Dear community members,

I was trying to verify the validation of "pnoise" analysis
for kT/C noise simulation. So, I read through the doc named
"Simulating SC filters with SpectreRF" by Ken, and
understood the noise folding and "maxsideband" param.

My test bench is very simple; a nmos switch and a capacitor.
(fsample=135MHz, C=1nF, Ron=100)
                 __
    _______|   |_______
                              __|__  
                              _____
                                  |            

Based on the theory, if the RC bandwidth is much smaller than
the sampling freq, noise folding is negligible. In this case,
PSD of "noise" analysis and "pnoise" should match, I think.

However, simulation shows that they don't (please see attched).
Actually, PSD of "pnoise" is about twice larger than that of "noise",
and bandwidth is about twice smaller than that of "noise".
These two weirdos compensate each other
so that the total integrated noise of sample & hold waveform
becomes kT/C (I am looking at sample & hold waveform,
but it should not affect PSD since RC bandwidth << sampling freq).

Is this a bug in the simulator or am I missing something?
I would appreciate if somebody can explain what's going on.

Thanks,