Spectre Phase Noise Simulation for Superharmonic Injection Locked Oscillator
I have two oscillators running at a nominal frequency of 10 GHz and 5 GHz respectively. The 10 GHz oscillator has a lower phase noise compared to the 5 GHz oscillator. I am trying to run the simulation where 10 GHz oscillator is acting as an injection oscillator to 5 GHz oscillator. I have set the Beat frequency as 5 GHz (since the 5 GHz oscillator is oscillating at 5 GHz) and declared the 10 GHz as the oscillator node in the pss analysis, and tried to see the phase noise at the output of 5 GHz and 10 GHz nodes using PNoise setup. The Phase Noise at the output node of the 5 GHz oscillator shows the phase noise which tracks the phase noise of the 10 GHz for frequency below the Jitter Tracking Bandwidth and this is what expected. But the phase noise at the output node of 10 GHz shows a flat region at High frequency which is even higher than the phase noise of 5 GHz whereas the phase noise of 10 GHz in standalone mode is lower and rolls-off with the increase in frequency.
Is there anybody who know why we get a flat noise floor at higher frequencies in the phase noise of 10 GHz oscillator? Is this a problem of Spectre or something I am doing wrong? Or the spectre is unable to calculate phase noise of oscillators with different tones in one simulation?
I will appreciate the feedback.
Sounds annoying but sometimes a higher frequency oscillator could have lower phase noise than a lower frequency oscillator, even they use the same active component.
Leeson equation stated that: doubling the operating frequency the phase noise increase by 6dB, but in the same time, doubling the loaded-Q of the resonator the phase noise decrease also by 6dB.
Phase noise improves with higher loaded-Q of the resonator, and because the Q factor of some reactive components increase with frequency, it is possible to get better phase noise at higher frequencies.
Usually can get lower oscillator phase noise at higher frequencies when the output signal is extracted from the resonator, in this way using the resonator response selectivity to filter the wideband noise spectrum.
is the flat phase noise region of the 10 GHz oscillator perhaps 6 db or so higher than that of the free running 5 Ghz oscillators? Maybe your 10 is locking to the 5, instead of the other way around?