MIDI recording late

[bayswater]

Should dp register MIDI in the track exactly at the time you played it or the time you heard it?

When there is a delay that large, it should record when I play it. When I hear after a 1024 buffer it is not a meaningful position in time. Just like recording audio. You want the signal recorded as if you are monitoring from the source, not delayed as if you are recording from the end of a processing chain with significant latency.

[dewdman42]

Which is why dp has that setting I believe and then people complain when they turn it off they are getting the MIDI registered in the track early. Most likely they is because try as they might to ignore the delayed sound of the instrument with a big buffer they are actually playing it earlier then they realize due to the way we are wired as humans

[bayswater]

Should dp register MIDI in the track exactly at the time you played it or the time you heard it?

When there is a delay that large, it should record when I play it. When I hear after a 1024 buffer it is not a meaningful position in time. Just like recording audio. You want the signal recorded as if you are monitoring from the source, not delayed as if you are recording from the end of a processing chain with significant latency.

BTW, I'd also argue it it not recording MIDI when I hear it either. If I play to a metronome or a recorded series of quarter notes, I can hear what I'm playing through the VI and can hear that it is in sync with the metronome. But after it is recorded, when I play it back I can hear the delay and see it in the MIDI editors.

[bayswater]

Which is why dp has that setting I believe and then people complain when they turn it off they are getting the MIDI registered in the track early. Most likely they is because try as they might to ignore the delayed sound of the instrument with a big buffer they are actually playing it earlier then they realize due to the way we are wired as humans

I tend to play guitar early because I learned in live bands where you stand 15 msec away from your amp. But I can't play 60 msec early.

[dewdman42]

Nobody can but it will still be a little early. I read reports that they were 20ms early

[dewdman42]

Should dp register MIDI in the track exactly at the time you played it or the time you heard it?

When there is a delay that large, it should record when I play it. When I hear after a 1024 buffer it is not a meaningful position in time. Just like recording audio. You want the signal recorded as if you are monitoring from the source, not delayed as if you are recording from the end of a processing chain with significant latency.

BTW, I'd also argue it it not recording MIDI when I hear it either. If I play to a metronome or a recorded series of quarter notes, I can hear what I'm playing through the VI and can hear that it is in sync with the metronome. But after it is recorded, when I play it back I can hear the delay and see it in the MIDI editors.

Alright well produce a better test to prove it! The iac loopback and MIDI clock tests have too many moving parts to prove anything. And also nobody is as good as they think they are at playing to the metronome.

I also want to say, that this problem cannot possibly be related to the CPU...I don't know where that theory is coming from, but its doubtful. If the Xeon could not "keep up" the result would be dropped audio or MIDI notes....not "late notes". That is just fundamental to how buffer based DAW's work.

it is certainly possible that MOTU has introduced errors how it calculates the offsets, but we need more conclusive tests then what has been presented here thus far...

[bayswater]

Alright well produce a better test to prove it! The iac loopback and MIDI clock tests have too many moving parts to prove anything. And also nobody is as good as they think they are at playing to the metronome.

I also want to say, that this problem cannot possibly be related to the CPU...I don't know where that theory is coming from, but its doubtful. If the Xeon could not "keep up" the result would be dropped audio or MIDI notes....not "late notes". That is just fundamental to how buffer based DAW's work.

it is certainly possible that MOTU has introduced errors how it calculates the offsets, but we need more conclusive tests then what has been presented here thus far...

I already did a test that I reported earlier. I get playback later than when I played it, or when I heard it. I’ve done enough testing and I’m moving on to ways to fix the problem in pieces with a lot of MIDI. MOTU has accepted that it happens, and I’ve done enough to convince myself, although I was skeptical when this thread started. If this gets fixed in DP 11 I might update, but otherwise I’ll probably be using Logic for MIDI heavy projects for a while.

I agree this is very unlikely to be CPU dependent. It’s conceivable that it is related to versions of OS that only run on specific CPUs, but even that seems to me to be a stretch. But i don’t have any M1 Macs so I can’t check this for myself.

[Kurt Cowling]

I did this test with my Intel Mac using a mic to record my fingers hitting the keys. MIDI data lined up with the audio when the track was assigned to a hardware MIDI device (Yamaha Steinberg USB driver). When the MIDI track was assigned to a VI the MIDI data was significantly behind the audio.

Just tried this on my M1-Max MBP. MIDI and audio lined up perfectly whether the track was assigned to hardware or a VI.

Take that for what it is worth.

[dewdman42]

I already did a test that I reported earlier. I get playback later than when I played it, or when I heard it. I’ve done enough testing and I’m moving on to ways to fix the problem in pieces with a lot of MIDI. MOTU has accepted that it happens, and I’ve done enough to convince myself, although I was skeptical when this thread started. If this gets fixed in DP 11 I might update, but otherwise I’ll probably be using Logic for MIDI heavy projects for a while.

I totally hear ya.

I'm trying some tests this evening myself with buffer set to max 2048. I don't have a mic at my disposal at the moment so I can't do it the way I would prefer, but I am trying some methodical things as well as I can. I'm getting sorry results also. I haven't recorded live MIDI in quite a while, so I'm kind of shocked by these results also...FWIW.

So far what I find is if I turn off "sync recorded MIDI to Patchthru", I turn down the VI so that I can barely hear it and then just try to hit notes on the metronome that I hear....I am able to get them pretty consistently 20ms early when it feels to me like I'm making the clank sound at the bottom of my keyboards travel line up with the metronome click.

so that agrees already with what someone posted earlier on this topic. But really that is not a good thing. My thought is that I should be able to get it pretty close to spank on the money with that preference disabled...give or take some player slop factor...

What's really interesting is that when I drop the buffer size to 128 using this same approach, attempting to match the metronome, the notes are recorded MORE, like 30-40ms early on the track.

So far this is not good, to me it looks as if the record offset is not factoring in the metronome's audio latency correctly. The calculation is obviously messed up, and the inverse relationship between that offset and the buffer size is interesting.

So then I turn that "sync recorded MIDI to Patchthru back on. Again I will attempt to match my physical keyboard clanking to the metronome I hear.

with a buffer of 128 they are ending up pretty much on the beat..my playing is sloppy, but on average they were landing on the beats when I try to match the clunk of my keyboard with the metronome. Now in theory DP should be nudging them a little later to match what I hear but with a small buffer setting, its too small to make any more difference then my own sloppy playing, within a few ms.

At buffer 2048, however, the notes are getting registered in the track 120-150ms late...the slop factor is me of course...but still...I think 120-150ms of latency is WAY more than would need to be corrected for a buffer size of 2048.

So yea, I think its broke too! I don't think it has anything to do with the CPU..this is a raw calculation bug somewhere related to buffer size and latency calculation.

I have never been able to figure out how to get a report of how much latency DP is detecting for my sound card and buffer. With LogicPro its shown to us in the preferences window. With DP...there is nothing shown. I used to be able to obtain this information from AULab, but that isn't running on Monterey for me, its just crashing, so I can't use that...I'd like to know what the buffer latency would be expected to be with a buffer size of 2048. DP used to double buffer and so forth..but still, 150ms is something like 6000 samples give or take...so that is way off...no doubt about it.

I agree this is very unlikely to be CPU dependent. It’s conceivable that it is related to versions of OS that only run on specific CPUs, but even that seems to me to be a stretch. But i don’t have any M1 Macs so I can’t check this for myself.

Unless MOTU has some code that is defined to run on one CPU and different code to run on a different CPU and those two versions of the code are not the same in their calculations. This is a calculation bug...not a CPU too slow bug.

[smashprod]

I did this test with my Intel Mac using a mic to record my fingers hitting the keys. MIDI data lined up with the audio when the track was assigned to a hardware MIDI device (Yamaha Steinberg USB driver). When the MIDI track was assigned to a VI the MIDI data was significantly behind the audio.

Just tried this on my M1-Max MBP. MIDI and audio lined up perfectly whether the track was assigned to hardware or a VI.

Take that for what it is worth.

YES! That s exactly it. No problem on an M1 computer, be it MacBook Pro, iMac or Mac Mini.

Which Intel Mac did you test with? We can add it to the list of CPUs that have the problem, along with all of the cheesegrater Mac Pros up through the 2012 5,1 model.


Sent from my iPad using Tapatalk

[dewdman42]

my Mac Pro specs are in my signature.

[smashprod]

I was replying to Kurt Cowling. Are you officially confirming the bug as well? I can’t see your specs in your signature (I use Tapatalk).


Sent from my iPad using Tapatalk

[Kurt Cowling]

My intel Mac is an iMac 18,3.

[dewdman42]

read my last long post. yes, there is clearly a problem. I'm on 5,1 with 3.46 12 core.

[smashprod]

read my last long post. yes, there is clearly a problem. I'm on 5,1 with 3.46 12 core.

Thanks, I have the same machine (2012).


Sent from my iPad using Tapatalk