Guitar amp modeling.

[mohogany]

Hello team,

I'm fairly new to the world of DSP. I know a good amount of theory but have no practical experience. I've always been passionate about guitar amps and any type of distortion sound. I was wondering if someone could guide me on how I could create my own guitar amp models, similar to what Line 6, Peavy Revalver, IK Amplitude and many more companies are doing. My goal is to eventually become a specialist with distortion shaping.

Thanks.

[Caco]

A good place to start is simulanalog who have some interesting articles on their website.

Also, the open source MDA plugins have an amp sim in them I think so that may help you as well

[robrokken]

Get ready for lots of testing. None of the VST plugins can get that real tube sound, and you will spend ur days obsessing over it All I can say is "transfer function". And my advice is not to try to emulate the hardware, but emulate the effects of the hardware.

Like, If a certain tube distorts more at some volume, don't try to emulate the tube with code, Use an envelope follower hooked to the gain of a waveshaper or something

Don't try to emulate the hardware, emulate the sound imo.

Good Luck!

~Rob.

[bmoviehorror]

I stumbled across David Yeh's thesis a little while back when looking for info on am[p modelling. You might find some useful stuff in there

https://ccrma.stanford.edu/~dtyeh/paper ... esided.pdf

...probably not the best starting point though...

[mohogany]

Thank you for all the great responses. Out of curiosity, do we have any guitar amp modeling fanatics in this forum?

[tjl100]

any guitar amp modeling fanatics in this forum?[/quote]

Um yes but I'm trying to quit, it's one of ,if not the hardest sound emulations, just try everything -slayer to the newest 64bit stuff, and slowly
go insane!!!!!!

[robrokken]

Oh, they are here I'm trying not to watch this thread lol, cause i know i'm gonna start going sleepless nights, trying to find the key. Like I said though, don't go tryin to replicate the input circuit, resistors, blah blah, Listen to the sound.

You have to read such things as Crossover distortion, Tube bias(and how it changes), power sag, transformer saturation, blow-by effect, compression.

One of your greatest center points imo is going to be the envelope follower, on that you can base quite a few things that are level(or power) dependent.

If you do get a good saturation model, that actually follows some sort of transfer curve, you then need to think about FM effect. Since you will have 1 impulse, you need to vary the high end and low end, to achieve realistic results. Again, think of a way to achieve that with what you have, (ex. 2 band eq that has the gain params linked to the follower for the high and low.

Eh, see! It's happening lol, I'm getting sucked back in!

~Rob.

[mohogany]

LOL. Didn't mean to trigger the obsession. The good news is that talking about it is therapy. My name is Mohogany and I'm a tone addict...

[onqel]

I'm doing guitar amp modeling.. my latest plugin is the TSE X30, it's done by emulating a schematic with the "resistor capacitor" approach robrokken is talking about.. I spent 6 months finding my solution

[tradivoro]

You may want to try the guitar amp modeling forum, at

http://www.guitarampmodeling.com/

There you can see what a lot of independents have been up to, just guys like you... Some of those people are doing great work...

[valhallasound]

I've done some in the past. My suggestion is to begin your work by modeling the various filters found in an amp: coupling capacitors, tone controls, speaker cabinet models. A lot of commercially successful emulators from a few years back got a lot of mileage from having filtering that sounded correct, even if the nonlinearities used were TOTALLY different from a tube amp (i.e. non-bandlimited hard clipping). Same goes for the gain staging - figure out what stages produce what amount of gain.

Another piece of advice: Taylor series polynomials only work within a given input range. You can "bound" the range by using hard clipping on the inputs of your polynomial, but once your input exceeds the clipping range, the harmonics predicted by the Taylor series go out the window.

One more thing: Your typical clean guitar signal has almost no energy above 10 KHz. Something to bear in mind when selecting your oversampling ratio. Test your algorithms with a guitar, not with random test signals.

Sean Costello

[mohogany]

All this is priceless information. Thank you all once again.

[duncanparsons]

I stumbled across David Yeh's thesis a little while back when looking for info on amp modelling. You might find some useful stuff in there

https://ccrma.stanford.edu/~dtyeh/paper ... esided.pdf

...probably not the best starting point though...

Don't think I've read that one before... thx for the link

[mohogany]

I stumbled across David Yeh's thesis a little while back when looking for info on am[p modelling. You might find some useful stuff in there

https://ccrma.stanford.edu/~dtyeh/paper ... esided.pdf

...probably not the best starting point though...

Got a chance to read David Yeh's thesis. Definitely not the best starting point... LOL Regardless, pretty impressive work. I need to hit the books and improve my math skills. Many new mathematical techniques (or old forgotten ones).

The rest of Yeh's papers can be found over here:
https://ccrma.stanford.edu/~dtyeh/papers/pubs.html

[robrokken]

Well, damnit, I tried to fight it, but I couldn't, heres the amp section of a new one I am messin with, still cannot figure out convolution for impulses though

http://stashbox.org/805276/guitartube.mp3
http://stashbox.org/805304/guitartube2.mp3

Close to a tube? Not even, but its getting the sound i want


~Rob.