Amp Modeling -- HOW is it Really Done?

[Bezmotivnik]

I've wasted a good deal of time asking this question elsewhere, so I figure I might as well give it a shot here.

To save time, I'm not asking for your opinions about modelers.

I'm asking if anyone actually KNOWS the technical process by which Line 6, Johnson, Digitech and many others produce their so-called models of existing amps and cabinets.

As time goes on, I am becoming more and more skeptical of it being anything nearly as exotic or substantive as the ad copy would have us believe.

Thanks to anyone who can provide some hard data or links to articles.

 

[tone4days]

Re: Amp Modeling -- HOW is it Really Done?

well, i'll tell ya what i know .. i have no idea what your educational background is, so i'll just shoot straight and you can ask follow up questions

any electrical circuit, be it an amp, an effects pedal, etc is made up of fixed and variable components that react to various input signals ... the values of the fixed components can be measured and the range of values for the variable components can be measured when an input signal is present ... these values can be put into a computer program that models the way the components interact ... the output of the original circuit can be measured too and this info can be fed into the model as well ... if you consider the circuit as a black box that takes an input signal, then does 'something' to that signal to change it, then produces the that changed signal as the output signal, then the model from the computer program is the 'something' that gets done ... a technical term for this is the 'transform' function ... it transforms the input signal into the output signal that corresponds to having gone through the 'equivalent' or 'approximation' (model) of the actual circuit that is desired ... there is a similar set of steps for modeling the speakers/cabinets used to make the sound ... cabinets have a certain response that can be measured and recreated ...

now, it isnt as simple as that oversimplified description makes it sound ... first of all, is the quality of the components that turn the analog signal into a digital signal and then turn the ouput signal from digital back to analog ... you need enough details (bits and sampling rate) to get as close to the real thing as possible ... .then you need a computer model that is very sophisticated .. the more, the better .... instead of modeling the components by themselves, more information can be measured to include the interaction of the components among themselves to capture more realism ... the more sophisticated the model, the more realistic the results ... further, the output measurements can be made over a wider range of input signal types (e.g. broad band input vs single tone input) ... the more powerful models lead to a need for the most powerful processors and max memory available ...

some folks believe / experience that absolutely complete modelling of even one actual amp has not yet been done ... in fact, some believe that it will be impossible to do so because there are too many interworking parts of the 'system' to completely model all of it (you cant digitize mojo) ... some say that the very act of digitizing destroys the relationship between phase and amplitude of the signal that can never be reconstructed ...

and that's all i have to say about that

t4d

 

[millsart]

Re: Amp Modeling -- HOW is it Really Done?

Thats certainly the problem with "mojo". I don't think anyone can really define what it is word for word, or exactly how you could create it in a lab, but we know it when we hear it don't we.


I sure would not want the task of trying to capture it and burn it into a microchip.

 

[Bezmotivnik]

Re: Amp Modeling -- HOW is it Really Done?

any electrical circuit, be it an amp, an effects pedal, etc is made up of fixed and variable components that react to various input signals ... the values of the fixed components can be measured and the range of values for the variable components can be measured when an input signal is present ... these values can be put into a computer program that models the way the components interact ...

I understand this, however what I'm asking is if this is indeed what is truly being done in the actual creation of the models in the amp modelers we buy.

I doubt it, if for no other reason that the results don't sound as though that is the case and I have never seen anything from the horse's mouth that indicates it is. Everything is "secret" and proprietary. When I hear that, I cannot help but think of the episode of The Simpsons in which we learn that the "secret sauce" at Crusty Burger is just mayonnaise left out in the sun.

What these modelers are sounding like to me is nothing more than a generic model with different EQ profiles and base gain settings added. In my experience, they do not behave like real amps in terms of the subtle , gradual changes that actual sophisticated circuit modeling of the sort you describe would be expected to produce. They just sound like someone sat and listened to a recording and tweaked digital mixer settings until the base amp somewhat resembled the sound of the recorded real amp.

Again, I don't know any of this for sure, I'm just (in the absence of first-hand authoritative information) going by what I hear.

Thanks for all ongoing interest and data. I've been wondering about this for a long time.

 

[david kornblatt]

Re: Amp Modeling -- HOW is it Really Done?

I'm of the opinion its a matter of time that the tube and transistor amp will be a thing of the past

 

[baritone]

Re: Amp Modeling -- HOW is it Really Done?

What about just an equalizer?

 

[FAC]

Re: Amp Modeling -- HOW is it Really Done?

When modeling natural phenomena in science, we compare the output of our computer models to the original data. It is posible to quantify the misfit, or error, between the model and the real thing, and give a "percent error" (this allows us to say things like "our model is 75% accurate at reproducing the data, etc...). The error function tells us how good we're doing at modeling the original data--or if you're a glass half empty kind of person, how badly we're modeling the data.

I wonder what the error functions look like for amp modelers...?? I wonder what kind of % error they consider "good enough" to sell to you?

 

[Gr8Scott]

Re: Amp Modeling -- HOW is it Really Done?

I figger (guess) that they put the amp of choice on a good dummy load like a weber mass etc and then proceed to run input into the amp and capture the sampled output. Do this with enough different input tones at different volumes and different amp settings and you'll have a map of sorts of what the amp does at a given setting, frequency and volume. You'll be calculating the "error" as FAC says between the input and output of the amp and building a formula for it using the data generated by the samples.

 

[PFDarkside]

Re: Amp Modeling -- HOW is it Really Done?

The other thing to consider is, the transfer function of tubes, especially in the way that guitar amps use them, is a pretty complicated, non-linear function. As processing power and memory becomes cheaper, and more time is spent on newer revisions of the models, they will invariably become more accurate. However, my fellings on the results are similar to yours, a few base models with extensive EQ. Not that they sound bad, we used one pretty successfully in my band, but comparing them to a tube amp still leaves something to be desired. The interesting thing is that on a recording they are sounding pretty accurate, but it's a feel thing where they come up short.

 

[Top Jimmy]

Re: Amp Modeling -- HOW is it Really Done?

I just wish they'd stop "modeling" and just make a hybrid with an attenuator that sounds real good and has it's own sound and vibe instead of trying to copy everyone else or at least just try and sound like one really good amp. Kinda like Tech 21 does.

Just a clean, crunch and lead hybrid, just reverb (no other crappy effects), an attenuator, and that's about all I want.

 

[LoFiBrian]

Re: Amp Modeling -- HOW is it Really Done?

I saw this youtube video where the plugin company waves and paul reed smith model some of his favorite botique amps...


doesnt show you much but I thought the final product sounded like poo.

 

[metalspider]

Re: Amp Modeling -- HOW is it Really Done?

ive got a modeler for over 2 years now,does it sound like the amps it pretends to?nope.
but i can get some nice tones out if it and its great for recording direct at home,so i like it.
i just dont expect it to sound like a real tube amp,which is a shame but hey i have a tube amp so this is just another kind of tone to play with.

 

[Robert Delahunt]

Re: Amp Modeling -- HOW is it Really Done?

I'd think the way it's done, at least with the Boss GT-8, is that you have COSM engines (their marketing term for some kind of processing unit, like maybe a Via Eden) that load a firmware or software modeling program. I've heard some that try to throw in special clauses and conditions like speaker breakup. This is how I think it is done at the hardware level. I doubt any of the stuff on the market has the processing power to try to simulate a virtual environment, so it's probably a combination of how the model reacts to the guitar and how it's EQ'd.

 

[Bezmotivnik]

Re: Amp Modeling -- HOW is it Really Done?

I'd think the way it's done, at least with the Boss GT-8, is that you have COSM engines...

I think (and again, I want to avoid useless speculation and opinions here, including my own) that once a model was determined in the Big Guitar Science Lab, the processing resources to reproduce that one profile as a preset would be manageable in a retail hardware unit.

I'm still looking for a comprehensive explanation from one of the producers of these devices of exactly what is modeled and more or less how.

After these many years, I'm sure that the competition is fully aware of what each other are doing in these processes and the "secrets" are only being kept from the buyers; truth is so precious it shouldn't be wasted on mere consumers. I find it quite suspicious that none of them actually say, and their language is very vague. I especially like the fine-print disclaimer on the M-Audio/Linn "Black Box" that says the models don't sound like the amps they've spent two or three pages listing and puffing, but were inspired by those amps!

What kind of weasel-worded evasion is that? What technical modeling process is "inspiration"? :?:

 

[Gamera]

Re: Amp Modeling -- HOW is it Really Done?

I was always under the impression that the models were built around a sampled sound from an amp. Kind of in the same way a keyboard can grab a sound sample, digitize it, and play it back in any key and played like a keyboard. Like the scene in Ferris Beuller's Day Off when he makes his keyboard cough.

If that's the case, it's only analyzing the end product and not the subtle intricacies that went into producing it and how those subtleties can be mastered and interpreted by the player.

 

[Gamera]

Re: Amp Modeling -- HOW is it Really Done?

Although come to think of it, if the above is the case, one could argue that a true artist could learn a new set of subtleties offered by the new medium and master those.