How does amp modelling work? (technical overview)

Discussion in 'Recording' started by altima_boy_2001, Sep 24, 2006.

  1. I was wondering how digital amp modelling works, ie. what is the process of creating POD/amplitube/guitar rig? Could anyone give me a technical overview of the subject?
    Google wasn't giving me any useful information...It doesn't need to be detailed, just need
    to know where to start.

    I guess I'd like to know more about what DSP topics I should be researching or
    any resources I should be looking at.

    For example, can I assume that if someone wants to model a Marshall amp that they:
    1. buy a marshall amp

    2. play sound through it at varying levels and frequencies and record the output
    --what sort of inputs are used? Function generators, guys playing guitars, etc...

    3. compare the input to the output
    --time domain? frequency domain? levels?

    4. create mathematical functions or approximations to convert input to output and
    program the HW/SW.

    (5. sell the results and make a bunch of money)


    do people take hardware schematics and create programmed equivalents of the capacitors/inductors/resistors/transistors/etc using the same layout in software?

    If anybody knows something I'd like to hear about it...
  2. anxious

    anxious Guest

    Per your question, more the first approach than the second. Commonly, the target system is broken down conceptually into distinct blocks in the signal path, each chosen to isolate one of the sonically important signal processing elements of complete system. These blocks are then modelled individually in whatever way is most appropriate and convenient for the function they perform. For example, a speaker might be "physically" modelled using a bunch of 2nd-Order DIF-EQ's and lumped mechanical parameters. A tube preamp stage might be modelled as a group of BIQUAD transfer functions that are parametrically selected in accordance with input signal level. A power amp might use a series of interconnecting sub-blocks, of both linear and non-linear type. Effects boxes are modelled with a non-linear transform and some feedforward and feedback, etc. Connect 'em together, add appropriate routing and controls and hope for the best...

    There have been a number of AES papers and US Patents filed with inventions in this area. I'll try to get you started, but I don't know what your expertise level is.


    I would personally say that you need to get some hardcore physical acoustics systems in, especially non-linear behavior, before hitting the DSP.
  3. Thanks for the info...sometimes looking for the place to start is the biggest obstacle...

    I have degrees in computer science (major), mechanical engineering (major), and math (minor) so I'm familiar with system dynamics (mainly spring-mass-damper) and associated DIF-EQ topics.
  4. anxious

    anxious Guest


    That's a good background. What you will need to take on next is, I think, non-linear systems dynamics. That's where the real progress is being made these days. There are some very smart people who have been investigating amp modelling most of their careers, and are still just scratching the surface, so don't get discouraged if a few months of study doesn't get you very far.

    For "just playing around" purposes, you can throw the usual battery of analytic signals at the amp: sines, chirps, impulses, noise, MLS, etc. Each will tend to give you better information about different things. Arguably, sine protocols are the only method that will yield the non-linear terms you want, but you need the others to understand time-domain behavior. You also need to figure out how and where you are going to measure your acoustic signals, in order to best represent the listener experience.

    If you are serious about pursuing this more deeply, you will need to understand more complete modern system identification techniques for non-linear processes. Basic analytic signals and inverse transforms don't get you very far, because the transforms are not deterministically reversible (as they would be in an LTI world.) The work of Wolfgang Klippel would be a reasonable place to start.

    Make sense? (This is a professional area of interest, so I am more than happy to respond to the occasional question as you progress, if you want.)
  5. mpd

    mpd Guest

    I can't comment on non-linear modelling, but the two standard texts on linear system modeling from a mathematical standpoint are are Kay's Modern Spectral Estimation, and Marple's spectral estimation book (title escapes me at the moment). Kay's book has been out of print for a while, as has Marple's first edition, but I think there is a new edition of the Marple book.

    The basics are rather simple. Pick or estimate a model. Run white noise through, measure, plug into algorithm, and get the model parameters (eg, coeficients for a biquad).

    The math behind it is rather complex, but the algorithms are available or can be implemented without too much work.
  6. This sounds like some good information. Looks like it's time to start reading. Thanks.
  7. Mike Papaleontiou

    Mike Papaleontiou Active Member

    my name is Mike Papaleontiou and I am studying music technology at UEL (University of East London)
    my graduation project is about amps and amp models.
    I will try to prove that amp simulation can work as well as the real amps would in a music production
    Could you please explain to me how exactly amp modeling/ sim works?
    Is there anywhere i can refer to, appart from the links above?
    Could you suggest amp models I should check?
    Any additional info would be appreciated!

    Thanks in advance
    bigtree likes this.
  8. audiokid

    audiokid Chris Well-Known Member

    BC, Canada
    Home Page:
    Mike Papaleontiou likes this.
  9. Mike Papaleontiou

    Mike Papaleontiou Active Member

    Thans audiokid, but I am looking for a more sientific approach from some more "reliable" sources than wikipedia. Unfortunately at this wiki's article there are not any references.
  10. DonnyThompson

    DonnyThompson Distinguished Member

    Akron/Cleveland, OH
    Home Page:
    You migh twant to read this... not sure it will help, but it's a White Paper released from M.I.T. from research done at both Stanford and Helsinki University:

    Link removed
    Mike Papaleontiou likes this.
  11. Mike Papaleontiou

    Mike Papaleontiou Active Member

    my graduation project is about amps and amp models. I will try to prove that amp simulation can work as well as the real amps would in a music production. Could you please send me the feared mix competition (2013) tracks so i can compare Randall Satan re-amped to the TH2 Randall simulation and to simulation Ola Englund’s inspired amp simulation ?
  12. Mike Papaleontiou

    Mike Papaleontiou Active Member

    Thanks Donny I think this will help a lot!

Share This Page