LC VCO's weird phase noise performance

The LC VCO is the popular cross-coupled N/PMOS type with NMOS tail current. The spectreRF noise summary shows me that the N/PMOS dominate the phase noise contribution(the four N/PMOS contributes >99% phase noie if I choose the integrated noise from 1khz to 10MHz) while the tail current contribute negiligible phase noise. Is this indicating the noise summary is something wrong? I am setting the maxsideband as only 5.

Moreover if I bias the NMOS tail current source in slightly linear region(Vds<Vdsat) through sizing. To my surprise the PNOISE is better than that if the tail NMOS is biased in saturation region.

Please help to analyze.

Thanks.

I've encountered the same problem before....i can't solve this weird problem till now...