Hi dear fellows.

I have not much experience designing comparators, so I need your help.

From the research I did, there are basically two types of comparators, those which use clock (latch type) and those who not, for example the basic 3-stage comparator (pre-amp, positive feedback, ouput drive) or the comparator with internal feedback (cross coupled bistable) and yet another variant known as Clamped CMOS voltage comparator (Allen's book).

Because I am not interested on using a clock, those last three are of my interest.

From what I read, I concluded that maybe the best one is the comparator with internal feedback (cross coupled bistable) whose circuit is this one:



Is fast, has less delay, relatively good gain;

Regarding this, I would like to hear from you, in your opinion which one is best.

The next question is: What consideration I must have in mind when designing the comparator? Should I use the lowest L in all transistors? The differential pair must have a large W? What current should I point? (I don't have restrictions) All the transistors must be in deep saturation or close to subthreshold? (100mV to 200mV(?)) The tech that I am using is UMC 130nm.

To finish, I tried to design that comparator with internal feedback (cross coupled bistable) based on some examples and the results were:

The circuit with dc operating points (both inputs with 1.65V):


The circuit operating as a comparator:


And the DC characteristic (don't know if it is a relevant detail):


The waveforms:


Last, the AC response, using the AC analysis on the Vin- pin, with a DC source around 1.65V:


The differential pair dimensions are 40um/0.3um.

The overdrive voltages I think are very high!  ::)

I would like to hear from you and I appreciate any tip.

Thank you very much in advance.

Best regards.

HI Electron and thank you for your answer.

I was predicting that what you said, that the AC analysis is not of great help/use.

I must say in first place that there is nothing defined. It's all open. That said, regarding the questions that you did I don't know if I can answer them all.

The comparator will be operating @ 500MHz. It will compare a sawtooth waveform with an almost DC value, almost because that DC value will have a small ripple.

what are your tolerable offsets?
Don't have clue, but I imagine that must be as low as possible.

What type of speed do you need?
500MHz.

how small of a voltage difference do you need to respond to?
This must have with the resolution of the comparator, right? Well, the lower as possible would be good, taking into account that the DC value that I referred before is coming from the output of a buck converter. So, in the limit, the comparator must be able to response to the rapid load transient.

do you need hysteresis or not allowed?
Because the DC reference has a small ripple, I don't know at which point it is necessary hysteresis.

power consumption?
Well, power consumption. That another issue. I don't have restrictions, however, knowing that the converter will supply 100mA, I think that 1% of that would be the limit (1mA current budget; The power supply is 3.3V), but if I cannot met the specs, I can go higher on that value.

common mode signal range?
As I said, the comparator will be comparing a sawtooth, which will vary from ~0V up to 2V, with a DC voltage, around ~1.2V, so I would say that the input common mode signal range should be around the minimum possible (at least > 400mV or so) and 1.9V.

Thanks for the help Electron!

Kind regards.

I see. However I didn't know what your comment can help me. Thanks anyway.

With the above mentioned comparator that I have done (in a first iteration), what is troubling me is the fact that he doesn't produce a rectangular waveform when he has to compare in a certain level (@ 1.9V):



I don't know if it might be related to the delay that he suffers.

The comparator will be used in the feedback loop of a buck converter.

The inputs are a sawtooth waveform and a "almost" dc voltage. Almost because it has a high frequency ripple.

That's the application.

I know that any kind of amplifier can be a comparator, but for this application I was thinking on that bi stable comparator because it has the cross couple which improves the gain (negative resistance).

Another option could be the Symmetrical OTA, because it can have a good slew rate.

Kind regards.