OPamp Compensation – Gain Reduction, Lead, Lead-Lag

by allenlu2007

本文主要參考 TI 的 Application Report “Stability Analysis of Voltage-Feedback Op Amps” 

Why Opamp Compensation

OPamp 基本上是所有 analog circuit design 的基礎。不論 amplifier, filter, 以至 bandgap, LDO, ADC 都需要各種不同的 opamp.  Opamp 是 high gain and feedback device, 因此 stability 是非常重要的考慮。Stability 有問題輕則 overshoot or ringing, 重則 oscillate 或 system crash.   Stability 最重要就是靠 compensation.  99% 的 opamp 都是 compensated to unit gain stable.  但也有例外的情況。

1. Some opamp is not fully compensated intentionally.

最常見的例子是為了 wide bandwidth application.  Well compensated (通常是 unit gain stable) opamp 一般會犧牲 opamp 的頻寬。例如 OPA643 stable in gain > 3.  

2. Circuit designer 需要 customized opamp 時。例如 opamp for ADC 或 high bandwidth amplifier 或 filter. 除了 unit gain stable compensation 之外,也可以用本文描述其他的 compensation 技巧。


我們先用以下的 two pole opamp 討論 compensation.  未 compensate 之前 phase margin 只有 25°, (unit gain) step response 有 overshoot 或 ringing.  



Dominant Pole Compensation

Dominant pole 大多用於 unit gain stable compensation.  Method 1: 就是大名鼎鼎的 Miller compensation, 歸類於 opamp internal (gain stage) compensation.  利用 gain stage 把 input stage 的 cap 放大 dominant pole, 也就是把上圖的 W push 到更低頻。Non-dominant pole 的位置基本上不變。所以 unit gain frequency 還是維持在 non-dominant pole 的位置 (to be safe, about 1/10 x 1/𝜏2).  


另一種方式是 external compensation, 就是外加一個 dominant pole, e.g. 下下圖的 CL, 讓新的 pole 成為 dominant pole.  此時的 non-dominant pole 變為 1/𝜏1.  所以 unit gain frequency 也許在 1/10 x 1/𝜏1.  比起 Miller compensation, external dominant pole compensation 的頻寬更低。



不論是 internal compensation 或是 external compensation, 就是 create 出一個 bottleneck, 讓 system 的反應被最慢的 dominant pole 限制。因此 dominant pole compensation 主要的缺點就是犧牲頻寬換取 stability.  接下再看其他的 compensation 方式。


Gain Reduction Compensation

Gain reduction compensation 和 dominant pole compensation 有點類似。只是把 gain 拉下來讓 unit gain frequency 變低來增加 phase margin.  比 dominant pole 更糟的是連 gain (含 DC) 都降低,有時會造成 offset 或 DC error 變大。有兩種使用模式。


(i) Gain attenuation 放在 opamp 之後 partial feedback, 實際上是 close loop amplifier 同時增加 phase margin (但減少頻寬以及 gain) 如 figure 13.  (Zg/Zg+Zf).   把 amplifier 和 compensation 結合在一起相當常見,如 OPA643.

(2) Gain attenuation 放在 opamp 之前 between virtue grounds.  可以看到 feedback ratio is reduced to (Rin//Rg) / (Rin//Rg+Rf)  compared with Rg/(Rg+Rf).  不過 close loop gain 仍然是 1+Rf/Rg regardless Rin.

這是一種使用 OPA643 在 Gain < 3 但仍 stable 的方式。但並不推薦。


那更好的 compensation 方式是什麼?


(Lead) Lag Compensation


首先是 (lead)-lag compensation.  就是保持 DC gain, 但讓 AC 的 gain reduction 以增加 phase margin.  為了達到這個目的,需要先 insert a pole then a zero.  Pole 的位置是在 1/RC;  zero 的位置是在 1/ReqC where Req 如下圖。下圖也可以把 C 的接地改成接 + 端的 virtue ground.  

Lead lag compensation 的好處是避免 DC offset or error 因為 opamp DC gain 和維持住。但是 opamp 的 ac 頻寬仍然被犧牲。在需要高速 settling 的應用 ( 如 ADC) 並不是好的 solution. 




Lead Compensation

Lead compensation 的觀念是先 zero 後 pole.  如下圖 zero = 1/RFC;  pole = 1/(RF//RIC).  第一個 zero 基本上 cancel opamp 的 non-dominant pole.  第二個 pole 變成新的 non-dominant pole.  因些新的 system 的頻寬比原來的頻寬大。同時保有原來的 high gain.

看似 lead compensation a perfect solution.  主要的問題是必須事先知道 opamp non-dominant pole 的位置。實務上 non-dominant pole 可能會隨著 loading 或製程 variation 變化。Compensator 要準確 cancel non-dominant pole 並不容易。有時可能會造成 doublet 或 not well settling 的行為。 

比較起來,gain reduction 或 lead-lag 對於 opamp 的變異比較不敏感。但在 low power 盛行的今天,lead compensation + some tracking 也許是不錯的 solution. 





本文1 and 本文2 有值得參考的 lead, lag, lead-lag compensators 的說明。

Lead and lag compensators are used quite extensively in control. A lead compensator can increase the stability or speed of reponse of a system; a lag compensator can reduce (but not eliminate) the steady-state error.

Lead compensator

Consider the first-order compensator shown below.

(1)$$ G(z) = K_{d}\frac{z-z_{o}}{z-z_{p}} $$

The gain (Kd) will be defined as follows so as not to affect the steady-state response.

(2)$$ K_{d} = \frac{1-z_{p}}{1-z_{o}} $$

The pole (zp) must be a real value inside the unit circle. For a lead compensator, the zero is greater than the pole (zo > zp) and the gain Kd is greater than 1.

Generally for the design of a lead compensator, the zero (zo) is placed close to the location of one of the plant’s poles to achieve an approximate pole-zero cancellation. The compensator pole (zp) is then placed to the left of the zero so that the root-locus will shift to the left. The figure below illustrates how the root-locus shifts by placing the pole and the zero.

Shifting the root-locus to the left results in faster response time.


除了上述 OPAMP 的 compensation 外。另一個標準例子是 LDO compensation.  請參考前文。

外掛 cap (>3uF) 的 dominant pole 在 output; non-dominant pole 在 pass PMOS gate.   一個常見的做法是在 cap 加 ESR.  產生的 zero 可以和 non-dominant pole cancel.  產生的 pole 在更高的 frequency.   


 Lag Compensataor

A first-order phase-lag compensator also can be designed using a frequency response approach. A lag compensator in frequency response form is given by the following.

(9)$$C(s)=\frac{1}{a}\left(\frac{1+aTs}{1+Ts}\right) \qquad [ a<1 ]$$

The phase-lag compensator looks similar to phase-lead compensator, except that a is now less than 1. The main difference is that the lag compensator adds negative phase to the system over the specified frequency range, while a lead compensator adds positive phase over the specified frequency. A Bode plot of a phase-lag compensator has the following form.

The two corner frequencies are at 1 / T and 1 / aT. The main effect of the lag compensator is shown in the magnitude plot. The lag compensator adds gain at low frequencies; the magnitude of this gain is equal to a. The effect of this gain is to cause the steady-state error of the closed-loop system to be decreased by a factor of a. Because the gain of the lag compensator is unity at middle and high frequencies, the transient response and stability are generally not impacted much.

The side effect of the lag compensator is the negative phase that is added to the system between the two corner frequencies. Depending on the value a, up to -90 degrees of phase can be added. Care must be taken that the phase margin of the system with lag compensation is still satisfactory. This is generally achieved by placing the frequency of maximum phase lag, wm as calculated below, well below the new gain crossover frequency.



The lag compensator is expressed in the same form as a lead compensator. The general form is repeated below.

(3)$$ G(z) = K_{d}\frac{z-z_{o}}{z-z_{p}} $$


(4)$$ K_{d} = \frac{1-z_{p}}{1-z_{o}} $$

The pole (zp) must again be a real value inside the unit circle. For a lag compensator, however, the zero is less than the pole (zo < zp) and the gain Kd is less than 1.

The philosophy of the design is the following. Consider the root-locus in the figure below as a starting point.

Suppose the root locations za and za_bar give a satisfactory transient response (such as overshoot, settling time, rise time), but that the gain (K) must be increased to reduce the steady-state error. We can add a compensator pole close to z = 1 and a compensator zero to the left of right-most plant pole as shown below.

Recall from your control textbook, the gain K goes to infinity as the closed-loop poles approach open-loop zeros and the gain goes to 0 as the closed-loop poles approach open-loop poles. By adding the compensator zero (zo) to the root-locus, the gain K at za and za_bar increases. Also the compensator zero and pole have been (or should have been) appropriately placed so that the desired root locations za and zabar remain in the same location.