I have a question about the “Simulating Switched Capacitor Filters with SpectreRF” paper.

For the simple track and hold, Sc(f) is listed as eq 16.  Next eq 17 which is the total noise is calculated by integrating eq 16 from 0 in infinity.  The result of eq 17 is kt/C.

In the results section eq 19 evaluates eq 16 at DC.  My simulation matches up for this.

Next I try to get the total noise by integrating over the output from unsmpledNoise^2 with Cadence’s AWD.  I’ve attached this waveform.  The result is 284p.

If I calculate kt/C I get: (13.81e-24)*300/10p=414.3p.

I would expect these numbers to be very close.  Is this a numerical problem or am I making a mistake?  I know I’m not going to infinity but I’m expecting the spectrum to be so flat that it doesn’t matter.

Thanks,
Brian