How if we use MOSCAP or varactor as the on-chip loop filter for a fractional-N PLL? It's OK for a integer-N PLL but I don't know if the nonliearity of the capacitor in loop filter will has impact on the phase noise? Any one know this clearly? Or some reference about this?
By nonlinearity, if you mean the capacitance vs. tuning voltage curve not being linear, it will have NO effect on phase noise.
If you mean structural defects, like traps, in the semiconductor, and you are tuning to the voltage extremes where those traps inject holes/electrons, then yes you can have some shot noise increase. Mostly, though, defects in the semiconductor influence the active device, not the tuning element.
For fractional-N PLL with sigma delta modulator, nonliearity of charge pump will cause high in-band quantization noise. So I think the nonliearity of loop filter (cap vs. voltage) will cause such problem too. Does anyone use MOSCAP of varactor in loop filter of a fractional-N PLL?
I see your point. Almost any varactor in a locked oscillator will tune in phase more in one tuning voltage direction than the other.
There are things called "linear tuning" diodes, with very careful varactor doping profiles to keep the C(V) curve fairly linear. Probably not useful for a monolithic system though.
How about just making the +going cp current pulse a little bigger than the -going current pulse? I will bet that the % difference between up and down is fairly constant over a wide tuning voltage range.