[time-nuts] Simulation tools for oscillators

[Dr Bruce Griffiths]

Tim Shoppa wrote:
> Are there any useful, free or near-free tools for simulating
> high-performance oscillators? I'm thinking particularly about
> making at least a little bit quantatitive statments
> about phase noise through simulation.
>
> SPICE will simulate an oscillator starting up, doesn't do a
> particularly good job when the nonlinearities of a simple
> oscillator kick in to limit the amplitude but most of the
> active device models can be tweaked to mimic the real world
> near saturation.
>
> If the active devices never go into saturation but there's
> an actual AGC loop then things can be done somewhat
> realistically without tweaking.
>
> But I have no idea how to properly model phase noise in even
> the most cartoon-like way with SPICE.
>
> Setting the time step to something ridiculously small and
> then FFTing to the frequency domain is almost completely
> hopeless due to all the numerical noise.
>
> SPICE old-timers will laugh at me because I even tried to
> simulate an oscillator in it, I'm sure :-). Hey, I actually
> started on a 11/780 with 50 users too!
>
> What do the pros use? I know there are goals (maximize
> power, minimize noise, at the same time minimize heating)
> and SPICE can vaguely help with them
> but common sense and cut-and-try seem to be the best tools
> I know of!
>
> Tim.
>
> _______________________________________________
> time-nuts mailing list
> time-nuts at febo.com
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>
>   
Tim

Most of the attempts I've seen to calculate oscillator phase noise, seem 
to rely on using Leeson's equation for oscillator phase noise 
supplemented, where possible, by measurement. However Leeson's equation 
doesn't allow the effect of amplifier flicker noise to be calculated 
from the known flicker noise characteristics of the active device.
Fortunately the latest extensions to Leeson's equation which use the 
fact that the noise in an oscillator is cyclostationary appear to work 
well in that the calculated 1/f^3 phase noise corner frequency that 
agrees calculation well with actual measurement.

Bruce