Recently I have tested a simple circuit which composed of a VCO and a gilbert cell based frequency doubler. When I use SpectreRF to check the phase noise both of the VCO and doubler output, the phase noise of the doubler is only 4dB larger than the VCO. I thought the multiplying circuit should give a 6dB rise in phase noise. What's the problem with this testfixture?  The phase noise becomes better than ideal one.

 I've checked my results, but the problem is still not solved.
What I thought is that when the freq. doubling function is from the multiplying of one signal and its 90 degree shifted version, the phase noise of the resultant output is not degraded by 6dB.

  Imagine a digital-like signal that is passed a polyphase shifter to generate its quadrature replica, and then the two signals are mixed to generate the freq. doubled signal. The jitter of the original signal should be equal to the variance of 2*period of the doubled signal. Is the corresponding phase noise still degraded by 6dB? or only 3dB??

Thanks a lot!

I use time domain pnoise analysis to find the noise around the
trigger threshold, and the calculate the equivalent phase noise from jitter and the results turn to be right.