Hi,

I am using Spectre to simulatd a LC VCO around 2GHz. According to advices from Ken, the number of output harmonics and maxsideband need to be increased until the noise result stop changing.

Typically, the phase noise regions divided into three regions, 1/f³, 1/f² and white noise floor. In my simulation, the phase noise keep increasing with increase of number of harmonic and maxsideband. specially, when I put a large number of output harmoninc in simulation, say 60, the white noise floor go to -80dbc at 500KHz offset.

The output node I selected is before the buffer, so any noise should be filted by LC low pass tank and the phase noise should follow 1/f² curve. Theoretically, the white noise floor should not be significant. I can not explain why this happen.

if this result is not expected, what's wrong with my simulation? I try a LC oscillator using idea component without any transistor and get the same result expect the white noise floor is lower than the VCO using real MOS (-130dbc at 23MHz). so I think this may not be a MOS model problem.

anyone had this problem before? any answers is highly appreciated.

thanks.

--zerk