how to calculate the phase noise of a oscillator?

[EMfox]

I am running simulations with a LC oscillator and trying to see the output phase noise due to a given noise source.

Here is how I have been doing:

1. run the transient simulation for a while

2. compute the spetrum of the output signal using FFT, and plot it in dB

3. read the power density (dB) at a given offset frequency from the first harmonic (ω0±10MHz), and substract the power density at ω0. And I think this is the phase noise expressed in "dB with respect to the carrier"

I don't know if this is correct. Please advise me if I am wrong.

Thank you!

 

[eecs4ever]

I'm not sure either.

But i do remember that Ali Niknejad, a Professor at berkeley, said something about this during one of his
lectures last semester. He said sometimes , people think they are looking at the phase
noise, but its not acutally the phase noise, its the distortion due to the windowing effect of running the simulation for only a finite amount of time.

This effect probably is lessened if you run your simulation long enough?

You can watch his lectures here. The Label on the lectures and the actual content
may be slightly off.

https://webcast.berkeley.edu/courses/archive.php?seriesid=1906978242

We also did homeworks on calculating/simulation the Phase noise. We used Leeson's equation to calculate the phase noise.

**broken link removed**

look at HW#9 , problem 1. BONUS.
The solutions are also posted.

I think the TA used Spetre RF to do the phase noise simulation.

If you have lots of time on your hands, you can try to simulate the phase noise using your method and see if its close to what he gets?

Good Luck

 

[zajbanlik]

Hi

In companny where I work we schould do such analyses but customer didn't specified such thing and we quited. I was starting to reading some papers and I can strongly recommend this one **broken link removed**
There you can found some theory, some instruction for simulator and refernces for further research

 

[montage2000]

if the noise is not white, the result is wrong