Latency measurements KPA (Constant Latency = ON), Focusrite 6i6, Reaper 44,1kHz

  • Hi, I did some tests with my KPA, since I had the feeling recordings could be a little tighter aligned to my drum-programmings. So, here are my findings:

    (Constant Latency = ON), Focusrite 6i6, Reaper 44,1kHz


    1) Recording (equivalent to a standard live recording Guitar->KPA Input connected to my 6i6):

    - KPA settings

    • KPA Input/Input Source = “Front Input”
    • KPA Output/Direct Output = “Guitar Analog” (->Bypass KPA)
    • KPA Output/SPDIF Output = “Git/Stack”
    • switch off Stomp/Amp/FX-sections (original signal stays unaffected)

    - Play Track 1: suitable sample (e.g. square wave or snare hit)

    • Send 6i6:LineOut1 -> KPA:FrontInput

    - Record on Track 2: KPA:DirectOut -> 6i6:LineIn3

    • Recording -13 (minus!) samples (early/in Reaper aligned before source-sample, probably 6i6-driver-calculation-error)

    - Record on Track 3: KPA:MainOut -> 6i6:LineIn4

    • Recording 203 Samples late (consistent with the 4,9ms latency pronounced by C.Kemper, considering the 6i6-driver-calculation-error of -13 Samples)

    - Record on Track 4: KPA:SPDIF_Out -> 6i6:SPDIF Left (Guitar Studio/DI)

    • Recording 362 Samples late

    - Record on Track 5: KPA:SPDIF_Out -> 6i6:SPDIF Right (Stack)

    • Recording 362 Samples late


    2) Reamping:

    - KPA settings

    • KPA Input/Input Source = “Front Input”

    - Play Track 1: suitable sample (e.g. square wave or snare hit)

    • Send to 6i6:SPDIF Out->KPA:SPDIF In (Reamp)

    - Record on Track 6: KPA:SPDIF_Out -> 6i6:SPDIF_In Right (Stack)

    • Recording 420 Samples late


    That means for me and my environment to get all KPA-recorded tracks aligned automatically:

    • To be save, Constant Latency always ON
    • Monitoring with KPA:MainOutput->available 6i6:LineIns – 4,9ms latency

    - Settings in Reaper to align recordings/reampings automatically

    • Options/Preferences/Audio/Recording/Use audio driver reported latency = ON
    • Output manual offset: 58 samples, compensating for the additional reamping-delay over 6i6:SPDIF Out->KPA:SPDIF In
    • Input manual offset: 362 samples

    --> Result, no more nudging necessary after Recording :)


    What do you think – I hope this makes sense and there’s no misconception on my side?

  • Can one hear a delay of 362 samples?...


    I like being on time and «tight», but this is midi-level tightness, isn’t it... =O


    Edit: did the math, 362 samples at 44.1 is 8,2 ms...

  • Can one hear a delay of 362 samples?...


    I like being on time and «tight», but this is midi-level tightness, isn’t it... =O


    Edit: did the math, 362 samples at 44.1 is 8,2 ms...

    And reamping is even 9,5ms - don't know why this takes even longer…

    Aligning the recordings sounds better and more pro to my ears - but maybe it's a Metal-thing where everything has to be as tight as possible ;)


    Btw, I never had any problems with my 6i6 or the 8i6 - happy Focusrite-User here. Buffer setting during the tests was 10ms (you only can set ms in Scarlett Mix Control). But as we can see, the aligning error in Reaper based on the audio driver calculation bypassing the KPA was only -13 samples, just -0,3ms - that's really not noticeable.

  • Constant latency adds latency like plugin delay compensation does in most DAWs. So for instance the pitch effects add a latency of a few mS. Enabling constant latency makes it so you don't feel that latency hit.


    Just disable it. I get roughly 3mS with constant latency disabled measured with both a Focusrite Clarett 4pre USB via spdif and a UA Apollo x6 spdif.


    Since the Kemper's spdif update there's no reason not to use spdif if your interface has it. The only minor annoyance is setting your interface to clock from the Kemper. Sound wise with good preamp/line inputs like the Clarett and Apollo there's no audible difference vs. spdif, but you shave a couple mS off the round trip.

  • Constant latency adds latency like plugin delay compensation does in most DAWs. So for instance the pitch effects add a latency of a few mS. Enabling constant latency makes it so you don't feel that latency hit.


    Just disable it. I get roughly 3mS with constant latency disabled measured with both a Focusrite Clarett 4pre USB via spdif and a UA Apollo x6 spdif.


    Since the Kemper's spdif update there's no reason not to use spdif if your interface has it. The only minor annoyance is setting your interface to clock from the Kemper. Sound wise with good preamp/line inputs like the Clarett and Apollo there's no audible difference vs. spdif, but you shave a couple mS off the round trip.

    Yes, that's correct, Constant Latency adds latency - but to both SPDIF and Main Out. The value added is dependent from the profile you use. So, if you got a profile that has 3ms, the difference is in this case 1.9ms - for me that's not noticeable. But, what's noticeable to me is the SPDIF-latency of 8.2ms (or 9.5ms for reamping), even if you'd substract the 1.9ms by switching CL off (then 7.6ms for reamping). You're right, there's still NO reason not using SPDIF for recording, because you can define an offset (at least in Reaper) in your DAW-preferences to compensate for that and get your item contents aligned automatically - that was just my tip here ...and I wanted to be sure my test cases where correct :)


    How did you measure? As you can see I recorded parallel 4 destination tracks from one source track (KPA bypassed via Direct Out, Main Out, SPDIF Git & Stack). Imho the type of audio interface makes no difference, because it can never know about the latency an external device adds and automatically compensate for that.

  • Gotta be honest, had a lot of problems with my Focusrite 6i6, eventually went with a Presonus 68 and had a lot better results.

    I have the polar opposite results. I had the Presonus 44vsl, and had nothing but problems with latency. I was in contact with Presonus tech support for 2 days, and finally returned it, and got a Focusrite 18i20, and never had any issues.

  • There are several USB removal utilities out that remove the USB reference from the system, and allows you to reconnect it like it's the first time. This allows the software to be loaded like new. Sometimes fixes some usb issues.


    USBOblivion is one that I have used in the past. It might be worth a shot. Just unplug the Focusrite (and any other usb devices other than mouse and keyboard). After it's done, reboot, and then plug in each device, one at a time to reload the drivers like a new device.


    Might be worth a shot.

  • I have the polar opposite results. I had the Presonus 44vsl, and had nothing but problems with latency. I was in contact with Presonus tech support for 2 days, and finally returned it, and got a Focusrite 18i20, and never had any issues.

    Yeah, the 18i20 is very nice. Head and shoulders above the 6i6. Don't get me wrong, I wanted the 6i6 to work, but guess it just wasn't in the cards.

  • How did you measure? As you can see I recorded parallel 4 destination tracks from one source track (KPA bypassed via Direct Out, Main Out, SPDIF Git & Stack). Imho the type of audio interface makes no difference, because it can never know about the latency an external device adds and automatically compensate for that.

    Just using the RTL utility from http://www.oblique-audio.com/free/rtlutility


    Since USB itself is a dynamic latency RTL utility should be fine for Focusrite interfaces. I didn't bother adjusting sample offsets in Reaper since I wouldn't likely use software monitoring with the Kemper. It's also handy for figuring out hidden buffers which mess up Reaper's latency calculation.


    The Apollo being thunderbolt 3 doesn't have the same problems with dynamic latency since it's basically external PCIe and you can also disable it's default 64 sample hidden buffer. Of course for the cost of this dang thing it better not have all those little annoyances.

  • Just using the RTL utility from http://www.oblique-audio.com/free/rtlutility


    Since USB itself is a dynamic latency RTL utility should be fine for Focusrite interfaces. I didn't bother adjusting sample offsets in Reaper since I wouldn't likely use software monitoring with the Kemper. It's also handy for figuring out hidden buffers which mess up Reaper's latency calculation.


    The Apollo being thunderbolt 3 doesn't have the same problems with dynamic latency since it's basically external PCIe and you can also disable it's default 64 sample hidden buffer. Of course for the cost of this dang thing it better not have all those little annoyances.

    I think we have a misunderstanding. The sample offset adjustments I do in Reaper are mainly for compensating the KPA-SPDIF-latency in the RECORDED items (375-13 In for the 1st recording + additonal 58 Out for a possible reamping loop, if needed). I also do not monitor through software, but through the KPA-Main Outs.


    Testing your Apollo with a suitable sample (square wave) playing as source on Track A, sent to Apollo/Line Out->KPA/Front_Input and recording on Track B from KPA/SPDIF_Out->Apollo:SPDIF_In would show what I mean (the pure KPA/SPDIF latency). Parallel recording (maybe on Track C) of the KPA/Output/Guitar_Analog is like bypassing the KPA and possibly shows a possible audio driver latency calculation error. That's the value reported to the DAW to align recorded audio automatically (but of course without considering latency of external devices like the KPA). For my 6i6 it's negligible, -13 samples - but since I have to compensate for the KPA/SPDIF-latency anyway, I include the -13... ;)


    I'm not an English native speaker and I wished I could explain better what I mean ;)

  • Great thread! I’m still a bit confused, though. There is a latency (4,9ms with constant latency) caused by Kemper to even hear the sound of the guitar. Doesn’t it make sense not to compensate that? Isn’t brain going to compensate for the sound you hear not when you plug the note?


    I’m using Focusrite Saffire Pro 24 DSP and based on my calculations, when using spdif for recording I have to add -139 samples to compensate the spdif slowness compared when I hear the analog signal i use for monitoring.

  • That's strange that the SPDIF total latency is higher than the one from the main outs.


    ckemper or Burkhard - do you have some light to shine on this?

    I can confirm based on my test (some time ago) two things:

    1) SPDIF Out add more or less 150 samples latency respect to Main Out

    2) REAMP add some latency respect to original Main Out Track Recorded (that i normally compensate manually each time I do some REAMP)

  • What can be a possible explanation for why SPDIF adds more latency? So that means, for live playing, analogue outs are more optimal? I just recently startef using SPDIF due to no AD/DA noise. Not sure it's worth it with added latency. I mean, I can't tell a difference, but I go wireless through a Helix into the Kemper and back into the Helix again and out to digital. Any latency reduction I can squeeze out is most wanted.

  • What can be a possible explanation for why SPDIF adds more latency? So that means, for live playing, analogue outs are more optimal? I just recently startef using SPDIF due to no AD/DA noise. Not sure it's worth it with added latency. I mean, I can't tell a difference, but I go wireless through a Helix into the Kemper and back into the Helix again and out to digital. Any latency reduction I can squeeze out is most wanted.

    I play with the Kemper and AxeFxII through SPDIF. I haven't tried much going into the Axefx, from there to the Kemper and back again, because I like the way the input on the Kemper reacts to my pedals.


    Anyhow, I don't notice any latency. I think I actually hear an improvement in sound, probably since the AxeFXII operates at 24bit / 48k, and the Kemper can output that via SPDIF.


    The weird thing is... I feel less latency than just playing through the AxeFXII on its own. Not sure if that's some psychological phenomenon or maybe latency was worse with a previous firmware... Can't tell.. Haven't played directly into the AxeFXII for over a year.


    C