do we really need high dc gain?.....

[sharpak]

Hi guys

It is known than telescopic or folded cascoded amplifiers bring a very high DC gain (gm²/gds²) when a "simple" differential pair gives a "gm/gds" dc gain.

In open loop, it is better to casAcde 2 stages than using a single cascode if we want to prevent from speed limitations. That is quite clear and OK.

In closed loop, we do have a choice. Let's suppose we are not swing limited (vdd = 3P3 volts for example). And let's suppose our system is reset every 10µs to be active for 10µs.

I may be wrong, but Fourier transformation say that we got a sort of door function. The frequency equivalent would be a sinus cardianal whom main energy would be large of 100 kHz (between 1/T and -1/T) .

DOes it mean that the system will see any signal from 0 to 100 khz and that so, the low frequency will affect the amplifier response quality?

Or should we understand the following: the system work at a frequency of 100 kHZ. So, there is no way the DC gain may influe on the closed loop response. Indeed, closing the loop, the pole are moving to cut at gm/CLoad for Both amplifiers. And we are then far above the cut frequancy of both amplifers, in a frequency area where folded cascoded and differential pair transfer function are exactly the same.

I'm very confused cause i tried some transient simulations, but i wouldn't see any effect between cascoded and fiderential, if not a better linearity for ..the differential pair.

---------- Post added at 16:38 ---------- Previous post was at 15:30 ----------

huuuh...nobody got any idea? Just even a clue?...

 

[steve_guo]

as a instance, u can try to solve step response of a feedback loop other than using a single tone input.
u are supposed to see some difference. DC gain acts as a static settling error in the final results.
It do make sense in a some cicuits e.g. switch cap,anyway,High dc gain benefits the linearity.

 

[sharpak]

as a instance, u can try to solve step response of a feedback loop other than using a single tone input.
u are supposed to see some difference. DC gain acts as a static settling error in the final results.
It do make sense in a some cicuits e.g. switch cap,anyway,High dc gain benefits the linearity.

You mean: using a step voltage as an input to watch the difference between vout and vin? That is what i precisely did. And still: it doesn't answer the fact it will takes a lot of time before thissignal is settling correctly. Switch op amp got µs operations.

 

[steve_guo]

And we are then far above the cut frequancy of both amplifers, in a frequency area where folded cascoded and differential pair transfer function are exactly the same

Why are they exactly the same?
Have u deduce the final solution yet ?
Let's check a single pole system,
A(s) = A/(1+s/p) --- Open Loop
I think the close loop should be like this,
(A/1+A)*(1/1+s/A*p) --- Close loop
As u can see,the former part in equation is affected by DC gain.

Maybe there is some misunderstanding between us.And I think in your transient simulation,
the DC gain of your OPA was set too high to miss the difference of two OPA system.
e.g 1000/1001 and 2000/2001, some time it's hard to track this.

 

[sharpak]

Why are they exactly the same?
Have u deduce the final solution yet ?
Let's check a single pole system,
A(s) = A/(1+s/p) --- Open Loop
I think the close loop should be like this,
(A/1+A)*(1/1+s/A*p) --- Close loop
As u can see,the former part in equation is affected by DC gain.

Maybe there is some misunderstanding between us.And I think in your transient simulation,
the DC gain of your OPA was set too high to miss the difference of two OPA system.
e.g 1000/1001 and 2000/2001, some time it's hard to track this.

Remember that differential pair got the same gain than folded cascoded at medium frequencies...
In fact, i do believe u'll get something like:
(1/B) x (1/(1+ 1/AB))=~ (1/B) x (1-1/AB)~= 1/B at low frequency if A is high enough)

But that is the whole point: are we at low frequnecy when they system only works during 10µs?.............. or do we cover the whole frequency range from 0 to 100khz when we operate at 10µs?