Any way to reduce quantization noise in the KPA?

  • My question is pretty much what it says on the tin. I think the KPA is the best piece of gear I've heard in regards to minimizing quantization noise/distortion and I've demoed most amp VSTs up to the new NeuralDSP and the Howard Benson Plugin, as well as my old Axe-Fx standard. So I'm actually applauding Kemper for having so little of it. But if it CAN be better, I wanna know how :)


    If you don't know what I mean by quantization noise, I mean a kind of bitcrushy sound in the high end of distorted profiles, kind of a "swoosh" that happens during the decay of a note, longer or shorter depending on how hard you play. Basically what the "Bit shaper" stomp does, but far more subtly. I recorded a little clip using an inferior VST amp sim (not the KPA) to demonstrate more clearly what it is if you don't know. Put on headphones and listen for the kind of sucking/swooping sound in the decay on these notes: SoundCloud


    In the Kemper, I've noticed the QN is made more apparent by turning the Tube Shape up, so I've started turning that down from 3.3 to 0 on all my patches. I think that makes some difference, at least it can't hurt. I just wish that knob went further down than 0 ^^ This makes me think: What other little parameters could be hidden in the Kemper that helps this?


    The only way I know of to combat quantization noise is dithering, but I can't really do that in realtime. But dithering is only used to mask existing QN, if one could keep it from appearing in the first place that would eliminate the need for dithering.

    So yeah, if you have any tricks similar to my Tube Shape discovery, please let me know!

    Edited once, last by JohnNT ().

  • Hi John. Welcome. You've identified one of the KPA's biggest tonal weaknesses, IMO. I've found that reducing the presence, definition, and gain by a few clicks can sometimes accomplish what you're after. Sometimes. If you're creating your own profiles, refining can sometimes help. Sometimes. Avoiding open-string, pedal tone-type riffs can also sometimes help. But only sometimes.


    Probably the best way to combat this is to find an appropriately-gained amp profile + IR combo that doesn't exhibit these type of artifacts based on the parts you're playing.


    I hope this helps. Please keep us posted if you come up with any solutions. I'll try your Tube Shape tip next time I encounter this.

  • Hi John. Welcome. You've identified one of the KPA's biggest tonal weaknesses, IMO. I've found that reducing the presence, definition, and gain by a few clicks can sometimes accomplish what you're after. Sometimes. If you're creating your own profiles, refining can sometimes help. Sometimes. Avoiding open-string, pedal tone-type riffs can also sometimes help. But only sometimes.


    Probably the best way to combat this is to find an appropriately-gained amp profile + IR combo that doesn't exhibit these type of artifacts based on the parts you're playing.


    I hope this helps. Please keep us posted if you come up with any solutions. I'll try your Tube Shape tip next time I encounter this.

    Thanks! I agree, this aspect of the Kemper is really the only thing that bothers me about it, but to it's credit it took me a few years to even notice it's there. When using the search function here I found some posts from ckemper himself where he seems to have a huge understanding about the subject, so I don't doubt the team has done/is doing everything they can to minimize the issue. But I couldn't find a thread specifically about reducing it on the KPA, so maybe this thread could help me find some answers on what you can do as a user :)

  • Given the ridiculously-high sample rate the algorithm for tube-amp simulation runs at (over 700kHz), I'm surprised to hear this.


    Quantisation noise at over 700kHz isn't going to be audible, leaving the conversion to 44.1kHz as a possible source, but even that's going to be super-high-frequency, well above where amp-cab rigs "operate".


    I would love to hear CK's thoughts on your observation, John.


    EDIT:

    In 7 years there's been no discussion about this that I'm aware of.


    CK mentioned it:

    Here 5 years ago, but not in a KPA context

    and

    Here 7 years ago, but again, not in a KPA context

  • Yeah I read those posts too, it's comforting to know that ckemper knows a lot about it, and also dislikes it when he hears it :D

    As far as high frequencies go, did you take a listen to the clip I posted? Like I said, that's not the Kemper, it's a sampled Electric Piano into the JST Toneforge plugin. That plugin produces a LOT of QN, so it demonstrates what I'm hearing in the KPA more clearly. I'm finding this in most digital distortion plugins, such as the highly regarded FabFilter Saturn too. I've kind of come to terms with it as one of those "This is where analog still beats digital" kinda things.


    I've narrowed down my profile collection to about 20 profiles now that after a bit of tweaking the gain, tube shape and clarity knobs have really, really negligible QN at least with the way I play. I find that on low gain sounds the clarity knob is useful, then you can compensate with the compression and sag parameters to get some "oomph" back.

    Edited once, last by JohnNT ().

  • I hear you mate.


    Have no means of hearing the QN on your clip (ZeroBass CrapMac™ speaker only ATM), but am super-familiar with the phenomenon, having worked with samplers since the '80s when it was par-for-the-course.


    Let's hope CK chimes in; I'm ultra-keen to hear what he has to say.

  • I hear you mate.


    Have no means of hearing the QN on your clip (ZeroBass CrapMac™ speaker only ATM), but am super-familiar with the phenomenon, having worked with samplers since the '80s when it was par-for-the-course.


    Let's hope CK chimes in; I'm ultra-keen to hear what he has to say.

    Cool :) I'm happy that my question is met with genuine interest, I was prepared for the worst :)

    I don't know much about this subject so it's fully possible that what I'm talking about isn't actually called Quantization Noise, in which case I apologize. Hence the clip I posted, if that is not actually QN, maybe someone who knows more can educate me so I get the terminology straight.

  • I want to chime in and say the artifacts I hear occasionally in the Kemper and in your clip, JohnNT , upon reflection, I would closer describe as aliasing. I may have spoken out of turn. I hope I have not misled you with my advice or confirmed in your mind something which may not be true.

  • I want to chime in and say the artifacts I hear occasionally in the Kemper and in your clip, JohnNT , upon reflection, I would closer describe as aliasing. I may have spoken out of turn. I hope I have not misled you with my advice or confirmed in your mind something which may not be true.

    Ah, that's interesting! Yes, it's definitely possible that I'm mislabeling this problem, calling it QN when it's actually aliasing that I'm hearing. I've heard of aliasing before but never understood what it is or what it sounds like.

  • Could it simply be the distortion of harmonics present in the source?


    Is it more-apparent when you put new strings on a guitar?


    If I'm right, this'd make sense. When you strike a string the harmonic-level envelope dies down more-quickly than the amplitude one, and I suspect you're hearing the former in distorted form.

  • Could it simply be the distortion of harmonics present in the source?


    Is it more-apparent when you put new strings on a guitar?


    If I'm right, this'd make sense. When you strike a string the harmonic-level envelope dies down more-quickly than the amplitude one, and I suspect you're hearing the former in distorted form.

    That's interesting! I have fairly old strings on all my guitars now embarrassingly enough, but I do have EMGs in some which retain a lot more high frequency content than the passives, and I can't say I hear a difference in regards to this particular issue.


    It does sound like you're onto something with the separate envelopes though, that's definitely how I'd describe it, it sounds like there's a layer of "something" tucked into the distortion that has a different envelope than the main signal. I hear this even on the noise of the guitar when I'm not playing, kind of like a modulation sweep going on.


    This obviously made me question my cables/electricity/interface etc., so to eliminate all of those variables I did a completely ITB test with my MacBook disconnected from its power supply, so entirely wireless. Running a DAW-generated sine wave into various distortion plugins, I noticed these artefacts all the same. So that made me come to the conclusion that it's not a problem with my setup, but rather a problem with digital distortions in general. I'd really prefer to be proven wrong though :)

  • The simplest way I can think of to look at it:


    If you pluck any stringed instrument, the harmonic content peaks shortly after or during the attack phase and dies down over time, the highest-frequency / lowest-energy ones being the first to disappear, right on down to the point that only the fundamental (essentially) is sounding.


    This applies to tine instruments as well, 'cause metal rods behave the same way harmonically, which would explain your hearing it with the EP you used as an example.

  • The simplest way I can think of to look at it:


    If you pluck any stringed instrument, the harmonic content peaks shortly after or during the attack phase and dies down over time, the highest-frequency / lowest-energy ones being the first to disappear, right on down to the point that only the fundamental (essentially) is sounding.


    This applies to tine instruments as well, 'cause metal rods behave the same way harmonically, which would explain your hearing it with the EP you used as an example.

    Yeah, I thought about that too, which is why I tried it with a sine wave later on. In this example, I used a sine wave: SoundCloud. There I also introduced white noise into the plugin, as a form of dithering, and it did indeed solve the issue but at the expense of getting really noisy instead.

  • It solved the issue 'cause there was no harmonic envelope.


    The "new" distortion envelope is caused by the amplitude decay of the sine sound.


    If you pass a non-decaying sine through you'll find there'll be no distortion envelope whatsoever.

  • If you can alleviate the issue by using tube shape, presence and other parameters on the Kemper, I don't see how it could be quantisation noise or aliasing.

    Interesting! I'm too uneducated about this to know, myself. I wouldn't say tube shape alleviates the issue, but rather it shortens the envelope of the problem, leaving the decay sounding smoother. More tube shape = longer "bitcrushing swoop". There was a post by ckemper here where he said "Did you know you can create QN on purpose in the KPA?". I'm assuming he was referring to the Bit Shaper stomp, so I'm playing around with that stomp now and it does seem to produce an exaggerated version of what I'm already hearing in the plain profiles. It's like I'm hearing a Bit Shaper but embedded in the far back of the sound, and Tube Shape really controls the envelope of it more so than reduces it. So I'm not sure I'm entirely off-base, but maybe you know more than I do about the science behind all this.

  • I suspect that what you are hearing in the tube shape parameters is a different distribution of higher order harmonics.


    My thinking is that quantisation noise is something that comes from AD/DA conversion, and I wouldn't expect internal parameters to affect this.


    But I'm no expert.

    If QN was produced at the AD conversion, that would mean that Bit Crushers couldn't exist, right? It should also mean that digitally distorting a synth generated within a DAW (i.e. no AD conversion) wouldn't produce QN. Yet it's been my experience that it does, as shown in the SoundCloud clips I posted. And the amount of artifacts you get basically depends on how "good" the distortion plugin is at emulating analog clipping as opposed to bit crushing.

    But yeah, I'm certainly no expert either so I'm speculating far and wide with all these posts.

  • I think you guys can both be right. Based on what CK said, the issue is that the low volume samples just get "quantizied out" to zero.


    The harmonics will definitely have lower amplitudes than the main signal so as things get quieter, the higher the harmonic, the earlier it disappears as the original note decays.


    Even though I don't think I heard it with the Kemper yet, this theoretically could sound like an analog synth getting its filter cutoff lowered - that removes the higher order harmonics too. This would also explain what Monkey_Man said, this does not work on a sine wave.


    Check out this video, shows it better than I can explain with a saw wave getting filtered. Naturally this effect would be MUCH less pronounced on the Kemper:


    External Content www.youtube.com
    Content embedded from external sources will not be displayed without your consent.
    Through the activation of external content, you agree that personal data may be transferred to third party platforms. We have provided more information on this in our privacy policy.


    Side-note: I want that synth!