Why does the "refine" step exist?

  • This is purely a technical question.


    Profiling works by the Kemper feeding a signal into an amp and listening to the responses to determine how to set the parameters of an internal model. I always presumed that the Kemper sweeps through the entire range of frequencies that guitars can produce and multiple volume levels to get a complete picture of how the amp will respond to various inputs.


    So why would subsequently feeding it a random guitar signal -- which should be less comprehensive coverage of amplitudes/frequencies than the original profiling signal -- make the profile better?


    Is it that the original profiling signal is less comprehensive than I imagined, and the guitar is filling in the blanks? Because that would indicate that a refined profile is only ever going to be truly accurate with that specific guitar. Is that the case?


    Why can't the original profiling signal be more comprehensive, so that it covers every possible guitar signal?


    I'm willing to hear people's speculation, but I'd be especially interested to hear if Kemper himself has every commented on this particular question.

  • Not sure, just my impression:


    The Kemper's UFO noises will give it a fairly accurate profile from a technical perspective, but may not capture everything going on in a device that's designed to operate musically. (read: harmonics and chords and intermodulation and stuff). Different amps and distortion circuits will react to chords and palm mutes differently, and bursts of almost-static don't necessarily stimulate the amp in the same way.

  • Grinch, here is a long interview with C.Kemper talking about profiling that might be helpful.
    http://www.guitar-muse.com/kemper-profiling-amp-2949-2949
    The Kemper patent info
    http://worldwide.espacenet.com…?CC=GB&NR=347206&KC=&FT=E
    It doesn't matter what guitar or pickups is used for refining, and sometimes results can be spot on without any refining.
    As far as I know Kemper have never fully explained how profiling works, in order to protect their technology.
    Where did you hear about any internal amp models? The first time I saw that mentioned was by Cliff Chase on his fractal forum (then others started that on TGP), saying the kemper has 9-11 or so amp models. He has a long post history of making up his own kemper "facts", lying about kemper and is proven wrong many times. He's always slinging mud at all competition gear so it's a bit funny and sad at the same time.

  • Quote from Lokasenna

    Different amps and distortion circuits will react to chords and palm mutes differently, and bursts of almost-static don't necessarily stimulate the amp in the same way.


    Yeah, I guess the question is why the Kemper doesn't itself produce noises that "stimulate the amp in the same way". I mean, if it doesn't matter what guitar is used, it could have built-in guitar vamping that it uses to profile. I just don't understand why a human player has to be involved at all. The device knows better than we what it's looking for to configure its model.


    ---------------------------------------------------------------------------------


    Quote from HappyKemper

    Where did you hear about any internal amp models?


    All digital amps use internal models. That's just the computer science term for a software simulation/representation of a real world thing/process.


    Chris talks about the developing models in the interview you linked to:


    Quote from Chris Kemper

    simply too many interdependent parameters, and it would have taken ages to model just one or two dozen amps. As a basically lazy person I spent my time trying to find an automated method, rather than modeling amps by hand.


    He defines "model" as virtual version of an amp:


    Quote from Chris Kemper

    By philosophy, “modeling” was used as a marketing term by some companies. It says: “Here is a valid virtual copy of a valuable original”.


    He then defines "profiling" as (1) process of creating that virtual copy automatically and (2) the ability to then A/B it against the original:


    Quote from Chris Kemper

    Profiling is an automated approach for reaching a result that is probably too complex and multidimensional to achieve by ear. [..] Profiling, in our sense of the word, is a promise to create a virtual version of your original, but with the ability to qualify the results by a fair A/B comparison. You get what you want, and you can check what you have just got.


    It doesn't necessarily make sense to talk about "models", plural. You can think of the code that simulates amps as the software model of an amp which is then configured by the profiling process. The different amp "models" in something like an AxeFX can be thought of in the same way.


    As for Cliff's opinion about how Kemper works, I just googled the name and see that he's the creator of AxeFX. So I'd say there's a 99.99% chance that his guess is better than yours or mine, because he's smarter than you or I and knows more about simulating guitar amp than most people on Earth.

  • quote:
    "As for Cliff's opinion about how Kemper works, I just googled the name and see that he's the creator of AxeFX. So I'd say there's a 99.99% chance that his guess is better than yours or mine, because he's smarter than you or I and knows more about simulating guitar amp than most people on Earth. "


    I noticed you also read the thread "What do you guys think of this from Cliff of Axe Fx?".
    https://www.kemper-amps.com/fo…his-from-Cliff-of-Axe-Fx/?
    There you see he frequently makes up lies to slam down on competing products, using silly FUD tactics trying to get users to doubt other gear. His claims were proven wrong and showed he wasn't so good at calculating. As consumers we can never be gullible and trust a salesman who completely ignores business ethics and frequently makes up stuff and lies about other gear so he can get advantage.

  • Not long ago, it has been said, that the refining is no longe crucial. Just search for refining and you will find that statement. I guess it was announced with the introduction of direct profiles.

  • You can think of the code that simulates amps as the software model of an amp which is then configured by the profiling process.


    Interesting way of putting it. I like it.


    I think in some of the later firmware versions, the refining process is usually redundant. But I don't have an explanation as to why it is necessary in the first place, other than the UFO noises not being sufficient to fully get a 100% accurate reading of all the parameters in all cases, at least originally. This may be due to some amps reacting in special ways - unpredictably.

  • One more thought: The manual states that during the refining process it is recommended that you play complex chords. This could mean that the refining process checks how the amp reacts to harmonic content. But this is mere speculation.

  • I always presumed that the Kemper sweeps through the entire range of frequencies that guitars can produce and multiple volume levels to get a complete picture of how the amp will respond to various inputs.


    this method is also known as dynamic convolution, but this is not how the Profiler works. ;)


    the Refining process makes a lot of sense in you think about it like this:
    due to the many variables involved a profile might sometimes turn out to be 'not quite there yet'. instead of having to profile the same setting again and again, the refining option gives the user the ability to give the Profiler more time and information about the setup, for as long as it's needed (you decide).


    it is true however that Refining isn't mandatory anymore, very often the profile is spot on right away.


    hth

  • Ripper. Thank you for that, Don.


    So, Kemperites, 2 posts in by the OP and we're into Axe territory. What's my prize for seeing this one before it even came over the hill? Do I get a free banana, or better still, a whole bunch?


    Perhaps it's just an unfortunate coincidence in this case, Grinch, but it's a statistical anomaly that it's rare in these parts that the Helix and other modellers aren't brought up in the context of picking the Kemper apart somehow, and that this appears to be the exclusive domain of Fractal. If anything, in the case of those other modellers, it's almost always a questioning of whether or not a particular product is nearing the Kemper's standard or not. Not so with Fractal gear. No, instead it's an inverted paradigm, one where the latter is somehow superior and various aspects of the Kemper are singled out for ridicule, for want of a better term. Think DA sample rate, IR length and CPU grunt and you're in the ballpark.


    Mini-rant over. I thought I smelled another rant, but it may have been my irantination.


  • All digital amps use internal models. That's just the computer science term for a software simulation/representation of a real world thing/process.


    Basically you're saying the Kemper has a base amp and all the other amps are derived from it.


    Do you own one? It's a bit different from what you think it is.



    As for Cliff's opinion about how Kemper works, I just googled the name and see that he's the creator of AxeFX. So I'd say there's a 99.99% chance that his guess is better than yours or mine, because he's smarter than you or I and knows more about simulating guitar amp than most people on Earth.


    The question is, is Cliff smarter than Mr CK?

  • LOL


    No harm intended, Grinchster; just a bit 'o fun, mate:


    Apple's OS10.9 built-in dictionary:
    Grinch
    noun N. Amer. informal
    a spoilsport or killjoy.


    Addendum: New to the font-size thingy. Don't know how many times I had to edit, but it was ridiculous - the sizes wouldn't stick, or spilled over to text they shouldn't have. Oh well.

  • Basically you're saying the Kemper has a base amp and all the other amps are derived from it.


    Do you own one? It's a bit different from what you think it is.



    That's not what he's saying :)


    Think of it this way: The kemper has a mathematical model which is then "populated" with parameters during the profiling process. This is not the same as saying there is a base amp all other profiles are built from.



    However, if the kemper contains several mathematical models (or parts of a model) of which ONE is chosen based on the results of the profiling process, and then populated with parameters, I guess that could be thought of as "base amps". But I doubt such models themselves would qualify as a profile in its own right.

  • I think I've read somewhere an interview with CK that the KPA have basically 2 algorithms, clean and overdriven (solid state vs. tube?), and that kind of make sense since You probably want to have different response to the gain structure and the type of harmonics generated by the algorithms.
    And the analysis of the amps response populates these algorithms with the parameters that tailor them to behave like the original amp.
    But I might have gotten it totally wrong... and I'm ok with that. :thumbup:

  • The speculum-titty-lactation ebbs and flows
    'Though none of us really ever knows
    The ins and outs of the genius' code
    That was borne out of CK's lazy mode*


    *CK has said many times that it was his laziness that motivated him to devise a simpler, more-convenient method for modelling valve amplifiers.



  • Actually, I think that is what he was driving at, i.e. there's no need for the plural "models". And he equates this as being similar to the AxeFX.


    Now there's obviously an algorithm of some kind that is involved in profiling. So a common formula.


    Now with something like an AxeFX, there is a "base" model, or rather multiple bass models, which can be modified to create various tones. This is understood, that is how they keep adding new amps. Through the inclusion of a new model. A new formula.


    With the Kemper, there is no base model per se. The amps can be infinite, whereas the AxeFX or another device's may seem limitless by virtue of tweaking options, but are indeed finite after a fashion.


    It does not seem logical to me to equate that as being the same thing that's going on with a GT-8 or a Line 6 or an AxeFX.


    I understand the semantics that were involved in describing a "base model". I just don't agree with the parallels that were drawn thereafter.

  • Could be; could well be, Mrs Zambesi. ;)


    @nightlight, I agree with you, AJ.


    Another way of putting it might be that conventional modellers "quantise" the possible parameter palette; we can only specify those variables that we conceive, after all, meaning that the spread of any given modeller's variables' set is by definition always going to be a subset of the full-spectrum, real-life original.


    As we refine our understanding of tube, circuit and acoustic behaviour, more variables (parameters) are added to the formulas used for conventional modelling, bringing us ever closer to 100% accuracy. This way of looking at it also tells us that technically we'd never quite make it to 100%. Instead, we'd chip away at it as more parameters are conceived and the accuracies of existing ones refined.


    In the case of the Kemper, I can't help but feel that the base model, code, or whatever you'd like to call it, allows the capture of that "full-spectrum" source in a less restricted manner, sans or with much less of this "parameter quantisation". In effect, it's therefore more able to capture the essence of the source due to the fact that it's not reducing it (the source) to as utilitarian a set of values.


    Well, as I suggested, that might be another way of putting it.

  • Quote from nightlight

    Basically you're saying the Kemper has a base amp and all the other amps are derived from it.


    No, I'm saying that all digital amplifiers use a computer model of an amplifier, or of some black box that transforms an input signal to an output signal. That's just what the term "model" means in computer science.


    Whether Kemper has 1 or 100 internal models is irrelevant. It's a digital model of a physical process that transforms a signal in a particular way.


    If you understand this, it's clear that this is what Chris is talking about in the linked interview. Rather than creating/parameterizing models by hand, he came up with a process of finding the parameters for a model programmatically by analyzing how an amplifier affects an input signal.


    I have no idea how this relates to AxeFx, and am not talking about nor am I interested in any fanboy partisan drama about competing products. I'm just relating, as a software engineer, how such software necessarily works.