The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> Pnoise Analysis (Pipeline ADC)
https://designers-guide.org/forum/YaBB.pl?num=1067444614

Message started by Pieter Berkelaar on Oct 29^th, 2003, 8:23am

Title: Pnoise Analysis (Pipeline ADC)
Post by Pieter Berkelaar on Oct 29^th, 2003, 8:23am

Hi,

I'm working on a 10-bit 100Msamples/sec Pipeline ADC in CMOS. The architecture is 1.5bit/stage with a front-end SHA. Each stage is essentially a switched-capacitor gain block. I would like to use the Pnoise analysis to predict the output-referred noise contribution of a single stage. However, I am not sure how to interpret the results I am getting. Having read Ken Kundert's papers, should I :
(a) use an ideal SHA at the output of the stage with the sources option in the Pnoise setup or,
(b) use the tdnoise option in the Pnoise setup and specify the sampling instant as the time of the falling edge of the sampling phase?

In both cases I am unsure what range of freq to integrate over because the output of one stage will be sampled by a next stage, which will in turn add noise and again cause folding of the noise into the baseband. The cct uses two non-overlapping phases of a 100MHz clock. A figure of merit for an individual stage would help me scale the stages appropriately down along the pipe!

I would really appreciate any help, tips, ideas or opinions,
Regards,
Pieter ???

Title: Re: Pnoise Analysis (Pipeline ADC)
Post by Ken Kundert on Oct 29^th, 2003, 2:54pm

Pieter,
The output of one stage is sampled by the next, so you need to include this effect. There are any number of ways of doing it, but if you are interested in noise the simplest way is to use the time-domain (or strobed) noise feature. This approach is documented in Simulating switched-capacitor filters with SpectreRF , which can be found at http://www.designers-guide.com/Analysis. See the section on noise (section 2.2).

The sampled noise spectrum is periodic with a period equal to the clock frequency. To compute the total noise power, you would integrate the spectral density from 0 to f_c/2.

This is a very cool application of the PNoise analysis of SpectreRF. Please report back on how well it works. Also, feel free to ask any questions you might have.

-Ken