[P4] 大注入 LC 模型與暫態行為:精確 pull-in 解、鎖定時間、APF 振幅暫態與 pulling 頻譜
先備 :paper_004 (APF 定義 [P4] Eq.(18)–(22)、ISF/APF quadrature Eq.(26)、augmented Adler Eq.(27))、lab_36 (鎖內 Adler 的 R R R 形式精確解、臨界慢化、cycle slips)、injection_locking_noise (Part A 一階 PLL、Part B 拍頻 ω b \omega_b ω b 與單邊梳)、phase_vs_amplitude_noise (τ 0 = 2 Q / ω 0 \tau_0=2Q/\omega_0 τ 0 = 2 Q / ω 0 的振幅恢復、OU 過程)|接下來 :paper_004 的 M:N/ILFD 段、lab_37(simulations/lab_37_ilfd_lock.py,收在 paper_004 頁內)、quadrature_and_coupled_oscillators 。
paper_004 已經把 [P4] 的 APF (amplitude
perturbation function,振幅擾動函數)定義、ideal-LC 的 ISF/APF quadrature 與 M:N 次諧波鎖定講完。
這一頁收 [P4] **Sec. III-E/F(大注入 LC 模型)與 Sec. V(暫態行為)**剩下的料,並把它們教完:
這頁要回答什麼 :
「振幅版 Adler」——Mirzaei 的 Generalized Adler's equation ([P4] Eq.(8))——為什麼恰好等於 [P4] 的
Γ ~ / ( 1 + A ) \tilde\Gamma/(1+A) Γ ~ / ( 1 + A ) 模型(Eq.(13)、(27))?大注入 lock range ω L = ω L 0 / 1 − a 2 \omega_L=\omega_{L0}/\sqrt{1-a^2} ω L = ω L 0 / 1 − a 2 (Eq.(9)、(23))
的 a a a 到底是 I i n j / I o s c I_{inj}/I_{osc} I inj / I osc 還是 I i n j / I m a x I_{inj}/I_{max} I inj / I ma x ?為什麼 a ≥ 1 a\ge1 a ≥ 1 時模型「無界」?
正弦 Adler 的精確暫態解 怎麼來的?[P4] Eq.(31) 的 tanh \tanh tanh (鎖內)與 Eq.(33) 的 tan \tan tan (鎖外)為何長得一樣?
鎖定要多久 ?有沒有閉式?為什麼靠近 lock range 邊緣時鎖得上卻鎖得極慢?[P4] Table I 那個星號
「用 lock characteristic 的斜率算 τ p \tau_p τ p 」是什麼配方?
鎖定捕獲期間振幅 在做什麼?APF 怎麼把相位暫態轉成振幅的 dip/overshoot?quasi-static 假設何時失效?
鎖不住(pulling)時,大注入的拍頻 還是 Δ ω 2 − ω L 2 \sqrt{\Delta\omega^2-\omega_L^2} Δ ω 2 − ω L 2 嗎?AM 對 pulling 頻譜做了什麼
([P4] Fig. 14(c) 的 "ISF + APF" vs "ISF Only")?
物理直覺(先講結論) :注入電流同時「推相位」(ISF,切向)與「改振幅」(APF,徑向)。振幅改了,ISF 就跟著
反比縮放 (Γ ~ L C = Γ ~ / ( 1 + A ) \tilde\Gamma_{LC}=\tilde\Gamma/(1+A) Γ ~ L C = Γ ~ / ( 1 + A ) :擺幅大 → 同一顆電荷推出的相位小)。這一個回饋就把 Adler 的
sin θ \sin\theta sin θ 變成 sin θ / ( 1 + a cos θ ) \sin\theta/(1+a\cos\theta) sin θ / ( 1 + a cos θ ) ——注入與振盪同相(θ ≈ 0 \theta\approx0 θ ≈ 0 )時振幅被撐大、恢復力變弱 、
鎖得慢 ;反相(θ ≈ ± π \theta\approx\pm\pi θ ≈ ± π )時振幅被吸乾、恢復力變強 、lock characteristic 被拉高——所以 lock range
從 ω L 0 \omega_{L0} ω L 0 撐到 ω L 0 / 1 − a 2 \omega_{L0}/\sqrt{1-a^2} ω L 0 / 1 − a 2 ,a → 1 a\to1 a → 1 時「振幅歸零」的非物理解讓它發散。至於暫態:正弦 Adler
用 tan 半角代換後只剩一條二次式 ,判別式的符號一翻,鎖內的 tanh \tanh tanh (指數收斂、率 ω p \omega_p ω p )就變成鎖外的
tan \tan tan (週期滑動、拍頻 ω b \omega_b ω b )——同一個畢氏根號 ∣ ω L 2 − Δ ω 2 ∣ \sqrt{\lvert\omega_L^2-\Delta\omega^2\rvert} ∣ ω L 2 − Δ ω 2 ∣ 。
本頁定位 :進階 deep-dive([P4] 剩料),非核心教學章節 。以下標「✓ 已核實」者皆由本人放大原始 PDF 逐字對照:
[P4] Eq.(5)–(9) 與 Sec. III-B 的 τ 0 = 2 Q / ω 0 \tau_0=2Q/\omega_0 τ 0 = 2 Q / ω 0 (p.2123)、Eq.(13)(p.2124)、Eq.(20)–(22) 與 footnote 7
(p.2126)、Eq.(23) 與 θ ∈ [ − 110 ∘ , 110 ∘ ] \theta\in[-110^\circ,110^\circ] θ ∈ [ − 11 0 ∘ , 11 0 ∘ ] 的經驗限制(p.2127)、Eq.(24)–(27) 與 Fig. 8 caption
(p.2128)、Eq.(31)–(34)、Table I 與其星號註(p.2130)、Fig. 13/Fig. 14 caption、Table II 與 amplitude-conscious
波形(p.2131–2132)、Eq.(35)–(38)(p.2132)。本站自行推導、不在 [P4] 內 的部分明標:Eq.(31)/(33) 的逐步推導
([P4] 只寫 "one can show that [29]")、大注入 pulling 的拍頻閉式(Sec. 5.2)、振幅一階遲滯模型(Sec. 4.3)。
外部文獻一律標「(外部文獻,非本站 5 篇 PDF)」 並附 [P4] 參考書目中逐字核對的出處。
0. 符號與慣例對帳(先把 2 與正負號釘死)
量 本站寫法 [P4] 寫法 對帳 失諧 Δ ω ≡ ω 0 − ω i n j \Delta\omega\equiv\omega_0-\omega_{inj} Δ ω ≡ ω 0 − ω inj Δ ω ≡ ω i n j / N − ω 0 \Delta\omega\equiv\omega_{inj}/N-\omega_0 Δ ω ≡ ω inj / N − ω 0 (p.2130)差一個整體正負號;本頁所有結果只依賴 Δ ω 2 \Delta\omega^2 Δ ω 2 或明寫分支 ISF-only 半 lock range ω L 0 ≡ I i n j 2 q m a x , 0 \omega_{L0}\equiv\dfrac{I_{inj}}{2q_{max,0}} ω L 0 ≡ 2 q ma x , 0 I inj ω L = 1 2 I i n j ∣ Γ ~ 1 ∣ \omega_L=\tfrac12 I_{inj}\lvert\tilde\Gamma_1\rvert ω L = 2 1 I inj ∣ Γ ~ 1 ∣ ,ideal LC ∣ Γ ~ 1 ∣ = 1 / q m a x , 0 \lvert\tilde\Gamma_1\rvert=1/q_{max,0} ∣ Γ ~ 1 ∣ = 1/ q ma x , 0 (Eq.(26))那個 1 2 \tfrac12 2 1 是積化和差的 1 2 \tfrac12 2 1 ([P3] Eq.(34)–(35), p.2114,本站 injection_locking_noise 已核實),與 phase-noise 的 SSB / 4 /4 /4 vs 時域 / 2 /2 /2 慣例無關 大注入半 lock range ω L ≡ ω L 0 / 1 − a 2 \omega_L\equiv\omega_{L0}/\sqrt{1-a^2} ω L ≡ ω L 0 / 1 − a 2 Eq.(9)=Eq.(23) at β = 90 ∘ \beta=90^\circ β = 9 0 ∘ 本頁凡寫 ω L \omega_L ω L 皆指含 APF 修正 的值;ISF-only 一律寫 ω L 0 \omega_{L0} ω L 0 注入強度(LC 專用) a ≡ I i n j I o s c a\equiv\dfrac{I_{inj}}{I_{osc}} a ≡ I osc I inj 1 2 I i n j ∣ Δ 1 ∣ = 1 2 τ 0 I i n j q m a x , 0 \tfrac12 I_{inj}\lvert\Delta_1\rvert=\tfrac12\tau_0\dfrac{I_{inj}}{q_{max,0}} 2 1 I inj ∣ Δ 1 ∣ = 2 1 τ 0 q ma x , 0 I inj 恆等式 ω 0 q m a x , 0 = Q I o s c \omega_0 q_{max,0}=Q\,I_{osc} ω 0 q ma x , 0 = Q I osc (p.2124)⟹ a = τ 0 ω L 0 a=\tau_0\,\omega_{L0} a = τ 0 ω L 0 (精確) 線性有效性 I i n j / I m a x I_{inj}/I_{max} I inj / I ma x ,I m a x ≡ ω 0 q m a x , 0 I_{max}\equiv\omega_0 q_{max,0} I ma x ≡ ω 0 q ma x , 0 footnote 11, p.2130;Eq.(35), p.2132 I o s c = I m a x / Q I_{osc}=I_{max}/Q I osc = I ma x / Q (p.2132):兩個不同的歸一化 ——I m a x I_{max} I ma x 管一階線性是否成立,I o s c I_{osc} I osc 管 LC 的振幅效應多大振幅記憶時間 τ 0 = 2 Q / ω 0 \tau_0=2Q/\omega_0 τ 0 = 2 Q / ω 0 [s]Sec. III-B, p.2123;Eq.(25), p.2128 是振幅 時間常數;能量時間常數是 Q / ω 0 Q/\omega_0 Q / ω 0 (差 2 倍,見 tank_Q ) 移位相位 ψ ≡ θ + π / 2 \psi\equiv\theta+\pi/2 ψ ≡ θ + π /2 N θ ~ ≡ N θ + ∠ Γ ~ N − ∠ I i n j N\tilde\theta\equiv N\theta+\angle\tilde\Gamma_N-\angle I_{inj} N θ ~ ≡ N θ + ∠ Γ ~ N − ∠ I inj ideal LC、餘弦注入、N = 1 N=1 N = 1 :∠ Γ ~ 1 = 90 ∘ \angle\tilde\Gamma_1=90^\circ ∠ Γ ~ 1 = 9 0 ∘ 、∠ I i n j = 0 \angle I_{inj}=0 ∠ I inj = 0 ⟹ θ ~ = θ + π / 2 = ψ \tilde\theta=\theta+\pi/2=\psi θ ~ = θ + π /2 = ψ
本頁除 Sec. 2 末段外全取 N = 1 N=1 N = 1 (基波注入)。dimension check 通用 :[ ω L 0 ] = [ A ] / [ C ] = [ C/s ] / [ C ] = rad/s [\omega_{L0}]=[\text{A}]/[\text{C}]=[\text{C/s}]/[\text{C}]=\text{rad/s} [ ω L 0 ] = [ A ] / [ C ] = [ C/s ] / [ C ] = rad/s ✓(rad 無因次);[ a ] = [ s ] ⋅ [ rad/s ] = [a]=[\text{s}]\cdot[\text{rad/s}]= [ a ] = [ s ] ⋅ [ rad/s ] = 無因次 ✓。
1. 論文原文:本單元新增核實的段落(逐字轉錄)
1.1 既有模型:Adler 與 Mirzaei 的 Generalized Adler([P4] Sec. III-A, p.2123 ✓)
Adler 方程([P4] Eq.(5))與 tank Q Q Q (Eq.(6)):
d θ d t = ω 0 − ω i n j − ω 0 2 Q I i n j I o s c sin θ , Q = R P ω 0 L = R P ω 0 C \frac{d\theta}{dt}=\omega_0-\omega_{inj}-\frac{\omega_0}{2Q}\frac{I_{inj}}{I_{osc}}\sin\theta,\qquad
Q=\frac{R_P}{\omega_0 L}=R_P\,\omega_0 C d t d θ = ω 0 − ω inj − 2 Q ω 0 I osc I inj sin θ , Q = ω 0 L R P = R P ω 0 C
[P4] 原文(p.2123):"A powerful improvement to Adler's equation was derived by Mirzaei et al. [11], where they forgo
the assumption of a weak injection signal. ... the oscillation amplitude under injection is roughly given by"
V o s c = ( I o s c + I i n j cos θ ) R P ([P4] Eq.(7)) V_{osc}=(I_{osc}+I_{inj}\cos\theta)\,R_P\qquad\text{([P4] Eq.(7))} V osc = ( I osc + I inj cos θ ) R P ([P4] Eq.(7))
"which leads to an augmented differential equation for the oscillator's phase:"
d θ d t = ω 0 − ω i n j − ω 0 2 Q I i n j sin θ I o s c + I i n j cos θ ([P4] Eq.(8)) \frac{d\theta}{dt}=\omega_0-\omega_{inj}-\frac{\omega_0}{2Q}\,\frac{I_{inj}\sin\theta}{I_{osc}+I_{inj}\cos\theta}\qquad\text{([P4] Eq.(8))} d t d θ = ω 0 − ω inj − 2 Q ω 0 I osc + I inj cos θ I inj sin θ ([P4] Eq.(8))
"The lock range associated with (8) was derived independently by a number of authors [9]–[12] to be"
ω L = ω 0 2 Q I i n j I o s c 1 1 − I i n j 2 I o s c 2 ([P4] Eq.(9)) \omega_L=\frac{\omega_0}{2Q}\frac{I_{inj}}{I_{osc}}\,\frac{1}{\sqrt{1-\dfrac{I_{inj}^2}{I_{osc}^2}}}\qquad\text{([P4] Eq.(9))} ω L = 2 Q ω 0 I osc I inj 1 − I osc 2 I inj 2 1 ([P4] Eq.(9))
[P4] 隨即列出這個模型的三個限制(p.2123):只處理正弦注入;Q Q Q 與 I o s c I_{osc} I osc 受寄生影響難以準確決定、現代積體振盪器
未必能用 Fig. 1 的電路建模;預測的 lock range 對稱 ,"which is not always the case [8], [12]"。同頁 Sec. III-B
的思想實驗明寫振幅時間常數:"we assume that any excess amplitude decays exponentially with a time constant of
τ 0 = 2 Q / ω 0 \tau_0=2Q/\omega_0 τ 0 = 2 Q / ω 0 in between successive injections due to the energetics of the oscillator."
1.2 反比於振幅的有效 ISF 與 augmented pulling equation([P4] Sec. III-C/E, p.2124 與 p.2126 ✓)
Γ ~ L C = Γ ~ 1 + A ([P4] Eq.(13), p.2124) \tilde\Gamma_{LC}=\frac{\tilde\Gamma}{1+A}\qquad\text{([P4] Eq.(13), p.2124)} Γ ~ L C = 1 + A Γ ~ ([P4] Eq.(13), p.2124)
p.2124 原文的物理:"the oscillation amplitude controls the slope of the waveform (for a fixed oscillation frequency), and a
steeper waveform corresponds to a proportionally smaller phase shift from the same injection of charge." footnote 4 誠實
註記:這個反比關係只在狀態變數互相正交(無 AM-to-PM)時精確成立 ,LC 與 Bose 振盪器滿足。
振幅偏差([P4] Eq.(20), p.2126)與 augmented pulling equation(Eq.(21)):
A = 1 T i n j ∫ T i n j Δ ( ω i n j t + θ ) i i n j ( t ) d t A=\frac{1}{T_{inj}}\int_{T_{inj}}\Delta\big(\omega_{inj}t+\theta\big)\,i_{inj}(t)\,dt A = T inj 1 ∫ T inj Δ ( ω inj t + θ ) i inj ( t ) d t
d θ d t = ω 0 − ω i n j + 1 T i n j ∫ T i n j Γ ~ ( ω i n j t + θ ) i i n j ( t ) d t 1 + 1 T i n j ∫ T i n j Δ ( ω i n j t + θ ) i i n j ( t ) d t \frac{d\theta}{dt}=\omega_0-\omega_{inj}+\frac{\dfrac{1}{T_{inj}}\displaystyle\int_{T_{inj}}\tilde\Gamma\big(\omega_{inj}t+\theta\big)\,i_{inj}(t)\,dt}{1+\dfrac{1}{T_{inj}}\displaystyle\int_{T_{inj}}\Delta\big(\omega_{inj}t+\theta\big)\,i_{inj}(t)\,dt} d t d θ = ω 0 − ω inj + 1 + T inj 1 ∫ T inj Δ ( ω inj t + θ ) i inj ( t ) d t T inj 1 ∫ T inj Γ ~ ( ω inj t + θ ) i inj ( t ) d t
[P4] 稱之為 "quasi-nonlinear" 模型:非線性只藏在 ISF 與 APF 的除法 裡,兩者各自仍線性於注入電流。正弦注入
i i n j = I i n j cos ( ω i n j t ) i_{inj}=I_{inj}\cos(\omega_{inj}t) i inj = I inj cos ( ω inj t ) 時只留基波(Eq.(22), p.2126;footnote 7:ideal LC 的 ISF/APF 是純正弦,其餘諧波
"effectively filtered out "):
d θ d t = ω 0 − ω i n j + 1 2 I i n j ∣ Γ ~ 1 ∣ cos ( θ + ∠ Γ ~ 1 ) 1 + 1 2 I i n j ∣ Δ 1 ∣ cos ( θ + ∠ Δ 1 ) \frac{d\theta}{dt}=\omega_0-\omega_{inj}+\frac{\tfrac12 I_{inj}\lvert\tilde\Gamma_1\rvert\cos(\theta+\angle\tilde\Gamma_1)}{1+\tfrac12 I_{inj}\lvert\Delta_1\rvert\cos(\theta+\angle\Delta_1)} d t d θ = ω 0 − ω inj + 1 + 2 1 I inj ∣ Δ 1 ∣ cos ( θ + ∠ Δ 1 ) 2 1 I inj ∣ Γ ~ 1 ∣ cos ( θ + ∠ Γ ~ 1 )
1.3 不對稱 lock range 與模型的「無界」([P4] Eq.(23), p.2127 ✓)
ω L ± = 1 2 I i n j ∣ Γ ~ 1 ∣ 1 2 I i n j ∣ Δ 1 ∣ cos β ± 1 − ( 1 2 I i n j ∣ Δ 1 ∣ sin β ) 2 , β ≡ ∠ Γ ~ 1 − ∠ Δ 1 \omega_L^{\pm}=\frac{\tfrac12 I_{inj}\lvert\tilde\Gamma_1\rvert}{\tfrac12 I_{inj}\lvert\Delta_1\rvert\cos\beta\pm\sqrt{1-\big(\tfrac12 I_{inj}\lvert\Delta_1\rvert\sin\beta\big)^2}},\qquad
\beta\equiv\angle\tilde\Gamma_1-\angle\Delta_1 ω L ± = 2 1 I inj ∣ Δ 1 ∣ cos β ± 1 − ( 2 1 I inj ∣ Δ 1 ∣ sin β ) 2 2 1 I inj ∣ Γ ~ 1 ∣ , β ≡ ∠ Γ ~ 1 − ∠ Δ 1
原文(p.2127):"This lock range is generally asymmetric, meaning ω L + ≠ − ω L − \omega_L^+\ne-\omega_L^- ω L + = − ω L − . ... Only in the specific
case of the ISF and the APF being in perfect quadrature with respect to each other (β = ± π / 2 \beta=\pm\pi/2 β = ± π /2 ) is the lock range
symmetric." 以及本頁最重要的警語:"the lock characteristic from (22) is no longer bounded for all θ \theta θ when
I i n j ∣ Δ 1 ∣ ≥ 2 I_{inj}\lvert\Delta_1\rvert\ge2 I inj ∣ Δ 1 ∣ ≥ 2 , resulting in an infinite lock range [i.e., (23) no longer holds]. Physically, this is
because the fractional amplitude change A A A is able to dip below − 1 -1 − 1 for certain values of θ \theta θ , corresponding to
the nonphysical scenario of an oscillation amplitude which is zero or negative." 處置:"roughly restricting
θ ∈ [ − 110 ∘ , 110 ∘ ] \theta\in[-110^\circ,110^\circ] θ ∈ [ − 11 0 ∘ , 11 0 ∘ ] for very large injection amplitudes usually results in reliable estimates of the lock
range."(footnote 8:Generalized Adler (8) 與 [9]–[12] 同樣在 I i n j ≥ I o s c I_{inj}\ge I_{osc} I inj ≥ I osc 時預測無限 lock range。)
1.4 Ideal LC:從 ISF/APF 回到 Generalized Adler([P4] Sec. III-F, p.2128 ✓)
Γ ~ ( φ ) = − 1 q m a x , 0 sin φ , Λ ~ ( φ ) = 1 q m a x , 0 cos φ (Eq.(24)) \tilde\Gamma(\varphi)=-\frac{1}{q_{max,0}}\sin\varphi,\qquad\tilde\Lambda(\varphi)=\frac{1}{q_{max,0}}\cos\varphi\qquad\text{(Eq.(24))} Γ ~ ( φ ) = − q ma x , 0 1 sin φ , Λ ~ ( φ ) = q ma x , 0 1 cos φ (Eq.(24))
d ( t , φ ) = e − t / τ 0 , τ 0 = 2 Q ω 0 , ∫ 0 ∞ d d t = τ 0 ⟹ Δ ( φ ) = τ 0 Λ ~ ( φ ) (Eq.(25)) d(t,\varphi)=e^{-t/\tau_0},\quad\tau_0=\frac{2Q}{\omega_0},\quad\int_0^\infty d\,dt=\tau_0\ \Longrightarrow\ \Delta(\varphi)=\tau_0\,\tilde\Lambda(\varphi)\qquad\text{(Eq.(25))} d ( t , φ ) = e − t / τ 0 , τ 0 = ω 0 2 Q , ∫ 0 ∞ d d t = τ 0 ⟹ Δ ( φ ) = τ 0 Λ ~ ( φ ) (Eq.(25))
Γ ~ 1 = 1 q m a x , 0 ∠ 90 ∘ , Δ 1 = τ 0 q m a x , 0 ∠ 0 (Eq.(26)) \tilde\Gamma_1=\frac{1}{q_{max,0}}\angle90^\circ,\qquad\Delta_1=\frac{\tau_0}{q_{max,0}}\angle0\qquad\text{(Eq.(26))} Γ ~ 1 = q ma x , 0 1 ∠9 0 ∘ , Δ 1 = q ma x , 0 τ 0 ∠0 (Eq.(26))
d θ d t = ω 0 − ω i n j − 1 2 I i n j q m a x , 0 sin θ 1 + 1 2 τ 0 I i n j q m a x , 0 cos θ (Eq.(27)) \frac{d\theta}{dt}=\omega_0-\omega_{inj}-\frac{\tfrac12\dfrac{I_{inj}}{q_{max,0}}\sin\theta}{1+\tfrac12\tau_0\dfrac{I_{inj}}{q_{max,0}}\cos\theta}\qquad\text{(Eq.(27))} d t d θ = ω 0 − ω inj − 1 + 2 1 τ 0 q ma x , 0 I inj cos θ 2 1 q ma x , 0 I inj sin θ (Eq.(27))
原文收尾:"Finally, if we use the identity ω 0 q m a x , 0 = Q I o s c \omega_0 q_{max,0}=QI_{osc} ω 0 q ma x , 0 = Q I osc shown in Fig. 2(a) ... to eliminate the maximum
charge swing q m a x , 0 q_{max,0} q ma x , 0 , we arrive at Generalized Adler's equation (8)." 驗算 :1 2 τ 0 I i n j / q m a x , 0 = 1 2 ⋅ 2 Q ω 0 ⋅ I i n j q m a x , 0 = Q I i n j ω 0 q m a x , 0 = I i n j I o s c = a \tfrac12\tau_0 I_{inj}/q_{max,0}=\tfrac12\cdot\dfrac{2Q}{\omega_0}\cdot\dfrac{I_{inj}}{q_{max,0}}=\dfrac{Q\,I_{inj}}{\omega_0 q_{max,0}}=\dfrac{I_{inj}}{I_{osc}}=a 2 1 τ 0 I inj / q ma x , 0 = 2 1 ⋅ ω 0 2 Q ⋅ q ma x , 0 I inj = ω 0 q ma x , 0 Q I inj = I osc I inj = a ✓、
1 2 I i n j / q m a x , 0 = ω 0 2 Q I i n j I o s c = ω L 0 \tfrac12 I_{inj}/q_{max,0}=\dfrac{\omega_0}{2Q}\dfrac{I_{inj}}{I_{osc}}=\omega_{L0} 2 1 I inj / q ma x , 0 = 2 Q ω 0 I osc I inj = ω L 0 ✓——Eq.(27) 逐項等於 Eq.(8)。
物理讀法:Mirzaei 用 V o s c = ( I o s c + I i n j cos θ ) R P V_{osc}=(I_{osc}+I_{inj}\cos\theta)R_P V osc = ( I osc + I inj cos θ ) R P 猜到的振幅,在 [P4] 框架裡就是 APF 的基波
A = a cos θ A=a\cos\theta A = a cos θ 。
Fig. 8 caption(p.2128 ✓)給了本頁 Table I 反推要用的實際數字:"Injection amplitudes of I i n j = 0.75 I_{inj}=0.75 I inj = 0.75 mA and
I i n j = 1.5 I_{inj}=1.5 I inj = 1.5 mA, respectively, for a CMOS differential LC oscillator with tank parameters L = 6 L=6 L = 6 nH, C = 4.15 C=4.15 C = 4.15 pF, and
Q = 15 Q=15 Q = 15 and biased at I t a i l = 1 I_{tail}=1 I t ai l = 1 mA, resulting in I o s c = ( 4 / π ) I_{osc}=(4/\pi) I osc = ( 4/ π ) mA and 2 / ∣ Δ 1 ∣ = 1.25 2/\lvert\Delta_1\rvert=1.25 2/ ∣ Δ 1 ∣ = 1.25 mA. (c) Bipolar
Colpitts oscillator shown in Fig. 6 subjected to an injection amplitude of I i n j = 7.5 I_{inj}=7.5 I inj = 7.5 mA." 同頁兩句話留給 Sec. 4:
"the stable mode must feature a negative lock characteristic slope, which also corresponds to the larger oscillation
amplitude";以及模型的極限——"the deviation between theory and simulation near the center of the oscillation amplitude
plot for larger injection strengths ... occurs since nonlinear amplitude restoring effects, which are not captured by
the APF, are more prominent at larger oscillation amplitudes."
1.5 Sec. V-A Pull-In Process(p.2130 ✓)