Hi

I have performed some simulations on a simple track and hold
circuit. I am not using the verilog-A model of a track and hold
as is done in the paper "simulating switched capacitor circuits with
spectre-RF". Instead I'm using the instance switch in analogLib,
an ideal resistor from analogLib to model the switch resistance and an ideal capacitor in analog lib as well.

I have confirmed the simulations in the paper for Ron = 2.3kOhm,
C = 10pF, m=0.4 and fc = 400kHz and I get the correct results.

However I have tried some other experiments that raises a few questions.

1. It is known that to run the PSS and Pnoise, the input signal can
be disabled. However I tried the simulation in figure 4 in the paper for maxsideband = 5 (Low value just to get fast simulations) and an input signal with frequency 2kHz. I believed that wth the input signal enabled the noise spectral density should not change but it was reduced and Sc(0) was about 6nV/sqrt(Hz). The same simulation with 1000 sidebands then returned the spectral density back to the correct level. I cannot understand why this is happening or maybe it should happend?

2. I tried a simulation without the input signal but changed the values to Ron = 240kOhm, C = 165fF, m=0.5 and fc = 4MHz. The reason that
I changed Ron to such a large value is that I will use this circuit to
generate switched exponential pulses. Off course now the capaciitor
won't charge fully during one half of the clock period but the noise
spectral density should not be unaffected. I plotted the theoretical noise PSD for the circuit with the values used above and simulated it in spectreRF but the results didn't match. The Graph showing both PSD:s is attached. The integrated noise seems to be the same but the shapes are clearly different. What might be the reason to this?  

Thanks for the reply