wierd VCO pnoise vs current

Dear All,
I am designning a PMOS cross-coupled VCO with PMOS current tail. When I sweep the current bias from 3mA to 4mA the output pnoise varies a lot. The VCO frequency is 2GHz. The pnoise at 1MHz offset with current of 3.8mA is -125dBc/Hz while the noise increases to -115dBc/Hz with a current of 3.9mA or 3.7mA. I don't know whether the model is accurate. Do you guys have any ideas? If I have to design the VCO with this technology, which simulation results should I choose? I am so confused and depressed.
Thank you.

Yes, I have met this question too.
I think we should not trust the simulation result so much.
And you can only trust the the average of the result.
So the one time reslut is not the fact.

maybe the models it not so accurate.

khouly

Thakns. It is possible that the model is not so accurate. The PDK provides only RF model for the channel length of 0.35um, so I can't change the length to reduce the phase noise. Is there any other good ways to push the 1/f^3 noise to less than 500Hz? Is there good refs on the reduction of flicker noise in VCO or reducing the 1/f^3 noise corner?

i think we need opinion of one of the VCO experts here

but if the technology is 0.35 , u can't go less than this length

khouly