Understanding pnoise PSD: a simple test with resistors

Hello,

I would like to understand the pnoise simulations.

Triggered by the Cadence Application note "Jitter Measurements Using Spectre RF", I tried to do a pnoise simulation on a simple circuit. I attached 2 slides to show the circuit, the pnoise settings and results. I did a .noise analysis first (slide 1), then a .pnoise (slide 2).

Circuit: The circuit is made up of a R1=1.57k resistor, an ideal switch with R2=32.5 Ohm series resistor and a C= 1.35p capacitor. 

The capacitor is charged through R1 and discharged through the reset switch+R2. The reset switch is driven by a reset clock. The switch is closed when the clock is high (=1V). The clock frequency is 2.5GHz

Slide 1 - .noise analysis: 

- If the reset clock has DC=0V, the branch with R2 is disconnected. The noise PSD=4KTR1=(5.09nV/sqrt(Hz))^2 and the bandwidth is BW=1/(2*pi*R1*C)=75MHz. These results are consistent with the simulated green PSD on slide 1

- If the reset clock has DC=1V, the capacitor sees the parallel R1||R2 that is ~R2 . The noise PSD=4KTR2=(733pV/sqrt(Hz))^2 and the bandwidth is BW=1/(2*pi*R2*C)=3.6GHz. Again, these results are consistent with the simulated pink PSD on slide 1

Slide 2 - .poise (tdnoise) analysis: I did a pss shooting, with 20 harmonics, then a .pnoise with the settings on slide 2 (sweeptipe relative, freq 100-1.25GHz). Then I plotted the PSD for all the different time instants.

I expected to find the same levels and BW of the .noise analysis, when the switch opens and closes, but the PSD I got are completely different: the low freq. values go from (1.59nV/sqrt(Hz))^2 to (1.77nV/sqrt(Hz))^2 and the BW is around 2.5MHz.

(Note that for .pnoise the plot is in V^2/Hz, while in the .noise plot is in V/Hz.)

I don't understand the values I got from the .pnoise analysis. Can anyone help me to understand why the PSD levels are lower and the BW narrower?

Thanks a lot!!!!

Test noise.pdf