[SOLVED] Offset compensation feedback of multistage-amplifier

[vaah]

Hi guys,

I know that a DC-feedback loop can reduce offset in a multi-stage amplifier (more specifically limiting amplifier for high-speed link)

I couldn't extract the equation how the offset is resourced by the feedback loop. I have seen from "http://ece.tamu.edu/~spalermo/ecen689_oi/lecture6_ee689_limiting_amps.pdf" the total offset is

sqrt[(Vos/AF)^2+(Vos1/A)^2]

From my calculation "Vos1/F" is true. Can someone please explain this?



Thanks.

 

[FvM]

The square root expression is describing a statistical quantity, typical absolute offset value. It has nothing to due with actual offset voltage of an individual amplifier.

 

[dick_freebird]

The trick is to find a feedback scheme that does
not compromise the forward-path performance.
You could sample, but this inserts "stuff" into a
continuous-time path that will need to be taken
out, and/or dealt with. You could use a very-low-
pass filter to try and pick up the DC error, but
this tends to pick up residue of dynamic activity
and input signal imbalance / history.

What does your application want? High or low speed,
sampled or continuous-time, ???

 

[vaah]

The trick is to find a feedback scheme that does
not compromise the forward-path performance.
You could sample, but this inserts "stuff" into a
continuous-time path that will need to be taken
out, and/or dealt with. You could use a very-low-
pass filter to try and pick up the DC error, but
this tends to pick up residue of dynamic activity
and input signal imbalance / history.

What does your application want? High or low speed,
sampled or continuous-time, ???

Thank you for your help.

What I want to know is that how the abovementioned equation is extracted not how the offset compensation is done.
Are "Vos1" and "Vos" output offset or input offset? How come the total output offset is "sqrt[Vos^2 + (F*Vos1)^2]".

- - - Updated - - -

Thank you for your help.

What I want to know is that how the abovementioned equation is extracted not how the offset compensation is done.
Are "Vos1" and "Vos" output offset or input offset? How come the total output offset is "sqrt[Vos^2 + (F*Vos1)^2]".

I figured it out.
Vos1 and Vos are input offset.
To obtain the total input offset we should calculate the total output offset and divide it by the signal path gain not the total gian.