Hi all,

I am trying to set up a common mode feedback circuit for a two stage fully differential op-amp.

My first stage is a simple p-input differential op-amp with biased nMOS loads.

The second stage is a simple common source (nMOS input) amplifier.

I am using the differential difference cmfb circuit to set the output common mode.

The issue with two stage fully differential op-amp is the fact that the outer feedback is +ve for common mode signals which makes it challenging for designers to attain stable cmfb operation.

I read a few literature on the same but could not find one where they have taken up this problem in details and explained all the caveats.

Can somebody throw some light on this or maybe point us to some good paper that talks about it?

Thanks in advance,
Bhupendra

PS.
I am sorry if the question is redundant.
I tried searching for all topics related to cmfb but didn't seem to find a thread that takes up this problem.

Hi subgold,

Yes I can do that but the positive feedback for the common mode signal still exist and would tend to make the system unstable.

I guess what I looking for are the conditions that one has to meet in order to insure that the -ve cmfb loop is 'stronger' than the +ve feedback loop thus ensuring stability over the whole common mode range.

Regards,
Bhupendra

I believe you might have misunderstood what subgold was saying.  When you have two cmfb circuits in this case, you do not use the same outputs to regulate both stages.  You should use the outputs of the first stage to regulate the first stage and the outputs of the second stage to regulate the second stage.  On nice thing about this is that you can have two CM levels for your two stages because in general, maximum signal swing is not reached when both the input and the output of an amp stage are at the same CM level.  Doing it this way, you will not have any +ve feedback at DC.  Of course, you have to check the stability at high freq.

Another way to do it is to use a CMFB circuit that has both a +ve tap and a -ve tap.  That way, you drive the output stage with the +ve tap (since there is another inversion from either the tail or the load) and the input stage with the -ve tap (since there will be two inversions in that path).  Most continuous time CMFB circuits give both taps.

Hi Bhupendra,

If I understood you correctly, you are worried about the positive CMFB due to the feedback network?
You are right as to if the output CM increases then the input common mode will increase. Therefore you are right you have a two stage +ve feedback whose loop gain needs to be overcome by the -ve feedback of the CMFB circuit.
However the (natural) +ve feedback for the CM has a relative low gain: the gain is set by the common mode gain of each of the two stages comprising the opamp. Such common mode gain is much lower than the differential gain of each stage (assuming proper biasing op points for the biasing current sources and small lenght modulation effect for MOS or large early voltage if bjts).

Therefore the +ve feedback should always have a relatively small loop gain, which is easily overcome by the -ve CMFB circuit whose gain is based on differential stages.

Bottom line: the +ve feedback for the CM is based on CM gain stages (typically low gain stages) while the -ve feedback is based on differential stages (typically you convert to a single-ended voltage that controls some biasing inside the opamp).

Hope this helps.
Tosei