external MIDI instrument timing

[RROY]

I think it is almost certain that "next gen pre gen" is the culprit when it comes to bad external MIDI timing, and that for whatever reason Audio I/O timing is not compensating the way it was in DP 9.02 and previous versions. From the testing I've done I think audio is still being recorded to where it should be, regardless of buffer size, unless that audio is being generated by an external MIDI sound source, then things get messy.

[dewdman42]

Can you please do the following test

setup a MIDI track sending to external MIDI device.

send the output of the external MIDI device directly to your speaker and not back into DP.

Setup the largest buffer size you can.

setup a quantized track.

leave metronome on and play it back with the MIDI going to the external device and listening to it directly through your mixer.

In that case, are the MIDI events being sounded off way late or right on time? Is there any difference with a small buffer size?

If they are right on time that way and no difference with buffer size, then nothing is wrong with the MIDI send.

The loopback test brings record offset into the mix so you need to determine which thing is the problem.

Sorry you're having so many problems, but if you want MOTU to fix it, they need the best information you can provide them.

I personally doubt that MIDI is being sent late and I'm much more skeptical that the record offset is not working properly...due to all the things they are doing messing around with the buffer handling. And if so, then all audio recording will be wrong, not just external MIDI instruments.

If on the other hand you can isolate the problem to the MIDI output through the above test, then that can give them better feedback for what to fix.

[RROY]

If you look back at some of my previous posts you will see the tests I conducted, with links to screen shots showing how buffer size effects external MIDI side stick vs a V.I. generated side stick. These were done in DP 9.1 and DP 9.02. All this info has been sent to Motu, via Tech Links.

[dewdman42]

Your test includes a loop back so you have not isolated MIDI from audio. Please try the test I mentioned and report the results and we will know if the problem is MIDI out or audio in, I suspect the latter.

[RROY]

Yo, dewdman42. I couldn't resist! I tried your test as described. At smaller buffer settings the timing is tighter, as the buffer is increased the flaming between the two sources becomes greater. This is just listening back to the 2 sources no recording of audio involved. To me this would indicate it is an external MIDI timing issue. Also, if you record this back into DP, the recording confirms what I am hearing.

[dewdman42]

One more thing to test. Can you determine which sound is late, the external sound or the meteonome? Larger buffer will make the metronome sound later and later. Again I am predicting the MIDI is being sent right on time and it's the metronome that is being delayed

[RROY]

For me it's the external MIDI gets later and later as you increase the buffer, the metronome stays on time at all buffer settings. Again, none of this happens in DP 9.02.

[dewdman42]

Ok send that info to motu. Doesn't make sense to me that MIDI would go later and later with larger buffer size, but please report it!

[OldTimey]

I've wrestled with the "fine tune audio timing" window for years and years. i've found that the audio loopback (send a track out of soundcard and then back in) only lines up when I have both values set to the inverse of each other (ie Playback Timing: -188 Record Offset 188) Or when leaving both at 0.

Some MIDI devices are so jittery that it's virtually impossible to get them to all play in perfect time. Take my MPC1000 for example, I can set a MIDI track in DP to trigger a kick drum sample on the MPC1000 on every beat (quantised) for 100 bars. If I print this to an audio track, the hits are all "pretty close" and some are "spot on" but from hit to hit there can be a variance of plus or minus 30 samples. At 441000 samples per second, this equates to less than one thousandth of a second. Kind of gives it some "life" i think. I have a hard time perceiving anything under 10 milliseconds.

In the end when printing external MIDI devices (I use a number of them) I use my ears and nudge tracks as needed.

[dewdman42]

30 samples is not enough jitter to worry about. Of course external MIDI is not sample accurate, but 30 samples is sub-milisecond. Not an issue.

If you have MIDI hardware that has MIDI timestamping, then it is going to be very tight enough to not be an issue for playback of MIDI tracks. When we see these much larger timing problems..there is something else going on...

[Maxxy]

I am also quite confused

Aren’t we using the Fine Tuning I/O to precisely ‘tune up’ our Audio Interface? …. so that the recorded audio will sit in DP with near sample sample accuracy.

I don’t mean to sound like a simpleton but can we go through it step by step?

I use an example of audio with a clear fast transient…our ‘Clik’

The Clik is sent out our ‘Audio Interface’ eg. analog Output #11…. and hard patched through either a patchbay or with a cable to , for example, ‘Audio Interface’ analog input #3

Let’s make it easy and call the audio interface MOTU 16A…(because I have one and it sounds absolutely stellar) ( :

In my case the recorded click sound is now 55 samples late in the time line. By using the Fine Tuning I/O settings I set the Recording offset to 55 samples exactly. Now, no matter which buffer setting I am using on the 16A itself…. from 32 to 2048….. each recorded click now comes back exactly where it should be. ….great result!!

The new Hardware Insert plug-in in DP 9.1 should accomplish the same thing in a nano second …. including compensating for the various delays of each external hardware processor on top of the audio interfaces inherent delays. However i think that by simply setting one’s own Interface’s basic delay aren’t we already most of the way there?

It’s not like this is entirely new….there always used to be the I/O tune up project that came bundled with each new version of DP….and then ….I haven’t heard or thought about it for years

In my testing it is the other setting…Playback setting…. that I am having difficulty with

Using any Playback setting in 9.02 other than zero just throws the recorded audio in the above example out of whack …forgetting about MIDI here for a moment….. as one suggested in this thread he was getting results with Playback setting being the opposite of Recording Offset….but for me…once the Recording offset is set at the default number for your interface (in my case 55 samples) as far as audio Out/In goes anything other than zero in the Playback just throws the recorded audio all about….

IOW set the Rec Offset and don’t touch the Playback

So what about MIDI?….as far as I can see so far….and I’m really sorry if i’m completely askew here…. but no matter what number of samples I might punch into Playback settings the synth blip I am triggering by MIDI and recording the audio of always come back *at the same place* !! IOW the Playback settings have no effect on where the recorded audio of the MIDI triggered synth blips ends up…they always seem to be at exactly the same late offset. As dewdman42 says it’s really just a matter of either sending the MIDI out early or shifting the recorded audio early …on a per synth basis.

different MIDI interfaces combined with different hardware synths, samplers etc would seem to be a very awkward thing for there to be just one all-purpose global setting

For sure one thing I have learned is to keep all plug ins off the Master Bus….

also Auxiliary Channels have an impact…. In starting this test I recorded the Metronome Click by directly bussing it into a mono audio channel….lines up perfectly. But piping the Met. click out the same bus to an Aux Channel with that bus as its input ..and then recording that click from a Send of that Aux and the latency compensation doesn’t work

I hope this may help our mutual knowledge here….and I really hope I haven’t added to anyone’s confusion (other than my own)

I’m going to try dewdman’s flamming test….

and will try more Magic Dave tests now on 9.12.

RRoy I think you are absolutely right with external MIDI timing and changing Buffer settings in 9.1…will get back

One thing for sure is that 9.12 feels really good when playing VIs and using MIDI controllers….

I’d like to get on top of this Fine Tuning thing so will press on

Early days with the new Pre-Gen engine… it’s exciting..and challenging….hats off the guys at MOTU…it’s obviously a huge task

Thanks a lot

I’ve actually got a lot of these compensation click tests in a DP project i started a long time ago…I may clean it up and put it on DropBox if anyone’s interested…..might save you a lot of time…will try to make it as clear as I can ( : …. good luck he says!

[dewdman42]

So the reason plugins have more of an effect on latency issues is because Aux and master tracks are not tacking their input from an actual track data, they are taking their input "live" from a send somewhere. Normal audio and instrument tracks can read the pre-recorded track data a little earlier in order to make sure the resulting audio lines up with the timeline. Plugin Delay Compensation (PDC) will take the MIDI early or the recorded audio early...based on whatever plugins are on that track, so that the processed sound lines up with the timeline.

AUX tracks, however, don't have pre recorded audio, they get their audio live...so...any plugins will create latent results, unless PDC is doing what it can for aux tracks too.

I don't now what DP does, because I can't find it documented, but other DAW's I know about deal with aux track PDC by delaying all other tracks by a certain amount so that the delayed aux track processing will line up with all other tracks. I'm not sure what happens to the timeline clock counter in that situation.

Now the question is, how does all of that relate to external MIDI tracks....and the actual clock counter? In theory, the PDC changes should all be handled in a way that the clock counter exactly matches what we hear. My experience is that external MIDI tracks tend to match the clock and don't have any latency compensation of any kind happening.

Large buffer or small buffer should not matter, as long as the clock matches what we hear. Large buffers really should only effect when you are attempting to actually human perform a part through plugins...then the large buffer is going to have latency compared to your performance and there is no way around that.

Back to the discussion of audio fine tuning, the loopback test is interesting because its measuring round trip latency. We don't really know for sure if the actual clock counter exactly matches what we hear, there is not a clear way to measure that. Let's assume for a minute that the playback sound was a little late compared to the clock counter. I will argue that DP should not let that happen if the soundcard is correctly reporting its own latency, but for a moment let's assume otherwise.

Then you can do the loopback test, which measures that, and also measures some latency during record, you end up with 55 samples. So you set record offset to 55, which makes sure that the late playback gets recorded to the track 55 samples early in order to line it up for that test...but what about when you use that setting for recording vocals? Well in theory if you are hearing the metronome and music by some late amount compared to the actual clock, you will sing the part late too so that 55 sample record offset would place the recorded vocal on the right point of the timeline. However.... we don't know if you were singing late compared to the timeline or not...we don't have a measurement for that. We just know you sang it on the audible metronome and the 55 record offset places it back in time with the audible metronome.

Why does that matter? Because we don't know what the MIDI is doing. MIDI is probably following the clock exactly though I think. The playback offset would then change the audio to be a little earlier so that it won't be late compared to the clock counter. But I don't see a way to measure that in order to set it. Using an external MIDI device, as Magic Dave suggested is perhaps one way. The thing is, that test also picks up delay from the MIDI line...so it doesn't really make playback match the counter either, plus if you're using more then one MIDI device, it would simply not be accurate. And if you change the playback offset, then I would expect the record offset loopback test to be effected and need to be redone.

Probably the only way to do it properly is to make a patch cable with MIDI on one end and audio on the other that will take the MIDI out signals and record as audio noise to the input in loopback. Also do an audio loopback at the same time. The difference in latency between the audio loopback and the MIDI patched through latency would be the playback latency. That might get closer to measuring the amount of playback offset needed, assuming MTS is operating.

After that use the audio loopback alone to determine if record offset is still needed.

And personally I would presume that these tests should not be neccessary at all with simple setups with a modern soundcard and no other digital processors in the signal path. If you're using a digital mixer in non-soundcard-mode to take your inputs, for example, then you might need to do this test. But if its straight through a soundcard, I would expect that DP can determine all the calculations it needs based on the soundcard's reported latency.

After that, dealing with different MIDI device latencies should be handled on a track by track basis.

[Maxxy]

Ahhh.... as usual Magic Daaaave

I hadn't read your revised instructions carefully..... now I get it....and it works perfectly...very interesting

Well done and thanks very much

It remains the case for me however that in 9.1 and 9.12 the MIDI driven returned audio is coming back at varying delays ....ever increasing with buffer size.... whereas it exhibited very good behaviour in 9.02

No biggie....I'm sure you'll work it out... thanks again for making sense of it.... actually I have recently been dusting off and re-integrating a number of old ...and new...outboard synths into the studio.... so this info is very timely for me

I have to be careful as trying to keep track of presets, banks , sysex etc might well do my head in....but at least I can now approach it with an understanding of how to keep it all in time ...

I wonder if the Fine Tuning I/O nomenclature there couldn't be improved? ... As you point out... the Recording Offset is for fine tuning the audio return of MIDI-driven hardware synths and samplers etc....whereas the Playback Offset is for fine tuning the audio return of Audio driven outboard processors

I think I get it now....

Will think about the further implications suggested by dewdman42 soon

for the moment....I learned something great today...thank you

[dewdman42]

I wonder if the Fine Tuning I/O nomenclature there couldn't be improved? ... As you point out... the Recording Offset is for fine tuning the audio return of MIDI-driven hardware synths and samplers etc....whereas the Playback Offset is for fine tuning the audio return of Audio driven outboard processors

The way I think I am wrapping my head around it, if I even have it right...is like this...

Playback offset changes the offset of audio playback relative to the clock counter. external MIDI follows that clock. So if you offset playback later, then MIDI will be early, relative to everything you hear, including the metronome..and the clock will also be ahead of the audio, including the metronome.

Recording offset changes where on the clock timeline incoming audio is registered on the track relative to when its received by DP.

In that sense, the nomenclature is fine. Its just a little confusing to understand how this can keep external MIDI gear sounding in sync with our audio tracks.

If you set playback offset to be a bit later, then in theory what happens is that playback of audio tracks (and MIDI through instrument tracks?) will have some additional latency added (relative to the clock). That means external MIDI goes out right on time with the clock(but early relative to the audible metronome). By the time the MIDI gets to the synth and the synth makes sound...then hopefully that sound will be right in sync with other audio sounds that are coming out delayed by the playback offset, if its set to the right value. Everything heard will be heard in sync, but the actual clock counter will be a little bit ahead of what you hear.

Then by using the audio loopback test to set the record offset (after you have established that playback offset), then the late incoming audio can be offset earlier onto the track so that it will be registered where you intended it to be on the clock timeline based on hearing the metronome and other sounds when you played whatever you played..

I think this should work fine but different synths may require different offsets to make that work right. Also if you are doing any midiclock or MTC sync with other devices, might run into some oddness, since this trick is intentionally delaying all the audio from DP in order to keep it in sync with the audio heard from the outputs of your synths after MIDI finally gets there. or maybe not, I'm not sure about that right now.

I'm also not sure what happens with those settings when I try to play live through a VI. Will the audio of that VI also be delayed with the playback offset? if so, that will be undesirable latency. If the playback offset only effects pre recorded audio tracks and pre recorded MIDI tracks feeding instruments tracks, then fine. But also if I am trying to perform the MIDI part for that instrument track...and record the MIDI...will the MIDI end up late on the MIDI track since I will be listening to a late metronome and playing against it but recording the MIDI without any offset on the MIDI track?

Also, when recording external MIDI tracks, recording the MIDI part, I prefer to have the MIDI go directly to the synth from the MIDI interface...I don't want it to route through DP on its way to the synth while actually playing it. I want as little latency between hitting the key and hearing the sound as possible while performing the part. So I'm not sure right now whether those audio timing offset values would be what I want while trying to do that.. maybe.. I have to think about it some more. The problem there is that you'd be recording your MIDI part as close as possible to what you hear...which may be offset-delayed compared to the clock counter..but the MIDI part would end up recorded on the track without record offset helping it to be in the correct place. I guess it would be late. I am not setup to test this right now.

If the above is true, then in my mind, I don't want playback offset. While performing MIDI parts I would want to minimize any and all latencies so that the audible metronome matches the actual clock counter as closely as possible, and the MIDI I play will go onto the MIDI track where I intended it to go.

I know some people prefer to have the latency baked in because they will adjust their playing to the latency...and they will subconsciously rush the parts so that the sounds are on time. in that case, the MIDI part going late to the clock would be ok, but I don't work that way at all, and especially with larger latencies..

For film work, its also not clear to me what is going to happen, will the video follow the clock or follow the audible metronome? Playback offset may cause confusion that way too?

I would prefer to avoid using playback offset at all. let the playback sound audibly as close to the actual clock counter as possible.... external MIDI tracks can be sent early on a track by track basis using the time shift plugin in order that external synths will sound when the rest of audio does. Performing MIDI parts to record them will be closer to the actual clock. Latency will be minimized throughout the system.

I just don't think the clock counter should ever be different then what we are hearing or seeing on video sync. It should be exactly the same. Its not clear to me right now what is happening in that regard, with or without the audio timing offsets set to zero or something else.

[Maxxy]

This has been a really interesting post to me… dewdman42 I think you’ve woken a number of us up about how to think about DP and the way it works under the hood… to be able to get better timing….

The way I see that part of the Getting Started manual….about Fine Tuning In/Out… is that it sort of indicates that it’s a good thing and to use it if you need it …..or leave it…. unless there’s a problem…. in which case please do use it…..just keep adjusting till it’s right………hmmmm……not sure if I really want to do that this month Clearly it isn’t …errrr…. clear

Magic Dave’s recent explanation was a revelation to me….. and IMO should go into the manual

I guess to date we have just worked around it using our best approaches to getting the MIDI played live into DP with good feel….. using methods and techniques …. like lower buffers , testing latencies of controllers, and giving the hairy eyeball to all those tiny errant ticks…. eg. Shift-L earlier by 7 ticks in the MIDI track itself and then another judgement of its partner audio track (recorded simultaneously or played back in later) ….. dig right in to determine the amount of shift with each new track/instrument ….. and all in order so the MIDI sits properly in the MIDI track…..and the MIDI -generated audio also has its rightful place in the sun.

One great trick I learned here is to simply switch to an alternative plug-in free Mix for tracking…….switching back to the plug-in and FX laden evolving mix once the tracking has been done

These are methods that work in our DAW….

I recently installed an iConnect MIDi4+ as the central hub of all my outboard stuff…. four MIDI synths fully plugged in and a number of USB controllers, sequencers and modern USB synths through a powered hub, ….it has 3 computer ports so that the three computers or iPads can access any part of the MIDI /audio setup and cross-communicate…. so far it’s been fun…I don’t believe it uses MIDI time stamping as such

It has had me thinking about IAC MIDI delay , and outboard MIDI and audio timing issues that you have brought up with this great post dewdman42

As I see it now, the key to thinking about this is to understand that “Fine tuning I/O” allows DP to get everything lined up correctly when sending MIDI or audio out of the box to be recorded back in again

The first and main job of DP in this sense is that DP does its whole comprehensive “internal* compensation ….ADC, PDC, VIs, for all audio tracks, auxes, plug-ins, virtual instruments, side-chains etc and plays all that sample-accurately calibrated and in sync…to the Master Output. DP knows how many samples each element needs to be adjusted in order for all the various output stages and the Master Output to be in sync …..nice trick. It does all this automatically and there is nothing in particular we need to do about this

Second job , now requiring Fine Tune, is if we need to record the output of an outboard synth being triggered by an internal *MIDI* track . In this case we have to give it an offset so that it sends the MIDI out *a little earlier* than the Master (totally compensated) Output….and from the adjustment we make to the “Record Offset” the outboard synths can now play in sync with the DP Master Output

Third job, also requiring Fine Tune, is that we need to compensate for any *audio* that is heading for outboard processing …it is not going directly to DPs Master Output…..similarly to MIDI this audio from the DAW also needs to be adjusted in order to leave the box *earlier* so as to be heard by us through the outboard processor at the same time as the DP Master Output…. that is a different calculation to the Record Offset….no MIDI involved….this is the “Playback Offset”. Question … Will this be needed if the Hardware Insert plug-in is used?

Your point about the vocal I think is that she is simply singing in time with the music she hears ….DP knows how late it is sending that to her….and it then automatically compensates so that her voice is in the right place….in this case the size of the buffer settings is not important as DP simply makes the adjustment automatically

We know that playing a keyboard live into DP will benefit from using the lowest buffer settings we are able…..the closer we can get the buffer to zero the snappier both VIs and outboard synths will sound… whether it is a VI or an outboard synth

I guess it then becomes a question of MIDI Patch Thru settings… let’s not go there right now

…. perhaps it should be called “Fine-tuning Recording of Outboard Audio Processors and MIDI devices when Sent from Internal Audio and MIDI tracks”….bit of a mouthful though! ( :

I hope I’m not too far off the mark here… like you, I do find it confusing…but am thankful for Magic Dave’s explanation

I like your idea of having a MIDI plug-in to handle this function … I don’t see how one setting in Fine Tuning can be used for all situations and devices