Subject: Re: Aylward, Engelhardt: Noise in a transient sim?
X-Newsreader: Microsoft Outlook Express 5.50.4807.1700
Date: Fri, 18 Oct 2002 16:21:23 GMT
NNTP-Posting-Date: Fri, 18 Oct 2002 12:21:23 EDT
Organization: Cox Communications
"Kevin Aylward" wrote in message
> I think I see what your getting at. You mean effectively having the real
> noise of a device that is usually only given in the frequency domain, be
> given in the time domain. This would seem to be a bit tricky, and
> probably more involved then what one really requires. i.e. I can't be
> bothered with all that effort:-)
That is what I mean, and yes, it is what I require...
> The deal is though, you can get a pretty good estimate by worst casing
> and erroring on the high side. I have simulated, e.g. VCO's and simple
> put various signals on the power supply to see the effect. Its often the
> case that it is extraneous noise that is the issue with high performance
> designs, such that the normal circuit noise is not that relevant. If you
> have digital switching in analogue circuits, it difficult to predict
> just what noise is going to be anyway.
I can isolate my circuits from supply and digital noise well enough that the
dominant contributors to my VCO jitter are the FET noise sources.
There are various estimation methods that yield reasonably accurate VCO
jitter estimates from AC simulations; Hajimiri's Impulse Sensitivity
Function is the one I see used most often. It works well, but to achieve the
best accuracy, you have to run separate simulations at different bias points
over a cycle to find the noise at each bias point, in addition to all the
transient simulations required to construct the sensitivity function. It's
significantly easier, and usually faster, to run a transient noise
simulation and extract the noise from that.
-- Mike --