extra latency using SPDIF

  • I had even higher latency using a Focusrite interface

    The important stuff is the "latency diff" and the "Net KPA latency" (see post #35 above).

    Audio interfaces and computers have different latencies depending on driver, buffer setting, sample rate, internal DSP, AD/DA hardware, etc.

    It's not possible to directly compare the total latency values (and Kemper can't do anything to improve your system latency anyway). :)

  • And I had a Focusrite interface on my last pc and firewire. I had no issue with latency.

    Again, do you refer to additional latency with spdif which you confirmed by measuring, or just generally a low perceived latency?


    We are discussing here 1-3 ms of additional latency with spdif compared to the analog out of the Kemper. This is absolutely not a deal-breaker and under normal circumstances cannot be perceived without measuring, just something that should (and probably can) be fixed.

  • The important stuff is the "latency diff" and the "Net KPA latency" (see post #35 above).

    Audio interfaces and computers have different latencies depending on driver, buffer setting, sample rate, internal DSP, AD/DA hardware, etc.

    It's not possible to directly compare the total latency values (and Kemper can't do anything to improve your system latency anyway). :)

    I had an even higher SPDIF latency difference compared to mains out :). I think it's clear what we are talking about in this thread.

  • Hi!


    Here's the story:

    Different AD converters, as well as different DA converters have different latencies. This is because internal filtering in the converters differs from type to type, and from manufacturer to manufacturer.

    Straight SPDIF does not need such filters.


    As a consequence it is impossible to match the latency of analog converters to SPDIF by default. Even if we compensated the latency for our outputs, we would not be able to anticipate the latency of the analog input of an arbitrary audio interface.


    For the variable sample rates in the Profiler we use a sample rate converter which is build in the DSP hardware, since 2018. This is again comparable to the filters of AD/DA converters. We have no access or control about its latency as well.


    In a future firmware we are planning to bypass the sample rate converter, when the sample rate is set to 44.1 and Profiler is Master. This would bring back the low latency of earlier firmwares, for that specific setting.



    Background story:

    The filters mentioned in the AD/DA converters are digital anti-aliasing filters. The latency is generated by the linear-phase design of these filters.

    To my knowledge, latency-free filters would have been very appropriate and would have allowed to time-align signals by only an error of a few samples, while bringing down the latency by a significant amount.

    But unfortunately audiophile cork sniffers have once critizised AD/DA manufactures for using non linear-phase filters and were afraid of phase distortions. As a consequence the manufacturers, serving the users, mostly use linear-phase filters since more than two decades, baked into their hardware.


    To my knowledge, such phase distortions could be seen on a scope or by other means, but cannot be perceived by ear.

    Same with 96 kHz ...

  • Thank you for the explanation but the question would be why is SPDIF signal delayed compared to Analog out? It doesn't pass through AD/DA so in theory it should be faster. I am asking because I love how it sounds over SPDIF and I'd like to use for monitoring when playing at home. Of course the latency over SPDIF is not a lot bigger than analog out.

  • Yes. I wondered why the digital signal is slower than the analogue?

    CK:

    "For the variable sample rates in the Profiler we use a sample rate converter which is build in the DSP hardware, since 2018. This is again comparable to the filters of AD/DA converters. We have no access or control about its latency as well."


    External hardware converters may be "faster" than sample rate converter. When CK implements 44.1 bypass, it will be lower latency on par with firmware from before variable sample rates were implemented. Only for 44.1 though.

  • CK:

    "For the variable sample rates in the Profiler we use a sample rate converter which is build in the DSP hardware, since 2018. This is again comparable to the filters of AD/DA converters. We have no access or control about its latency as well."


    External hardware converters may be "faster" than sample rate converter. When CK implements 44.1 bypass, it will be lower latency on par with firmware from before variable sample rates were implemented. Only for 44.1 though.

    Yes, I read that from the statement thanks. It seems like even the SPDIF bypassed signal will only match the analogue converted signal though.


    It is not a criticism, as I am happy using the Kemper like a real amp with XLR outs. I just expected like others that the straight digital path would be quicker.

    Karl


    Kemper Rack OS 9.0.5 - Mac OS X 12.6.7

  • In a future firmware we are planning to bypass the sample rate converter, when the sample rate is set to 44.1 and Profiler is Master. This would bring back the low latency of earlier firmwares, for that specific setting.

    This is interesting, as I remember the low latency of the early firmware and figured the change in latency had to do with an audio interface upgrade our studio did when we re-measured during lockdown, which had us scratching our head...

    Please keep us posted on when you will implement the old low-latency back into the 44.1 SPdif FW

  • Yes, I read that from the statement thanks. It seems like even the SPDIF bypassed signal will only match the analogue converted signal though.


    Not exactly.


    The bypass of the SRC will make the SPDIF a bit faster than analog, as it was reported in this thread. But that is less than a millisecond.

  • Less than a milli sec ? People will still complain :)

    (Joke !!!!)

    To be more precise: The difference (improvement if you will) to the analog outputs is less than a millisecond.


    Some could say that we could synchronize it to the analog outputs by delaying the SPDIF signal by this small amount.

    But we cannot anticipate additional differences of the connected audio interface.

  • To be more precise

    One more question ... you wrote that you can reduce the latency for the Profiler at 44.1kHz when it's master.

    Does this mean that the Stage in slave mode at 44.1kHz still runs through the sample rate conversion filters (linear phase) although there is no conversion required for a synchronization ... due to a hardware limitation?

  • One more question ... you wrote that you can reduce the latency for the Profiler at 44.1kHz when it's master.

    Does this mean that the Stage in slave mode at 44.1kHz still runs through the sample rate conversion filters (linear phase) although there is no conversion required for a synchronization ... due to a hardware limitation?

    That should be possible.

    The first idea is created and tested already.

    For the 44.1 slave solution, some more work and team intercommunication would be required.


    Please don't expect a quick solution for the second. The respective team members are working on some more exiting features, I do not want to interrupt them for a while.

  • I'm a bit confused and I'm hoping someone can help me understand.

    From what I was seeing in this thread, the issue at hand is that people are experiencing more latency with SPDIF, than with Analog Outputs, right? Which is weird because SPDIF requires one less conversion or something right?


    And Christoph Kemper hopped in the thread and kindly explained why latency is variable and not always in their control when it comes to AD/DA converters and that they're looking into getting the latency for the analog outs, back down to what it was at its prime back in another firmware. Which is super awesome and I'm glad he chimed in. It's really cool to see the Big CK himself in the threads here on the forum! ^^


    But I didn't see any explanation for why SPDIF is experiencing more latency and I'm wondering if someone from Kemper has said whether or not it's being looked into. I thought the extra latency through SPDIF was the issue at hand?


    Am I missing something?

    I'm feeling like I'm not understanding all of the thread and I'm very interested in using SPDIF instead of the main outs for recording but, I'd like to avoid using it if the latency is so much more when using SPDIF.