[sdiy] simulation a relaxation type oscillator : was the jitter business

[Czech Martin]

i.e. simulating a saw core.

In normal time simulations one is interested in waveshape
and overall timing. But not in jitter or spectrum or amplitude error.
Actually they are minimised by the difference equation solver
by minimising the integration error. This means non equidistant time
step , of course.

For having an audible result equidistant time step is a must.
Then I fear that this has to be very fine, i.e. extreme oversampling.
E.g. 128x oversampling @ 44100 gives only 177ns time resolution,
in terms of a simulator for electronic stuff, this is very poor.
The adaptive algorithm would normaly switch to 1ns or so during
saw discharge.

So, massive computational load.
I can not think of any other method to simulate a saw core
without too much artefacts that could hide the things we are after.

Any hints?

Just in this very minute I have an idea: what about to
build an idea s&h plus alias filter into the SPICE type simulation?
It does not need to be SPICE really, because the difference equation
is quite simple and linear, so a special C program could be done.

MfG
Martin Czech

CAD/CAE Library Group
Micronas GmbH
Hans-Bunte-Str 19
D 79108 Freiburg
Tel.: +49-761-517-2422
FAX.: +49-761-517-2258
http://www.micronas.com