The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Class AB opamp
https://designers-guide.org/forum/YaBB.pl?num=1283179383

Message started by nobody on Aug 30^th, 2010, 7:43am

Title: Class AB opamp
Post by nobody on Aug 30^th, 2010, 7:43am

Hello, all

I have a question on the size of some transistos and marked them in red.
The question is why they are not equal and why they are sized like that way(50/2, 24/2).
What I thought is try to set sources of M20 and M28 equal but that is not true.
Thanks a lot

Title: Re: Class AB opamp
Post by AnalogDE on Aug 30^th, 2010, 10:56am

I'll take a stab at this -- M19 and M20 set the bias voltages on output transistors M25/M26.. It looks like classical AB biasing to reduce cross over distortion.

M28/M27 are used to isolate the two diode connected devices -- M28/M27 are larger because the biasing leg requires more headroom (it has 2*Vth less saturation margin vs the other leg). So you would want those larger to keep the biasing leg alive.

Title: Re: Class AB opamp
Post by nobody on Aug 30^th, 2010, 7:29pm

Hello, AnalogDE

Thanks for your explanation. You are right about the distortion and the function of M19 and M20.
However, I do not get your point about why M28/M27 are larger.
A plot is attached to show biasing voltages. M15,M16,M17 and M18 are
high swing cascode devices. The souce of M27, which is Va, is Vov,M17+Vth,M7. Vc is set by M23,M24 and M19, which means that the swing is set by the same way. Therefore it is possible for two biasing legs to have the same swing but it is not that way according the sizes of M27 and M28.
Best

Title: Re: Class AB opamp
Post by Alexandar on Aug 31^st, 2010, 1:03am

M16 and M18 need 2x a certain VDS, which is probably lower than 1x VGS needed for M17. So M16 and M18 have more margin. The nonequal W/L for M27 and M19 compensates this to some extent. So by less drain modulation the currents in each branch will match better.

Title: Re: Class AB opamp
Post by nobody on Aug 31^st, 2010, 1:50am

Hello, Alexandar

It seems that the souce voltage of M27 is biased by two branches.
One is through M23, M24 and M27.
Vs27= Vov,M24+Vov,M23+Vth,M23+Vth,M24-Vov,M27-Vth,M27
The other one is M15 and M17.
Vs27=Vg17=Vov,M17+Vth,M17
So these two branches are fighting each other. Is that right ?

Title: Re: Class AB opamp
Post by thechopper on Sep 1^st, 2010, 7:08pm

nobody wrote on Aug 30^th, 2010, 7:29pm:

However, I do not get your point about why M28/M27 are larger.
Therefore it is possible for two biasing legs to have the same swing but it is not that way according the sizes of M27 and M28.

It is in fact possible to make the same swing in both branches, but in that way full swing on the output branch would not be used. In order to fully use the output branch swing, different sizes are required as Analog DE suggested.

nobody wrote on Aug 30^th, 2010, 7:29pm:

So these two branches are fighting each other. Is that right ?

The answer is no. M23 and M24 set the gate voltage of M27, but not its current. Such current is set by M15 and M17, which set the source voltage of M27. Thus the two branches are not fighting each other.

Best
Tosei

Title: Re: Class AB opamp
Post by nobody on Sep 1^st, 2010, 8:47pm

The answer is no. M23 and M24 set the gate voltage of M27, but not its current. Such current is set by M15 and M17, which set the source voltage of M27. Thus the two branches are not fighting each other.

M23 and M24 set the gate voltage of M27 and the current through of M27 is set by M15 and M17. Therefore, the source voltage can be derived because of the size and current of M27.
However, the source voltage is also set by M15 and M17. So I think they are fighting to each other.