 2 ポートは天生隔離inherently isolated本質的にアイソレート
$S_{23}=S_{32}$ $0$ T ↔ B 兩個輸出埠互相隔離The two output ports are isolated from each other2 つの出力ポートは互いにアイソレート
$S_{21}=S_{12}$ $\dfrac{-j}{\sqrt{2}}$ Σ → T Σ 同相分配到 T:Σ splits in phase to T: Σ は同相で T に分配:半功率half power半分の電力 and $-90°$
$S_{31}=S_{13}$ $\dfrac{-j}{\sqrt{2}}$ Σ → B Σ 同相分配到 B:Σ splits in phase to B: Σ は同相で B に分配:半功率half power半分の電力 and $-90°$(與上一格((前の欄と同相in phase同相)
$S_{24}=S_{42}$ $\dfrac{+j}{\sqrt{2}}$ Δ → T(RF 走這條)Δ → T (the RF takes this route)Δ → T(RF はこの経路を通る) Δ 反相分配到 T:Δ splits in anti-phase to T: Δ は逆相で T に分配:半功率half power半分の電力 and $+90°$
$S_{34}=S_{43}$ $\dfrac{-j}{\sqrt{2}}$ Δ → B(RF 走這條)Δ → B (the RF takes this route)Δ → B(RF はこの経路を通る) Δ 反相分配到 B:Δ splits in anti-phase to B: Δ は逆相で B に分配:半功率half power半分の電力 and $-90°$(與上一格((前の欄と反相in anti-phase逆相!)!)!)

核心非對稱性:The core asymmetry:核心となる非対称性: $S_{24}=+j/\sqrt{2}$ 但 $S_{34}=-j/\sqrt{2}$ — Δ 輸入時,兩個輸出埠收到的訊號 $S_{24}=+j/\sqrt{2}$ but $S_{34}=-j/\sqrt{2}$ — with an input at Δ, the signals received at the two output ports are $S_{24}=+j/\sqrt{2}$ だが $S_{34}=-j/\sqrt{2}$ — Δ から入力したとき、2 つの出力ポートが受け取る信号は相位差 180°180° apart in phase位相が 180° 異なります。這就是 Δ port 的「difference」分配,也是 ring mixer 把 RF 注入二極體環 T 和 B 時讓兩個二極體看到 $\pm V_{RF}$(差動)的根本原因。. This is the “difference” split of the Δ port, and it is the fundamental reason the ring mixer's diodes at T and B see $\pm V_{RF}$ (differentially) when the RF is injected into the ring.。これが Δ port の「difference」分配であり、RF をダイオードリングに注入したときに ring mixer の T と B のダイオードが $\pm V_{RF}$(差動)を見る根本的な理由です。

幺正性檢查:Unitarity check:ユニタリ性の検査: $[S]^\dagger[S] = [I]$(lossless);矩陣對稱(reciprocal);對角線全 0(全埠匹配)。三項一起鎖死為理想 180° hybrid。 $[S]^\dagger[S] = [I]$ (lossless); the matrix is symmetric (reciprocal); the diagonal is all zeros (every port matched). Together these three pin it down as an ideal 180° hybrid. $[S]^\dagger[S] = [I]$(lossless);行列は対称(reciprocal);対角要素はすべて 0(全ポート整合)。この 3 点が合わさって、理想的な 180° hybrid であることが確定します。

⚠ 為什麼把 arm 上的相位延遲加起來(-270° + -90° = 0°)≠ port 3 的輸出相位 (±90°)?⚠ Why does adding up the phase delays along the arms (−270° + −90° = 0°) ≠ the output phase at port 3 (±90°)?⚠ なぜ arm 上の位相遅延を足し合わせた値(-270° + -90° = 0°)が port 3 の出力位相 (±90°) と一致しないのか?

圖上的每個 arm 標籤(λ/4 (−90°)、3λ/4 (−270°))標的是Each arm label in the figure (λ/4 (−90°), 3λ/4 (−270°)) gives 図中の各 arm ラベル(λ/4 (−90°)、3λ/4 (−270°))が示しているのは該 arm 上的傳播相位延遲the propagation phase delay along that armその arm 上の伝搬位相遅延 — 訊號穿過這段傳輸線時會多走多少電氣長度,對應多少相位。這個數值對「單條路徑」是對的。 — how much extra electrical length, and hence phase, the signal picks up crossing that section of line. That number is correct for a *single* path. — 信号がその伝送線区間を通過する際に増える電気長、およびそれに対応する位相です。この数値は「単一の経路」については正しいものです。

但 port 3 的輸出相位But the output phase at port 3 is しかし port 3 の出力位相は不是notけっして沿單條路徑相加的結果。 the result of adding up along a single path.単一の経路に沿って足し合わせた結果ではありません。

看起來合理的單路徑算法:The seemingly reasonable single-path calculation:一見もっともらしい単一経路の計算:

$$\Delta \xrightarrow{3\lambda/4\;(-270°)} \Sigma \xrightarrow{\lambda/4\;(-90°)} \text{port 3} \quad\Rightarrow\quad -270° + (-90°) = -360° \equiv 0°$$

但實際 port 3 的相位是 But the actual phase at port 3 is しかし実際の port 3 の位相は ±90°(依符號慣例),不是 0°。為什麼?(depending on sign convention), not 0°. Why?(符号の慣習による)であり、0° ではありません。なぜでしょうか?

真相 ① — RF 在 Δ 同時往「兩個方向」傳播:Truth ① — the RF at Δ propagates in *both* directions at once:真相 ① — Δ の RF は同時に「両方向」へ伝搬する:

  • 順時針(CW):Δ → 上臂 → port 2 → 右臂 → port 3,總長 $\lambda/2$,相位 $-180°$Clockwise (CW): Δ → top arm → port 2 → right arm → port 3, total length $\lambda/2$, phase $-180°$時計回り(CW):Δ → 上側アーム → port 2 → 右側アーム → port 3、全長 $\lambda/2$、位相 $-180°$
  • 逆時針(CCW):Δ → 長臂 → Σ → 底臂 → port 3,總長 $\lambda$,相位 $0°$Anticlockwise (CCW): Δ → long arm → Σ → bottom arm → port 3, total length $\lambda$, phase $0°$反時計回り(CCW):Δ → 長いアーム → Σ → 下側アーム → port 3、全長 $\lambda$、位相 $0°$

兩條波抵達 port 3 後The two waves arriving at port 3 then port 3 に到達した 2 つの波は相干疊加add coherentlyコヒーレントに重なり合います:

$$V_3 \propto e^{j(-180°)} + e^{j(0°)} = -1 + 1 = 0 \;? \quad\text{(看起來該對消...)}$$ $$V_3 \propto e^{j(-180°)} + e^{j(0°)} = -1 + 1 = 0 \;? \quad\text{(so they ought to cancel…)}$$ $$V_3 \propto e^{j(-180°)} + e^{j(0°)} = -1 + 1 = 0 \;? \quad\text{(so they ought to cancel…)}$$

真相 ② — 每個 ring junction 還有「散射」:Truth ② — every ring junction also *scatters*:真相 ② — 各 ring junction にはさらに「散乱」がある:

每個 port 節點其實是一個 3-port tee junction。波到達節點時不會「全部走過去」,而是同時:Each port node is really a 3-port tee junction. A wave reaching the node does not simply “all carry on through”; instead it simultaneously:各 port ノードは実際には 3-port tee junction です。ノードに到達した波は「すべて通り抜ける」わけではなく、同時に:

  • 部分穿過partly transmits一部が透過し(沿 ring 繼續走)(carrying on around the ring)(ring に沿って進み続ける)
  • 部分從這個埠出去partly couples out一部がこのポートから外へ出て(被該埠的負載吸走)(absorbed by that port's load)(そのポートの負荷に吸収される)
  • 部分反射回來partly reflects一部が反射して戻ります(沿 ring 倒退)(travelling back around the ring)(ring に沿って逆向きに進む)

所以兩條路徑到 port 3 時,So when the two paths reach port 3 their したがって 2 つの経路が port 3 に到達したとき、振幅不相等amplitudes are not equal振幅は等しくなく,相位也會因 junction 處的散射而, and their phases have picked up 、位相も junction での散乱によって多加幾度a few extra degrees数度余分に加わります。整個 ring 的完整行為要靠 from the scattering at the junctions. The full behaviour of the whole ring has to be 。ring 全体の完全な挙動は [S]-matrix 一次性算出來。 computed in one go. によって一度に計算する必要があります。

真相 ③ — [S]-matrix 給出的最終答案:Truth ③ — the final answer from the [S] matrix:真相 ③ — [S]-matrix が与える最終的な答え:

$$V_3 = S_{31}\,V_{RF} = \tfrac{-j}{\sqrt{2}}\bigl(\pm 1\bigr)V_{RF} \;\Rightarrow\; \pm 90°\;\text{相位}$$ $$V_3 = S_{31}\,V_{RF} = \tfrac{-j}{\sqrt{2}}\bigl(\pm 1\bigr)V_{RF} \;\Rightarrow\; \pm 90°\;\text{phase}$$ $$V_3 = S_{31}\,V_{RF} = \tfrac{-j}{\sqrt{2}}\bigl(\pm 1\bigr)V_{RF} \;\Rightarrow\; \pm 90°\;\text{phase}$$

port 2 和 port 3 相位差 $180°$ — 這就是 Δ 端口輸入該產生的差動 (anti-phase) 輸出。Ports 2 and 3 are $180°$ apart in phase — exactly the differential (anti-phase) output that a Δ-port input should produce.port 2 と port 3 の位相差は $180°$ — これこそ Δ ポート入力が生じるべき差動 (anti-phase) 出力です。

單路徑相位相加Adding phases along a single path単一経路の位相加算完整 ring 分析([S]-matrix)Full ring analysis ([S] matrix)完全な ring 解析([S]-matrix)
✅ 對「propagation phase delay」正確✅ Correct for the “propagation phase delay”✅「propagation phase delay」については正しい✅ 加上 junction 散射✅ Includes junction scattering✅ junction の散乱を含む
❌ 沒考慮反向波❌ Ignores the reverse wave❌ 逆方向の波を考慮していない✅ 加上反向波相干疊加✅ Includes coherent addition of the reverse wave✅ 逆方向の波のコヒーレントな重ね合わせを含む
❌ 沒考慮每個 junction 的分流❌ Ignores the split at each junction❌ 各 junction での分流を考慮していない✅ 給出最終振幅 + 相位✅ Gives the final amplitude and phase✅ 最終的な振幅 + 位相を与える
結果:誤導Result: misleading結果:誤解を招く結果:±90°(正確)Result: ±90° (correct)結果:±90°(正しい)

那 arm 上的 -90° / -270° 標籤還有用嗎?So are the −90° / −270° arm labels of any use?では arm 上の -90° / -270° ラベルはまだ役に立つのか? 有 — 它們的意義是「每段 arm 的電氣長度」,幫你理解: Yes — they mean “the electrical length of each arm”, which helps you understand: 役に立ちます — その意味は「各 arm の電気長」であり、次のことを理解する助けになります:
  • 為什麼短臂和長臂差 $\lambda/2 = 180°$ → 這是 rat-race 能做 Σ/Δ 分離的   • why the short and long arms differ by $\lambda/2 = 180°$ → this is the   • なぜ短いアームと長いアームが $\lambda/2 = 180°$ 異なるのか → これが rat-race で Σ/Δ 分離ができる根本原因fundamental reason根本的な理由
  • 為什麼這顆元件叫「180° hybrid」 a rat-race can separate Σ from Δ;   • why the component is called a “180° hybrid” です;  • なぜこの部品が「180° hybrid」と呼ばれるのか
但它們 But they are しかしそれらは不是notけっして沿某條路徑累積到輸出埠的相位 — 那需要 [S]-matrix。 the phase accumulated along some path to an output port — that requires the [S] matrix.ある経路に沿って出力ポートまで累積した位相ではありません — それには [S]-matrix が必要です。

📚 名詞解釋:什麼是「tee junction」?為什麼它讓 ring 分析變複雜?📚 Terminology: what is a “tee junction”, and why does it complicate the ring analysis?📚 用語解説:「tee junction」とは何か?なぜ ring の解析を複雑にするのか?

Tee junction(T 接點)是 RF 電路中(a T-junction) is a 3-port component in RF circuits where (T 接点)は RF 回路において三條傳輸線匯合於一點three transmission lines meet at a point3 本の伝送線が 1 点で交わる的 3-port 元件,形狀像字母「T」:, shaped like the letter “T”:3-port 部品で、形は文字「T」に似ています:

Port 1(外部 Z₀)Port 1 (external Z₀)Port 1(外部 Z₀) Port 2 (ring CW arm) Port 3 (ring CCW arm) 穿透 (transmit)transmit透過 (transmit) 耦合出去 (couple out)couple out外へ結合 (couple out) 反射 (reflect)reflect反射 (reflect) incident wave 3-port tee junction (三條傳輸線匯合於一點)3-port tee junction (three transmission lines meeting at a point)3-port tee junction (3 本の伝送線が 1 点で交わる)

回到 rat-race:每個 ring 角落都是這個 tee junctionBack to the rat-race: every corner of the ring is one of these tee junctionsrat-race に戻ると:ring の各コーナーはこの tee junction です

rat-race ring 的四個角落(Δ、port 2、port 3、Σ)The four corners of the rat-race ring (Δ, port 2, port 3, Σ) rat-race ring の 4 つのコーナー(Δ、port 2、port 3、Σ)は都對應到上面那個 3-port tee junctionall correspond to the 3-port tee junction aboveすべて上記の 3-port tee junction に対応します,只是「外部 $Z_0$ 那個 port」連接到不同的東西:, the only difference being what “the external $Z_0$ port” connects to:。違いは「外部の $Z_0$ のポート」が何に接続されるかだけです:

Ring 上的位置Position on the ringRing 上の位置 Tee Port 1(外部 $Z_0$)Tee port 1 (external $Z_0$)Tee Port 1(外部 $Z_0$) Tee Port 2(ring CW 方向)Tee port 2 (CW around the ring)Tee Port 2(ring の CW 方向) Tee Port 3(ring CCW 方向)Tee port 3 (CCW around the ring)Tee Port 3(ring の CCW 方向)
Δ 角The Δ cornerΔ コーナー(左上)(top left)(左上) RF 輸入RF inputRF 入力 頂部 $\lambda/4$ → ring port 2top $\lambda/4$ → ring port 2頂部 $\lambda/4$ → ring port 2 長臂 $3\lambda/4$ → Σlong arm $3\lambda/4$ → Σ長いアーム $3\lambda/4$ → Σ
Ring port 2(右上)(top right)(右上) 到 diode to the diode ダイオード T(RF+)(RF+)(RF+)へ 右側 $\lambda/4$ → ring port 3right-hand $\lambda/4$ → ring port 3右側 $\lambda/4$ → ring port 3 頂部 $\lambda/4$ → Δtop $\lambda/4$ → Δ頂部 $\lambda/4$ → Δ
Ring port 3(右下)(bottom right)(右下) 到 diode to the diode ダイオード B(RF−)(RF−)(RF−)へ 底部 $\lambda/4$ → Σbottom $\lambda/4$ → Σ底部 $\lambda/4$ → Σ 右側 $\lambda/4$ → ring port 2right-hand $\lambda/4$ → ring port 2右側 $\lambda/4$ → ring port 2
Σ 角The Σ cornerΣ コーナー(左下)(bottom left)(左下) $Z_0$ 終結$Z_0$ termination$Z_0$ 終端 長臂 $3\lambda/4$ → Δlong arm $3\lambda/4$ → Δ長アーム $3\lambda/4$ → Δ 底部 $\lambda/4$ → ring port 3bottom $\lambda/4$ → ring port 3底部 $\lambda/4$ → ring port 3

放大來看 Δ 這個角落,就是一個 tee junction:Zoom in on the Δ corner and it is exactly a tee junction:Δ のこの角を拡大して見ると、ちょうど一つの tee junction です:

RF input (Tee Port 1 ≡ 外部 Z₀)(Tee port 1 ≡ external Z₀)(Tee Port 1 ≡ 外部 Z₀) Δ 頂部 λ/4top λ/4頂部 λ/4 (Tee Port 2 → ring port 2) 2 長臂 3λ/4long arm 3λ/4長アーム 3λ/4 (Tee Port 3 → Σ) Σ Δ 角的 zoom:RF 入射時,部分穿透到 port 2、部分繞長臂到 Σ、部分反射回 RF inZoom on the Δ corner: when the RF is incident, part transmits through to port 2, part goes the long way round to Σ, and part reflects back into the RF inputΔ 角の zoom:RF 入射時、一部は port 2 へ透過、一部は長アームを回って Σ へ、一部は RF in へ反射

每個 ring 角落都這樣分析 — 把 4 個 tee 串起來,加上兩道波(CW/CCW)的相干疊加,整顆 4-port hybrid 的 [S]-matrix 就出來了(見下一個 details 區塊的完整推導)。Analyse every corner of the ring this way — chain the 4 tees together, add the coherent superposition of the two waves (CW/CCW), and the [S] matrix of the whole 4-port hybrid falls out (see the full derivation in the next details block).リングの各角をこのように解析します — 4 個の tee をつなぎ、2 つの波(CW/CCW)の干渉重ね合わせを加えると、4-port hybrid 全体の [S]-matrix が得られます(次の details ブロックの完全な導出を参照)。

在 rat-race ring 上,On a rat-race ring, rat-race ring 上では、每個外部埠(port 1, 2, 3, 4)都是一個 tee junctionevery external port (ports 1, 2, 3, 4) is a tee junction各外部ポート(port 1, 2, 3, 4)はいずれも tee junction です — 一邊接外部負載 ($Z_0$),另外兩邊接 ring 上的兩段傳輸線。 — one side connects to an external load ($Z_0$) and the other two to the two sections of ring transmission line. — 一方は外部負荷($Z_0$)に接続し、他の二方は ring 上の 2 本の伝送線路に接続します。

當訊號到達 tee junction 會發生什麼?What happens when a signal reaches a tee junction?信号が tee junction に到達すると何が起こるでしょうか?

三件事同時發生:Three things happen at once:三つのことが同時に起こります:

  1. Transmit through(穿透)— 部分能量沿 ring 繼續往下一段 arm 走。(transmit) — part of the energy carries on around the ring into the next arm.(透過)— エネルギーの一部は ring に沿って次のアームへ進みます。
  2. Couple out(耦合出去)— 部分能量從外部 port 流出去(被 $Z_0$ 負載吸收)。(couple out) — part of the energy flows out of the external port (absorbed by the $Z_0$ load).(結合出力)— エネルギーの一部は外部 port から流れ出ます($Z_0$ 負荷に吸収されます)。
  3. Reflect back(反射)— 部分能量沿 ring 倒回原來的方向。(reflect) — part of the energy travels back around the ring the way it came.(反射)— エネルギーの一部は ring を元の方向へ戻ります。

每項的比例由The proportions are set by 各項の比率は三條線的特性阻抗the characteristic impedances of the three lines3 本の線路の特性インピーダンス決定。對 rat-race 而言,環上 $Z_R = \sqrt{2}Z_0$、外部 port 是 $Z_0$,每個 tee 有特定散射係數 — 不能當「直線穿過」處理。. For a rat-race, the ring is $Z_R = \sqrt{2}Z_0$ and the external ports are $Z_0$, so each tee has particular scattering coefficients — you cannot treat it as “straight through”.によって決まります。rat-race では、リング上は $Z_R = \sqrt{2}Z_0$、外部 port は $Z_0$ なので、各 tee は固有の散乱係数を持ちます — 「直線的に通過する」として扱うことはできません。

⚠ 重要定理:「ideal lossless matched 3-port」不存在⚠ An important theorem: an “ideal lossless matched 3-port” does not exist⚠ 重要な定理:「ideal lossless matched 3-port」は存在しない

RF 教科書經典定理(Pozar §7.1):A classic theorem of RF textbooks (Pozar §7.1): RF 教科書の古典的定理(Pozar §7.1):沒有任何 3-port 元件能同時滿足下列三件事No 3-port component can satisfy all three of the following at onceいかなる 3-port 素子も次の三つを同時に満たすことはできません:

  • (a) 完美匹配(三個 port 都 $S_{ii}=0$)(a) perfectly matched (all three ports have $S_{ii}=0$)(a) 完全整合(三つの port すべてが $S_{ii}=0$)
  • (b) Lossless($[S]^\dagger[S]=[I]$)(b) lossless ($[S]^\dagger[S]=[I]$)(b) Lossless($[S]^\dagger[S]=[I]$)
  • (c) Reciprocal($S_{ij}=S_{ji}$)(c) reciprocal ($S_{ij}=S_{ji}$)(c) Reciprocal($S_{ij}=S_{ji}$)

證明:寫下三條件約束,發現 unitarity 條件無解 → 矛盾。Proof: write down the three constraints and find that the unitarity condition has no solution → contradiction.証明:三つの条件を書き下すと、unitarity 条件に解がないことが分かります → 矛盾。

→ 任何 3-port tee → Any 3-port tee → いかなる 3-port tee も 至少要犧牲一個性質must sacrifice at least one property少なくとも一つの性質を犠牲にしなければなりません:

  • 失去匹配Give up the match整合を捨てる:普通的 transmission-line tee(三個 port 都有反射): an ordinary transmission-line tee (all three ports reflect):普通の transmission-line tee(三つの port すべてに反射がある)
  • 加上損耗Add loss損失を加える:電阻式 Wilkinson、衰減型功分器: a resistive Wilkinson or an attenuating power divider:抵抗型 Wilkinson、減衰型電力分配器
  • 失去 reciprocalGive up reciprocityreciprocal を捨てる:必須用鐵氧體 circulator(單向元件): you need a ferrite circulator (a unidirectional component):フェライト circulator(単方向素子)を使う必要があります

那 4-port hybrid 為什麼能突破限制?So how does a 4-port hybrid get around the restriction?では 4-port hybrid はなぜ制限を突破できるのでしょうか?

Rat-race / branch-line / Wilkinson 等 4-port 元件其實是把4-port components such as the rat-race, branch-line and Wilkinson actually rat-race / branch-line / Wilkinson などの 4-port 素子は、実は多個 tee junction 串起來chain several tee junctions together複数の tee junction をつないで,再藉由 and then use 、さらに波的干涉抵消wave interference and cancellation波の干渉による打ち消し讓每個 tee 的反射在外部 port 對消。 to make each tee's reflection cancel at the external port.によって、各 tee の反射を外部 port で相殺させています。

具體機制:The mechanism in detail:具体的な機構:

  • 單看一個 tee → 不匹配(會反射)Taken on its own, one tee is not matched (it reflects)一つの tee だけを見ると → 整合しません(反射します)
  • 沿著兩條 ring 路徑(CW 和 CCW)走過的反射都會回到外部 portThe reflections travelling along both ring paths (CW and CCW) return to the external port2 本の ring 経路(CW と CCW)を通った反射はいずれも外部 port に戻ってきます
  • 路徑長度差(在 rat-race 是 λ/2)讓兩道反射The path-length difference (λ/2 in a rat-race) leaves the two reflections 経路長差(rat-race では λ/2)により、2 つの反射は相位剛好差 180°exactly 180° apart in phase位相がちょうど 180° ずれます → 相消 → they cancel → 打ち消し合います
  • 從外面看:matched ✓、lossless ✓、reciprocal ✓ — 三項都成立!Seen from outside: matched ✓, lossless ✓, reciprocal ✓ — all three hold!外から見ると:matched ✓、lossless ✓、reciprocal ✓ — 三つとも成立します!

這正是 hybrid 的設計核心:This is the heart of hybrid design: これこそが hybrid の設計の核心です:用 4-port 結構 + 干涉抵消use a 4-port structure plus interference cancellation4-port 構造 + 干渉による打ち消しを用いて,繞過 3-port 元件的物理限制。 to sidestep the physical limitation on 3-port components.、3-port 素子の物理的制限を回避します。

為什麼這個概念在 rat-race [S]-matrix 推導中關鍵?Why is this concept crucial to deriving the rat-race [S] matrix?なぜこの概念が rat-race [S]-matrix の導出において重要なのでしょうか?

簡單的 "path-summing" 算法(把每段 arm 的相位延遲加起來)失敗,是因為它The naive “path-summing” calculation (adding up the phase delay of each arm) fails because it 単純な「path-summing」の計算(各アームの位相遅延を足し合わせる方法)が失敗するのは、それが把每個 ring 節點當成「直通」treats every ring node as a straight-through connection各 ring ノードを「直通」とみなし,沒考慮 tee junction 的散射。 and ignores the scattering at the tee junctions.、tee junction の散乱を考慮していないからです。

正確分析(even-odd mode 或 ABCD matrix):The correct analysis (even-odd mode or ABCD matrices):正しい解析(even-odd mode または ABCD matrix):

  • 把每個 tee 視為一個會散射的 3-portTreat each tee as a 3-port that scatters各 tee を散乱する 3-port とみなす
  • 追蹤 CW 與 CCW 兩道波在 ring 中的多次反射Track the multiple reflections of both the CW and CCW waves around the ringCW と CCW の 2 つの波の ring 内での多重反射を追跡する
  • 讓所有反射相干疊加 → 得出 4-port [S]-matrixLet all the reflections add coherently → out comes the 4-port [S] matrixすべての反射を干渉重ね合わせする → 4-port [S]-matrix が得られる

結果就是下一個 details 區塊裡展示的完整推導。The result is the full derivation presented in the next details block.その結果が、次の details ブロックで示す完全な導出です。

其他 tee 變體Other tee variantsその他の tee の変形

類型Type種類結構Structure構造用途Use用途
E-plane tee波導在 E 場方向相交Waveguides meeting in the E-field direction導波管が E 場方向で交わる同相分配in-phase splitting同相分配
H-plane tee波導在 H 場方向相交Waveguides meeting in the H-field direction導波管が H 場方向で交わる反相分配anti-phase splitting逆相分配
Magic-TE-plane + H-plane 結合的 4-portA 4-port combining an E-plane and an H-plane teeE-plane + H-plane を結合した 4-port波導版本的 180° hybridThe waveguide version of a 180° hybrid導波管版の 180° hybrid
微帶 teeMicrostrip teeマイクロストリップ tee三條微帶線匯合Three microstrip lines meeting3 本のマイクロストリップ線路が合流RF/IC 上的通用節點The general-purpose node in RF/IC workRF/IC における汎用ノード
🔬 完整 ring 分析(even-odd mode decomposition)— 詳細計算公式🔬 The full ring analysis (even-odd mode decomposition) — detailed derivation🔬 完全な ring 解析(even-odd mode decomposition)— 詳細な計算式

下面用 Below we use 以下では even-odd mode analysis(偶/奇模分解,Pozar §7.8)算出整顆 rat-race 的 [S]-matrix。這個方法同時處理:(even/odd mode decomposition, Pozar §7.8) to compute the [S] matrix of the whole rat-race. This method handles all of the following at once:(偶/奇モード分解、Pozar §7.8)を用いて rat-race 全体の [S]-matrix を求めます。この方法は次を同時に扱えます:

  • ✅ 每個 junction 的散射(不只是「直走」)✅ the scattering at each junction (not just “straight through”)✅ 各 junction での散乱(「直進」だけではない)
  • ✅ CW 與 CCW 兩道波的相干疊加✅ the coherent superposition of the CW and CCW waves✅ CW と CCW の 2 つの波の干渉重ね合わせ
  • ✅ 給出最終 amplitude + phase✅ it gives the final amplitude and phase✅ 最終的な振幅と位相を与える

設定(geometry)Setup (geometry)設定(geometry)

  • 四個外部埠都以 $Z_0$ 終結(典型 50 Ω)。All four external ports are terminated in $Z_0$ (typically 50 Ω).四つの外部ポートはすべて $Z_0$ で終端します(典型的には 50 Ω)。
  • 環的The ring's リングの特性阻抗 $Z_R = \sqrt{2}\,Z_0$characteristic impedance is $Z_R = \sqrt{2}\,Z_0$特性インピーダンス $Z_R = \sqrt{2}\,Z_0$(≈ 70.7 Ω)— 這個選擇使每個 tee junction 在分流後從外部看仍匹配。(≈ 70.7 Ω) — a choice that leaves each tee junction matched as seen from outside once the split is taken into account.(≈ 70.7 Ω)— この選択により、各 tee junction は分流後も外部から見て整合したままになります。
  • 環上有三段 $\lambda/4$(埠 1→2→3→4,順時針)和一段 $3\lambda/4$(埠 4→1,長臂)。The ring has three $\lambda/4$ sections (ports 1→2→3→4, clockwise) and one $3\lambda/4$ section (ports 4→1, the long arm).リング上には 3 段の $\lambda/4$(ポート 1→2→3→4、時計回り)と 1 段の $3\lambda/4$(ポート 4→1、長アーム)があります。
  • 對稱平面:通過 port 1–port 2 短臂的中點,以及 port 3–port 4 短臂的中點。把 port 1 ↔ port 4、port 2 ↔ port 3 對應到鏡像對。Plane of symmetry: through the midpoint of the port 1–port 2 short arm and the midpoint of the port 3–port 4 short arm. It maps port 1 ↔ port 4 and port 2 ↔ port 3 onto mirror pairs.対称平面:port 1–port 2 の短アームの中点、および port 3–port 4 の短アームの中点を通ります。port 1 ↔ port 4、port 2 ↔ port 3 を鏡像対に対応させます。
Rat-race ring geometry (with symmetry plane) 對稱平面 (symmetry plane)symmetry plane対称平面 (symmetry plane) → Δ ↔ Σ; port 2 ↔ port 3 4 Δ port 2 port 2 3 port 3 1 Σ port λ/4 短臂λ/4 short armλ/4 短アーム $Z_R = \sqrt{2}Z_0$ λ/4 短臂λ/4 short armλ/4 短アーム λ/4 短臂λ/4 short armλ/4 短アーム $Z_R = \sqrt{2}Z_0$ 3λ/4 長臂3λ/4 long arm3λ/4 長アーム $Z_R = \sqrt{2}Z_0$ $Z_0$ $Z_0$ $Z_0$ $Z_0$ 外部 50 Ωexternal 50 Ω外部 50 Ω 系統阻抗system impedanceシステムインピーダンス → CW ↓ CW ← CW 紫色虛線 = 對稱平面(horizontal symmetry axis),鏡射 port 1↔4、port 2↔3The purple dashed line = the plane of symmetry (horizontal symmetry axis), mirroring port 1↔4 and port 2↔3紫色の破線 = 対称平面(horizontal symmetry axis)、port 1↔4、port 2↔3 を鏡映

圖示說明:Key to the figure:図の説明:

  • 綠線Green lines緑の線:三段 $\lambda/4$ 短臂(環上每對相鄰埠之間): the three $\lambda/4$ short arms (between each adjacent pair of ports on the ring):3 段の $\lambda/4$ 短アーム(リング上の隣接する各ポート対の間)
  • 紅線Red line赤の線:一段 $3\lambda/4$ 長臂(port 4 經左側繞回 port 1)— 視覺上畫得比較長以對應實際電氣長度: the single $3\lambda/4$ long arm (from port 4 round the left back to port 1) — drawn visually longer to reflect its actual electrical length:1 段の $3\lambda/4$ 長アーム(port 4 から左側を回って port 1 へ)— 実際の電気長に対応するよう視覚的に長めに描いています
  • 環的特性阻抗Ring characteristic impedanceリングの特性インピーダンス:所有四段都用 $Z_R = \sqrt{2}Z_0$ 的傳輸線: all four sections use $Z_R = \sqrt{2}Z_0$ line:4 段すべてに $Z_R = \sqrt{2}Z_0$ の伝送線路を使用
  • 外部埠External ports外部ポート(黑色短引線):四個埠都以 $Z_0$(50 Ω 系統阻抗)終結(the short black leads): all four ports are terminated in $Z_0$ (the 50 Ω system impedance)(黒い短い引き出し線):四つのポートはすべて $Z_0$(50 Ω システムインピーダンス)で終端
  • 紫色虛線Purple dashed line紫色の破線:對稱平面(穿過上臂中點與下臂中點的水平線),對稱性把 port 1 ↔ port 4 (Σ ↔ Δ) 對映、port 2 ↔ port 3 對映 — 這個對稱性是 even-odd mode 分析的關鍵: the plane of symmetry (the horizontal line through the midpoints of the upper and lower arms). The symmetry maps port 1 ↔ port 4 (Σ ↔ Δ) and port 2 ↔ port 3 — and it is the key to the even-odd mode analysis:対称平面(上アームの中点と下アームの中点を通る水平線)。この対称性により port 1 ↔ port 4 (Σ ↔ Δ)、port 2 ↔ port 3 が対応します — この対称性が even-odd mode 解析の鍵です
為什麼 $Z_R = \sqrt{2}Z_0$?Why $Z_R = \sqrt{2}Z_0$?なぜ $Z_R = \sqrt{2}Z_0$ なのでしょうか?
若環的特性阻抗剛好是 $\sqrt{2}Z_0$,則每個 tee junction 從外部 port 看進去時:兩條相鄰的 $\lambda/4$ 線各把對面的 $Z_0$ 變換成 $Z_R^2/Z_0 = 2Z_0$;兩個 $2Z_0$ 並聯得到 $Z_0$ — If the ring's characteristic impedance is exactly $\sqrt{2}Z_0$, then looking into any tee junction from its external port: each of the two adjacent $\lambda/4$ lines transforms the $Z_0$ at the far end into $Z_R^2/Z_0 = 2Z_0$; two $2Z_0$ in parallel give $Z_0$ — 環の特性インピーダンスがちょうど $\sqrt{2}Z_0$ であれば、各 tee junction を外部 port から見たとき:隣接する 2 本の $\lambda/4$ 線路がそれぞれ反対側の $Z_0$ を $Z_R^2/Z_0 = 2Z_0$ に変換し、2 つの $2Z_0$ の並列で $Z_0$ が得られます — 剛好匹配外部exactly matching the outsideちょうど外部と整合します。這就是 rat-race 在. That is the physical mechanism by which a rat-race achieves $S_{ii}=0$ at 。これが、rat-race が $S_{ii}=0$ を実現する物理的機構であり、それが成り立つのは每個埠every port各ポート都做到 $S_{ii}=0$ 的物理機制。 . においてです。

Step 1 — 在 port 1(Σ)以振幅 $a_1=1$ 激發;對稱化分解Step 1 — excite port 1 (Σ) with amplitude $a_1=1$; decompose symmetricallyStep 1 — port 1(Σ)を振幅 $a_1=1$ で励振し、対称的に分解する

把單一激發 $a_1 = 1$ 寫成兩個對稱激發的疊加:Write the single excitation $a_1 = 1$ as the superposition of two symmetric excitations:単一の励振 $a_1 = 1$ を、二つの対称励振の重ね合わせとして書きます:

$$\underbrace{a_1 = 1, a_4 = 0}_{\text{原始激發}} = \underbrace{\tfrac{1}{2}(a_1=1,\,a_4=+1)}_{\text{even mode (e)}} + \underbrace{\tfrac{1}{2}(a_1=1,\,a_4=-1)}_{\text{odd mode (o)}}$$ $$\underbrace{a_1 = 1, a_4 = 0}_{\text{original excitation}} = \underbrace{\tfrac{1}{2}(a_1=1,\,a_4=+1)}_{\text{even mode (e)}} + \underbrace{\tfrac{1}{2}(a_1=1,\,a_4=-1)}_{\text{odd mode (o)}}$$ $$\underbrace{a_1 = 1, a_4 = 0}_{\text{original excitation}} = \underbrace{\tfrac{1}{2}(a_1=1,\,a_4=+1)}_{\text{even mode (e)}} + \underbrace{\tfrac{1}{2}(a_1=1,\,a_4=-1)}_{\text{odd mode (o)}}$$

每個模式都有自己的反射係數 $\Gamma$ 與透射係數 $T$。我們分別計算。Each mode has its own reflection coefficient $\Gamma$ and transmission coefficient $T$. We compute them separately.各モードは固有の反射係数 $\Gamma$ と透過係数 $T$ を持ちます。これらを別々に計算します。

Step 2 — Even mode:對稱平面 = magnetic wall(開路)Step 2 — even mode: the plane of symmetry is a magnetic wall (open circuit)Step 2 — Even mode:対称平面 = magnetic wall(開放)

偶模時,對稱平面上電流為零(如同開路)。把環沿著對稱平面剖開,得到兩個獨立的半環,每個由In the even mode the current on the plane of symmetry is zero (as if it were an open circuit). Cutting the ring along the plane of symmetry gives two independent half-rings, each consisting of 偶モードでは、対称平面上の電流はゼロです(開放と同じ)。リングを対称平面に沿って切り開くと、二つの独立した半リングが得られ、それぞれは一段 $\lambda/8$ 開路線a $\lambda/8$ open-circuited line1 段の $\lambda/8$ 開放線路(從 port 1 到對稱面)(from port 1 to the symmetry plane) (port 1 から対称面まで)串聯一段 $3\lambda/8$ 開路線in series with a $3\lambda/8$ open-circuited lineと 1 段の $3\lambda/8$ 開放線路を直列接続したもの(從 port 1 經長臂到另一個對稱面點)構成。(from port 1 round the long arm to the other point on the symmetry plane).(port 1 から長アームを経て対称面上のもう一点まで)から構成されます。

🤔 為什麼「偶模時對稱平面上電流為零」?🤔 Why is the current on the plane of symmetry zero in the even mode?🤔 なぜ「偶モードでは対称平面上の電流がゼロ」なのでしょうか?

這不是巧合,而是This is no coincidence but これは偶然ではなく、對稱性的直接結果a direct consequence of the symmetry対称性の直接的な帰結です。讓我用幾個角度解釋:. Let me explain it from several angles:。いくつかの角度から説明します:

① 從「對稱激發」的定義出發① Starting from the definition of “symmetric excitation”① 「対称励振」の定義から出発する

偶模 (even mode)The even mode偶モード (even mode) 定義為:對稱面兩側施加 is defined as an excitation applied on both sides of the symmetry plane with の定義は:対称面の両側に等振幅、同相位equal amplitude and the same phase等振幅・同位相的激發。例如同時從 port 1 和 port 4(兩個鏡像對應的埠)注入相同的訊號 $V_0$:. For example, injecting the same signal $V_0$ into port 1 and port 4 (the two mirror-image ports) simultaneously:の励振を加えることです。例えば、port 1 と port 4(互いに鏡像となる二つのポート)から同時に同じ信号 $V_0$ を注入します:

$$a_1 = a_4 = V_0 \quad\text{(同相、同振幅)}$$ $$a_1 = a_4 = V_0 \quad\text{(same phase, same amplitude)}$$ $$a_1 = a_4 = V_0 \quad\text{(same phase, same amplitude)}$$

由電路的鏡像對稱性,整個解(電壓、電流分佈)也必須鏡像對稱。By the mirror symmetry of the circuit, the entire solution (the voltage and current distributions) must also be mirror-symmetric.回路の鏡像対称性により、解全体(電圧・電流分布)もまた鏡像対称でなければなりません。

② 在對稱平面上會發生什麼?② What happens on the plane of symmetry?② 対称平面上では何が起こるか?

把任何一點 P 和它的鏡像點 P' 放在對稱平面兩側:Place any point P and its mirror image P' on either side of the symmetry plane:任意の点 P とその鏡像点 P' を対称平面の両側に置きます:

  • 電壓Voltage電圧是「scalar 量」 — 鏡像下 $V(\text{P}) = V(\text{P}')$(純粹相等) is a scalar — under the mirror $V(\text{P}) = V(\text{P}')$ (simply equal)は「スカラー量」です — 鏡像のもとで $V(\text{P}) = V(\text{P}')$(単純に等しい)
  • 電流Current電流是「vector 量」 — 鏡像下方向會「反過來」。如果 P 的電流朝向對稱面為正方向,P' 那邊的電流也朝向對稱面為正方向(因為對稱)。但 is a vector — under the mirror its direction is reversed. If the current at P is positive when directed towards the symmetry plane, then by symmetry the current at P' is likewise positive towards the plane. But は「ベクトル量」です — 鏡像のもとで向きが「反転」します。P の電流が対称面へ向かう向きを正とすると、対称性より P' の電流も対称面へ向かう向きが正です。しかし它們流動的「絕對方向」是相反的their absolute directions of flow are oppositeそれらが流れる「絶対的な向き」は逆です(一個向右、一個向左,沿著 ring)(one to the right, one to the left, along the ring)(リングに沿って一方は右向き、他方は左向き)
對稱平面plane of symmetry対称平面 Port 1 $+V_0$ Port 4 $+V_0$(同相)$+V_0$ (in phase)$+V_0$(同相) I (向右)I (to the right)I (右向き) I (向左)I (to the left)I (左向き) I = 0 在對稱平面I = 0 on the plane of symmetry対称平面上で I = 0 (兩側電流方向相反,匯合處 = 0)(the currents on the two sides flow in opposite directions, so where they meet = 0)(両側の電流の向きが逆なので、合流点 = 0) Even mode:兩側電流相向流入 → 在對稱面強制為零(KCL)Even mode: the currents on both sides flow towards each other → forced to zero on the symmetry plane (KCL)Even mode:両側の電流が向かい合って流入 → 対称面で強制的にゼロ(KCL)
③ KCL 強制:電流不能在某點累積③ KCL enforces it: charge cannot accumulate at a point③ KCL の強制:電荷はある点に蓄積できない

對稱平面上的某一點 X:Take a point X on the plane of symmetry:対称平面上のある点 X:

  • 左側流向 X 的電流 = $I_L$the current flowing into X from the left = $I_L$左側から X へ流れ込む電流 = $I_L$
  • 右側流向 X 的電流 = $I_R$the current flowing into X from the right = $I_R$右側から X へ流れ込む電流 = $I_R$
  • 由偶模對稱性:$I_L = I_R$(大小相等、都朝向 X)by even-mode symmetry: $I_L = I_R$ (equal in magnitude, both directed at X)偶モードの対称性より:$I_L = I_R$(大きさが等しく、ともに X へ向かう)
  • by さらに、Kirchhoff 電流定律 (KCL)Kirchhoff's current law (KCL)キルヒホッフの電流則 (KCL):穩態下電荷不能累積,所以 $I_L + I_R = 0$: charge cannot accumulate in the steady state, so $I_L + I_R = 0$より:定常状態では電荷は蓄積できないため、$I_L + I_R = 0$
  • 聯立兩條件 → $2 I_L = 0$ → $\boxed{I_L = I_R = 0}$Combining the two conditions → $2 I_L = 0$ → $\boxed{I_L = I_R = 0}$二つの条件を連立 → $2 I_L = 0$ → $\boxed{I_L = I_R = 0}$
④ 邊界條件 = open circuit (magnetic wall)④ The boundary condition = an open circuit (magnetic wall)④ 境界条件 = open circuit (magnetic wall)

「電流 = 0」就是「斷路」的定義。所以在對稱平面上插入“Current = 0” is the definition of an open circuit. So inserting an 「電流 = 0」は「開放」の定義です。したがって対称平面に開路 (open circuit) 邊界條件open-circuit boundary condition開放 (open circuit) 境界条件等價於原始的偶模對稱激發。這稱為 on the plane of symmetry is equivalent to the original symmetric even-mode excitation. It is called a を挿入することは、元の偶モード対称励振と等価です。これを magnetic wall,因為傳輸線中 $I=0$ 對應到磁場為零(電場可以存在 → 開路)。, because $I=0$ on a transmission line corresponds to zero magnetic field (an electric field may still exist → an open circuit).と呼びます。伝送線路中では $I=0$ が磁界ゼロに対応するためです(電界は存在しうる → 開放)。

⑤ 奇模 (odd mode) 為什麼是「短路」?⑤ Why is the odd mode a “short circuit”?⑤ 奇モード (odd mode) はなぜ「短絡」なのでしょうか?

對偶定理:把激發改成 $a_1 = +V_0, a_4 = -V_0$(反相)。同樣對稱性分析:By duality: change the excitation to $a_1 = +V_0, a_4 = -V_0$ (anti-phase). The same symmetry argument gives:双対性より:励振を $a_1 = +V_0, a_4 = -V_0$(逆相)に変えます。同じ対称性の議論から:

  • 對稱平面上的電壓:左側 $V_L$,右側 $V_R = -V_L$(反相對稱)the voltage on the plane of symmetry: $V_L$ on the left, $V_R = -V_L$ on the right (antisymmetric)対称平面上の電圧:左側は $V_L$、右側は $V_R = -V_L$(反対称)
  • 在對稱面這同一個點上,$V$ 必須既等於 $V_L$ 又等於 $V_R$,但 $V_R = -V_L$at the same point on the symmetry plane, $V$ must equal both $V_L$ and $V_R$, yet $V_R = -V_L$対称面上の同じ点で、$V$ は $V_L$ と $V_R$ の両方に等しくなければなりませんが、$V_R = -V_L$
  • → $V_L = -V_L$ → $V_L = 0$ → 對稱平面上電壓為零 = short circuit (electric wall)the voltage on the plane of symmetry is zero = a short circuit (electric wall)対称平面上の電圧はゼロ = short circuit (electric wall)
⑥ 把兩個 mode 加起來 = 還原原始問題⑥ Adding the two modes back together recovers the original problem⑥ 二つのモードを足し合わせる = 元の問題の復元

關鍵的線性疊加:任何單一埠的激發(例如只從 port 1 注入 $V_0$、port 4 不接)都可以拆成:The crucial linear superposition: any single-port excitation (say injecting $V_0$ into port 1 only, with port 4 unconnected) can be split as:重要な線形重ね合わせ:任意の単一ポート励振(例えば port 1 だけから $V_0$ を注入し、port 4 は接続しない)は次のように分解できます:

$$(a_1=V_0,\;a_4=0) = \underbrace{\tfrac{V_0}{2}(a_1=1,\,a_4=+1)}_{\text{even mode}} + \underbrace{\tfrac{V_0}{2}(a_1=1,\,a_4=-1)}_{\text{odd mode}}$$

每個 mode 分別計算(套用開路/短路邊界),最後線性疊加就得到單埠激發的完整解 — 而且Compute each mode separately (applying the open-/short-circuit boundary) and superpose linearly at the end to obtain the complete solution for the single-port excitation — and 各モードを別々に計算し(開放/短絡境界を適用)、最後に線形に重ね合わせれば単一ポート励振の完全な解が得られます — しかも每個 mode 的分析複雜度只有原問題的一半each mode is only half as complex to analyse as the original problem各モードの解析の複雑さは元の問題の半分しかありません(因為我們只看半個對稱電路)。這就是 even-odd mode 分析的威力。(because we only look at half of the symmetric circuit). That is the power of even-odd mode analysis.(対称回路の半分だけを見るため)。これが even-odd mode 解析の威力です。

類比:兩列等大火車相向而行An analogy: two identical trains running towards each other類比:同じ大きさの二つの列車が向かい合って走る
想像兩列等質量、等速度的火車從兩側朝同一個點開過去(偶模)。在那個點上,兩列火車的「淨運動量」對消,那個點看起來是靜止的(電流 = 0)。 Imagine two trains of equal mass and equal speed driving towards the same point from either side (the even mode). At that point the net momentum of the two trains cancels, and the point appears to be at rest (current = 0). 質量も速度も等しい二つの列車が、両側から同じ点へ向かって走ってくると想像してください(偶モード)。その点では二つの列車の「正味の運動量」が打ち消し合い、その点は静止しているように見えます(電流 = 0)。
若兩列火車是同方向、同速度(奇模),它們會在那個點「無縫接續」,就像沒有界面(電壓 = 0,無電位差)。 If the two trains travel in the same direction at the same speed (the odd mode), they join up seamlessly at that point as though there were no interface at all (voltage = 0, no potential difference). もし二つの列車が同じ方向・同じ速度であれば(奇モード)、それらはその点で「継ぎ目なく接続」し、界面がないかのようになります(電圧 = 0、電位差なし)。

從 port 1 看進去,這兩條 stub 的並聯輸入阻抗:Looking in from port 1, the parallel input impedance of these two stubs is:port 1 から見たとき、この 2 本の stub の並列入力インピーダンスは:

$$Z_e^{in} = \bigl[\,Z_R\,\tfrac{1}{j}\cot(\beta\!\cdot\!\lambda/8) \,\parallel\, Z_R\,\tfrac{1}{j}\cot(\beta\!\cdot\!3\lambda/8)\,\bigr]$$
📐 詳細展開這個公式(含 Smith chart 視覺化)📐 Expanding the formula in detail (with a Smith chart visualization)📐 この式を詳しく展開する(Smith chart による可視化つき)
① 公式從哪來?— 開路 stub 的輸入阻抗① Where does the formula come from? — the input impedance of an open stub① 式はどこから来るのか?— 開放 stub の入力インピーダンス

一段For a length 一本の長度 $L$、特性阻抗 $Z_R$、末端開路$L$ of line with characteristic impedance $Z_R$, open-circuited at the far end長さ $L$、特性インピーダンス $Z_R$、終端開放的傳輸線,從另一端看進去的輸入阻抗是:, the input impedance seen from the other end is:の伝送線路について、他端から見た入力インピーダンスは:

$$Z_{in,oc}(L) = -jZ_R\cot(\beta L) = Z_R\,\tfrac{1}{j}\cot(\beta L)$$

其中 $\beta = 2\pi/\lambda$ 是傳播常數。公式可由 transmission-line 通用式 $Z_{in} = Z_0\,\dfrac{Z_L+jZ_0\tan\beta L}{Z_0+jZ_L\tan\beta L}$ 取 $Z_L \to \infty$ 得到。where $\beta = 2\pi/\lambda$ is the propagation constant. The formula follows from the general transmission-line expression $Z_{in} = Z_0\,\dfrac{Z_L+jZ_0\tan\beta L}{Z_0+jZ_L\tan\beta L}$ by taking $Z_L \to \infty$.ここで $\beta = 2\pi/\lambda$ は伝搬定数です。この式は伝送線路の一般式 $Z_{in} = Z_0\,\dfrac{Z_L+jZ_0\tan\beta L}{Z_0+jZ_L\tan\beta L}$ で $Z_L \to \infty$ とすれば得られます。

② 對應到 rat-race 的兩條 stub② Applying it to the rat-race's two stubs② rat-race の 2 本の stub への対応

從 port 1 (Σ) 看進 ring。在這裡我們用 Look into the ring from port 1 (Σ). Here we use the port 1 (Σ) から ring を見ます。ここでは rat-race 的雙對稱(bisymmetric)bisymmetric property of the rat-racerat-race の双対称(bisymmetric)性質 — 環本身對— the ring is symmetric about 性を利用します — リング自体が兩個正交對稱平面two orthogonal planes of symmetry二つの直交する対称平面都對稱,所以可以同時在所有 4 段 arm 的中點切開(偶模 → 全部變成 open)。從 port 1 看,, so we can cut it at the midpoints of all 4 arms at once (in the even mode they all become opens). Seen from port 1, the two のいずれについても対称なので、4 段すべてのアームの中点で同時に切ることができます(偶モード → すべて open になります)。port 1 から見て、緊鄰immediately adjacentすぐ隣接する的兩條 stub 是: stubs are:2 本の stub は:

  • 短路徑The short path短い経路:port 1 → 沿: port 1 → along half of the :port 1 → 左下短臂 (port 1 → port 2)lower-left short arm (port 1 → port 2)左下の短アーム (port 1 → port 2)的一半 → 切點(open)。長度 → the cut (open). Length の半分に沿って → 切断点(open)。長さ $\lambda/8$
  • 長路徑The long path長い経路:port 1 → 沿: port 1 → along half of the :port 1 → 長臂long arm長アーム的一半 → 切點(open)。長度 → the cut (open). Length の半分に沿って → 切断点(open)。長さ $3\lambda/8$

這兩條 stub 都直接從 port 1 出發到最近的切點,所以從 port 1 看是Both stubs run directly from port 1 to the nearest cut, so from port 1 they appear この 2 本の stub はいずれも port 1 から最も近い切断点へ直接つながるため、port 1 から見て互いに並聯in parallel with並列.

Bisymmetric decomposition: 兩個正交對稱平面 → 切 4 個 arm 中點Bisymmetric decomposition: two orthogonal symmetry planes → cut at the midpoints of all 4 armsBisymmetric decomposition: 二つの直交対称平面 → 4 つの arm の中点で切断 水平對稱面horizontal symmetry plane水平対称面 (1↔4, 2↔3) 垂直對稱面 (1↔2, 3↔4)vertical symmetry plane (1↔2, 3↔4)垂直対称面 (1↔2, 3↔4) λ/4 短臂λ/4 short armλ/4 短アーム λ/4 短臂λ/4 short armλ/4 短アーム λ/4 短臂 (port 1 → port 2)λ/4 short arm (port 1 → port 2)λ/4 短アーム (port 1 → port 2) 3λ/4 長臂long arm長アーム 4 Δ 3 2 1 Σ cut (open) cut cut (open) cut (open) λ/8 stub 3λ/8 stub port 1 看到兩條相鄰的 open stub:λ/8(短臂半邊)並聯 3λ/8(長臂半邊)Port 1 sees two adjacent open stubs: λ/8 (half of a short arm) in parallel with 3λ/8 (half of the long arm)port 1 は隣接する 2 本の open stub を見る:λ/8(短アームの半分)と 3λ/8(長アームの半分)の並列 注意:切 4 個中點後,只有緊鄰長臂的 Σ/Δ 埠 (port 1, 4) 看到 (λ/8, 3λ/8);port 2, 3 夾在兩短臂間 → (λ/8, λ/8)Note: after cutting all 4 midpoints, only the Σ/Δ ports adjacent to the long arm (ports 1 and 4) see (λ/8, 3λ/8); ports 2 and 3 sit between two short arms → (λ/8, λ/8)注意:4 つの中点を切った後、長アームに隣接する Σ/Δ ポート (port 1, 4) だけが (λ/8, 3λ/8) を見る。port 2, 3 は二つの短アームに挟まれる → (λ/8, λ/8) 水平對稱面horizontal symmetry plane水平対称面 垂直對稱面vertical symmetry plane垂直対称面 arm 中點 (open)arm midpoint (open)arm 中点 (open)
⚠ 但是直覺上「port 1 經短臂到對稱面」應該是 3λ/8 才對啊?⚠ But intuitively, “port 1 through the short arm to the symmetry plane” ought to be 3λ/8, surely?⚠ しかし直感的には「port 1 から短アームを経て対称面まで」は 3λ/8 のはずでは?

你的直覺沒錯!只用Your intuition is right! With only あなたの直感は正しいです!一個one一つ對稱平面(例如水平面,只 map port 1↔4 和 port 2↔3)時: plane of symmetry (say the horizontal one, mapping only port 1↔4 and port 2↔3):の対称平面(例えば水平面、port 1↔4 と port 2↔3 だけを対応させる)だけを使う場合:

  • 水平面只切兩條 arm 的中點:長臂中點、右臂中點the horizontal plane cuts the midpoints of just two arms: the long arm and the right arm水平面は 2 本のアームの中点しか切りません:長アームの中点と右アームの中点
  • 從 port 1 (BL) 經短臂到右臂中點:底臂 (λ/4) + 右臂一半 (λ/8) = from port 1 (BL) through the short arm to the midpoint of the right arm: bottom arm (λ/4) + half the right arm (λ/8) = port 1 (BL) から短アームを経て右アームの中点まで:底アーム (λ/4) + 右アームの半分 (λ/8) = 3λ/8
  • 從 port 1 經長臂到長臂中點:from port 1 through the long arm to its midpoint: port 1 から長アームを経てその中点まで:3λ/8
  • 兩條都 3λ/8 → 並聯後變成 $+jZ_R/2$,both are 3λ/8 → in parallel they give $+jZ_R/2$, どちらも 3λ/8 → 並列にすると $+jZ_R/2$ になり、不是 ∞not ∞∞ ではありません

所以「單一對稱」確實算不出 $\Gamma_e=+1$。要得到漂亮的答案必須用So a single symmetry really does not yield $\Gamma_e=+1$. To get the elegant answer you must use つまり「単一の対称」では確かに $\Gamma_e=+1$ は得られません。美しい答えを得るために必要なのは雙對稱 (bisymmetric)bisymmetry双対称 (bisymmetric) — 同時利用環的兩個正交對稱性,這樣可以在所有 4 段 arm 中點同時切開,從 port 1 看到的兩條相鄰 stub 才會是 (λ/8, 3λ/8)。 — exploiting both orthogonal symmetries of the ring at once, which lets you cut all 4 arm midpoints simultaneously, so that the two adjacent stubs seen from port 1 really are (λ/8, 3λ/8). — リングの二つの直交対称性を同時に利用することで、4 段すべてのアームの中点で同時に切断でき、port 1 から見える隣接する 2 本の stub がはじめて (λ/8, 3λ/8) になります。

本文用的版本是 bisymmetric 簡化版The version used here is the simplified bisymmetric one本稿で用いているのは bisymmetric の簡略版です — 這是 Pozar §7.8 隱含但沒明講的捷徑。嚴格用單一對稱平面 + ABCD matrix 也能算出 $\Gamma_e=+1$,但代數複雜得多。 — a shortcut that Pozar §7.8 implies but does not spell out. You can also obtain $\Gamma_e=+1$ rigorously with a single symmetry plane plus ABCD matrices, but the algebra is far messier. — これは Pozar §7.8 が暗に示しているものの明言していない近道です。単一の対称平面 + ABCD matrix を厳密に用いても $\Gamma_e=+1$ を求められますが、代数ははるかに煩雑になります。

⚠ 重要更正:port 2、port 3 看到的 stub 跟 port 1、4 ⚠ An important correction: the stubs seen by ports 2 and 3 are ⚠ 重要な訂正:port 2、port 3 が見る stub は port 1、4 とは不一樣not the same同じではありません

切完 4 個 arm 中點後,as those seen by ports 1 and 4. After cutting all 4 arm midpoints, 。4 段すべての arm の中点を切った後、不是每個 port 都看到 (λ/8, 3λ/8)not every port sees (λ/8, 3λ/8)すべての port が (λ/8, 3λ/8) を見るわけではありません。關鍵差別在於:哪些埠「緊鄰那條 3λ/4 長臂」。. The crucial difference is which ports are “immediately adjacent to the 3λ/4 long arm”.。重要な違いは、どのポートが「3λ/4 長アームにすぐ隣接しているか」です。

每個埠的相鄰半臂(cut 在每段中點)The adjacent half-arms of each port (cut at the midpoint of each section)各ポートの隣接する半アーム(各段の中点で切断) 1 Σ 2 3 4 Δ λ/4 λ/4 λ/4 3λ/4 每個埠的兩條相鄰半臂The two adjacent half-arms of each port各ポートの 2 本の隣接する半アーム Portポート 相鄰兩臂Two adjacent arms隣接する 2 本のアーム 看到 stubStubs seen見える stub 1 (Σ) λ/4 短臂 + 3λ/4 長臂λ/4 short arm + 3λ/4 long armλ/4 短アーム + 3λ/4 長アーム (λ/8, 3λ/8) 2 λ/4 短臂 + λ/4 短臂λ/4 short arm + λ/4 short armλ/4 短アーム + λ/4 短アーム (λ/8, λ/8) 3 λ/4 短臂 + λ/4 短臂λ/4 short arm + λ/4 short armλ/4 短アーム + λ/4 短アーム (λ/8, λ/8) 4 (Δ) λ/4 短臂 + 3λ/4 長臂λ/4 short arm + 3λ/4 long armλ/4 短アーム + 3λ/4 長アーム (λ/8, 3λ/8) 只有 Only この中で Σ(1) and Δ(4) 緊鄰 3λ/4 長臂, are adjacent to the 3λ/4 long arm, so only they だけが 3λ/4 長アームに隣接するため、 才會出現 3λ/8 那條 stub。produce that 3λ/8 stub.3λ/8 の stub が現れます。 那 port 2,3 的 (λ/8, λ/8) 共振嗎?So do ports 2 and 3, with (λ/8, λ/8), resonate?では port 2,3 の (λ/8, λ/8) は共振するのでしょうか? 偶模開路:(−jZ_R) ∥ (−jZ_R) = −jZ_R/2Even mode, open-circuited: (−jZ_R) ∥ (−jZ_R) = −jZ_R/2偶モード開放:(−jZ_R) ∥ (−jZ_R) = −jZ_R/2 不是 open(不是並聯諧振)Not an open circuit (not a parallel resonance)open ではありません(並列共振ではありません). 所以「並聯諧振=open」這招是針對So the “parallel resonance = open” trick applies specifically to theつまり「並列共振 = open」という手法は被激發的excited励振された Σ 埠(port 1)。port 2,3 要從它們自己的激發分析。Σ port (port 1). Ports 2 and 3 must be analysed from their own excitation.Σ ポート(port 1)に対するものです。port 2,3 はそれぞれ自身の励振から解析する必要があります。
沿環一圈:三段 λ/4 短臂(藍)+ 一段 3λ/4 長臂(紅)。在每段中點切開後,Round the ring: three λ/4 short arms (blue) plus one 3λ/4 long arm (red). Once each section is cut at its midpoint, リングを一周:3 段の λ/4 短アーム(青)+ 1 段の 3λ/4 長アーム(赤)。各段を中点で切断した後、埠所夾的兩條半臂長度the lengths of the two half-arms flanking a portポートが挟む 2 本の半アームの長さ取決於它鄰接哪些臂:Σ(1)、Δ(4) 鄰接長臂 → (λ/8, 3λ/8);port 2、3 兩側都是短臂 → (λ/8, λ/8)。 depend on which arms it adjoins: Σ(1) and Δ(4) adjoin the long arm → (λ/8, 3λ/8); ports 2 and 3 have short arms on both sides → (λ/8, λ/8).は、そのポートがどのアームに隣接するかで決まります:Σ(1)、Δ(4) は長アームに隣接 → (λ/8, 3λ/8);port 2、3 は両側とも短アーム → (λ/8, λ/8)。

為什麼分析只看 port 1?Why does the analysis look only at port 1?なぜ解析は port 1 だけを見るのでしょうか?因為這個 even-odd 推導是 Because this even-odd derivation starts from なぜなら、この even-odd の導出は從 Σ 埠(port 1)激發an excitation at the Σ port (port 1)Σ ポート(port 1)からの励振出發、要算 $S_{11}$(與經由對稱得到的 $S_{41},S_{21},S_{31}$)。port 1 剛好緊鄰長臂,所以它看到的 (λ/8, 3λ/8) 在偶模下並聯成 open($+jZ_R \parallel -jZ_R = \infty$)。port 2、3 的匹配($S_{22}=S_{33}=0$)要改從 port 2 或 port 3 激發、用它們各自的 (λ/8, λ/8) 幾何(或直接靠整體對稱性 + 互易)來證明,不能套用同一張 (λ/8, 3λ/8) 圖。 and is out to compute $S_{11}$ (and, via symmetry, $S_{41},S_{21},S_{31}$). Port 1 happens to be adjacent to the long arm, so the (λ/8, 3λ/8) it sees combine in parallel to an open in the even mode ($+jZ_R \parallel -jZ_R = \infty$). The match at ports 2 and 3 ($S_{22}=S_{33}=0$) has to be proved by exciting port 2 or port 3 and using their own (λ/8, λ/8) geometry (or directly from overall symmetry plus reciprocity); the same (λ/8, 3λ/8) picture cannot simply be reused.から出発し、$S_{11}$(および対称性から得られる $S_{41},S_{21},S_{31}$)を求めるものだからです。port 1 はちょうど長アームに隣接しているため、そこから見える (λ/8, 3λ/8) は偶モードで並列に open となります($+jZ_R \parallel -jZ_R = \infty$)。port 2、3 の整合($S_{22}=S_{33}=0$)は、port 2 または port 3 から励振してそれぞれの (λ/8, λ/8) の幾何を用いる(あるいは全体の対称性 + 相反性から直接)ことで証明する必要があり、同じ (λ/8, 3λ/8) の図をそのまま使うことはできません。

🧮 完整推導:怎麼證明 $S_{22} = S_{33} = 0$?— 從 port 2 激發🧮 The full derivation: how do you prove $S_{22} = S_{33} = 0$? — exciting port 2🧮 完全な導出:$S_{22} = S_{33} = 0$ をどう証明するか?— port 2 から励振する

先說結論為什麼 (λ/8, λ/8) 並聯First, why (λ/8, λ/8) in parallel is まず、なぜ (λ/8, λ/8) の並列では不夠not enough不十分:なのか:把 port 2 兩側兩條 λ/8 開路 stub 直接並聯,偶模下得 $(-jZ_R)\parallel(-jZ_R) = -jZ_R/2$(純電抗,不是 open,也不是 $Z_0$)。所以「bisymmetric 雙對稱切 4 中點」這招對 port 2 不能直接給 $\Gamma=0$。要真正算出 $S_{22}=0$,必須用 Putting the two λ/8 open stubs on either side of port 2 straight in parallel gives $(-jZ_R)\parallel(-jZ_R) = -jZ_R/2$ in the even mode (a pure reactance — neither an open nor $Z_0$). So the “bisymmetric cut at all 4 midpoints” trick cannot give $\Gamma=0$ at port 2 directly. To genuinely compute $S_{22}=0$ you must use port 2 の両側にある 2 本の λ/8 開放 stub をそのまま並列にすると、偶モードで $(-jZ_R)\parallel(-jZ_R) = -jZ_R/2$ が得られます(純リアクタンスであり、open でも $Z_0$ でもありません)。したがって「bisymmetric に 4 つの中点を切る」という手法は port 2 に対して直接 $\Gamma=0$ を与えることはできません。$S_{22}=0$ を本当に求めるには、單一對稱平面 + 完整的 half-circuit(包含 λ/4 主線與鄰埠的 $Z_0$ 載入)a single plane of symmetry plus the complete half-circuit (including the λ/4 main line and the $Z_0$ loading of the neighbouring ports)単一の対称平面 + 完全な half-circuit(λ/4 主線路と隣接ポートの $Z_0$ 負荷を含む)。下面是完整步驟。. Here are the full steps.を用いなければなりません。以下が完全な手順です。

Step A — 用單一對稱平面做 even-odd 分解Step A — even-odd decomposition using a single plane of symmetryStep A — 単一の対称平面を用いた even-odd 分解

rat-race 的對稱平面把 port 2 對映到它的「鏡像埠」(依埠號慣例可能是 port 3 或 port 4)。激發 port 2、其鏡像埠用 even/odd 分解:The rat-race's plane of symmetry maps port 2 onto its “mirror port” (which may be port 3 or port 4 depending on the numbering convention). Excite port 2 and decompose it and its mirror port into even/odd modes:rat-race の対称平面は port 2 をその「鏡像ポート」(ポート番号の慣例により port 3 または port 4)に対応させます。port 2 を励振し、port 2 とその鏡像ポートを even/odd モードに分解します:

$$\begin{pmatrix}a_2\\a_{\text{mirror}}\end{pmatrix} = \tfrac{1}{2}\begin{pmatrix}1\\+1\end{pmatrix}_{\text{even}} + \tfrac{1}{2}\begin{pmatrix}1\\-1\end{pmatrix}_{\text{odd}}$$ $$S_{22} = \tfrac{1}{2}(\Gamma_2^e + \Gamma_2^o)$$
Step B — port 2 的半電路(剖開後)Step B — the half-circuit at port 2 (after cutting)Step B — port 2 の半回路(切断後)

對稱平面剖開後,port 2 所在的半電路長這樣(從 port 2 看進去):Once the symmetry plane has been cut, the half-circuit containing port 2 looks like this (seen from port 2):対称平面を切断した後、port 2 を含む半回路は次のようになります(port 2 から見て):

Port 2 (輸入 a₂)(input a₂)(入力 a₂) Γ₂ (e/o) λ/8 stub (open in even / short in odd) λ/8 λ/4 線, Z_R = √2·Z₀λ/4 line, Z_R = √2·Z₀λ/4 線路, Z_R = √2·Z₀ λ/4 鄰埠 (port 1)neighbouring port (port 1)隣接ポート (port 1) terminated in Z₀ 3λ/8 stub (open in even / short in odd) 3λ/8 Z₀ 外部負載external load外部負荷 從 port 2 看:(λ/8 stub) ∥ [λ/4 線 → port 1 (Z₀ ∥ 3λ/8 stub)]Seen from port 2: (λ/8 stub) ∥ [λ/4 line → port 1 (Z₀ ∥ 3λ/8 stub)]port 2 から見て:(λ/8 stub) ∥ [λ/4 線路 → port 1 (Z₀ ∥ 3λ/8 stub)]
Port 2 的半電路:port 2 端有 λ/8 stub,經 λ/4 主線連到鄰埠(port 1),鄰埠端有 3λ/8 stub 並聯 $Z_0$。注意這跟 port 1 的半電路是The port 2 half-circuit: a λ/8 stub at the port 2 end, connected through the λ/4 main line to the neighbouring port (port 1), which has a 3λ/8 stub in parallel with $Z_0$. Note that this is port 2 の半回路:port 2 側に λ/8 stub があり、λ/4 主線路を経て隣接ポート(port 1)につながり、隣接ポート側では 3λ/8 stub が $Z_0$ と並列になっています。これは port 1 の半回路と同一個 2-portthe very same 2-portまったく同じ 2-port,只是從另一端看。 as the port 1 half-circuit, merely viewed from the other end.であり、単に反対側から見ているだけである点に注意してください。
Step C — Even mode(stub 末端 O.C.)的 $\Gamma_2^e$Step C — $\Gamma_2^e$ in the even mode (stub ends open-circuited)Step C — Even mode(stub 末端 O.C.)の $\Gamma_2^e$

各 stub 的The 各 stub の導納admittanceアドミタンス(歸一化到 $Y_0 = 1/Z_0$,記 $z_R = \sqrt2$): of each stub (normalized to $Y_0 = 1/Z_0$, writing $z_R = \sqrt2$):($Y_0 = 1/Z_0$ で規格化し、$z_R = \sqrt2$ とおく):

  • port 2 的 λ/8 開路 stub:$Z = -jZ_R\cot(\pi/4) = -j\sqrt2$,所以 $y_{\text{stub2}} = +j/\sqrt2$the λ/8 open stub at port 2: $Z = -jZ_R\cot(\pi/4) = -j\sqrt2$, so $y_{\text{stub2}} = +j/\sqrt2$port 2 の λ/8 開放 stub:$Z = -jZ_R\cot(\pi/4) = -j\sqrt2$、したがって $y_{\text{stub2}} = +j/\sqrt2$
  • 鄰埠的 3λ/8 開路 stub:$Z = -jZ_R\cot(3\pi/4) = +j\sqrt2$,所以 $y_{\text{stub1}} = -j/\sqrt2$the 3λ/8 open stub at the neighbouring port: $Z = -jZ_R\cot(3\pi/4) = +j\sqrt2$, so $y_{\text{stub1}} = -j/\sqrt2$隣接ポートの 3λ/8 開放 stub:$Z = -jZ_R\cot(3\pi/4) = +j\sqrt2$、したがって $y_{\text{stub1}} = -j/\sqrt2$

鄰埠看回去(stub 並聯 $Z_0$):Looking back from the neighbouring port (the stub in parallel with $Z_0$):隣接ポートから見返すと(stub と $Z_0$ の並列):

$$y_{\text{port1}} = 1 + (-j/\sqrt2) = 1 - j/\sqrt2 \;\Rightarrow\; z_{\text{port1}} = \tfrac{2}{3}\!\left(1 + j/\sqrt2\right) = \tfrac{2}{3} + j\tfrac{\sqrt2}{3}$$

經 λ/4 變換到 port 2 那一側:Transformed through the λ/4 line to the port 2 side:λ/4 線路を経て port 2 側へ変換すると:

$$z_{\text{line}} = \frac{z_R^2}{z_{\text{port1}}} = \frac{2}{\tfrac{2}{3}+j\tfrac{\sqrt2}{3}} = 2 - j\sqrt2 \;\Rightarrow\; y_{\text{line}} = \tfrac{1}{3} + j\tfrac{\sqrt2}{6}$$

並上 port 2 自己的 λ/8 stub:In parallel with port 2's own λ/8 stub:port 2 自身の λ/8 stub と並列にすると:

$$y_2^e = y_{\text{line}} + y_{\text{stub2}} = \tfrac{1}{3} + j\tfrac{\sqrt2}{6} + \tfrac{j}{\sqrt2} = \tfrac{1}{3} + j\tfrac{2\sqrt2}{3}$$ $$z_2^e = \tfrac{1}{y_2^e} = \tfrac{1}{3} - j\tfrac{2\sqrt2}{3}$$ $$\Gamma_2^e = \frac{z_2^e - 1}{z_2^e + 1} = \frac{-\tfrac{2}{3} - j\tfrac{2\sqrt2}{3}}{\tfrac{4}{3} - j\tfrac{2\sqrt2}{3}} = \boxed{\;-\frac{j}{\sqrt2}\;}$$
Step D — Odd mode(stub 末端 S.C.)的 $\Gamma_2^o$Step D — $\Gamma_2^o$ in the odd mode (stub ends short-circuited)Step D — Odd mode(stub 末端 S.C.)の $\Gamma_2^o$

所有 stub 換成短路:$y_{\text{stub2}}=-j/\sqrt2$、$y_{\text{stub1}}=+j/\sqrt2$。重複同樣的代數(所有虛部變號):All the stubs become short circuits: $y_{\text{stub2}}=-j/\sqrt2$ and $y_{\text{stub1}}=+j/\sqrt2$. Repeating the same algebra (every imaginary part changes sign):すべての stub が短絡になります:$y_{\text{stub2}}=-j/\sqrt2$、$y_{\text{stub1}}=+j/\sqrt2$。同じ代数を繰り返すと(すべての虚部の符号が変わります):

$$z_{\text{port1}} = \tfrac{2}{3} - j\tfrac{\sqrt2}{3}, \quad z_{\text{line}} = 2+j\sqrt2, \quad y_{\text{line}} = \tfrac{1}{3} - j\tfrac{\sqrt2}{6}$$ $$y_2^o = \tfrac{1}{3} - j\tfrac{2\sqrt2}{3}, \quad z_2^o = \tfrac{1}{3} + j\tfrac{2\sqrt2}{3}$$ $$\Gamma_2^o = \boxed{\;+\frac{j}{\sqrt2}\;}$$
Step E — 合成 → $S_{22} = 0$Step E — combine → $S_{22} = 0$Step E — 合成 → $S_{22} = 0$
$$S_{22} = \tfrac{1}{2}(\Gamma_2^e + \Gamma_2^o) = \tfrac{1}{2}\!\left(-\tfrac{j}{\sqrt2} + \tfrac{j}{\sqrt2}\right) = 0 \;\checkmark$$

$\Gamma_2^e$ 與 $\Gamma_2^o$ 是$\Gamma_2^e$ and $\Gamma_2^o$ are $\Gamma_2^e$ と $\Gamma_2^o$ は純虛數共軛purely imaginary conjugates純虚数の共役(如同 port 1 的情形),相加為 0 → port 2 完全匹配。由 port 2 ↔ port 3 的對稱性,$S_{33} = 0$ 自動成立。(just as for port 1), so they sum to 0 → port 2 is perfectly matched. By the port 2 ↔ port 3 symmetry, $S_{33} = 0$ follows automatically.であり(port 1 の場合と同様)、和は 0 になります → port 2 は完全に整合します。port 2 ↔ port 3 の対称性より、$S_{33} = 0$ は自動的に成り立ちます。

關鍵觀察The key observation重要な観察
  • Port 2 的半電路The port 2 half-circuit port 2 の半回路は恰好就是 port 1 半電路is precisely the port 1 half-circuitちょうど port 1 の半回路そのもの,只是從另一端看;所以兩者的 $\Gamma^e, \Gamma^o$ 數值, merely viewed from the other end; hence the values of $\Gamma^e, \Gamma^o$ for the two are であり、単に反対側から見ているだけです。したがって両者の $\Gamma^e, \Gamma^o$ の値は完全相同exactly identical完全に同一です($\mp j/\sqrt2$)。這保證了 $S_{11}=S_{22}=0$ 同時成立。($\mp j/\sqrt2$). That guarantees $S_{11}=S_{22}=0$ hold together.($\mp j/\sqrt2$)。これにより $S_{11}=S_{22}=0$ が同時に成り立つことが保証されます。
  • $\frac{1}{2}(\Gamma_2^e - \Gamma_2^o) = -j/\sqrt2$ 不為零 — 這是 port 2 到$\frac{1}{2}(\Gamma_2^e - \Gamma_2^o) = -j/\sqrt2$ is not zero — it is the coupling from port 2 to $\frac{1}{2}(\Gamma_2^e - \Gamma_2^o) = -j/\sqrt2$ はゼロではありません — これは port 2 から它鏡像埠its mirror portその鏡像ポート的耦合(rat-race 的功率分配,$|S_{2,\text{mirror}}| = 1/\sqrt2$),不是 isolation。Port 2 與 port 3 之間的隔離($S_{23}=0$)來自(the rat-race's power split, $|S_{2,\text{mirror}}| = 1/\sqrt2$), not isolation. The isolation between ports 2 and 3 ($S_{23}=0$) comes from a への結合(rat-race の電力分配、$|S_{2,\text{mirror}}| = 1/\sqrt2$)であり、isolation ではありません。port 2 と port 3 の間のアイソレーション($S_{23}=0$)は不同different異なる的分析(兩條路徑相消干涉)。 argument (destructive interference of two paths).議論(2 つの経路の相殺干渉)によるものです。
  • 之所以「(λ/8, λ/8) 並聯」算不出 $\Gamma=0$,是因為它The reason “(λ/8, λ/8) in parallel” fails to give $\Gamma=0$ is that it 「(λ/8, λ/8) の並列」で $\Gamma=0$ が求められないのは、それが缺少了鄰埠那段 λ/4 線和 $Z_0$ 載入omits the contribution of the neighbouring port's λ/4 line and its $Z_0$ loading隣接ポートの λ/4 線路と $Z_0$ 負荷の寄与を欠いている的貢獻。把它們算進來就回到上面這個完整的半電路。. Include those and you are back to the complete half-circuit above.からです。それらを算入すれば、上記の完全な半回路に戻ります。

更簡單的另一條路:對稱 + 互易。A simpler alternative route: symmetry plus reciprocity.より簡単なもう一つの道:対称性 + 相反性。rat-race 是被動互易、無耗的四埠網路。已證 $S_{11}=0$ 和 isolation/分配比例。把 $[S]$ 完整寫出來、套用 $[S]^*[S] = U$(unitarity,無耗)與 $[S]^T=[S]$(reciprocity)兩條件,剩下的元素被唯一決定 — 直接給出 $S_{22}=S_{33}=0$。這條路不用再算半電路,但要先有 unitarity/reciprocity 框架。 The rat-race is a passive, reciprocal, lossless four-port. We have already proved $S_{11}=0$ along with the isolation and split ratios. Write out $[S]$ in full and impose $[S]^*[S] = U$ (unitarity, losslessness) and $[S]^T=[S]$ (reciprocity), and the remaining elements are uniquely determined — giving $S_{22}=S_{33}=0$ directly. This route needs no further half-circuit algebra, but it does require the unitarity/reciprocity framework first.rat-race は受動・相反・無損失の四ポート回路網です。すでに $S_{11}=0$ と isolation・分配比を証明しました。$[S]$ を完全に書き下し、$[S]^*[S] = U$(unitarity、無損失)と $[S]^T=[S]$(reciprocity)の二条件を課すと、残りの要素は一意に決まり — 直接 $S_{22}=S_{33}=0$ が得られます。この道はもう半回路の計算を必要としませんが、先に unitarity/reciprocity の枠組みが必要です。

③ 代入 $\beta L$③ Substituting $\beta L$③ $\beta L$ を代入する

計算每條 stub 的電氣長度:Compute the electrical length of each stub:各 stub の電気長を計算します:

$$\beta\!\cdot\!\tfrac{\lambda}{8} = \tfrac{2\pi}{\lambda}\!\cdot\!\tfrac{\lambda}{8} = \tfrac{\pi}{4} = 45°$$ $$\beta\!\cdot\!\tfrac{3\lambda}{8} = \tfrac{2\pi}{\lambda}\!\cdot\!\tfrac{3\lambda}{8} = \tfrac{3\pi}{4} = 135°$$

查 cot 值:Look up the cot values:cot の値を調べます:

$$\cot(\pi/4) = +1, \qquad \cot(3\pi/4) = -1$$
④ 結果:兩條 stub 各自的阻抗④ The result: the impedance of each stub④ 結果:2 本の stub それぞれのインピーダンス
$$Z_{\lambda/8} = -jZ_R\cot(\pi/4) = \boxed{-jZ_R}\quad\text{(電容性)}$$ $$Z_{3\lambda/8} = -jZ_R\cot(3\pi/4) = -jZ_R\cdot(-1) = \boxed{+jZ_R}\quad\text{(電感性)}$$ $$Z_{\lambda/8} = -jZ_R\cot(\pi/4) = \boxed{-jZ_R}\quad\text{(capacitive)}$$ $$Z_{3\lambda/8} = -jZ_R\cot(3\pi/4) = -jZ_R\cdot(-1) = \boxed{+jZ_R}\quad\text{(inductive)}$$ $$Z_{\lambda/8} = -jZ_R\cot(\pi/4) = \boxed{-jZ_R}\quad\text{(capacitive)}$$ $$Z_{3\lambda/8} = -jZ_R\cot(3\pi/4) = -jZ_R\cdot(-1) = \boxed{+jZ_R}\quad\text{(inductive)}$$

兩個阻抗都是純虛數,大小相等、符號相反 — 一個是電容、一個是電感。Both impedances are purely imaginary, equal in magnitude and opposite in sign — one capacitive, one inductive.二つのインピーダンスはいずれも純虚数で、大きさが等しく符号が反対です — 一方は容量性、他方は誘導性です。

⑤ Smith chart 視覺化⑤ Smith chart visualization⑤ Smith chart による可視化

把這兩條開路 stub 在 Smith chart 上走一遍:Walk these two open stubs around the Smith chart:この 2 本の開放 stub を Smith chart 上でたどってみます:

Smith chart: even-mode open-circuit stub rotations Open (Z=∞) Γ = +1 起點 (start)start起点 (start) +jZ_R (inductive) Γ = +j −jZ_R (capacitive) Γ = −j Short (Z=0) Γ = −1 Z = Z_R λ/8 stub 轉 90° CWrotate 90° CW90° CW 回転 3λ/8 stub 轉 270° CWrotate 270° CW270° CW 回転 (= 90° CCW) 兩個阻抗在虛軸上對稱 → 並聯時分母 = 0 → $Z_{in}$ = ∞The two impedances are symmetric about the imaginary axis → in parallel the denominator = 0 → $Z_{in}$ = ∞二つのインピーダンスは虚軸に関して対称 → 並列時に分母 = 0 → $Z_{in}$ = ∞

圖示解讀:Reading the figure:図の読み方:

  • 起點starting point起点:兩條 stub 末端都是: both stubs are terminated in :どちらの stub も末端は開路open-circuit開放(綠點,Smith chart 最右端 $\Gamma=+1$)(the green dot, at the far right of the Smith chart, $\Gamma=+1$)(緑点、Smith chart の最右端 $\Gamma=+1$)
  • 紅色路徑(λ/8 stub)The red path (the λ/8 stub)赤い経路(λ/8 stub):沿 Smith chart 順時針旋轉 $2\beta L = 90°$(朝向 generator),終點在 $-j$(最下面),對應阻抗 $-jZ_R$ — : rotates clockwise around the Smith chart by $2\beta L = 90°$ (towards the generator), ending at $-j$ (the very bottom), corresponding to an impedance of $-jZ_R$ — :Smith chart 上を時計回りに $2\beta L = 90°$ 回転し(generator に向かって)、終点は $-j$(最下部)、対応するインピーダンスは $-jZ_R$ — 純電容purely capacitive純容量性
  • 藍色路徑(3λ/8 stub)The blue path (the 3λ/8 stub)青い経路(3λ/8 stub):沿 Smith chart 順時針旋轉 $2\beta L = 270°$(= 逆時針 90°),終點在 $+j$(最上面),對應阻抗 $+jZ_R$ — : rotates clockwise by $2\beta L = 270°$ (= 90° anticlockwise), ending at $+j$ (the very top), corresponding to an impedance of $+jZ_R$ — :時計回りに $2\beta L = 270°$ 回転し(= 反時計回りに 90°)、終点は $+j$(最上部)、対応するインピーダンスは $+jZ_R$ — 純電感purely inductive純誘導性
  • 並聯結果The parallel combination並列の結果:兩個阻抗在虛軸上對稱($+jZ_R$ 與 $-jZ_R$),形成: the two impedances are symmetric about the imaginary axis ($+jZ_R$ and $-jZ_R$), forming a :二つのインピーダンスは虚軸に関して対称($+jZ_R$ と $-jZ_R$)であり、並聯 LC 共振parallel LC resonance並列 LC 共振,總阻抗 $\to \infty$, so the total impedance $\to \infty$を形成するため、全インピーダンス $\to \infty$
⑥ 物理直覺:為何剛好是 LC 共振?⑥ Physical intuition: why does it land exactly on LC resonance?⑥ 物理的直感:なぜちょうど LC 共振になるのか?

這不是巧合 — $\lambda/8$ 加上 $3\lambda/8$ 剛好等於 $\lambda/2$(總路徑長為半波長)。沿著環走完半圈時的相位累積是 $\beta\cdot\lambda/2 = \pi = 180°$。所以兩條 stub 的反射波到達 port 1 時剛好相位差 $180°$ → This is no coincidence — $\lambda/8$ plus $3\lambda/8$ is exactly $\lambda/2$ (a total path of half a wavelength). The phase accumulated over half a turn around the ring is $\beta\cdot\lambda/2 = \pi = 180°$. So the reflected waves from the two stubs arrive back at port 1 exactly $180°$ apart in phase → これは偶然ではありません — $\lambda/8$ に $3\lambda/8$ を加えるとちょうど $\lambda/2$(全経路長が半波長)になります。リングを半周する間に累積する位相は $\beta\cdot\lambda/2 = \pi = 180°$ です。したがって 2 本の stub からの反射波が port 1 に戻るとき、位相差はちょうど $180°$ となり → 對消干涉destructive interference打ち消し干渉.

從 Smith chart 上看,這就是:起點同位置(都從開路出發),但兩條路徑長度相差 $\lambda/2$(對應 Smith chart 上 $360°$ 的相對旋轉差),所以兩個終點恰好On the Smith chart this reads as: the two paths start at the same place (both from an open circuit) but differ in length by $\lambda/2$ (a relative rotation of $360°$ on the Smith chart), so their two end points are exactly Smith chart 上で見ると、これは:出発点が同じ位置(どちらも開放から出発)で、二つの経路の長さが $\lambda/2$ 異なる(Smith chart 上で $360°$ の相対的な回転差に対応)ため、二つの終点はちょうど在 Smith chart 上對稱於實軸symmetric about the real axis of the Smith chartSmith chart 上で実軸に対称(一個在 $+j$、一個在 $-j$)→ 形成 LC 共振 → 並聯 = ∞。(one at $+j$, one at $-j$) → they form an LC resonance → in parallel = ∞.(一方は $+j$、他方は $-j$)→ LC 共振を形成 → 並列 = ∞。

並聯諧振動畫 — 為什麼 $+jZ_R \parallel -jZ_R \to \infty$(open)Parallel-resonance animation — why $+jZ_R \parallel -jZ_R \to \infty$ (an open circuit)並列共振アニメーション — なぜ $+jZ_R \parallel -jZ_R \to \infty$(open)なのか
無耗 tank:$\ddot v + \tfrac{1}{LC}v = 0 \Rightarrow v(t)=V_0\cos\omega_0 t$,$\omega_0=1/\sqrt{LC}$,電壓與電流相差 90°。共振時 $i_L$ 與 $i_C$ 大小相等、方向相反 → 端點淨電流 = 0 → 阻抗 $\to\infty$ = open。A lossless tank: $\ddot v + \tfrac{1}{LC}v = 0 \Rightarrow v(t)=V_0\cos\omega_0 t$ with $\omega_0=1/\sqrt{LC}$, and voltage and current 90° apart. At resonance $i_L$ and $i_C$ are equal in magnitude and opposite in direction → the net current at the terminals = 0 → the impedance $\to\infty$ = an open circuit.無損失 tank:$\ddot v + \tfrac{1}{LC}v = 0 \Rightarrow v(t)=V_0\cos\omega_0 t$、$\omega_0=1/\sqrt{LC}$、電圧と電流は 90° ずれます。共振時に $i_L$ と $i_C$ は大きさが等しく向きが逆 → 端子の正味電流 = 0 → インピーダンス $\to\infty$ = open。
對應 rat-race:$\lambda/8$ stub 的 $-jZ_R$(容性)與 $3\lambda/8$ stub 的 $+jZ_R$(感性)並聯,正是這個並聯 LC 共振 → 每個埠看到 open。Applied to the rat-race: the $-jZ_R$ (capacitive) of the $\lambda/8$ stub in parallel with the $+jZ_R$ (inductive) of the $3\lambda/8$ stub is exactly this parallel LC resonance → every port sees an open circuit.rat-race に対応:$-jZ_R$ stub の $\lambda/8$(容量性)と $+jZ_R$ stub の $3\lambda/8$(誘導性)の並列が、まさにこの並列 LC 共振です → 各ポートは open を見ます。
C + + + − − − L 能量Energyエネルギー E_C E_L E_total (固定)E_total (constant)E_total (一定) 波形 v(t), i(t)Waveforms v(t), i(t)波形 v(t), i(t) v i
這就是 rat-race 的「魔法」This is the rat-race's “magic”これが rat-race の「魔法」です
$\sqrt{2}Z_0$ 的環阻抗 + 三段 $\lambda/4$ + 一段 $3\lambda/4$ 的不對稱設計,讓 A ring impedance of $\sqrt{2}Z_0$ plus the asymmetric design of three $\lambda/4$ sections and one $3\lambda/4$ section makes $\sqrt{2}Z_0$ の環インピーダンス + $\lambda/4$ 三段 + $3\lambda/4$ 一段という非対称設計により、每個 mode 在每個埠都看到「並聯諧振 = open」every mode see a “parallel resonance = open circuit” at every port各 mode が各ポートで「並列共振 = open」を見ます。再經過外部 $\lambda/4$ 線段的阻抗變換後,就剛好匹配到 $Z_0$(外部 50 Ω)。這就是為什麼 rat-race 能做到「四個埠都 $S_{ii}=0$」這個 3-port tee 做不到的事。 . After the impedance transformation of the external $\lambda/4$ sections, that comes out matched to $Z_0$ (the external 50 Ω). This is how a rat-race achieves “$S_{ii}=0$ at all four ports”, something a 3-port tee cannot do. 。さらに外部の $\lambda/4$ 線路区間によるインピーダンス変換を経ると、ちょうど $Z_0$(外部の 50 Ω)に整合します。これが、rat-race が 3-port tee にはできない「4 つのポートすべてが $S_{ii}=0$」を実現できる理由です。
λ/4 阻抗變換:為什麼每個埠剛好匹配到 Z₀The λ/4 impedance transformation: why every port comes out matched to Z₀λ/4 インピーダンス変換:なぜ各ポートがちょうど Z₀ に整合するのか ① 原理:λ/4 線段是「阻抗反相器」① The principle: a λ/4 section is an “impedance inverter”① 原理:λ/4 線路区間は「インピーダンス反転器」 Z_in λ/4 線,Z_Rλ/4 line, Z_Rλ/4 線路、Z_R λ/4 Z_L Z_in = Z_R² / Z_L 負載大 → 輸入小,負載小 → 輸入大(反相)A large load → a small input; a small load → a large input (inverted)負荷が大 → 入力が小、負荷が小 → 入力が大(反転) ② 套用到 rat-race 埠:兩條相鄰 λ/4 臂並聯② Applied to a rat-race port: two adjacent λ/4 arms in parallel② rat-race のポートへの適用:隣接する 2 本の λ/4 アームの並列 埠 (外部 Z₀ = 50 Ω)Port (external Z₀ = 50 Ω)ポート (外部 Z₀ = 50 Ω) 看進環 = ?Looking into the ring = ?環を見込むと = ? λ/4, Z_R = √2·Z₀ λ/4 Z₀ λ/4, Z_R = √2·Z₀ λ/4 Z₀ Z_R² / Z₀ = 2Z₀² / Z₀ = 2Z₀ Z_R² / Z₀ = 2Z₀ 節點看進環 = 2Z₀ ∥ 2Z₀ = Z₀Looking into the ring from the node = 2Z₀ ∥ 2Z₀ = Z₀節点から環を見込む = 2Z₀ ∥ 2Z₀ = Z₀ 與外部 Z₀ 完全相同 → 無反射 → S_ii = 0 ✓Identical to the external Z₀ → no reflection → S_ii = 0 ✓外部の Z₀ と完全に同じ → 反射なし → S_ii = 0 ✓
$\lambda/4$ 線是「阻抗反相器」:$Z_\text{in}=Z_R^2/Z_L$。在 rat-race 的每個埠,兩條相鄰 $\lambda/4$ 臂各把對面的 $Z_0$ 變換成 $Z_R^2/Z_0 = 2Z_0$(因為 $Z_R=\sqrt2 Z_0$);兩個 $2Z_0$ 並聯回到 $Z_0$,剛好等於外部饋入線阻抗 → 該埠匹配($S_{ii}=0$)。這就是「外部 $\lambda/4$ 線段阻抗變換 → 匹配 $Z_0$」的具體機制。A $\lambda/4$ line is an “impedance inverter”: $Z_\text{in}=Z_R^2/Z_L$. At each rat-race port, the two adjacent $\lambda/4$ arms each transform the $Z_0$ at their far end into $Z_R^2/Z_0 = 2Z_0$ (since $Z_R=\sqrt2 Z_0$); two $2Z_0$ in parallel come back to $Z_0$, exactly equal to the external feed-line impedance → that port is matched ($S_{ii}=0$). That is the concrete mechanism behind “the external $\lambda/4$ impedance transformation → matched to $Z_0$”.$\lambda/4$ 線路は「インピーダンス反転器」です:$Z_\text{in}=Z_R^2/Z_L$。rat-race の各ポートでは、隣接する 2 本の $\lambda/4$ アームがそれぞれ反対端の $Z_0$ を $Z_R^2/Z_0 = 2Z_0$ に変換し($Z_R=\sqrt2 Z_0$ のため)、2 つの $2Z_0$ を並列にすると $Z_0$ に戻り、ちょうど外部の給電線インピーダンスに等しくなります → そのポートは整合します($S_{ii}=0$)。これが「外部の $\lambda/4$ 線路区間によるインピーダンス変換 → $Z_0$ に整合」の具体的な仕組みです。
❓ 為什麼 $Z_R$ 在公式裡是「平方」?(點擊展開)❓ Why is $Z_R$ squared in the formula? (click to expand)❓ なぜ $Z_R$ は式の中で「二乗」なのか?(クリックで展開)

$Z_R^2$ 的「平方」可以從三個角度看,每個都揭示同一件事的不同面相。The “square” in $Z_R^2$ can be seen from three angles, each revealing a different facet of the same fact.$Z_R^2$ の「二乗」は三つの角度から見ることができ、それぞれが同じ事実の異なる側面を明らかにします。

① 從一般傳輸線公式取極限① Take the limit of the general transmission-line formula① 一般の伝送線路公式から極限をとる

任意長度 $\ell$ 的無耗傳輸線(特性阻抗 $Z_R$、負載 $Z_L$):A lossless line of arbitrary length $\ell$ (characteristic impedance $Z_R$, load $Z_L$):任意の長さ $\ell$ の無損失伝送線路(特性インピーダンス $Z_R$、負荷 $Z_L$):

$$Z_\text{in}(\ell) = Z_R\cdot\frac{Z_L + jZ_R\tan(\beta\ell)}{Z_R + jZ_L\tan(\beta\ell)}$$
📐 這條公式怎麼來?(5 步推導,點擊展開)📐 Where does this formula come from? (a 5-step derivation, click to expand)📐 この公式はどこから来るのか?(5 ステップの導出、クリックで展開)

Step 1 — 無耗線上的電壓與電流Step 1 — voltage and current on a lossless lineStep 1 — 無損失線路上の電圧と電流(前進波 $V^+$ + 反射波 $V^-$ 疊加):(the superposition of a forward wave $V^+$ and a reflected wave $V^-$):(進行波 $V^+$ と反射波 $V^-$ の重ね合わせ):

$$V(z) = V^+e^{-j\beta z} + V^- e^{+j\beta z}, \qquad Z_R\,I(z) = V^+e^{-j\beta z} - V^- e^{+j\beta z}$$

$I$ 前面的 $Z_R$ 是 Ohm's law 在行進波上的版本:前進波載電流 $V^+/Z_R$。The $Z_R$ in front of $I$ is Ohm's law in its travelling-wave form: the forward wave carries a current $V^+/Z_R$.$Z_R$ の前にある $I$ は Ohm's law の進行波版です:進行波は電流 $V^+/Z_R$ を運びます。

❓ 為什麼 $V$ 是「+」、$I$ 是「−」?— 互動展示❓ Why is $V$ a “+” and $I$ a “−”? — an interactive demonstration❓ なぜ $V$ は「+」で $I$ は「−」なのか?— インタラクティブな実演

物理本源:The physical root:物理的な根源:

  • 特性阻抗 $Z_R$ 定義為「the characteristic impedance $Z_R$ is defined as “特性インピーダンス $Z_R$ は「波本身行進方向上的 V/I 比the V/I ratio in the direction the wave itself travels波自身が進む方向における V/I 比」。前進波(往 +z 走)的電流順著 +z 流:$I^+ = +V^+/Z_R$。”. The forward wave (travelling towards +z) carries current along +z: $I^+ = +V^+/Z_R$.」と定義されます。進行波(+z へ進む)の電流は +z に沿って流れます:$I^+ = +V^+/Z_R$。
  • 反射波(往 −z 走)的「自然」電流方向是 −z;但我們用The reflected wave (travelling towards −z) has a “natural” current direction of −z; but we measure current against 反射波(−z へ進む)の「自然な」電流の向きは −z ですが、電流を測るのに同一條 +z 參考線the same +z reference同じ +z の基準線量電流,要加負號:$I^- = -V^-/Z_R$。, so it takes a minus sign: $I^- = -V^-/Z_R$.を用いるため、マイナス符号が必要です:$I^- = -V^-/Z_R$。
  • 電壓Voltage電圧是兩導線間的電位差,是純量 → 兩波直接相加 (+)。 is the potential difference between the two conductors, a scalar → the two waves simply add (+).は二導体間の電位差であり、スカラーです → 二つの波は単純に加算されます (+)。電流Current電流有方向 → backward wave 相對 +z 參考是反向,要相減 (−)。 has direction → the backward wave runs opposite to the +z reference, so it must be subtracted (−).は向きを持ちます → backward wave は +z 基準に対して逆向きなので、減算されます (−)。

下方互動:The interactive below:下のインタラクティブ:切換顯示前進波/反射波/兩者;拖動「時間」滑桿讓波移動,觀察上方導線的 switch between showing the forward wave, the reflected wave, or both; drag the “time” slider to make the waves move and watch the 進行波/反射波/両方の表示を切り替え、「時間」スライダーをドラッグして波を動かし、上側の導体の電流箭頭方向direction of the current arrows電流矢印の向き:前進波 → 永遠向右,反射波 → 永遠向左。 on the upper conductor: the forward wave always points right, the reflected wave always points left.を観察します:進行波 → 常に右向き、反射波 → 常に左向き。

電壓 V(z, t)Voltage V(z, t)電圧 V(z, t) z → 0 +1 −1 V⁺ (forward, →) V⁻ (backward, ←) 傳輸線Transmission line伝送線路 top return I 方向Direction of II の向き V = … Z_R·I = … probe @ 中點probe @ midpointprobe @ 中点

觀察重點:What to look for:観察のポイント:切到「兩個都顯示」,找一個 V⁺ 與 V⁻ Switch to “show both” and find a position where V⁺ and V⁻ have 「両方を表示」に切り替え、V⁺ と V⁻ が 同號的位置the same sign同符号になる位置 → V 相加變大、$Z_R\!\cdot\!I$ 相減趨近 0(電壓駐波 → the V terms add and grow while the $Z_R\!\cdot\!I$ terms subtract towards 0 (a voltage → V の項は加算されて大きくなり、$Z_R\!\cdot\!I$ の項は相殺されて 0 に近づきます(電圧定在波のantinode、電流 and a current 、電流のnode)。再找一個 V⁺ 與 V⁻ ). Then find a position where V⁺ and V⁻ have )。次に V⁺ と V⁻ が 異號的位置opposite signs異符号になる位置 → V 相消、$Z_R\!\cdot\!I$ 相加變大(電壓 → the V terms cancel while the $Z_R\!\cdot\!I$ terms add and grow (a voltage → V の項は打ち消し合い、$Z_R\!\cdot\!I$ の項は加算されて大きくなります(電圧のnode、電流 and a current 、電流のantinode)。這就是 $V$ 用「+」、$I$ 用「−」差別的最直接證據。). That is the most direct evidence of the difference between the “+” for $V$ and the “−” for $I$.)。これが $V$ は「+」、$I$ は「−」という違いの最も直接的な証拠です。

Step 2 — 在 $z=0$(負載端)套邊界條件,定義反射係數 $\Gamma$:Step 2 — apply the boundary condition at $z=0$ (the load end) and define the reflection coefficient $\Gamma$:Step 2 — $z=0$(負荷端)で境界条件を適用し、反射係数 $\Gamma$ を定義します:

$$Z_L = \frac{V(0)}{I(0)} = Z_R\cdot\frac{V^+ + V^-}{V^+ - V^-} \;\Rightarrow\; \Gamma \equiv \frac{V^-}{V^+} = \frac{Z_L - Z_R}{Z_L + Z_R}$$

Step 3 — 在輸入端 $z = -\ell$ 算 $Z_\text{in}$,分子分母同除 $V^+e^{j\beta\ell}$:Step 3 — compute $Z_\text{in}$ at the input $z = -\ell$, dividing numerator and denominator by $V^+e^{j\beta\ell}$:Step 3 — 入力端 $Z_\text{in}$ で $z = -\ell$ を計算し、分子分母を $V^+e^{j\beta\ell}$ で割ります:

$$Z_\text{in} = Z_R\cdot\frac{V^+e^{j\beta\ell} + V^- e^{-j\beta\ell}}{V^+e^{j\beta\ell} - V^- e^{-j\beta\ell}} = Z_R\cdot\frac{1 + \Gamma\,e^{-j2\beta\ell}}{1 - \Gamma\,e^{-j2\beta\ell}}$$

Step 4 — 代入 $\Gamma=(Z_L-Z_R)/(Z_L+Z_R)$,分子分母再乘 $e^{j\beta\ell}$:Step 4 — substitute $\Gamma=(Z_L-Z_R)/(Z_L+Z_R)$ and multiply numerator and denominator by $e^{j\beta\ell}$:Step 4 — $\Gamma=(Z_L-Z_R)/(Z_L+Z_R)$ を代入し、分子分母にさらに $e^{j\beta\ell}$ を掛けます:

$$\text{分子}=(Z_L+Z_R)e^{j\beta\ell}+(Z_L-Z_R)e^{-j\beta\ell} = Z_L\underbrace{(e^{j\beta\ell}{+}e^{-j\beta\ell})}_{2\cos\beta\ell}+Z_R\underbrace{(e^{j\beta\ell}{-}e^{-j\beta\ell})}_{2j\sin\beta\ell}$$ $$\quad= 2\bigl[\,Z_L\cos\beta\ell + jZ_R\sin\beta\ell\,\bigr]$$ $$\text{分母}=2\bigl[\,Z_R\cos\beta\ell + jZ_L\sin\beta\ell\,\bigr]\quad\text{(同樣展開)}$$ $$\text{numerator}=(Z_L+Z_R)e^{j\beta\ell}+(Z_L-Z_R)e^{-j\beta\ell} = Z_L\underbrace{(e^{j\beta\ell}{+}e^{-j\beta\ell})}_{2\cos\beta\ell}+Z_R\underbrace{(e^{j\beta\ell}{-}e^{-j\beta\ell})}_{2j\sin\beta\ell}$$ $$\quad= 2\bigl[\,Z_L\cos\beta\ell + jZ_R\sin\beta\ell\,\bigr]$$ $$\text{denominator}=2\bigl[\,Z_R\cos\beta\ell + jZ_L\sin\beta\ell\,\bigr]\quad\text{(expanded the same way)}$$ $$\text{numerator}=(Z_L+Z_R)e^{j\beta\ell}+(Z_L-Z_R)e^{-j\beta\ell} = Z_L\underbrace{(e^{j\beta\ell}{+}e^{-j\beta\ell})}_{2\cos\beta\ell}+Z_R\underbrace{(e^{j\beta\ell}{-}e^{-j\beta\ell})}_{2j\sin\beta\ell}$$ $$\quad= 2\bigl[\,Z_L\cos\beta\ell + jZ_R\sin\beta\ell\,\bigr]$$ $$\text{denominator}=2\bigl[\,Z_R\cos\beta\ell + jZ_L\sin\beta\ell\,\bigr]\quad\text{(expanded the same way)}$$

Step 5 — 分子分母同除 $\cos\beta\ell$:Step 5 — divide numerator and denominator by $\cos\beta\ell$:Step 5 — 分子分母を $\cos\beta\ell$ で割ります:

$$\boxed{\;Z_\text{in} = Z_R\cdot\frac{Z_L + jZ_R\tan\beta\ell}{Z_R + jZ_L\tan\beta\ell}\;}\quad\text{QED.}$$

檢驗(用三個極限):Checks (three limiting cases):検証(三つの極限で):

  • $\ell\to 0$(沒有線):$\tan\to 0 \Rightarrow Z_\text{in}=Z_R\cdot Z_L/Z_R = Z_L$ ✓$\ell\to 0$ (no line at all): $\tan\to 0 \Rightarrow Z_\text{in}=Z_R\cdot Z_L/Z_R = Z_L$ ✓$\ell\to 0$(線路なし):$\tan\to 0 \Rightarrow Z_\text{in}=Z_R\cdot Z_L/Z_R = Z_L$ ✓
  • $\ell=\lambda/4$:$\tan\to\infty$,分子分母同除 $\tan$ → $Z_\text{in}=Z_R\cdot(jZ_R)/(jZ_L)=Z_R^2/Z_L$ ✓(阻抗反相器)$\ell=\lambda/4$: $\tan\to\infty$; dividing numerator and denominator by $\tan$ → $Z_\text{in}=Z_R\cdot(jZ_R)/(jZ_L)=Z_R^2/Z_L$ ✓ (an impedance inverter)$\ell=\lambda/4$:$\tan\to\infty$、分子分母を $\tan$ で割ると → $Z_\text{in}=Z_R\cdot(jZ_R)/(jZ_L)=Z_R^2/Z_L$ ✓(インピーダンス反転器)
  • $\ell=\lambda/2$:$\tan=0 \Rightarrow Z_\text{in}=Z_L$ ✓(半波線「透明」)$\ell=\lambda/2$: $\tan=0 \Rightarrow Z_\text{in}=Z_L$ ✓ (a half-wave line is “transparent”)$\ell=\lambda/2$:$\tan=0 \Rightarrow Z_\text{in}=Z_L$ ✓(半波長線路は「透明」)

幾何意義:Geometric meaning:幾何学的な意味:分子分母裡的兩項 $(Z_L, jZ_R\tan)$ 與 $(Z_R, jZ_L\tan)$ 正是行進波展開成 cos 與 sin 後的「實部 + 虛部」 — $\cos\beta\ell$ 那項保留了「直接看到負載」的成分,$\sin\beta\ell$ 那項是「線本身的特性阻抗反映」。整個分式可看成 the two terms $(Z_L, jZ_R\tan)$ and $(Z_R, jZ_L\tan)$ in numerator and denominator are precisely the “real + imaginary parts” of the travelling waves expanded into cos and sin — the $\cos\beta\ell$ term retains the “looking straight at the load” component, while the $\sin\beta\ell$ term is “the line's own characteristic impedance showing through”. The whole fraction can be read as 分子分母にある二つの項 $(Z_L, jZ_R\tan)$ と $(Z_R, jZ_L\tan)$ は、まさに進行波を cos と sin に展開したときの「実部 + 虚部」です — $\cos\beta\ell$ の項は「負荷を直接見ている」成分を保ち、$\sin\beta\ell$ の項は「線路自身の特性インピーダンスが現れたもの」です。この分数全体は$Z_L\leftrightarrow Z_R$ 在分子分母間互相交換$Z_L\leftrightarrow Z_R$ swapping places between numerator and denominator$Z_L\leftrightarrow Z_R$ が分子と分母の間で入れ替わる,這正是線的「阻抗反相」性質的母體。, which is the parent of the line's impedance-inverting property.ものと見なすことができ、これが線路の「インピーダンス反転」性質の母体です。

令 $\ell = \lambda/4 \Rightarrow \beta\ell = \tfrac{\pi}{2} \Rightarrow \tan(\beta\ell)\to\infty$。把分子分母同除以 $\tan$,沒有 $\tan$ 那兩項丟掉:Set $\ell = \lambda/4 \Rightarrow \beta\ell = \tfrac{\pi}{2} \Rightarrow \tan(\beta\ell)\to\infty$. Dividing numerator and denominator by $\tan$ discards the two terms without a $\tan$:$\ell = \lambda/4 \Rightarrow \beta\ell = \tfrac{\pi}{2} \Rightarrow \tan(\beta\ell)\to\infty$ とします。分子分母を $\tan$ で割ると、$\tan$ を含まない二つの項は落ちます:

$$Z_\text{in} = Z_R\cdot\frac{jZ_R}{jZ_L} = \boxed{\;\frac{Z_R^2}{Z_L}\;}$$

第一個 $Z_R$The first $Z_R$1 つ目の $Z_R$ 是公式前面的固定係數(任何 T-line 公式都有); is the fixed prefactor of the formula (present in every T-line formula); は公式の前にある固定係数です(どの T-line 公式にもあります);第二個 $Z_R$the second $Z_R$2 つ目の $Z_R$ 是分子裡唯一沒被丟掉的項。兩個 $Z_R$ 相乘 → $Z_R^2$。 is the only surviving term in the numerator. Multiply the two $Z_R$ together → $Z_R^2$. は分子で唯一落ちずに残る項です。二つの $Z_R$ を掛け合わせると → $Z_R^2$。

② 物理直覺:幾何平均② Physical intuition: the geometric mean② 物理的直感:幾何平均

反過來解出讓 $Z_L$ 完美匹配到 $Z_\text{in}$ 所需的 $Z_R$:Turn it round and solve for the $Z_R$ needed to match $Z_L$ perfectly to $Z_\text{in}$:逆に、$Z_R$ を $Z_L$ に完全に整合させるために必要な $Z_\text{in}$ を解きます:

$$Z_R = \sqrt{Z_\text{in}\cdot Z_L}$$

$\lambda/4$ 線是A $\lambda/4$ line is a $\lambda/4$ 線路は「幾何平均匹配段」“geometric-mean matching section”「幾何平均整合区間」。要把 $Z_L$ 變成 $Z_\text{in}$,中間插一段阻抗剛好. To turn $Z_L$ into $Z_\text{in}$, insert a length of line whose impedance lies exactly です。$Z_L$ を $Z_\text{in}$ に変えるには、間にインピーダンスがちょうど位於兩者對數中點at the logarithmic midpoint of the two両者の対数的な中点にある的傳輸線 — 那個中點就是幾何平均 $\sqrt{Z_\text{in}Z_L}$。兩邊平方就得到 $Z_R^2 = Z_\text{in}Z_L$。— and that midpoint is the geometric mean $\sqrt{Z_\text{in}Z_L}$. Square both sides and you get $Z_R^2 = Z_\text{in}Z_L$.伝送線路を挿入します — その中点が幾何平均 $\sqrt{Z_\text{in}Z_L}$ です。両辺を二乗すると $Z_R^2 = Z_\text{in}Z_L$ が得られます。

③ 波傳播觀點:兩端各貢獻一個 $Z_R$③ The wave-propagation view: each end contributes one $Z_R$③ 波の伝搬の観点:両端がそれぞれ $Z_R$ を一つずつ寄与する

$Z_R$ 線上電壓 $V$ 與電流 $I$ 由前進波 $V^+$ 與反射波 $V^-$ 組成:$V = V^+ + V^-$,$Z_R\cdot I = V^+ - V^-$。On a $Z_R$ line, the voltage $V$ and current $I$ are made up of the forward wave $V^+$ and the reflected wave $V^-$: $V = V^+ + V^-$ and $Z_R\cdot I = V^+ - V^-$.$Z_R$ 線路上の電圧 $V$ と電流 $I$ は、進行波 $V^+$ と反射波 $V^-$ から構成されます:$V = V^+ + V^-$、$Z_R\cdot I = V^+ - V^-$。

位置Location位置$V$$Z_R\cdot I$
負載端 ($z=0$)Load end ($z=0$)負荷端 ($z=0$)$V^+ + V^-$$V^+ - V^-$
輸入端 ($z=-\lambda/4$,相移 $\pm 90°$)Input end ($z=-\lambda/4$, a phase shift of $\pm 90°$)入力端 ($z=-\lambda/4$、位相シフト $\pm 90°$)$j(V^+ - V^-)$$j(V^+ + V^-)$

走完 $\lambda/4$ 後,After travelling $\lambda/4$, $\lambda/4$ を進んだ後、$V$ 與 $Z_R\cdot I$ 的角色對調the roles of $V$ and $Z_R\cdot I$ are exchanged$V$ と $Z_R\cdot I$ の役割が入れ替わります。所以:. So:。したがって:

$$Z_L = Z_R\cdot\frac{V^+ + V^-}{V^+ - V^-}, \qquad Z_\text{in} = Z_R\cdot\frac{V^+ - V^-}{V^+ + V^-}$$ $$\Rightarrow\; Z_\text{in}\cdot Z_L = Z_R^2 \;\Rightarrow\; Z_\text{in} = \frac{Z_R^2}{Z_L}$$

負載端用一個 $Z_R$ 把電流換成電壓At the load end one $Z_R$ converts current into voltage負荷端では一つの $Z_R$ が電流を電圧に変換し(得到 $Z_L$);(giving $Z_L$); ($Z_L$ が得られる);輸入端再用一個 $Z_R$ 換回來at the input end another $Z_R$ converts it back入力端ではもう一つの $Z_R$ がそれを戻します(得到 $Z_\text{in}$)。中間那段 $\lambda/4$ 把分子分母對調 → $Z$ 反轉。兩個 $Z_R$ 各來自一端,乘起來 = $Z_R^2$。(giving $Z_\text{in}$). The $\lambda/4$ section in between swaps numerator and denominator → $Z$ is inverted. Each $Z_R$ comes from one end, and multiplied together they give $Z_R^2$.($Z_\text{in}$ が得られる)。間の $\lambda/4$ 区間が分子分母を入れ替える → $Z$ が反転します。二つの $Z_R$ はそれぞれ一端から来ており、掛け合わせると = $Z_R^2$ になります。

套用到 rat-raceApplied to the rat-racerat-race への適用

環阻抗 $Z_R = \sqrt2\,Z_0$、負載 $Z_L = Z_0$:Ring impedance $Z_R = \sqrt2\,Z_0$ and load $Z_L = Z_0$:環インピーダンス $Z_R = \sqrt2\,Z_0$、負荷 $Z_L = Z_0$:

$$Z_\text{in} = \frac{Z_R^2}{Z_L} = \frac{(\sqrt2\,Z_0)^2}{Z_0} = \frac{2Z_0^2}{Z_0} = 2Z_0$$

兩條這樣的臂並聯:$2Z_0 \parallel 2Z_0 = Z_0$ — 剛好匹配外部 50 Ω。$\sqrt2$ 之所以是 $\sqrt2$,就是因為它要被「Two such arms in parallel: $2Z_0 \parallel 2Z_0 = Z_0$ — exactly matching the external 50 Ω. The reason $\sqrt2$ is $\sqrt2$ is that it has to be このようなアームを 2 本並列にすると:$2Z_0 \parallel 2Z_0 = Z_0$ — ちょうど外部の 50 Ω に整合します。$\sqrt2$ が $\sqrt2$ であるのは、それが「平方掉squared away二乗されて」變成 2,才能配合「兩臂並聯」的因子 1/2 → 還原成 $Z_0$。 into a 2 in order to pair with the factor of 1/2 from “two arms in parallel” → returning to $Z_0$.」2 になり、「2 本のアームの並列」による 1/2 の因子と組み合わさって → $Z_0$ に戻るためです。

一句話:In one sentence:一言で言うと:$Z_R^2$ 是「$\lambda/4$ 線兩端各一個 $Z_R$」的乘積。$\lambda/4$ 不是讓 $Z_R$「自己平方」,而是讓 $Z_R$ 在兩端各扮演一次「電壓⇄電流轉換係數」的角色,乘起來就出現 $Z_R^2$。$Z_R^2$ is the product of “one $Z_R$ at each end of the $\lambda/4$ line”. The $\lambda/4$ does not make $Z_R$ “square itself”; rather, it makes $Z_R$ play the role of “voltage⇄current conversion factor” once at each end, and multiplying those gives $Z_R^2$.$Z_R^2$ は「$Z_R$ 線路の両端にそれぞれ一つずつある $\lambda/4$」の積です。$\lambda/4$ は $Z_R$ に「自分自身を二乗させる」のではなく、$Z_R$ が両端でそれぞれ一度ずつ「電圧⇄電流変換係数」の役割を果たすようにし、それを掛け合わせると $Z_R^2$ が現れます。

偶模的輸入阻抗為The even-mode input impedance is 偶モードの入力インピーダンスは無限大infinite無限大 → 從 port 1 看就是 open。所以: → from port 1 it is simply an open circuit. Hence: → port 1 から見れば単に open です。したがって:

$$\Gamma_e = +1$$

偶模的傳輸 — 對稱地,port 2 與 port 3 接收同樣的訊號。算 $T_e$ 需要追蹤 $\lambda/4$ 線段的傳輸:Even-mode transmission — by symmetry, ports 2 and 3 receive the same signal. Computing $T_e$ requires tracking transmission along the $\lambda/4$ sections:偶モードの伝送 — 対称性により、port 2 と port 3 は同じ信号を受け取ります。$T_e$ を計算するには $\lambda/4$ 線路区間の伝送を追う必要があります:

$$T_e = \tfrac{-j}{\sqrt{2}}\;(\text{至 port 2 與 port 3 各一份})$$ $$T_e = \tfrac{-j}{\sqrt{2}}\;(\text{one share each to ports 2 and 3})$$ $$T_e = \tfrac{-j}{\sqrt{2}}\;(\text{one share each to ports 2 and 3})$$
偶模傳輸:為什麼 port 2 與 port 3 拿到相同的 $-j/\sqrt{2}$Even-mode transmission: why ports 2 and 3 both receive $-j/\sqrt{2}$偶モード伝送:なぜ port 2 と port 3 が同じ $-j/\sqrt{2}$ を得るのか 對稱平面plane of symmetry対称面 1 Σ in 2 3 4 Δ (isolated) λ/4 → −90° 對稱路徑symmetric paths対称な経路 $-j/\sqrt2$ $-j/\sqrt2$ $T_e$ 拆成兩個因子$T_e$ splits into two factors$T_e$ は二つの因子に分解される ① 等功率分配 (−3 dB)① Equal power split (−3 dB)① 等電力分配 (−3 dB) 輸入功率分到 2 個輸出 → 每個 $\tfrac12$ 功率The input power is divided between 2 outputs → each gets $\tfrac12$ the power入力電力が 2 つの出力に分配される → 各々が $\tfrac12$ の電力 電壓係數 = $\sqrt{1/2} = 1/\sqrt2$Voltage coefficient = $\sqrt{1/2} = 1/\sqrt2$電圧係数 = $\sqrt{1/2} = 1/\sqrt2$ ② $\lambda/4$ 線相位延遲 90°② The λ/4 line's 90° phase delay② λ/4 線路の 90° 位相遅延 傳輸相位因子 $e^{-j\beta\ell}$,$\beta\ell=\tfrac{\pi}{2}$Transmission phase factor $e^{-j\beta\ell}$ with $\beta\ell=\tfrac{\pi}{2}$伝送位相因子 $e^{-j\beta\ell}$、$\beta\ell=\tfrac{\pi}{2}$ $\Rightarrow e^{-j\pi/2} = -j$ $T_e = \dfrac{1}{\sqrt2}\times(-j) = \dfrac{-j}{\sqrt2}$ 至 port 2 與 port 3 各一份One share each to ports 2 and 3port 2 と port 3 に一つずつ 為什麼 port 2 = port 3?Why is port 2 = port 3?なぜ port 2 = port 3 なのか? 偶模 = 對稱激發,整個電路對The even mode is a symmetric excitation, so the whole circuit is偶モード = 対称励振なので、回路全体が 對稱平面鏡像 → port 2 與 port 3 是mirror-symmetric about the symmetry plane → ports 2 and 3 are a対称面に対して鏡像対称 → port 2 と port 3 は 鏡像對 → 收到mirror pair → they receive 鏡像対 → 受け取るのは完全相同exactly identical完全に同一(同相同幅)的訊號。 signals (same phase, same amplitude).(同相同振幅)の信号です。
偶模(Σ 激發)下,輸入從 port 1 經 $\lambda/4$ 臂傳到 port 2 與 port 3。傳輸係數 = In the even mode (Σ excitation) the input travels from port 1 along the $\lambda/4$ arms to ports 2 and 3. The transmission coefficient = 偶モード(Σ 励振)では、入力は port 1 から $\lambda/4$ アームを経て port 2 と port 3 に伝わります。伝送係数 = 等功率分配 $1/\sqrt2$an equal power split of $1/\sqrt2$等電力分配 $1/\sqrt2$ × $\lambda/4$ 相位延遲 $e^{-j\pi/2}=-j$and the $\lambda/4$ phase delay $e^{-j\pi/2}=-j$$\lambda/4$ の位相遅延 $e^{-j\pi/2}=-j$ = $-j/\sqrt2$。因為偶模是對稱激發、port 2 與 port 3 對對稱平面互為鏡像,所以兩者收到 = $-j/\sqrt2$. Because the even mode is a symmetric excitation and ports 2 and 3 are mirror images about the symmetry plane, they receive = $-j/\sqrt2$ です。偶モードは対称励振であり、port 2 と port 3 は対称面に対して互いに鏡像なので、両者が受け取るのは同相同幅identical signals in phase and amplitude同相同振幅的訊號 — 這正是 Σ 埠「同相分配」的行為。— which is exactly the “in-phase split” behaviour of the Σ port.の信号です — これがまさに Σ ポートの「同相分配」の振る舞いです。

Step 3 — Odd mode:對稱平面 = electric wall(短路)Step 3 — odd mode: the plane of symmetry is an electric wall (short circuit)Step 3 — Odd mode:対称面 = electric wall(短絡)

奇模時,對稱平面上電壓為零(如同短路)。兩個半環變成:In the odd mode the voltage on the plane of symmetry is zero (as if it were a short circuit). The two half-rings become: 奇モードでは、対称面上の電圧はゼロになります(短絡と同じ)。二つの半環は次のようになります:$\lambda/8$ 短路 stuba $\lambda/8$ short-circuited stub$\lambda/8$ の短絡 stub 並聯 in parallel with $3\lambda/8$ 短路 stuba $3\lambda/8$ short-circuited stub$3\lambda/8$ の短絡 stub の並列.

$$Z_o^{in} = \bigl[\,jZ_R\tan(\pi/4) \,\parallel\, jZ_R\tan(3\pi/4)\,\bigr] = \bigl[\,jZ_R \,\parallel\, -jZ_R\,\bigr] = \infty$$

同樣是 open!所以:which is again an open circuit! Hence:これも同じく open です!したがって:

$$\Gamma_o = +1$$

但奇模的傳輸方向相反(兩邊反相):But the odd mode transmits with the opposite sign (the two sides are in anti-phase):ただし奇モードの伝送は符号が逆になります(両側が逆相):

$$T_o = \tfrac{-j}{\sqrt{2}}\;\text{at port 2}, \quad T_o = \tfrac{+j}{\sqrt{2}}\;\text{at port 3}$$

等等!這裡得到的 $\Gamma_e = \Gamma_o = +1$(兩種模式從 port 1 都看到開路),這代表整顆 port 1 在沒有對稱激發時看到的並非匹配。實際分析要更小心:需要把Hold on! We have found $\Gamma_e = \Gamma_o = +1$ (both modes see an open circuit from port 1), which would mean that port 1 as a whole is not matched in the absence of a symmetric excitation. The real analysis has to be more careful: it must include 待ってください!ここで得られた $\Gamma_e = \Gamma_o = +1$(どちらのモードも port 1 から開放を見る)は、対称励振がない場合に port 1 全体が整合していないことを意味します。実際の解析はもっと慎重に行う必要があります:$\lambda/4$ 線段對 $Z_0$ 終結的影響the effect of the $\lambda/4$ sections terminated in $Z_0$$Z_0$ で終端された $\lambda/4$ 線路区間の影響納入計算。.を計算に入れる必要があります。

Step 4 — 更嚴謹的做法:每段 $\lambda/4$ 線做阻抗轉換Step 4 — a more rigorous approach: each $\lambda/4$ section performs an impedance transformationStep 4 — より厳密な方法:各 $\lambda/4$ 線路区間がインピーダンス変換を行う

$\lambda/4$ 變壓器把終端阻抗 $Z_L$ 轉成 $Z_R^2/Z_L$。port 2、3 看出去都是 $Z_0$,所以從環的「另一端」看回每個 port 是:A $\lambda/4$ transformer turns a terminating impedance $Z_L$ into $Z_R^2/Z_L$. Ports 2 and 3 both look out into $Z_0$, so looking back at each port from “the other side” of the ring gives:$\lambda/4$ 変成器は終端インピーダンス $Z_L$ を $Z_R^2/Z_L$ に変換します。port 2、3 は外側にどちらも $Z_0$ を見るので、環の「反対側」から各 port を見返すと:

$$Z_R^2 / Z_0 = (\sqrt{2}Z_0)^2 / Z_0 = 2 Z_0$$

在偶模/奇模分析裡,這些反射成份疊加後給出整顆 ring 的 $[S]$ matrix。Pozar §7.8 經完整計算後得到:In the even/odd mode analysis, superposing these reflected components yields the $[S]$ matrix of the entire ring. Pozar §7.8 works this through in full and obtains:偶モード/奇モード解析では、これらの反射成分を重ね合わせると ring 全体の $[S]$ matrix が得られます。Pozar §7.8 では完全な計算を行い、次を得ています:

$$\boxed{\;[S]_{180°} = \frac{-j}{\sqrt{2}}\begin{bmatrix} 0 & 1 & 1 & 0 \\ 1 & 0 & 0 & -1 \\ 1 & 0 & 0 & 1 \\ 0 & -1 & 1 & 0 \end{bmatrix}\;}$$

Step 5 — 驗證:對 Δ port (port 4) 激發給出 $\pm 90°$ 輸出Step 5 — verification: exciting the Δ port (port 4) gives $\pm 90°$ outputsStep 5 — 検証:Δ port (port 4) を励振すると $\pm 90°$ の出力が得られる

對 Δ 激發 $a_4 = V_{RF}$,其他 $a_i = 0$:Excite Δ with $a_4 = V_{RF}$ and all other $a_i = 0$:Δ を $a_4 = V_{RF}$ で励振し、他はすべて $a_i = 0$:

$$\begin{aligned} b_2 &= S_{24}\,a_4 = \tfrac{-j}{\sqrt{2}}\!\cdot\!(-1)\!\cdot\!V_{RF} = \tfrac{+j}{\sqrt{2}}V_{RF} \;\Rightarrow\; |b_2|=\tfrac{V_{RF}}{\sqrt{2}},\;\angle b_2 = \mathbf{+90°}\\ b_3 &= S_{34}\,a_4 = \tfrac{-j}{\sqrt{2}}\!\cdot\!(+1)\!\cdot\!V_{RF} = \tfrac{-j}{\sqrt{2}}V_{RF} \;\Rightarrow\; |b_3|=\tfrac{V_{RF}}{\sqrt{2}},\;\angle b_3 = \mathbf{-90°}\\ b_1 &= S_{14}\,a_4 = 0 \quad\text{(Σ 隔離)}\\ b_4 &= S_{44}\,a_4 = 0 \quad\text{(Δ 自我匹配)} \end{aligned}$$ $$\begin{aligned} b_2 &= S_{24}\,a_4 = \tfrac{-j}{\sqrt{2}}\!\cdot\!(-1)\!\cdot\!V_{RF} = \tfrac{+j}{\sqrt{2}}V_{RF} \;\Rightarrow\; |b_2|=\tfrac{V_{RF}}{\sqrt{2}},\;\angle b_2 = \mathbf{+90°}\\ b_3 &= S_{34}\,a_4 = \tfrac{-j}{\sqrt{2}}\!\cdot\!(+1)\!\cdot\!V_{RF} = \tfrac{-j}{\sqrt{2}}V_{RF} \;\Rightarrow\; |b_3|=\tfrac{V_{RF}}{\sqrt{2}},\;\angle b_3 = \mathbf{-90°}\\ b_1 &= S_{14}\,a_4 = 0 \quad\text{(Σ is isolated)}\\ b_4 &= S_{44}\,a_4 = 0 \quad\text{(Δ is self-matched)} \end{aligned}$$ $$\begin{aligned} b_2 &= S_{24}\,a_4 = \tfrac{-j}{\sqrt{2}}\!\cdot\!(-1)\!\cdot\!V_{RF} = \tfrac{+j}{\sqrt{2}}V_{RF} \;\Rightarrow\; |b_2|=\tfrac{V_{RF}}{\sqrt{2}},\;\angle b_2 = \mathbf{+90°}\\ b_3 &= S_{34}\,a_4 = \tfrac{-j}{\sqrt{2}}\!\cdot\!(+1)\!\cdot\!V_{RF} = \tfrac{-j}{\sqrt{2}}V_{RF} \;\Rightarrow\; |b_3|=\tfrac{V_{RF}}{\sqrt{2}},\;\angle b_3 = \mathbf{-90°}\\ b_1 &= S_{14}\,a_4 = 0 \quad\text{(Σ is isolated)}\\ b_4 &= S_{44}\,a_4 = 0 \quad\text{(Δ is self-matched)} \end{aligned}$$

所以 port 2 = $+90°$,port 3 = $-90°$,So port 2 is at $+90°$ and port 3 at $-90°$, したがって port 2 = $+90°$、port 3 = $-90°$ となり、相差 $180°$$180°$ apart$180°$ の差 — 正是 Δ-port 激發應該得到的差動輸出。 — exactly the differential output a Δ-port excitation should produce. — まさに Δ-port 励振で得られるべき差動出力です。

Step 6 — 驗證 [S] 滿足 unitarity(lossless)與 reciprocityStep 6 — verify that [S] satisfies unitarity (losslessness) and reciprocityStep 6 — [S] が unitarity(lossless)と reciprocity を満たすことを検証

Reciprocity: $[S]^T = [S]$ ✓ (矩陣對稱): $[S]^T = [S]$ ✓ (the matrix is symmetric): $[S]^T = [S]$ ✓ (行列は対称)

Unitarity: $[S]^\dagger[S] = [I]$。逐行檢查,例如第 2 列與自己內積:: $[S]^\dagger[S] = [I]$. Check it row by row; for example, the inner product of row 2 with itself:: $[S]^\dagger[S] = [I]$。行ごとに確認します。例えば第 2 行と自分自身との内積:

$$\sum_k |S_{2k}|^2 = \Bigl|\tfrac{-j}{\sqrt{2}}\Bigr|^2 + 0 + 0 + \Bigl|\tfrac{+j}{\sqrt{2}}\Bigr|^2 = \tfrac{1}{2} + \tfrac{1}{2} = 1 \;\checkmark$$

不同列的內積(驗證正交性),如第 2 列 vs 第 3 列:And the inner product of different rows (verifying orthogonality), such as row 2 against row 3:異なる行同士の内積(直交性の検証)、例えば第 2 行 vs 第 3 行:

$$\sum_k S_{2k}^* S_{3k} = (\tfrac{+j}{\sqrt{2}})(\tfrac{-j}{\sqrt{2}}) + 0 + 0 + (\tfrac{-j}{\sqrt{2}})^*\!(\tfrac{-j}{\sqrt{2}}) = \tfrac{1}{2} - \tfrac{1}{2} = 0 \;\checkmark$$

三個 sanity check 通通通過 → 矩陣正確。All three sanity checks pass → the matrix is correct.三つの sanity check がすべて通過 → 行列は正しいです。

總結對照Summary comparisonまとめの対照

分析方式Method of analysis解析方法處理Handles扱う対象port 3 結果Result at port 3port 3 の結果
單路徑相位相加Adding phases along a single path単一経路の位相の加算只看 $3\lambda/4 + \lambda/4$Only considers $3\lambda/4 + \lambda/4$$3\lambda/4 + \lambda/4$ のみを考慮$0°$ ❌
CW+CCW 簡單疊加Naive superposition of CW+CCWCW+CCW の単純な重ね合わせ$e^{-j180°} + e^{j0°}$(忽略 junction)$e^{-j180°} + e^{j0°}$ (ignoring the junctions)$e^{-j180°} + e^{j0°}$(junction を無視)對消 = $0$ ❌Cancellation = $0$ ❌打ち消し = $0$ ❌
Even-odd mode (完整)Even-odd mode (complete)Even-odd mode (完全)含 junction 散射 + 兩道波相干Includes junction scattering + coherence of the two wavesjunction 散乱 + 二つの波の干渉を含む$\pm 90°$ ✓

📖 完整推導📖 The full derivation📖 完全な導出:Pozar, : Pozar, :Pozar, Microwave Engineering, §7.8 "The 180° Hybrid",第 4 版 pp. 343–349。本節省略了 $\lambda/4$ 阻抗變壓的詳細阻抗計算,但結果與完整推導一致。, §7.8 “The 180° Hybrid”, 4th edition, pp. 343–349. This section omits the detailed impedance arithmetic of the $\lambda/4$ transformers, but the result agrees with the complete derivation., §7.8 “The 180° Hybrid”、第 4 版 pp. 343–349。本節では $\lambda/4$ インピーダンス変換の詳細な計算を省略していますが、結果は完全な導出と一致します。

解釋:圖中 Σ 埠那個 Explanation: what does the 解説:図中の Σ ポートにある Z₀ + 電阻 + 接地Z₀ + resistor + groundZ₀ + 抵抗 + 接地 符號是什麼意思? symbol at the Σ port in the figure mean? という記号は何を意味するのか?

它是 Σ 埠的It is the Σ port's これは Σ ポートの匹配終端(matched load termination)matched load termination整合終端(matched load termination) — 三個符號疊在一起: — three symbols stacked together: — 三つの記号が重なったものです:

符號Symbol記号意義Meaning意味
Z₀(標籤)(label)(ラベル)電阻值,等於系統的特性阻抗,the resistance value, equal to the system's characteristic impedance, 抵抗値、システムの特性インピーダンスに等しく、典型為 50 Ωtypically 50 Ω典型的には 50 Ω.
鋸齒形(紅)The zigzag (red)ギザギザ形(赤)一個阻值為 $Z_0$ 的電阻。a resistor of value $Z_0$.抵抗値 $Z_0$ の抵抗。
⏚(橫線)⏚ (the horizontal bars)⏚(横線)接地參考。the ground reference.接地基準。

整體就是:「Altogether it means: “全体としては:「一顆 $Z_0$ 的電阻,從 Σ 埠連到地a $Z_0$ resistor connected from the Σ port to groundΣ ポートから接地へつながる $Z_0$ の抵抗」。”.」。

那 Σ 埠到底是哪兩個訊號的「和」?So the Σ port is the “sum” of which two signals exactly?では Σ ポートは、いったいどの二つの信号の「和」なのでしょうか?

180° hybrid 的四個埠對應這個電路:The four ports of the 180° hybrid map onto this circuit as:180° hybrid の 4 つのポートはこの回路に次のように対応します:

Portポート位置Location位置連接到Connected to接続先
Port 1 (Δ)左上top left左上RF 輸入RF inputRF 入力
Port 2右上top right右上二極體環頂點 top vertex of the diode ring ダイオードリングの頂点 T
Port 3右下bottom right右下二極體環底點 bottom vertex of the diode ring ダイオードリングの底点 B
Port 4 (Σ)左下bottom left左下$Z_0$ 終結$Z_0$ termination$Z_0$ 終端

Σ 埠看到的是 port 2 + port 3 兩端訊號的「和」The Σ port sees the “sum” of the signals at ports 2 and 3Σ ポートが見るのは port 2 と port 3 の信号の「和」(共模 / common-mode 成份):(the common-mode component):(同相 / common-mode 成分):

$$v_\Sigma = \tfrac{1}{\sqrt{2}}\bigl(v_T + v_B\bigr) \qquad\text{相對地}\qquad v_\Delta = \tfrac{1}{\sqrt{2}}\bigl(v_T - v_B\bigr)$$ $$v_\Sigma = \tfrac{1}{\sqrt{2}}\bigl(v_T + v_B\bigr) \qquad\text{as against}\qquad v_\Delta = \tfrac{1}{\sqrt{2}}\bigl(v_T - v_B\bigr)$$ $$v_\Sigma = \tfrac{1}{\sqrt{2}}\bigl(v_T + v_B\bigr) \qquad\text{as against}\qquad v_\Delta = \tfrac{1}{\sqrt{2}}\bigl(v_T - v_B\bigr)$$

就是 Σ/Δ 名字的由來 — which is where the names Σ/Δ come from — これが Σ/Δ という名前の由来です — sum port / difference port。 port. port。

為什麼 Σ 埠通常沒訊號(理想情況)?Why does the Σ port normally carry no signal (ideally)?なぜ Σ ポートには通常(理想的には)信号が来ないのか? RF 從 Δ 進來後,hybrid 做反相分配:port 2 拿到 $+V_{RF}/\sqrt{2}$、port 3 拿到 $-V_{RF}/\sqrt{2}$。從二極體回看 Σ: Once the RF enters at Δ the hybrid splits it in anti-phase: port 2 receives $+V_{RF}/\sqrt{2}$ and port 3 receives $-V_{RF}/\sqrt{2}$. Looking back at Σ from the diodes: RF が Δ から入ると、hybrid は逆相で分配します:port 2 は $+V_{RF}/\sqrt{2}$ を、port 3 は $-V_{RF}/\sqrt{2}$ を受け取ります。ダイオード側から Σ を見返すと:

$$v_\Sigma = \tfrac{1}{\sqrt{2}}\Bigl[\underbrace{+V_{RF}/\sqrt{2}}_{v_T} \;+\; \underbrace{-V_{RF}/\sqrt{2}}_{v_B}\Bigr] = 0$$

兩個訊號剛好反相相加 → the two signals add exactly in anti-phase → 二つの信号がちょうど逆相で加算される → 完美對消perfect cancellation完全な打ち消し,Σ 埠 RF 能量 = 0。這就是 $S_{14}=0$(RF↔LO 完美隔離)的物理意義。, so the RF energy at the Σ port = 0. That is the physical meaning of $S_{14}=0$ (perfect RF↔LO isolation).、Σ ポートの RF エネルギー = 0 になります。これが $S_{14}=0$(RF↔LO の完全なアイソレーション)の物理的な意味です。

那 Σ 埠真正吃到的是什麼?So what does the Σ port actually absorb?では Σ ポートが実際に受け取るものは何か? 當兩條路徑因不對稱而產生「共模殘餘」時 — 例如: Whatever “common-mode residue” the two paths produce through asymmetry — for example: 二つの経路が非対称によって生み出す「同相残留」です — 例えば:

不對稱來源Source of asymmetry非対称の原因對 Σ 的影響Effect on ΣΣ への影響
二極體 RF 反射 $\Gamma_T \ne \Gamma_B$ 的共模分量the common-mode part of the diodes' RF reflections when $\Gamma_T \ne \Gamma_B$$\Gamma_T \ne \Gamma_B$ のときのダイオードの RF 反射の同相成分$v_T^{refl}+v_B^{refl}\ne 0$ → Σ 吸收掉$v_T^{refl}+v_B^{refl}\ne 0$ → absorbed by Σ$v_T^{refl}+v_B^{refl}\ne 0$ → Σ が吸収
LO 不對稱漏到兩個二極體asymmetric LO leakage into the two diodesLO が二つのダイオードへ非対称に漏れる共模 LO 跑到 Σ,the common-mode LO goes to Σ and 同相の LO は Σ へ行き、不會回流到 RFdoes not flow back into the RFRF へ戻らない
環境耦合、雜訊的共模成份the common-mode part of environmental coupling and noise環境結合や雑音の同相成分同樣被 Σ 的 $Z_0$ 吃掉likewise swallowed by the $Z_0$ at Σ同様に Σ の $Z_0$ に吸収される
直觀類比:Σ 埠是「共模垃圾桶」An intuitive analogy: the Σ port is the “common-mode waste bin”直感的な類比:Σ ポートは「同相のゴミ箱」
任何兩端訊號 $(v_T, v_B)$ 都可以拆成 Any pair of signals $(v_T, v_B)$ can be decomposed into a 任意の二端子の信号 $(v_T, v_B)$ は、共模common mode同相モード($v_{cm}=\tfrac{v_T+v_B}{2}$)和($v_{cm}=\tfrac{v_T+v_B}{2}$) and a ($v_{cm}=\tfrac{v_T+v_B}{2}$)と差模differential mode差動モード($v_{dm}=\tfrac{v_T-v_B}{2}$)。180° hybrid 把這兩個分量($v_{dm}=\tfrac{v_T-v_B}{2}$). The 180° hybrid ($v_{dm}=\tfrac{v_T-v_B}{2}$)に分解できます。180° hybrid はこの二つの成分を物理分離到兩個埠physically separates these two components into two ports物理的に二つのポートへ分離します:
  • Δ 埠The Δ portΔ ポート = 差模出口(RF 走這裡) = the differential-mode exit (the RF goes through here) = 差動モードの出口(RF はここを通ります)
  • Σ 埠The Σ portΣ ポート = 共模出口(不要的東西全往這裡丟,被 $Z_0$ 吸收) = the common-mode exit (everything unwanted is dumped here and absorbed by the $Z_0$) = 同相モードの出口(不要なものはすべてここに捨てられ、$Z_0$ に吸収されます)
這就是為什麼把 Σ 接 $Z_0$ 終結會讓整顆 mixer「 That is why terminating Σ in $Z_0$ makes the whole mixer “ これが、Σ を $Z_0$ で終端すると mixer 全体が「所有不要的共模反射、共模雜訊都自動被處理掉dispose of every unwanted common-mode reflection and common-mode noise automatically不要な同相反射や同相雑音がすべて自動的に処理される」 — 它就像差動放大器的 tail current source 一樣,提供共模消除的物理出口。 ” — it works just like the tail current source of a differential amplifier, providing a physical exit for common-mode rejection. 」ようになる理由です — 差動増幅器の tail current source と同じように、同相除去のための物理的な出口を提供します。

為什麼要終結 Σ 埠?Why terminate the Σ port at all?なぜ Σ ポートを終端する必要があるのか?

180° hybrid 是個 A 180° hybrid is a 180° hybrid は 4-port 元件,但在 ring mixer 拓樸下只用 Δ 埠來接 RF — 剩下的 Σ 埠用不到。但你 component, but in the ring-mixer topology only the Δ port is used for the RF — the remaining Σ port is unused. Yet you 素子ですが、ring mixer トポロジーでは RF に Δ ポートしか使わず、残る Σ ポートは使いません。しかし不能把它直接開路或短路cannot simply leave it open- or short-circuitedそのまま開放や短絡にしておくことはできません放著,原因有三:, for three reasons:。理由は三つあります:

  1. 避免反射。To avoid reflections.反射を避けるため。Σ 埠開路/短路會造成全反射;二極體環產生的雜散訊號跑到 Σ 埠後會反彈回去汙染 RF 路徑。 An open or short at the Σ port causes total reflection; spurious signals generated by the diode ring that reach the Σ port would bounce back and contaminate the RF path.Σ ポートが開放/短絡だと全反射が生じます。ダイオードリングが生成したスプリアス信号が Σ ポートに達すると跳ね返り、RF 経路を汚染します。
  2. 維持 hybrid 的行為。To preserve the hybrid's behaviour.hybrid の振る舞いを保つため。$[S]$ 矩陣的推導假設 The derivation of the $[S]$ matrix assumes $[S]$ 行列の導出は所有埠都以 $Z_0$ 終結every port is terminated in $Z_0$すべてのポートが $Z_0$ で終端されている。未終結的埠破壞這個前提,Σ/Δ 分離就失準。. An unterminated port breaks that premise and the Σ/Δ separation goes off.ことを前提とします。終端されていないポートはこの前提を壊し、Σ/Δ の分離が狂います。
  3. 吸收雜散能量。To absorb spurious energy.スプリアスエネルギーを吸収するため。因二極體不平衡而漏到 Σ 埠的訊號,會被電阻轉成熱消耗掉,不會再回流干擾。 Signals that leak to the Σ port because of diode imbalance are turned into heat by the resistor and dissipated, so they cannot return to cause interference.ダイオードの不平衡によって Σ ポートに漏れた信号は、抵抗で熱に変換されて消費され、戻って干渉することがありません。

為什麼是 $Z_0$、且接地?Why $Z_0$, and why grounded?なぜ $Z_0$ で、しかも接地なのか?

  • 阻值 $Z_0$:The value $Z_0$:抵抗値 $Z_0$:反射係數 $\Gamma = (Z_L - Z_0)/(Z_L + Z_0) = 0$ → the reflection coefficient $\Gamma = (Z_L - Z_0)/(Z_L + Z_0) = 0$ → 反射係数 $\Gamma = (Z_L - Z_0)/(Z_L + Z_0) = 0$ → 完美吸收,零反射perfect absorption, zero reflection完全吸収、反射ゼロ。任何其他阻值都會部分反射回 hybrid。. Any other value reflects part of the energy back into the hybrid.。他の抵抗値ではエネルギーの一部が hybrid に反射して戻ります。
  • 接地:Grounding:接地:電阻需要兩端形成完整迴路才能消耗能量;接地提供電位參考。 a resistor needs a complete loop across its two ends to dissipate energy, and the ground provides the potential reference.抵抗はエネルギーを消費するために両端で完全な閉回路が必要であり、接地が電位の基準を与えます。

其他相同套路The same trick elsewhere他の同じ手法

所有「閒置的耦合器埠」幾乎都這樣處理:Almost every “idle coupler port” is handled this way:ほとんどすべての「未使用のカプラポート」はこのように扱われます:

  • Wilkinson 功分器的隔離埠 → $Z_0$ 接地the isolated port of a Wilkinson divider → $Z_0$ to groundWilkinson 電力分配器の隔離ポート → $Z_0$ で接地
  • 方向耦合器的隔離埠 → $Z_0$ 負載the isolated port of a directional coupler → a $Z_0$ load方向性結合器の隔離ポート → $Z_0$ 負荷
  • Balun / N-port 元件閒置埠 → $Z_0$ 終結the idle port of a balun / N-port component → terminated in $Z_0$バラン / N-port 素子の未使用ポート → $Z_0$ で終端

在 datasheet 上常稱為 On datasheets this is usually called a datasheet では通常 "dummy load" or または "termination resistor".と呼ばれます。

② Balun 把 LO 分成 +LO 與 −LO(並提供 IF 取出點)② The balun splits the LO into +LO and −LO (and provides the IF take-off point)② バランが LO を +LO と −LO に分ける(さらに IF の取り出し点を提供する)

Balun = balanced-to-unbalanced transformer。它的工作:balanced transformer. Its job:balanced transformer です。その役割は:

為什麼 balun 中心抽頭天生抑制 LO?Why does a balun centre tap inherently suppress the LO?なぜバランのセンタータップは本質的に LO を抑圧するのか?
LO 在初級兩半線圈的電壓 The LO voltages in the two halves of the primary winding are 一次側巻線の二つの半分における LO 電圧は反相in anti-phase逆相,中心抽頭剛好是兩個訊號的中點 = 0(理想情況下)。所以 , and the centre tap sits exactly at the midpoint of the two signals = 0 (ideally). Hence であり、センタータップはちょうど二つの信号の中点 = 0(理想的には)になります。したがって LO → IF 隔離度天生很好LO → IF isolation is inherently excellentLO → IF のアイソレーションは本質的に非常に良好です,不靠額外濾波。 , without any additional filtering. 、追加のフィルタリングに頼りません。

③ 二極體環的切換邏輯③ The switching logic of the diode ring③ ダイオードリングのスイッチング論理

假設 $|V_{LO}|\gg|V_{RF}|$(LO 強到能主導二極體的偏壓):Assume $|V_{LO}|\gg|V_{RF}|$ (the LO is strong enough to dominate the diode bias):$|V_{LO}|\gg|V_{RF}|$ と仮定します(LO はダイオードのバイアスを支配できるほど強い):

狀態State状態L 端電位Potential at LL 端の電位R 端電位Potential at RR 端の電位導通Conducting導通切斷Cut off遮断等效路徑Equivalent path等価経路
Phase A
(LO > 0)
+$V_{LO}$−$V_{LO}$D1, D2D3, D4T → D1 → R 和 R → D2 → BT → D1 → R and R → D2 → BT → D1 → R と R → D2 → B
(R 變成「短路節點」)(R becomes the “short-circuit node”)(R が「短絡節点」になる)
Phase B
(LO < 0)
−$V_{LO}$+$V_{LO}$D3, D4D1, D2B → D3 → L 和 L → D4 → TB → D3 → L and L → D4 → TB → D3 → L と L → D4 → T
(L 變成「短路節點」)(L becomes the “short-circuit node”)(L が「短絡節点」になる)

④ 為什麼 IF = RF × sgn(LO)?④ Why is IF = RF × sgn(LO)?④ なぜ IF = RF × sgn(LO) なのか?

因為 R 與 L 是接到 balun 初級的兩端,而 IF 取自Because R and L connect to the two ends of the balun primary, while the IF is taken from the R と L はバラン一次側の両端に接続されており、IF は中心抽頭centre tapセンタータップ:から取り出されるためです:

所以隨著 LO 在正負之間切換,IF 在 +RF 與 −RF 之間切換。數學上:So as the LO alternates between positive and negative, the IF switches between +RF and −RF. Mathematically:したがって LO が正負の間で切り替わるにつれ、IF は +RF と −RF の間で切り替わります。数学的には:

$$i_{IF}(t) \propto v_{RF}(t)\cdot \text{sgn}\bigl[v_{LO}(t)\bigr]$$

把方波 $\text{sgn}[\cos\omega_{LO}t]$ 做 Fourier 展開,主項是 $(4/\pi)\cos\omega_{LO}t$,所以:Taking the Fourier expansion of the square wave $\text{sgn}[\cos\omega_{LO}t]$, whose leading term is $(4/\pi)\cos\omega_{LO}t$:方形波 $\text{sgn}[\cos\omega_{LO}t]$ を Fourier 展開すると、主項は $(4/\pi)\cos\omega_{LO}t$ なので:

$$i_{IF}(t) \approx \tfrac{4}{\pi} v_{RF}(t)\cos\omega_{LO}t = \tfrac{2}{\pi}V_{RF}\bigl[\cos(\omega_{RF}-\omega_{LO})t + \cos(\omega_{RF}+\omega_{LO})t\bigr]$$

後面接低通濾波器 → 留下差頻 $\omega_{IF}=\omega_{RF}-\omega_{LO}$ 的 IF 訊號。A low-pass filter then follows → leaving the IF signal at the difference frequency $\omega_{IF}=\omega_{RF}-\omega_{LO}$.後段に低域通過フィルタを接続 → 差周波数 $\omega_{IF}=\omega_{RF}-\omega_{LO}$ の IF 信号が残ります。

⑤ 為什麼叫「雙平衡 (double-balanced)」?⑤ Why is it called “double-balanced”?⑤ なぜ「二重平衡 (double-balanced)」と呼ぶのか?

三埠互相隔離 — 這就是 double-balanced 的招牌All three ports are isolated from one another — that is the hallmark of a double-balanced mixer三つのポートが互いにアイソレートされている — これが double-balanced ミキサの看板です
洩漏路徑Leakage path漏れ経路對消機制Cancellation mechanism打ち消しの仕組み
RF → IF180° hybrid 在 T 和 B 給出反相 RF;經過 ring 的對稱結構後,在 balun 中心抽頭上對消。The 180° hybrid delivers anti-phase RF to T and B; after passing through the ring's symmetric structure, they cancel at the balun centre tap.180° hybrid は T と B に逆相の RF を与えます。ring の対称構造を通った後、バランのセンタータップで打ち消されます。
LO → IFBalun 初級兩半線圈反相,The two halves of the balun primary are in anti-phase, so they バラン一次側の二つの半分の巻線は逆相なので、天生在中心抽頭對消cancel inherently at the centre tapセンタータップで本質的に打ち消し合います.
RF ↔ LO180° hybrid 的 Σ-Δ 隔離 + ring 對稱性 → RF 入射不會洩漏到 LO 端。The Σ-Δ isolation of the 180° hybrid plus the symmetry of the ring → incident RF cannot leak to the LO port.180° hybrid の Σ-Δ アイソレーション + ring の対称性 → 入射 RF は LO ポートへ漏れません。
所以這顆混頻器三個埠(RF / LO / IF) So all three ports of this mixer (RF / LO / IF) have したがってこのミキサの三つのポート(RF / LO / IF)は互相都有 20+ dB 隔離20+ dB of mutual isolation互いに 20+ dB のアイソレーションを持ち,是所有混頻拓樸中隔離度最好的。也, the best of any mixer topology. It also 、あらゆるミキサトポロジーの中で最良です。さらに抑制偶次諧波suppresses even-order harmonics偶数次高調波を抑圧し($2\omega_{RF}$、$2\omega_{LO}$、$2\omega_{RF}\pm 2\omega_{LO}$ 等),spurs 表現最乾淨。 ($2\omega_{RF}$, $2\omega_{LO}$, $2\omega_{RF}\pm 2\omega_{LO}$ and so on), giving the cleanest spur performance. ($2\omega_{RF}$、$2\omega_{LO}$、$2\omega_{RF}\pm 2\omega_{LO}$ など)、最もクリーンなスプリアス性能を実現します。

⑥ 三句話帶走⑥ Three sentences to take away⑥ 三つの文で持ち帰る

  1. 180° hybrid + balun 把 RF、LO 都做成 Turn both the RF and the LO into RF と LO をどちらも差動信號differential signals差動信号注入二極體環的對角頂點。 and inject them at diagonally opposite vertices of the diode ring.にして、ダイオードリングの対角の頂点に注入します。
  2. 4 顆 ring 二極體The 4 ring diodes4 個の ring ダイオード當作「LO 控制的方波斬波器」— 每半週只有 2 顆導通,把 RF 切成 ±RF 交替輸出。 act as a “square-wave chopper controlled by the LO” — only 2 conduct on each half-cycle, chopping the RF into an alternating ±RF output.は「LO に制御される方形波チョッパ」として働き — 半周期ごとに 2 個だけが導通し、RF を ±RF の交互出力に切り刻みます。
  3. 結果:The result: 結果:IF = RF × sgn(LO),三埠互相隔離,雜訊與 spurs 都最低。, with all three ports mutually isolated and the lowest noise and spur levels of any topology.、三つのポートは互いにアイソレートされ、雑音と spurs はいずれも最低です。