The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Charge injection in PXF analysis
https://designers-guide.org/forum/YaBB.pl?num=1225990810

Message started by Tlaloc on Nov 6^th, 2008, 9:00am

Title: Charge injection in PXF analysis
Post by Tlaloc on Nov 6^th, 2008, 9:00am

I have a switched cap integrator and am trying to determine the CMRR from a variety common mode sources. Can PXF (or PAC) take into account the variations in charge injection due to the changing impedances as a function of the common mode voltages? From my understanding, PXF is a linearized, small-signal analysis around the large signal bias point found by PSS, but is charge injection a bias change for the switches? If that is the case, then it would seem like PXF cannot include those changes.

Title: Re: Charge injection in PXF analysis
Post by Ken Kundert on Nov 6^th, 2008, 10:58am

Yes.

Try it and see for yourself.

-Ken

Title: Re: Charge injection in PXF analysis
Post by Tlaloc on Nov 6^th, 2008, 11:24am

I tried to verify it using a tran analysis. The circuit is a swcap integrator. PXF reported the common mode rejection from the CM input to the differential output as 80dB at DC (really 0.1Hz). With a 1V change, then, I should see a differential output shift of 100uV. I ran three different trans, all with the other sources zeroed out and the common mode voltage at 0V, 1.8V, and 3.6V. Based off of the PXF I would expect the differences in the outputs to be hundreds of uV's, but they were identical to 500nV. These two results don't seem to match, but I don't know how to resolve the discrepancy.

Title: Re: Charge injection in PXF analysis
Post by Ken Kundert on Nov 7^th, 2008, 2:00pm

You do not have to avoid DC in PXF. All of the small signal analyses in Spectre support the frequency being 0.

I did not understand your statement that "they were identical to 500nV", but in generally trying to get uV accuracy out of transient analysis is dubious. Generally small signal analyses are much more accurate than transient when only small signals are involved.

-Ken

Title: Re: Charge injection in PXF analysis
Post by Tlaloc on Nov 10^th, 2008, 3:20pm

I start the freq at a non-zero value since the waveform tools can't display a log scale x axis with 0 in the mix.

There are two reasons that I am looking at the tran analysis.
1) I am new to PXF, so I needed to verify the results with a known simulator to prove to myself that it was running correctly.
2) Two transient sims with a difference only in the common-mode voltage takes about 13 minutes for me to run. One PXF with the necessary PSS takes about 45 minutes. Since I am only debugging the rejection of one of the independent sources, the PXF's extra information is not helpful during this stage in the design.

Having said all of that, I am now seeing agreement between the tran and the PXF. It is looking like the absolute accuracy is at least as small as 1uV, which is well below the 30uV or so of CM error that I'm seeing. Thanks for your inputs.