Dear all,

Right now I am confused by the time domain noise simulation for phase noise.
I setup is that I only compute the phase noise at 1MHz offset

All voltage is taken as differential signal.
I first plot 3 items: phase noise, the I plot the output noise, and the rms spectrum of the signal.
It agrees that

  phase noise = 20*log10(pnoise output noise in V/sqrt(Hz) / PSS differential rms voltage )
                    = 20*log10(pnoise output noise in V/sqrt(Hz) / PSS differential peak voltage ) + 3dB


And I switched to time domain noise simulation, also only compute 1MHz offset.
Then I get a noise power over an oscillation cycle.
I take the average power, and divided by the sqaure of the rms PSS amplitude.

10*log10(tdnoise average power in V^2/Hz) - 20*log10(PSS differential rms voltage), and this number is 3dB higher then the previous phase noise result at 1MHz offset. So it turns out to be that:

phasenoise = 10*log10(tdnoise/2 in V^2/Hz) - 20*log10(PSS differential rms voltage)
                = 10*log10(tdnoise/2 in V^2/Hz) - 20*log10(PSS differential peak voltage) + 3dB

Rather than simply use the average power, an extra 0.5 factor should be added. I do not have any idea about it.

Did anybody experienced the similar problem?