[sdiy] Analog Modeling, with a computer!

[Richard Wentk]

At 19:44 14/09/2005, Harry Bissell Jr wrote:

>I'd disagree (in part). Modeling a capacitor or
>resistor is unlikely to result in any change in
>sound..unless that component was AWFUL  (the parasitic
>elements are very large).

Well - no.

There's no such thing as an ideal component in the real world, and some 
capacitors particularly are far away from being ideal components.

This *will* affect the sound, often noticeably. So the only way to really 
model a physical circuit is to replace all the ideal components with 
nominal R+L+C+noise network models, and then add some extra mostly-C, some 
R and perhaps a little L to simulate PCB parasitics, and finally calculate 
possible crosstalk between all of the circuit elements where it's likely to 
matter.

It should be obvious that even for a simple circuit you immediately have 
the network from hell. Too bad. This is the only way it can be done properly.

>Modeling the semiconductors WOULD make a large
>difference and a better model would be noticible.

And that too.

True component level modelling is the ONLY way - really, I mean this 
literally - to accurately model analogue sound quality in software. 
Functional-level models are only going to be viable for the control path. 
Hence an ADSR or LFO can be modelled pretty simply without too much effort.

But anything in the audio path will HAVE to be modelled accurately at the 
component and board level to be truly indistinguishable from the real deal. 
The reason being that for most components there really is a significant gap 
between the notional model and actual performance.

You can certainly simulate analogue sound using functional level 
descriptions. But this approach is guaranteed not to capture the nuances of 
the best examples of the real thing. Put simply, there is a *big* and very 
audible difference between simulation and accurate modelling.

I have no idea how much computer power is needed to do this properly in 
real time. But I'd guess we're about ten years away.

Richard