about comparator design - compensation, parameters

[lhlbluesky]

comparator no feedback

if i use a two-stage opamp as a comparator(such as folded cascode + common source or telescopic cascode + common source) to get a high gain, then is compensation needed? if PM is small(smaller than 45 deg), is that ok? if not ok,then what's the effect and how to improve it?

besides, for the comparator design(two stage opamp), what parameters are important? and how to ensure that?

thanks for reply.

 

[haff99]

two-stage comparator design

An unstable opamp is a comparator. You don't need any compensation if you want to use your op amp as a comparator. You want the comparator to peg rail to rail, very quickly, with a very small delta at the input.

If you left the compensation in, the circuit would work as a comparator except it would be very slow. So unless you have some specific reason for a very slow comparator, I would not use any sort of compensation.

 

[JPR]

comparator bode plot

The compensation of an amplifier is needed for it to operate in a negative feedback orientation. Phase margin is really only a measure of stability when an amplifier is used in a unity gain configuration using negative feedback. Feedback is used to provide a specific gain (or gain vs frequency in the case of a filter topology).

The purpose for a comparator is to provide so much gain that there really is no appreciable linear region between high output and low output states. Adding negative feedback here would tend to provide a larger range where the output is between high and low. For this reason, the comparator is usually used without any feedback. No feedback means that there is no stability issues, which means that there is no need for compensation. Any compensation that is added would tend to slow down the output.

 

[lhlbluesky]

auto zero of comparator

i know that, but for a sc comparator, in the first phase, output and negetive input is shorted to reset the output of comparator, so it need feedback also.then, is PM important? and how to design a sc comparator?any notice or ideas?

 

[sriramachandra]

gain of a comparator with no feedback

As far as i know compensation is required only for an amplifier because of its operation in negative feedback. The purpose for a comparator is to give huge gain.

So there is no need of any compensation for comparators

 

[leehying]

about comparator design

Some comparators used in switched cap circuit should stable, too.

 

[jcpu]

Re: about comparator design

Haff99 is right.
Bode plot is must for amplifier because they go with feedbacks.
When we use comparator, it goes beyond linear domain.
No feedback needed, hence no stability criterion.

In cases where system gets feeback like in sigma-delta convertors,
that's system issue, and architect should consider stability carefully in Z domain.
Not comparator designer.

Enjoy.

 

[JPR]

Re: about comparator design

It sounds like you are using the comparator in an "auto-zero" type configuration. That changes everything, since in the autozero phase, it is essentially an amplifier in unity gain feedback.

What I would suggest in order to maintain auto-zero capability is to consider auto-zeroing using the first stage in order to maintain stability. The first stage should have high enough gain to reduce offsets (perhaps a gain of 100-200 or more in the first stage is probably sufficient, depending upon the application...)

If you auto-zero the entire comparator, then there are a couple of things that might be possible. The first would be that you might be able to switch in a compensation capacitor when you have the comparator in feedback, then switch it out when you are running open loop and don't care about feedback stability. The second is that you would need to make the comparator, with feedback, fast enough to provide sufficient phase margin AND sufficient comparator speed.