new here plus a question...

[Martin Czech]

:::Now, I know *something* about digital filter design, but sadly enough I
:::know next to nothing about analysing analog electronics schemes, nor about
:::converting them to digital forms. :(
:::
:::So here's my question: would it be possible to convert these schematics to
:::the digital domain using bilinear transforms? If so, how should I try
:::this? Perhaps some of you could help me figure out the transfer functions
:::on these schematics, as that would help me a lot too.
:::
:::Andy and me have been pondering about specialised methods that could model
:::these frequency curves in the digital domain using weird non-linear stuff
:::like genetic algorithms, but then I'd need the transfer functions again :)

I'm no expert for discrete computation and I know little about analog
circuitry so here's my 0.02$:

Discrete systems have limited bandwith (Nyquist etc.)  Analog systems
tend to have wide bandwidth. Therefore some warping hase to be made. This
will destroy phase characteristics.  So the bilinear mapping does not
even work if there would be no nonlinear distortions, unless you use
pretty high internal sampling rate.

And this is the next keyword:

Nonlinear distortion:

The recent Moog discussion shows how important this is for "vintage"
sound.  And the moog example also shows that there is a complex
interaction between nonlinear circuit elements and more or less linear
filtering elements.

So , maybe you could modell each transistor, each cap and so on  in a
spice like format, but this will mean 10s of MHz internal sampling rate.
What an awfull waste of computation time. And still it has to be proven
if the spice modells incorporate all the audio effects.

I mean the models for a single transistor have up to 40 parameters and
most of them are not physical but ad hoc inventions to force better
matching between simulation and reality. And this development was made
to achive curve matching and stability, audio was never thought of.

The nonlinearity also demands a high internal sampling rate, because
(inaudible) partials would be aliased otherwise. Listen to cheap guitar
dsp fx stuff, and you'll know what I mean.

Transistors are well known and well explored. Loudspeakers, audio
transformers, Leslie cabinets and other strange stuff may still need such
good theory.  The guitar speaker simulation of most fx boxes I know is
simply a IIR filter that shows more or less the frequency plot of
the original. Phase? Nonlinearities?

I will not talk about tubes, because I know very little about that.
But it seems that the characteristics of tubes change pretty much 
with different signal strength (tend to get slower when overdriven).
Could any tube guru please confirm?


There is lots of investigations about violin or trumpet sound,
but so far I've seen less stuff about these above mentioned things.
Mostly for the perspective of makings things better: "How to avoid
loudspeaker distortion" and not "Investigation on Greenback distortion".
But perhaps I haven't looked to hard for that.


A physical mapping is not possible not desireable and also not needed.
A psychoacoustic mapping would be adequate.  And this is the main
question: How can we achive the wanted psycho acoustic effect with a
combination of simple elements like digital filters, function tables
etc etc. I guess nobody knows. There exists no theory for this problem.

Do you remember "Frequency Modulation"? This is a relatively simple
thing, simple math. Still we have no closed theory on how to emulate
a given sound with a couple of operators. And heuristical searching
can be a very long process.

So, to summarise, there's lots of stuff to be explored, theory needed,
measurements etc. etc. This is way to much for an individual, and
also for a company like Roland or Yamaha allone. And academic research
is always short of money and still sticks mostly on "trumpet" sound.

But sure enough: I wait for the day that I could have the same sound
out of a 19" brick as a 4x12 with tube amp will give me. Same applies
for a Mogg wall or Leslie cabinet.  That would be a very good thing.

So far any machine I tested has failed the test, for the reasons I
explained above.

m.c.