The Watering Hole

I have been curious for quite some time why instead of modeling an amp, designers don't model the components that make the amp.

For instance, start with a given, the input from the guitar. From there take anhy amp, say a JCM800. I know squat about the path through the amp, but instead of modeling the end result, why not send the signal through a virtual circuit including all of the components? If there is a specific capacitor in the signal path, model it's effects on the signal.

Wouldn't, in the end, this give a more realistic and better end result? Also, then users could design their own virtual amps. If the modeling was accurate, a wrongly designed amp just would not work. But for those who actually know their way around schematics, circuits, signal path, etc... this could be the coolest thing ever made. Imagine being able to make virtual custom amps.

Is this just not realistic? Could one of you tech gurus explain to me why this would not work? A friend of mine said his dad teaches an electronics course and before the students actually assemble real circuits, they use a program to create fake ones. Maybe an advanced, elaborate version of such software would be useable for amp modeling purposes.

Revalver takes that approach.

And it is decent.

It uses IR convolutions modelling for speaker/mic -
Not sure what modelling technique it uses for the rest - but you can tweak individual components on a number of "real" parameters.

SynthEdit uses a similar approach except of course it is for synthesizers.  It can get quite time consuming and complex starting from the ground up.  Not to mention frustrating when it doesn't work.  The ability to add pre-built components or sections helps tremendously though.

Like any other program, if a modeler that modeled amps might have a couple hundred presets, so you could always sart your designs with an already configured amp, and mod the hell out of it one component at a time if you like. I imagine it would be quite fun to play with, maybe so much so that one might forget that they were actually building patches to play their instrument through.

SPA,

Good question!  I'm glad NFinga chimed in.  IMO, it would take someone with computer knowledge and someone with amp design knowledge to fully answer this question.  Hopefully Dar can add something when he gets home.   I know the amp design thing fairly well, and I know how modeling works fairly well, but I can't answer with authority on the modeling side.  Actually Noel has provided much of my knowledge on how they do that.  But with a decent understanding how how actual components in the amp design side of it work, this is a VERY good question.  And the answer probably is, it would take way too much computing power.  Too many variables.  But as computers advance as rapidly as they do, I would not be surprised if this was not available in great detail and with great results, in a couple years.  

I seem to recall that the Line 6 Vetta was supposed to be 'point-to-point' modelling

Rich  ;)

spaivxx — Sep 18, 2008Like any other program, if a modeler that modeled amps might have a couple hundred presets, so you could always sart your designs with an already configured amp, and mod the hell out of it one component at a time if you like. I imagine it would be quite fun to play with, maybe so much so that one might forget that they were actually building patches to play their instrument through.

Well grab a copy of revalver and play around with it.

I think that  modelling with  convolutions is  the way forward right through the components. i.e. don't try to understand the physics/algorithms of what is happening so you can model it - just model it by replicating what it does to a signal.

The clever modelling involved in that will be about the handling a whole mapping of convolutions across settings.

And it might require what seems like a shit load of processing power now - but it is an inherently parallel architecture and 64 processor boxes will be common place in 5 years.