Ken,

Let me rephrase the question for better clarity. I ran two pnoise sims, with a simplified generic NMOS LC VCO. The gm pair was replaced with an ideal -ve R using a VCCS. Tank has ideal L, ideal fixed C, and a non-ideal varactor across tank to capture the analog Kv of the VCO.

The 2 cases were:
1. Simulation with noise from a 1MOhm resistor injected into the tuning line using a VCVS. The resistor is not noiseless.
2. Same as (1), except with a 100fF included in parallel with the 1MOhm, giving a corner of about 1.7MHz for noise injected into the tuning line.

Expected Results: (1) and (2) have the same phase noise profile, up to the 1.7MHz offset. Beyond this frequency offset, (2) should have lower phase noise than (1) due to the roll-off in the injected noise.

Observed Results: The phase noise profile of (2) is same as (1), but shifted down by 22dB. The noise of (2) is much less than calculated, using the narrow band FM approximation from the Vtune line to the VCO frequency.

This is troubling as the noise injected within the bandwidth of the 1.7MHz pole is seeing different transfer functions in the two cases. I can't think of how noise at, say, 200kHz would change as a result of adding that 1.7MHz pole. The PNOISE sim result does not make sense in this regard.

Thank you in advance for your response,
paidian