MIDI recording late

[bayswater]

No way changing buffer size is a reasonable solution to this problem, and I wouldn’t let the ticket close at that point.

Pretty sure this is an issue with DP (I'm on 10.13). Although I haven't tested lately, my recollection is consistent with what smashprod is reporting. It needs to be addressed.

Fair enough, it clearly happens, and there are pre V10 reports of it here. Too bad it hasn't been fixed in V11. But it is not happening on my setup, so I suspect it is specific to particular configurations of DP and equipment. I'm in the middle of a project with dozens of tracks, lots of them MIDI, with thousands of notes, using buffers from 128 to 2048, and nothing like this has happened yet. Short of replicating a setup where it happens regularly, it's going to be difficult for MOTU to diagnose this.

[smashprod]

No way changing buffer size is a reasonable solution to this problem, and I wouldn’t let the ticket close at that point.

Pretty sure this is an issue with DP (I'm on 10.13). Although I haven't tested lately, my recollection is consistent with what smashprod is reporting. It needs to be addressed.

Fair enough, it clearly happens, and there are pre V10 reports of it here. Too bad it hasn't been fixed in V11. But it is not happening on my setup, so I suspect it is specific to particular configurations of DP and equipment. I'm in the middle of a project with dozens of tracks, lots of them MIDI, with thousands of notes, using buffers from 128 to 2048, and nothing like this has happened yet. Short of replicating a setup where it happens regularly, it's going to be difficult for MOTU to diagnose this.

That is really frustrating, because if this is not happening with everyone then it might be related to other software on my machine, the OS, or the machine itself. Much harder to track down, but I have to figure this out before my next project starts. As you say, pads are one thing, but rhythmic staccato sounds are another. Any more troubleshooting ideas?

[HCMarkus]

I can't play rhythmic staccato sounds with a buffer of 1028. If I wasn't listening to the VI, maybe monitoring from an external synth, it would work. But I'm interested to see how my Mac, specs in my signature below, performs under DP10.13

Surprising results. given smashprod's report...

Record enable a MIDI Track AND an Audio Track.
Route a Mic to the Audio Track and place it near your keyboard, so it picks up the sound the keyboard makes as it is struck by your fingers
Important: Record-disable each track before changing buffer size, then re-enable to do the test. In the past, I have noted that changing buffer size with tracks record-enabled can mess up timing. Didn't test that this time.

I tried this with the following results (not perfectly accurate, but you'll see a pattern here):

Sync Recorded MIDI to Patch Thru OFF
Buffer Size - MIDI Delayed relative to AUDIO in Samples @48k

1024 - 170
512 - 700
128 - 1050

Sync Recorded MIDI to Patch Thru ON
Buffer Size - MIDI Delayed from AUDIO

1024 - 400
512 - 700
128 - 1200

Tempo was set at 200. I tried at 118, results were similar, with tighter timing @ the 1024 buffer, about 30 samples delay. Slightly less delay at the smaller buffers, too.

The pattern established above (smaller buffer = later MIDI) was maintained. The absolute difference in timing includes MIDI delay from the keyboard, but the relative difference in MIDI delay is clearly not consistent.

I got these results using my standard template. Now I'm going to test with a fresh project, no plugins running anywhere. It is possible something with Look-Ahead is in my template somewhere, and that could throw off results. Stay tuned!

EDIT: Tried this experiment on a New Project with no plugins. The pattern is maintained... smaller buffer = later MIDI.

So what are we talking here in real-time?

170 Samples @ 48k = 0.0035 seconds or 3.5 Milliseconds

700 Samples @ 48k = 0.0146 seconds or 14.6 Milliseconds

1200 Samples @ 48k = 25 Milliseconds.

[smashprod]

Am I right in understanding that in your test, the lower the buffer setting, the later the MIDI, and the more out of sync with the audio (or, in my case, click)? That's the opposite of what's happening to me, where the higher settings make the MIDI later. Also, even though the Sync pref works the same way for you it did for me (with it off, MIDI is recorded earlier, on, MIDI is recorded later), there's a big difference-- when I turn it off, the MIDI notes are recorded before the beat.

Curioser and curioser.

This didn't used to happen-- years ago I used to use DP for my pop stuff, where timing was critical. Now I use Pro Tools for anything pop, and DP only for scoring gigs.

[HCMarkus]

Am I right in understanding that in your test, the lower the buffer setting, the later the MIDI, and the more out of sync with the audio (or, in my case, click)? That's the opposite of what's happening to me, where the higher settings make the MIDI later. Also, even though the Sync pref works the same way for you it did for me (with it off, MIDI is recorded earlier, on, MIDI is recorded later), there's a big difference-- when I turn it off, the MIDI notes are recorded before the beat.

You are correct, sir!

Although I have a session this morning, as soon as I get a chance, I'm going to open up my fresh partition with DP11, update to 11.03, and try the test.

On that partition, everything has been installed fresh. I have barely used it to date, just haven't had time, so it is a good place to see if similar results are obtained with my aging 5,1 Mac Pro

[smashprod]

I also have a 5,1 (with a Metal card running Mojave).

[smashprod]

Since it seems not everybody has this problem, I'm wondering if there might be some KEXTs or other system-level software that are affecting things? Any likely culprits to look for?

[mikehalloran]

Since it seems not everybody has this problem, I'm wondering if there might be some KEXTs or other system-level software that are affecting things? Any likely culprits to look for?

Certain Soundflower but not all, old NI from OS 10.6, old HP print drivers were culprits on my system. Apple Engineering helped me find and remove them but they were causing many problems.

[smashprod]

OK, I went through and deleted what I could of my KEXTs, launch items, etc.; I can't use Safe Mode because then DP doesn't have what it needs to run. My new test file doesn't use VEPro, I'm just running Kontakt as part of the main sequence. MIDI still records late, and only reducing the buffer size helps. The biggest thing is that "Fine-Tune Audio I/O Timing" doesn't work at all. That's a red flag. MoTU can't figure it out either. ANYONE else having this problem?? Wondering if after all these years I need to switch to Logic...

[smashprod]

Oh, and by the way, seems obvious, but no, this doesn't happen in Pro Tools, not even with a 1024 buffer.

[HCMarkus]

Problems Continue

I just looked thru this thread... did you ever identify what computer/OS version you are using?

I've been really busy and haven't had a chance to investigate further, but based on my recollections and recent testing, my 2010 12 Core Mac Pro may be exhibiting similar behavior to that which you are experiencing; I'm running 10.13 for daily use.

So ready to go M1 Pro.

[smashprod]

Have a 5,1 Mac Pro running Mojave (with a Metal graphics card). But MIDI in Pro Tools is fine.

[smashprod]

OK, I installed 11.04 today, and I created a new sequence based on the DP "Empty Document" template. I am not using VEPro, just one Instrument track with one instance of Kontakt (loaded with a NI grand piano), one MIDI track, and one audio track, where I have built an audio click (so I'm not reliant on the built-in click for these experiments). Still the same situation. Also still the same situation with "Fine-Tune Audio I/O Timing"; I've tried values of 10,000 and -10,000 in both Recording and Playback offset, nothing makes a difference. And 10,000 samples at 48K is over 200 milliseconds!

On the advice of a MoTU tech I tried "Live Performance Mode" (no help), using the built-in audio under "Configure Hardware" (no help), and "MIDI Only" (which helped). There's gotta be a conflict with a KEXT or something, but I don't have enough room on my SSD for a second, clean system install.

By the way, my guess is that my MIDI is late by about 40 milliseconds. This is not subtle!

Has anybody had this problem?

[smashprod]

We found a reproducible problem! I'ver been working with Matt Batson at MoTU and he reproduced the problem on an Intel MBP he has. It's Intel-specific; there is no problem with an M1 system or an AMD (PC) system. There is a consistent MIDI delay (about 40-50 ms) when a MIDI track is connected to a VI-- any VI. Remove the VI and MIDI records perfectly in sync with the click, regardless of the hardware buffer setting (which is how it works on an M1 or AMD machine).

This is being brought up at today's MoTU meeting, I should have more info soon. At least I should know whether the programmers are aware of the problem, and, if they are, whether they have a fix planned.

I wish I could just jump ship and go to the M1, but the Mac Minis don't have enough memory for serious scoring work yet.

[James Steele]

This is great news that they have been able to reproduce this. Thank you for being persistent about it and helping MOTU find this. I hope they feel there’s enough of an installed Intel user base that this needs to addresses. DP 11.05, anyone?