Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Feedforward: What am I missing?

Status
Not open for further replies.

exp

Full Member level 1
Joined
Jan 28, 2013
Messages
97
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,286
Activity points
2,570
Hi,

I wanna build 2-stage OpAmp with feedforward compensation.

Let me start with the basics: Let's assume I have two ideal transistors with intrinsic gains a1,a2 and certain GBW1,GBW2. Let's assume both parameters are comparable and I connect them in series then I have gain a1*a2 and bandwidth sqrt(sqrt(2)-1)*GBW1 = 0.64*GBW1. Nothing will beat this. The problem: 40dB/dec and hence no phase margin.

Common approaches for compensation are narrowbanding by making one pole dominant and pushing the other beyond the unity crossover. Bandwidth will be a1*a2 times lower, far from 0.64*GBW1. Or use Miller compensation.

Or: Feedforward compensation!This introduces a zero which, when placed on top of th second pole creates a 90deg shift at unty crossover.

Now the problem is that the zero is set at:

wz = wp1 * (1 + a1 * gm2/gm3)

It is reasonable to assume gm2 ~gm3 which means that the zero is roughly a1 times AFTER the first pole. So the second pole must be a1 times apart of the first pole.

The one way this can be mitigated is to make one pole a1 times smaller than the maximum possible.

Some numbers to clarify: I want to have a1*a2 > 100 and BW > 800 MHz. I can get intrinic gain on the order of 10 and GBW on the order of 80 GHz.

However, due to loading (Cgg of the second stage and miller with Cgd) the dominant pole of the first stage can be at maximum 1.5 GHz.

Having wp1=1.5 GHz means that the zero wz is roughly around wz = wp1*(1+a1*gm2/gm3) ~ wp1*(1+a1) ~ wp1*11 ~17 GHz.

But placing the second pole at such a high frequency is not possible. The maximum is 8 GHz based on GBW.

The one option I have is to choose wp1 based on the maximum achievable wp2 which is around 8 GHz:

wz = wp2 = 8 GHz = (1+a1)*wp1

-> wp1 = 8GHz/11 = 727 MHz

But wp1 = gds1/Cdd1 (to first order). Since gds1 is set to the minimum possible to achieve a1=10, the one knob is to purposely connect a capacitor in series with Cdd1, effectively narrowbanding one stage.

This really sounds weird to me and I have not seen such capacitors in feed-forward compensated circuits.




Do I miss something?
 
Last edited:

Remarks:
* If my post is too confusing please let me know
* If anyone has already used feedforward compensation and what I write sounds correct ... knowing this would already help me too!
 

Weekdays is when our daily regular members spend more hours here. You're more likely to get replies then. A few posts have brought up the same terminology you use.

An expert (not me) will probably ask whether you have solved the issue of gain-bandwidth product in op amps, at such high frequencies.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top