I’m not an expert by any means, but in reading interviews or listening to podcasts/videos with people who work at places like Line 6, Positive Grid, etc., amp modelling can get incredibly complex.
While it sounds like there are various techniques, one is component modelling (and it sounds like there’s techniques within that). For a pre-amp tube, for example, your’re not just describing how it compresses/distorts. You’re modelling how those characteristics change at various operating points, how it interacts with other components within the amp. You’re modelling how the power tubes interact with the output transformers in the circuit, how the power section would interact with the varying impedance of a speaker load. Tube biasing, bias excursion, sag/squish; there’s all sorts of complex things going on within the amp that simulations like these try to model.
You might be interested by this thread :
I’ve managed to get a usable tone with only a x42-eq - Parametric Equalizer and a distortion plugin that’s on the bright side, not muddy (eg : CollisionDrive or GxSD2Lead). As with most amp sims, it sounds quite horrible until you add a cab sim. I’m not saying it does exactly the same as an amp simulator but the whole sounds good. The Pareto Principle seems to work here.
However, I don’t use this pedalboard because the Onyx amp sim is versatile, sounds good and doesn’t use more CPU than the Eq alone.
Thanks for the replies, it’s very informative and that’s a very cool video. I also had a discussion with chat gpt today to learn more about amp modeling. It sounds like it’s a very complicated task with state space modeling and a lot of differential equations.
There are things happening that aren’t as easy to describe or conceptualise as other effects. One example I can think of now is how it reacts to palm mutes. In my experience amp sims have a bigger change in sound between palm mute and no palm mute than standalone effects. I can hear it, but I have no way of conceptualising what is actually happening to the signal the same way I can for other effects.
Basically you could break it down to exactly that. While the saturation/clipping/compression mostly is non-linear. That could become very complex to cover the filter response of a real amp in digital dsp.
I use in most cases a IIR Filter to cover the linear response of a Amp, (EQ’ing) and a lookup table were I interpolate the non-linear response from. The table I generate from the amplitude response of the Amp (circuit simulation), the IIR Filter I generate from the Frequency Response. In most cases I use
IIR Filter → non linear transpose → IIR filter.