Hi,
I have a question regarding the pnoise simulation when the noisetype is chosen as "pmjitter".

My understanding is that, it samples the output noise when the output voltage crosses the "threshold value" and calculate the total voltage as well as the phase noise spectral density.

However, it seems to be more than just that. Take an example of driving an inverter with an 25MHz square wave (with finite rising/falling time) and adding some band limited noise (say, nri=4nV/rtHz with 1MHz 10th order Butterworth filter) to this voltage source.  

Now if I run the pss+pnoise(pmjitter) simulation, I will get the output noise psd of this input noise source, e.g. nro=16.7nV/rtHz. This indicates the gain of the inverter is about 4.1 times (Note that the noise is <<f/2 so there is no noise folding in this case).
I can also get the input voltage when the output passes the threshold point from the PSS result, say, vdd=1.1V, vout=0.55V, vin=0.31V. Now if I do DC or AC simulation, the simulated gain with this input/output combination is 6.3. Which means a 4nV/rtHz input noise results in  25.7nV/rtHz output noise, and this is HIGHER than reported by the PNOISE.

So what have I missed? Thanks a million for help!

Regards,
Siwen

Hi Shico,

The inverter is made of real CMOS devices. However in the output noise, I am just simply checking the output noise due to the ideal noise source (i.e. 1kOhm resistor which is 4nV/rtHz).

The bias point I used in the DC/AC simulation is obtained from the previous SpectreRF PSS simulation, so it is exactly when the PNoise calculates the total noise.

What I am trying to understand is the method of calculating the output voltage noise spectrum density, which is essentially how to calculate the noise transfer function. Thanks.

Regards,
Siwen