The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Switch Cap pnoise question
https://designers-guide.org/forum/YaBB.pl?num=1380837074

Message started by prestonee on Oct 3^rd, 2013, 2:51pm

Title: Switch Cap pnoise question
Post by prestonee on Oct 3^rd, 2013, 2:51pm

I have been trying to verify correct usage of the spectre rf pnoise sim in my circuit by doing a simple switch cap sim, and have ran into issues(even after reading the switch cap paper). :-/

If i set noise type=sources, run the frequency well beyond the noise bandwidth set by the rc, I get the correct sqrt(kt/c) value for integrated noise,20uV.
If I then change setting to the timedomain and run again the noise is much higher. It is clear when i view the n() plot that the noise is spiking at intervals of the pss period. in the paper i did not see the setup of the sim but noticed the clock freq was 1mhz, and it appeared that the flat thermal noise was measured in sim and then multiplied by the 4rc noise bandwidth to get the total noise as opposed to integrating the full sweep to verify total noise? I would normally ask if I am missing something, but I know that I am so I will start off by asking, what am I missing? how can i accurately measure switch cap circuit noise if artificial noise is being injected at sample frequency?

I have attached a screenshot of the two sims and total noise results.

Title: Re: Switch Cap pnoise question
Post by Ken Kundert on Oct 3^rd, 2013, 11:12pm

That 'artificial noise' is flicker noise.

-Ken

Title: Re: Switch Cap pnoise question
Post by prestonee on Oct 4^th, 2013, 6:17am

Ken,
Thank you for the response, but I am not referring to the low frequency 1/f but the noise spikes occurring at the sample frequencies in the time domain sim(brown).

I am thinking this is aliasing, which causes noise folding, and that if the sample freq is 100e6 then I should observe the total integrated noise of the time domain noise at 50e6 and compare that to the noise type sources at end point(10e9), doing so gives me two numbers very similar. If what I stated is true, when would I use noise type sources vs noise type time domain?
The circuit I am interested in applying this to contains circuitry to allow a sc amplifier to be either chopped or auto zeroed. Initial noise sims make it seem chopping should be done in the sources mode while auto zero needs to be done in the time domain?

Title: Re: Switch Cap pnoise question
Post by Ken Kundert on Oct 4^th, 2013, 10:06am

I believe those noise spikes are aliases of the flicker noise. Examining them in the noise contribution summary should confirm that.

-Ken

Title: Re: Switch Cap pnoise question
Post by prestonee on Oct 7^th, 2013, 1:00pm

When would I use noisetype= sources, over noisetype=timedomain?
and was i correct about when to stop integrating the noise in time domain(pss sample frequency /2)

Title: Re: Switch Cap pnoise question
Post by Ken Kundert on Oct 8^th, 2013, 10:36pm

The essential question is what is expected to observe the output of the DUT? If the stage that will follow the DUT is a sampling circuit, then you need the noise at the sampling point. If the DUT will be followed with a continuous time filter, then you need the time-averaged noise (for reasons I do not understand, Cadence named the time-averaged noise 'sources').

-Ken