Hey Everyone --

I'm working on designing a dynamic latched comparator (for an ADC) and am trying to test the noise response (in spectre).  My current simulation methodology is to run an AC simulation of a zero input comparator (that is on) and find the approximate bandwidth of the device.  Then I switch to a transient analysis with noisefmax = AC bandwidth.  I then find the BER of the output with an LSB/2 input and if I get above 99% accuracy I figure I'm ok.

My general question is whether this is accurate or not.  While I haven't been on the forum for very long, I did see some posts questioning the accuracy of transient noise and suggesting using Pnoise to get the same information.  

Even if Pnoise is more accurate, I'm still unsure of how to test the comparator.  I would like to get the input referred integrated noise, but I'm not sure where to strobe the output (ie right before the comparator turns on or right after) and how the input referred calculation is made.  

Thanks for the help and if someone could point me to a past post that I might have missed that would be great too!

Hi Sheldon --

Thanks for the post.  I've been reading the paper recommended and will try to get Pnoise simulations for my comparator parameters.

On the transient noise front, why is it not as push button as you say?  From the set-up it seems that most of the parameters are pretty simple, with "noisefmax" being the only possible question mark.  Are there any references on doing "BER" type analysis using transient noise?  Thanks a ton!