I has a question about the digital output of comparator.
Like attached file , the comparator has two output Vo+ and Vo-
Then i want to get the PCM signal , is only get Vo+ to do next step(FFT analysis)?
or get the differential output (Vo+) - (Vo-) to do next step?

I had already checked that , there are have some difference.
Which are the correct output , single ended or differential ?
or just see what application i used?
thanks~~!!

i say differential, as only then will you be cancelling common-mode effects of oversampling clocks etc.
i think if you do the fft of single-ended vs differential (assumming you have real transistors below the symbols!) you WILL see a difference, especially in even-order harmonics.

Hi ACWWong,

I don't quite understand the point you make. It is clear that one needs to look at differential
outputs when dealing with analog signals such as the integrator outputs to cancel the common-mode
and even harmonics,

However, if a comparator is working correctly, then Vo+ is always the complement of Vo-, and these
2 signals are digital, and contain exactly the same information, except for an inversion. So, there can be
no common-mode or even harmonics on these. To have even harmonics on differential comparator outputs,
you would need a situation where sometimes Vo+ = Vo- and not its complement. This would imply
improper comparator operation. So I really wonder how one would see errors unless one was doing the
calculations incorrectly.

Could you elaborate your point further?

Regards
Vivek

Yes, ideally Vout+ and Vout- are complementary, especially when it come to the signal being processed. Yes the signal is also digital.
Unfortunately other signals (unwanted) also appear, such as clock feedthrough, power supply/substrtae noise etc. and then can add to both Vout+ and Vout- simultaneously, therefore making Vout+ not the exact complement of Vout-. These effects also result due to non-ideal comparator design, rather than improper design.
Therefore to help cancel these unwanted (mostly even order, of which DC can be though of "0" order, whcih is the point sheldon made) effects, and get the best out of the fft, the fft should be done differentially.

Of course i am an RF/Analog IC person, so my judgement is always somewhat blurred  :P


vivkr wrote on Apr 21^st, 2006, 1:48am:

Hi,

Most differential output signal is centered around near zero. However, the single ended output signal will be centered around a DC voltage. Therefore, my opinion is that, if you are doing FFT on the single ended output, you might found that it will has a very high DC component where the FFT of the differential signal will not shown. Depending on what kind of window you are using, you might find that the component at DC actually leak to other low frequency component, which might cause your output DFT to looks differently. Since FFT/DFT is actually sampling at a precise timing, it does not matter what your transient output looks like, what is important is that at the moment the FFT samples, your output is at the right value to be sampled, therefore unless the output is heavily contaminated with high frequency noise, otherwise it should give you the same result for using either kind of signals except for the DC part. For DS modulator kind of circuits, the absolute value of the output is somehow not important as long as they are in the range to properly define a logic state. For example, you can use an ideal amplifier to amplify the output of a modulator 100 times, so that your output is let say. 200V, take a DFT, yet you will find that the SNR, SFDR will remain the same. If your circuit will be working in baseband, I will suggest you to take FFT differentially since it is easier.

Regards,
Chase