Read section 30.0 in http://www.designers-guide.org/Theory/cyclo-preso.pdf. The basic idea is that in practice there is a low frequency corner that is caused by large signal effects and the simulators cannot predict because they are using small-signal analysis. Clearly the noise cannot be 90dB at 1Hz because that would imply that the total noise power in the normalized noise curve is much greater than one. There must be a corner, and the frequency must be such that the total area under the normalized noise L is 1.

-Ken

Hi Kappy,
Phase noise spectrum, L(fm), is the power spectral density of the noisy oscillator waveform, measured around the fundamental, divided by the power of the noiseless oscillator waveform, at the fundamental. Adding noise to an oscillator does little to change the total signal power. So the area under the first power spectral density is the same as the area under the second power density, i.e. power conserves. Therefore, the area under the phase noise curve is always one. For typical oscillators, phase noise is below 0 dBc/Hz around the fundamental and away from the fundamental. For ultra-low-noise oscillators, phase noise is above 0 dBc/Hz at offsets less than 1 Hz. Measured phase noise can never be above 0 dBc/Hz at offsets larger than 1 Hz. If this were the case, the area under phase noise curve would exceed one. Carefully note that measured phase noise at carrier frequency is rarely 0 dBc/Hz.
David

Dear Ken Kundert, David Lee and Frank Wiedmann

Thank you very much for your kind answers.
I didn't know my question was repeated one.
I should check the previous answers like you said.
I think it will take some more time for me to understand clearly.

 Before that, just one more question.
I modified my VCO and I want to check if the modified VCO is
really more improved than previous one by simulating phase noise.  
And these two VCO have different positive values at 1 Hz offset.

I measure the phase noise difference (A) of the previous VCO between 1Hz and 1kHz offset
and then I checked the phase noise difference (B) of the modified VCO between 1Hz and 1kHz offset.
If A < B, Does it mean that modified VCO is better than previous VCO considering phase
noise?  if this method is not correct, please, let me know the correct method.