pnoise/pss question

[ljp2706]

I have a design that is an autonomous frequency circuit. For it to initialize properly in the transient domain analysis I have a dynamic parameter set to force a reset. So technically I don’t have any driven frequency sources in my schematic. I am having trouble getting pss to show the correct transient state. It looks like the signal is stuck in reset which is the global parameter default. Does pss bypass this dynamic parameter, is there a way to include it if it does?


It is not an oscillator but it behaves like a relaxation oscillator, so I am thinking I can pretend like it is for simulation purposes in pss/pnoise.

 

[dick_freebird]

Well, you do have to get to a transient solution which
has a periodic behavior, and then run the exact length
of that cycle. Right?

You may need to impose an intial condition which "runs
clean from zero". I'm not sure if running a "preamble"
(to get to clean cyclic behavior) and then a nonzero
start time, can be used for these analyses (not my
thing, watched others do it and that's all the advice I've
got).

A relaxation oscillator is not small signal continuous so
these analyses are probably what you need to be doing
(if you care about noise and harmonics and distortion,
but do you? What's the harmonic distortion of a sawtooth
anyway?).

Do you truly need what these "meta-analyses" produce,
that a simple transient (or N of 'em, with care-abouts
picked off by Calculator) wouldn't?

In enther case I think the first order is to get a single cycle
transient to run (next would be that second cycle is
dead-equal to the first - not to be assumed, there can be
"history effects" in logic, sampling, even too-slow-for-the-
cycle "stuff" outside the main path of interest that needs
cycles of pumping before it gets right).

 

[ljp2706]

Thanks for the input. I did what you said and I was able to get pss to converge on the proper solution. The issue I am seeing now is that the noise plot is much higher than what I expected and does not match up to the normal noise analysis. One thing that helped was not sweeping frequency to a really high value, I truncated at 1/2 the fundamental frequency which is based on suggestions I saw online. I suppose in pnoise going much beyond that upper frequency starts folding noise back and corrupts the results. That knocked the noise into a more believable range but still too high (1V/sqrt(hz) spot noise to ~mV/sqrt(Hz)). I am wondering how do I select the lower frequency bound for pnoise analysis. Currently I have arbitrarily set it to 1Hz. Or are there other parameters I may have set improperly that could cause this to occur.


To comment on the “relaxation oscillator” note you made, it’s not exactly an oscillator and because of that, the periodic signal does has a linear portion that I am concerned about the impact of noise. That’s why I am event triggering pnoise in those regions but not the areas where the signal behaves like a digital pulse.