A Question on Direct Profiles and Sampling Rates

  • So recently I stumbled on this forum post from another forum. Its from an amp sim developer showing some interesting things about how the Kemper is working. It appears that it's profiling at 22.05kHz unless you are capturing a direct profile, in which case its doing it at 44.1kHz.

    its quite clearly no issue and produces a satisfactory result, hence the success of the profiler. It is interesting though, that its capturing a more accurate profile (in that its capturing above 11kHz) when done direct.

    It would be great to hear why this was chosen, and whether its something we'll see upgraded in the future - even if a lot of content that high is filtered out, it would be reassuring to know that it's being profiled and sampled accurately.

  • did you guys even look at the post I linked from gearslutz?


    The spectrogram plot shows a clear difference between the different capture methods. One behaves like a guitar amp does above 11k, and the other does some strange frequency mirroring that can only happen in the digital realm and is not representative of what we hear in the real world.


    stratdude - do you actually think that a guitar speaker can’t produce signals above 11kHz?


    A guitar speaker does roll off gradually at frequencies in that range (actually rolls from about 6kHz), but there is still harmonic content up there - that’s why we can hear a difference if we add a LPF to a guitar cab once it’s recorded. It’s not like a cab doesn’t produce frequencies in that range, it’s just that they are quieter than the lower range.

  • did you guys even look at the post I linked from gearslutz?

    Yeah, nah, sort of, not really, short attention span.


    I recall reading that a profile isn't so much created from scratch, but an alteration of a default amp model (Crunch?) already in the Kemper. It would make sense that there would have to be a different default model for D.I. profiles with all the additional high frequencies that none of us want to hear.

    I guess it would also make sense if the profiling process is also different when creating a D.I. profile.


    Either way, I just like to think of it as witchcraft, I can live with that.

  • I read it, but, I don’t think looking at a waveform is a direct indication of how the software that created it functions.

    Just looking of course tells you very little, but if you do the test for yourself - you can hear the difference in the top end being different, with the direct being more accurate to a real amp. The analysis of the spectrogram just helps to explain why the sound is different (and somewhat unnatural compared to the real thing).

    It does quite clearly show that the signal above 11kHz is being mirrored back rather than capturing the information in that region, which points to the sample rate being halved.

    I can of course live with it (I bought a Kemper based on how it sounds which Ive always been very happy with), but knowing this information is helpful for me in that I'll be making direct profiles more often now to get the most out of my unit.

    I'm assuming I'm not the only one who cares about this kind of stuff, but I also appreciate that for many its nothing to worry about. Its interesting enough to share the information and discuss IMO.

  • So recently I stumbled on this forum post from another forum. Its from an amp sim developer showing some interesting things about how the Kemper is working. It appears that it's profiling at 22.05kHz unless you are capturing a direct profile, in which case its doing it at 44.1kHz.

    Nope, sorry! Don't know what's wrong with the analysis you mentioned, but attached you can find what I get out of my precious Profiler using the stock studio profile MB - /13 JRT915 84 3 with the effects block switched off. No weird mirroring found.

  • Ooh that is interesting - I'd definitely like to know what's going on there as something clearly doesn't check out (I pm'd the person who made that post on GS and told him about this thread).

    I just did a really quick comparison of running a DI through various chains:

    Marshall JMP->Mesa V30->57->Chandler TG2->UA Apollo
    Marshall JMP->React IR (on through so the cab load is consistent for all)-> UA Apollo

    Kemper Direct Profile as captured from the React IR (but with the Mesa cab as the load for the amplifier)
    Kemper Studio Profile of the first chain
    Kemper Studio Profile of the first chain but with the cab disabled

    I also made an impulse response of the mic position by deconvolving the amp DI against the mic'd signal, and included prints of any direct amp sound through the same impulse.

    I've included the Kemper profiles used, as well as the impulse response of the Mesa cab, and the DI too.

    They can be downloaded here: https://www.dropbox.com/s/rt0m…rq/Kemper%20Test.zip?dl=0

    Photo of Amp+Cab Setup
    Photo of Rack Gear Used

    I've never done any spectral analysis before but I do have the standard version of RX7 so Ill see what I can do.

  • OK, so it looks like both the direct and studio profiles show that mirroring happening here (which isn't happening on the direct DI from the amp). These are all signals with no cabinet or IR's - I can upload screenshots of the cabinet processed files too - they are bit harder for me to gauge anything from though

  • OK, so it looks like both the direct and studio profiles show that mirroring happening here (which isn't happening on the direct DI from the amp).

    Thanks for your in-depth analysis and for sharing the results -- now things start to get interesting. I've analysed the audio samples you've provided, and can actually confirm your results (see attached plot)!


    Apparently the main difference between our setups is the sampling rate of the interface. I recorded via S/[email protected] while your recording appears to be Analogue@48kHz. So my assumption: The artefacts are some sort of aliasing effect caused by the different sampling rates of the Profiler and the interface. In short: It's not the fault of the Profiler!


    If you want to dig deeper into it and (dis)prove my assumption: Repeat the analysis with your interface and DAW set to 44.1kHz. Reamping with just the direct profile should do.

    EDIT: And make sure to use a DI recorded @44.1kHz so no resampling on both ends.


    That was fun. Cheers-

  • I actually recorded the Kemper via spdif - it seemed a fairer test to keep the ad/da consistent (and less conversions) and as everything would be used in normal circumstances. I’ve realised the way I worded the chains wasn’t the clearest, but my kemper is always hooked up by spdif (I also used it as a reamp device to keep gainstaging consistent).


    I’m curious to try some more tests and work out what’s going on - I didn’t have any FX in use, but maybe I need to disengage the entire block? I could also record the analogue and digital outs simultaneously.


    and I can do a test at 44.1kHz - it would be very strange if working at higher sample rates caused the kemper to work at lower ones....


    and FWIW, I think all the methods included in the zip do a reasonable job at capturing what they’re supposed to. Very important to understand the gear and know what they’re strongest and weakest links are (and how to get the best useage out of them).

  • I can do a test at 44.1kHz - it would be very strange if working at higher sample rates caused the kemper to work at lower ones....

    My understanding: The Profiler always works @44.1kHz. If you opt for a different S/PDIF rate it does the usual thing: (higher-oder) upsampling, then downsampling. The artefacts you get are probably just the result of the imperfections of the (every) resampling algorithm.

  • Redid some of the tests at 41kHz, and also recorded the analog output from the Kemper as well as SPDIF. There doesn't appear to be the obvious mirror line at 11.025kHz as before, but there is a definite change in the signal vs the direct amp signal above there - check out the sections where its black/blue.

  • Thanks for your confirmation. No hidden mirrors in KPA profiles.

    Regarding the difference in noise floor: Whatever it is, it's about 80dB below the hottest frequencies.


    EDIT: Apparently the cause is your Test DI (see attached plot). The SNR of the Profiler out just seems to be superior to your amp signal chain.

  • makes sense with the noise floor, profiles are always more quiet than a loud amp rig. so I guess there is an argument that if working at sample rates higher than 44.1, it may be preferable to use the analogue outputs (except for recording a DI for instance) - it avoids some of the internal resampling weirdness