stereo image skew due to unexplained 3-sample extra latency in firewire channel 4

Home Forums Products Rackmount stereo image skew due to unexplained 3-sample extra latency in firewire channel 4

  • This topic is empty.
Viewing 3 reply threads
  • Author
    Posts
    • #106472
      jwhiting
      Participant

      hello,

      i've recently discovered a latency problem with my H8000FW unit. it's not that the total latency is a problem, it's that the latency differs relative to the channel the audio is being routed through, which creates a stereo imaging problem.

      this is a bit hard to explain, but with a very straightfoward routing setup, a stereo signal ends up being skewed, where a 3-sample (exactly 3 samples) latency is present in the right channel but not in the left.

      so that you don't call me crazy or claim i've done something weird, i've created the simplest test case i could to illustrate the problem, and have linked detailed screenshots with explanations below. this example shows the problem when using firewire channels 3 and 4, but I can reproduce this latency mismatch using channels 5/6 or 7/8 as well.

      my system stats include all the latest software: an imac running OS X v10.5.8, the latest H8000FW driver (downloaded today and reinstalled just to be sure), H8000FW firmware version 5.3, logic pro v8.0.2.

      the screenshots/details:

      1. this shows the basic problem: http://69.93.127.137/h8kimg/h8k-phaseproblem.jpg – what you see here is the phase incoherence introduced by the 3 sample latency in channel 4. there is nothing fancy happening here. the routing looks like this: ES P mono synth -> firewire out 3/4 -> h8k in 3/4 -> h8k out3/4 -> firewire in 3/4 -> goniometer plugin showing the misaligned signals. (note that i have the h8k routing utility opened in a remote desktop session so you can see both my logic pro configuration alongside the h8k routing.)

      2. this shows the synth and goniometer signal with the h8k bypassed. there is no latency difference between the channels, they are perfectly aligned: http://69.93.127.137/h8kimg/h8k-bypass.jpg

      3. going back to the problem setup, this shows how i can correct the issue by manually inserting a 3 sample delay into the left channel, aliging the phase perfectly again: http://69.93.127.137/h8kimg/h8k-3sampdelay.jpg

      4. to try to isolate where in the h8k signal path the latency was introduced, i did some tests by duplicating the single input to both l/r outputs in a couple different places. what i found, by process of elimination, was that the latency appears to be introduced at the "main in" routing block of the h8k: http://69.93.127.137/h8kimg/h8k-in3dupout34.jpg versus http://69.93.127.137/h8kimg/h8k-fw3dupin34.jpg

      5. there is one final, possibly related issue, which is that the firewire input channel 1 does not pass signal back to my DAW. inputs 2 thru 8 do pass signal: http://69.93.127.137/h8kimg/h8k-nosignal.jpg

      thanks
      -Josh

    • #119143
      DGillespie
      Moderator
      Eventide Staff

      Hi Josh,

      Thanks for the phenomenally detailed bug report.  We'll take a look at this and see if we can reproduce it and get back to you.  If we can't reproduce it, we may email you because we want to get to the bottom of this.

      Thanks,

      Dan

    • #119144
      nickrose
      Moderator
      Eventide Staff

      Thanks for a very full and thoughtful report. This is rare and appreciated.

      We will look into this and hopefully include a solution in the next release. I say hopefully, because this may be a PC issue that is beyond our ken.

      Is there any possibility that the delay is 4 samples rather than 3 ?

    • #119146
      jwhiting
      Participant

      It's no trouble regarding the report's detail. I'm a software engineer myself, albeit in web development and not DSP, so I appreciate specificity in bug reports, too 🙂

      As for your question about a possible 4-sample latency, that might be possible if the plugins I'm using in Logic Pro to observe this are not telling me the whole truth. I'm relying on the "sample delay" and "multi-meter" goniometer plugins to measure the issue, but I do need to set the sample delay plugin to "3" in order to get a fully vertical goniometer readout. I've never attempted to measure so precisely at the sample level – there are probably better approaches.

      Let me know if there's anything specific you'd like me to test, I'd be happy to be of more help. I also have an ADAT interface on my PC and could connect and test the unit that way if you think this might be firewire specific, though I use firewire as my workhorse via my imac.

Viewing 3 reply threads
  • You must be logged in to reply to this topic.