Friday, November 6, 2009

Tube Modeling diversion

This has nothing to do with Rakarrack hacking, but just thought I'd put it out there:

undefined muse reason = none;
undefined muse insanity = reason - insanity_threshold;

undefined muse model_tube_amp()

if (insanity > reason)
return 0;

...and my thoughts about amp modeling:
It seems the most reliable way is to use convolution for frequency response and a look-up table for nonlinearities.

These can be measured directly from the tube amp. Most of the pre-amp filtering & tone controls can be accurately modeled with regular digital 1rst order filters.

It gets the most sticky when you are dealing with output power amp stage to the speakers. It's hard to decouple the impulse response from the nonlinear effects of transformer saturation, so the lookup table + convolution solution is only an approximation. A fair mathematical model requires a noteworthy DSP engine to process in real time.

Though, honestly, the sounds I love from my tube amp come from the preamp. I rarely turn it up to the point where power amp + transformer become notably nonlinear, and I don't actually prefer the resulting loose and mushy sound. One could make a fine FOSS tube preamp with some accurately modeled tone/frequency shaping filters and a look-up table generated from a trace of a 12AX7 tube.

Convolution with a good cabinet IR would be the final stage, simply ignoring nonlinear effects in the power amp & transformer stage. This is reasonable, since higher-end amps are designed to be rather hi-fidelity on the power amp stage. It's funny how the analog engineers are trying to minimize these kinds of distortion and the DSP engineers are trying introduce it into the system.

The result may not produce a perfect replica of amp "X", but it could produce the same nice interactive characteristic. It all depends whether the goal is a good sound, or a convincing model of something specific.

In the FOSS world, there's much more flexibility since you don't have pressure from a marketing department to produce a sound that sells. You can apply some creativity to the software development process and create something that sets its own precedent.

Most of my complaint about tube amp simulators is not so much how well the amp is simulated, but it's about what amps are modeled. I have a Peavey Duel 212, and have modified the circuitry to my taste. Line6 has yet to produce a product that behaves in any way like my amp, even though their modeling algorithms are perfectly capable. Only, I can't open up the hood on a POD and change a few capacitor and resistor values...

So you say, "well they have the DSP development kit for the ToneCore docks". They don't give you their proprietary amp modeling algorithms to you have to develop your own, and here's what I have to say about that:

No comments:

Post a Comment