external MIDI instrument timing

[dewdman42]

The "audio timing offset" sends MIDI early? Really?

[magicd]

Edited for accuracy.


See post for accurate info:

viewtopic.php?f=26&t=62230&start=30#p531139

Dave

[Robert Randolph]

The "audio timing offset" sends MIDI early? Really?

No. The record offset calibration places recorded audio in the correct place in relation to previously recorded audio.

setting the playback offset

Playback offset is not the same thing as the record offset. Playback offset sends out MIDI either earlier or later in relation to the timeline.

Dave

Hey Dave, have you checked the playback offset recently? Negative values don't seem to do anything for me here.

I've always just used record offset, since I'm usually re-recording MIDI>-synth recorded from live performances. I have that setup to work pretty much perfect for most things, so I've not encountered this before.

[James Steele]

You know, I don't think I've ever set the playback offset. I wish actually someone gave very simple, step-by-step instructions on this because I read the Getting Started section on it, and was confused. Actually it would be great to have the entire procedure and CORRECT order outlined. Audio I/O calibration (first playback offset then record?) and then the MIDI timing adjustment.

So first step would be to take something like a recorded click audio track, direct it to an audio output on my interface, take a cable and patch that output back into an input on my interface and record a new track, then see what the difference is, type in the value, re-record and see if they line up? Confused.

[James Steele]

Okay. I re-read Dave's post. Do the record offset first. Hmmmm... I'll have to mess with this.

[dewdman42]

I was messing with it last night in 9.1 and also not getting any results. It might be broken in 9.1

Md, the setting in question is in the "audio" timing fine tuning dialog. No mention of MIDI there. In one sense playback audio timing can be made earlier or later relative to MIDI and it makes no difference if you say the MIDI goes earlier or the audio goes later, it's maybe the same difference. or maybe not. The question is relative to the actual timeline, what is happening by that audio playback offset. Is the MIDI sticking to the timeline and adjusting the audio relative to that, which seems intuitive from the dialog box labels, or is the MIDI being sent early or late and the audio sticking exactly to the timeline? If the MIDI is going early truly then I like it, though it's still a global value but probably can be set to something close enough for government work most of the time. But with certain situations it will still be off. I would personally rather use that setting comprensate for inaccurate signal chain latency reporting In a general sense and then any additional MIDI instrument add in latencies handled individually.

I think the dialog labeling is a little confusing if what it is really doing is changing the MIDI offset globally, and this could only apply to pre recorded MIDI tracks too as far as sending early.

I understand about record offset perfectly. It's not working in 9.1 though.

[RROY]

Okay, so this looks like a serious issue. I was listening to a project with external MIDI and virtual instrument in DP 9.1 and thought it didn't sound right. Timing was loose, sloppy sounding. Played it in DP 9.02 and it sounded tight again. I came up with a way to illustrate the problem. Created a new blank project with two bars of quantized sidestick hits, triggering an external sound source (Korg TR-88). Duplicated the MIDI to play a sidestick in EZ Drummer 2. Panned EZ Drummer hard left and the Korg hard right. This was recorded back in to DP through my mixer (A&H Zed R 16). I recorded a few passes in both versions of DP with different Record Offset settings. The 9.1 tracks are quite sloppy no matter what offset I used. The 9.02 tracks are quite tight even with different offsets. I will try to include a screen shot to illustrate the difference. Hope it works. I will also send a report to MOTU. If someone could duplicate this test that would great. So for now I am going back to DP 9.02 and hope the next update sorts this out. If the screen shot works the red tracks are 9.1 and the orange are 9.02
https://farm9.staticflickr.com/8534/290 ... cc67_b.jpg

[RROY]

Did a bit more looking into this timing problem with DP 9.1 and if you lower the buffer to 128 the timing discrepancy between external MIDI and virtual instruments pretty much disappears. This does NOT happen in DP 9.02. No matter which buffer setting you choose in 9.02 the timing remains tight. So if your mixing a song with a larger buffer in DP 9.1 and you are combining live external MIDI sounds as part of that mix, timing will NOT be tight. As a matter of course I think we all mix with a higher buffer setting. Could some one confirm this issue or state otherwise, as I believe it's not just my system causing the problem.

[RROY]

Got a quick reply to my TechLink from Motu. "Thanks for the information. I will present this case to the DP Development Team for further investigation. Please stand by for an update on Monday, and let me know if you need anything else in the meantime."

[magicd]

AAARGH!!! DANG DANG DANG. I owe the forum a huge apology. I got my basic explanation of the timing offset wrong. The good news is that DP works. The bad news is that I explained it incorrectly. Anybody who has been following my posts in this thread should now be about right confused if they've tried to follow my instructions. I'm going to go back and edit my posts to get rid of the incorrect information.

I am so sorry for getting it wrong and wasting peoples time.

Robert Randolph tipped me. He said he wasn't getting the right result from my test. So I set my gear up and did the testing. That's where I found my mistake.

The basic theory is all the same. We are talking about record and playback offsets in relation to both audio and MIDI. What I got wrong was which function to calibrate first. Either the record or playback offset will change the calibration of recorded audio in relation to recorded audio. I haven't used external MIDI hardware in over ten years so I've always done the audio I/O calibration with the Record offset. As mentioned the Playback offset will also change audio I/O alignment so if you don't use external MIDI use can use either calibration and you'll have accurate audio timing.

Playback offset does not change MIDI timing in relation to audio. This is what I got wrong. Record timing does change where a returning audio signal from an external MIDI module will be placed in relation to the timeline.

Obviously I should have done this test before my first post...

I set up DP with an UltraLite interface and a JV2080 sound module. I recoded a woodblock sound in from the 2080 as my audio reference. I added a MIDI note to a MIDI track and lined up the audio and MIDI to the sample. I set up two additional audio tracks. Track 1 records the audio after it is sent out of the interface and back in. Track 2 records the audio return from the JV2080. What we want to see is when we record on track 1 and 2 the original MIDI note, original audio signal, returning audio signal, and returning audio from the JV all line up.

On the first test pass the audio return was late and the MIDI return was later. Set Recording offset to 203. Now the MIDI return is lined up with the original audio and MIDI but the audio return is early. Set the Playback offset to -144. Now the two new tracks line up with two original tracks. If I do repeated record passes I see an error in the MIDI return track of 3-5 samples.

At this point you know that MIDI playback and audio playback match both on output and also on input when overdubbing or recording the external MIDI track to audio.

Whew!

I'm reasonably confident that if you try the test as described in this post you will get the same results as myself. Again I sincerely apologize for getting it backwards. I will do better next time.

Dave

[dewdman42]

So questions,

Without these settings does DP do any automatic offsetting? For example let's say I leave everything at zero and record some vocals, will the vocals end up on the track a tad late due to sound card input latency or does DP factor that in already and register the vocals on the track when I actually sang them with some automatic record offset based on one way latency detection as reported by the sound card(not round trip)?

In your above scenario would recorded vocals end up in the timeline early due to the large record offset setting which is based on a MIDI instrument round trip latency?

Is live MIDI still immediate and MIDI track playback still right in line with the timeline regardless of these settings?

Would all audio playback from existing audio tracks and existing VI's be delayed 144 samples? Of particular interest is the responsiveness of Vi's to live MIDI but also from MIDI tracks.

Where I'm going with all this is that it seems to me that record offset should be set to make sure any incoming audio is recorded on the track timeline exactly at the point in time the analog audio hits the analog side of the sound card. I would have guessed that DP is already attempting to do that based on the reported latency from the soundcard and that the record offset would allow us to fine tune that record offset a little earlier or later if we desire so that any other unreported input latency or misreported input latency will be accounted for. I'm not sure how we could do a calibration test to determine if record offset is actually needed because the audio loop back test measures round trip latency. But nonetheless if we could figure that out and adjust the record offset then we would know that all incoming audio gets registered on the track at the moment in time it hit the analog side of the soundcard, we'd know that vocals, incoming instruments of any kind, etc are all being accurately recorded.

On the other hand if we add a bunch of record offset to make up for MIDI instrument latency.... Then right off the bat, all other incoming audio will be registered on the track too early on the timeline.

Delayiing the playback by 144 seems to add 144 samples of global latency to all outgoing audio which has its own drawbacks and as far as the process of hearing what you're doing while you're playing the MIDI part live to the external inst relative to playback of other recorded tracks I'm not sure right now at this hour of the morning what the above would mean but globally adding any more latency then I need doesn't sound good to me,

[magicd]

Without these settings does DP do any automatic offsetting? For example let's say I leave everything at zero and record some vocals, will the vocals end up on the track a tad late due to sound card input latency or does DP factor that in already and register the vocals on the track when I actually sang them with some automatic record offset based on one way latency detection as reported by the sound card(not round trip)?

DP does an automatic offset based on the latency reported by the audio driver. This means you can change the buffer size and not have to change the record offset. However there are things happening beyond the audio driver that can not be reported back to DP. So with no further calibration yes, audio recorded back in to DP will be late.

In your above scenario would recorded vocals end up in the timeline early due to the large record offset setting which is based on a MIDI instrument round trip latency?

Not if the calibration is done correctly. In my most recent post I did the step by step testing. The result was that a MIDI note triggering an external synth got the synth record on time in DP. After the secondary calibration the looped audio was then in sync also. Changing buffer sizes or adding plug-ins in DP will not change the calibration.

Is live MIDI still immediate and MIDI track playback still right in line with the timeline regardless of these settings?

Yes that's correct.

Would all audio playback from existing audio tracks and existing VI's be delayed 144 samples? Of particular interest is the responsiveness of Vi's to live MIDI but also from MIDI tracks.

The actual amount of delay compensation can vary throughout the program. For example if you add a plug-in that introduces delay to the audio track DP will make that part of the overall latency compensation throughout the program. But internal latency compensation is not the same thing as record/playback offset calibration. These offsets have to do with what is happening outside the computer.

Where I'm going with all this is that it seems to me that record offset should be set to make sure any incoming audio is recorded on the track timeline exactly at the point in time the analog audio hits the analog side of the sound card. I would have guessed that DP is already attempting to do that based on the reported latency from the soundcard and that the record offset would allow us to fine tune that record offset a little earlier or later if we desire so that any other unreported input latency or misreported input latency will be accounted for.

Yes that's all correct.

I'm not sure how we could do a calibration test to determine if record offset is actually needed because the audio loop back test measures round trip latency. But nonetheless if we could figure that out and adjust the record offset then we would know that all incoming audio gets registered on the track at the moment in time it hit the analog side of the soundcard, we'd know that vocals, incoming instruments of any kind, etc are all being accurately recorded.

That's the purpose of the loop test. You need to know that when you are overdubbing against pre-recorded tracks that what you hear for playback and what gets recorded are exactly in sync. That's why you test with the loop. When playback from the computer looped from output back to input lines up exactly, you've got your overdub calibration.

On the other hand if we add a bunch of record offset to make up for MIDI instrument latency.... Then right off the bat, all other incoming audio will be registered on the track too early on the timeline.

You are correct and that's why there are two separate calibrations. In my earlier posts I got the sequence of calibration wrong (Very sorry for that!) When I hooked up all the gear and did the actual test I realized my mistake. Once I figured that out I was able to get DP properly calibrated. Again the final result is that MIDI playback and audio playback line up with recorded audio and recorded MIDI.

Delayiing the playback by 144 seems to add 144 samples of global latency to all outgoing audio which has its own drawbacks and as far as the process of hearing what you're doing while you're playing the MIDI part live to the external inst relative to playback of other recorded tracks I'm not sure right now at this hour of the morning what the above would mean but globally adding any more latency then I need doesn't sound good to me,

There is no added latency to a live MIDI signal from your controller back out to the sound module. That hasn't changed. What these calibrations do is make sure that the audio recorded back into DP lines up. You can't do any kind of latency calibration to the MIDI as you play in real time but the returning audio will be correctly lined up in the track.

It's not difficult to set up the test. Give it a try.

Dave

[monkey man]

Thank you for clarifying that, Davo!

Either the record or playback offset will change the calibration of recorded audio in relation to recorded audio.

You've got me there, brother.

EDIT:
Disregarding Dave's little loop-back thingy I quoted, am I the only one here who's confused?

Thankfully I won't have to deal with this stuff for quite a while yet, but I'm starting to dread the day. I'll be using a mix of outboard and "inboard" MIDI, re-recording both audio and MIDI parts at all stages of production, and not rendering the outboard MIDI parts as audio until right before mixdown (by switching the audio tracks used to monitor them from the "blue-button" to record mode), so trying to figure all this latency stuff out without actually being able to test it in the real world has really done my head in. Brain's turned to mush and is begging me to leave this alone!

Any simple cures for this malady?

[dewdman42]

Without these settings does DP do any automatic offsetting? For example let's say I leave everything at zero and record some vocals, will the vocals end up on the track a tad late due to sound card input latency or does DP factor that in already and register the vocals on the track when I actually sang them with some automatic record offset based on one way latency detection as reported by the sound card(not round trip)?

DP does an automatic offset based on the latency reported by the audio driver. This means you can change the buffer size and not have to change the record offset.

Right so under most circumstances it should not be necessary to change the record offset. Leaving it at zero, recording vocals into my 2408 should be registered to the timeline at the correct time, and not the late time due to input latency. Correct?

However there are things happening beyond the audio driver that can not be reported back to DP. So with no further calibration yes, audio recorded back in to DP will be late.

Take MIDI out of it for the moment. What other things are happening that would make vocals late?

In your above scenario would recorded vocals end up in the timeline early due to the large record offset setting which is based on a MIDI instrument round trip latency?

Not if the calibration is done correctly. In my most recent post I did the step by step testing. The result was that a MIDI note triggering an external synth got the synth record on time in DP. After the secondary calibration the looped audio was then in sync also. Changing buffer sizes or adding plug-ins in DP will not change the calibration.

This is where we are missing each other. I say vocals should be accurate with record offset at zero and when you do a calibration involving a MIDI instrument to set your record offset, now the external MIDI synth might be recording in the track on time but if you try to record vocals with that setting they would end up on the track early.

Would all audio playback from existing audio tracks and existing VI's be delayed 144 samples? Of particular interest is the responsiveness of Vi's to live MIDI but also from MIDI tracks.

The actual amount of delay compensation can vary throughout the program. For example if you add a plug-in that introduces delay to the audio track DP will make that part of the overall latency compensation throughout the program. But internal latency compensation is not the same thing as record/playback offset calibration. These offsets have to do with what is happening outside the computer.

I wasn't referring at all to PDC that is a separate issue. Some clarity on PDC and how it works would also be helpful though. What I was referring to is that in your example you are setting the playback offset to 144 samples late, partly in order to compensate for the fact you set record offset 203 early. But if the playback offset plays back audio 144 samples later then the actual timeline and MIDI, then that is effectively adding latency above and beyond and PDC corrections.

Where I'm going with all this is that it seems to me that record offset should be set to make sure any incoming audio is recorded on the track timeline exactly at the point in time the analog audio hits the analog side of the sound card. I would have guessed that DP is already attempting to do that based on the reported latency from the soundcard and that the record offset would allow us to fine tune that record offset a little earlier or later if we desire so that any other unreported input latency or misreported input latency will be accounted for.

Yes that's all correct.

I'm not sure how we could do a calibration test to determine if record offset is actually needed because the audio loop back test measures round trip latency. But nonetheless if we could figure that out and adjust the record offset then we would know that all incoming audio gets registered on the track at the moment in time it hit the analog side of the soundcard, we'd know that vocals, incoming instruments of any kind, etc are all being accurately recorded.

That's the purpose of the loop test. You need to know that when you are overdubbing against pre-recorded tracks that what you hear for playback and what gets recorded are exactly in sync. That's why you test with the loop. When playback from the computer looped from output back to input lines up exactly, you've got your overdub calibration.

It seems to me that record offset should effect only half of round trip latency and playback offset should effect the other half, with no consideration for MIDI whatsoever. And that should only be necessary if the soundcard is not reporting its latency correctly or if there are external A/D conversions in the signal path creating additional latencies beyond the sound card but are part of the regular sound path used by everything. The audio loop back can't be used to set the record offset because it presumes both input and output latencies need to be accounted for in the record offset. The playback offset should account for the output latency, record offset for input latency and for most people it should probably remain at zero. If you put the entire round trip overdub latency into record offset then live instruments would be early it seems to me

On the other hand if we add a bunch of record offset to make up for MIDI instrument latency.... Then right off the bat, all other incoming audio will be registered on the track too early on the timeline.

You are correct and that's why there are two separate calibrations. In my earlier posts I got the sequence of calibration wrong (Very sorry for that!) When I hooked up all the gear and did the actual test I realized my mistake. Once I figured that out I was able to get DP properly calibrated. Again the final result is that MIDI playback and audio playback line up with recorded audio and recorded MIDI.

I don't see how the playback offset would magically record vocals correctly

Delayiing the playback by 144 seems to add 144 samples of global latency to all outgoing audio which has its own drawbacks and as far as the process of hearing what you're doing while you're playing the MIDI part live to the external inst relative to playback of other recorded tracks I'm not sure right now at this hour of the morning what the above would mean but globally adding any more latency then I need doesn't sound good to me,

There is no added latency to a live MIDI signal from your controller back out to the sound module. That hasn't changed. What these calibrations do is make sure that the audio recorded back into DP lines up. You can't do any kind of latency calibration to the MIDI as you play in real time but the returning audio will be correctly lined up in the track.

I wasn't referring to MIDI I was referring to audio and VI's

Yes this thread has become way confusing. An ext inst plugin would be simpler and I still think it's better for now to use the time shift plugin for that for now.



Sent from my iPhone using Tapatalk

[magicd]

Right so under most circumstances it should not be necessary to change the record offset. Leaving it at zero, recording vocals into my 2408 should be registered to the timeline at the correct time, and not the late time due to input latency. Correct?

No that's not correct. There are external factors that will delay the incoming audio in relation to playback.

Take MIDI out of it for the moment. What other things are happening that would make vocals late?

Converter delay. Any delay in digital processing outside the driver. External digital mixers. Again anything that can't report it's delay back through the driver will add latency to the record signal.

In your above scenario would recorded vocals end up in the timeline early due to the large record offset setting which is based on a MIDI instrument round trip latency?

Not if you do the calibration correctly.

This is where we are missing each other. I say vocals should be accurate with record offset at zero

That's an incorrect assumption and has nothing to do with MIDI.

and when you do a calibration involving a MIDI instrument to set your record offset, now the external MIDI synth might be recording in the track on time but if you try to record vocals with that setting they would end up on the track early.

Not if you first calibrate MIDI and then calibrate audio. I urge you to set up the test and follow my (revised) instructions for calibration.

Would all audio playback from existing audio tracks and existing VI's be delayed 144 samples? Of particular interest is the responsiveness of Vi's to live MIDI but also from MIDI tracks.

These calibrations have nothing to do with live triggering of VIs or external MIDI instruments. The calibrations have to do with playback of recorded audio and MIDI tracks and the relation of signals being recorded back in to DP.

I wasn't referring at all to PDC that is a separate issue. Some clarity on PDC and how it works would also be helpful though. What I was referring to is that in your example you are setting the playback offset to 144 samples late, partly in order to compensate for the fact you set record offset 203 early. But if the playback offset plays back audio 144 samples later then the actual timeline and MIDI, then that is effectively adding latency above and beyond and PDC corrections.

Yes but that is not a problem. What the calibrations do is make sure that incoming audio is recorded to the correct location, whether that is live audio or playback from a MIDI module.

Where I'm going with all this is that it seems to me that record offset should be set to make sure any incoming audio is recorded on the track timeline exactly at the point in time the analog audio hits the analog side of the sound card. I would have guessed that DP is already attempting to do that based on the reported latency from the soundcard and that the record offset would allow us to fine tune that record offset a little earlier or later if we desire so that any other unreported input latency or misreported input latency will be accounted for.

Yes that's all correct.

I'm not sure how we could do a calibration test to determine if record offset is actually needed because the audio loop back test measures round trip latency.

"Round trip latency" typically describes the time it takes to send a signal into, through, and out of the computer. That is not what we are measuring here. We are comparing an audio signal that is sent out of the computer and back in. We are comparing newly recorded audio to a pre recorded audio track. The measurement includes both output delay and input delay in terms of external factors such as converters and any additional output processing. A simple example: You have a click track playing back from DP. You play along with that signal and record back in to the computer. You needed to know that what you are recording lines up with what you are hearing. A more precise test is to just record the click output back in to a track in DP. When you see the recorded audio line up with the original you know that there is no offset in recorded audio in relation to playback from the computer.

On the other hand if we add a bunch of record offset to make up for MIDI instrument latency.... Then right off the bat, all other incoming audio will be registered on the track too early on the timeline.

We're back to page one. You're making it more complicated than it needs to be.

Please do this test:

Set up a MIDI track and an audio track. Trigger a transient sound (snare, woodblock, etc) and record that sound into the audio track. Zoom down to sample level and align the audio transient with a MIDI note that will trigger the same sound on the external module.

Set up two additional audio tracks. Track one records the looped output from the original audio track. Track two records the audio output from the external module. Hit record.

What I expect you will see is that the looped audio is late and the triggered module audio is later.

Use the Playback calibration to line up the external module triggered audio to the original MIDI note. The looped audio will now be early. Use the record offset to realign the looped audio.

For at least this synth you now have proper calibration and no further adjustment will be needed. If you have other synths that have a different response you can use a MIDI time shift plug for those devices.

Dave