Midi accuracy, midi jitter.

[lagoausente]

Hello, I have searched on this on the past. When record midi tracks on Cubase the timing is not accurate. Some google searching and I found a web (don´t remember what one) that described the reason, midi timing is taken from the midi driver looking into the mainboard internal clock. Result it that if you record some midi events and quantize them, than send them to the physical midi-out, then connect midi-out to midi-in and re-record it, will find that each time you re-record the same thing the events timing will be a little diferent each time, that´s called midi-jitter.
I´m guessing now if there is some sequencer on Scope that can record midi-tracks something like VDAT but with midi. Does it exist?
thanks a lot.

[garyb]

no. have you tried the "use system timestamp" option in Cubase?

[lagoausente]

no. have you tried the "use system timestamp" option in Cubase?

Yes, there was some tricks, combination of the correct driver (windows or native) with the timestamp checked/unchecked helped to avoid the big latencies. There was a tool out there to find out the correct option for your pc. That corrected big latencies but still total timing acuracy was not posible so I decided to record directly Scope audio output instruments instead of midi tracks, much more realistic.
I was guessing about SC can do some type of midi recorder or some type of midi-driver that sync the Asio and Midi events in the DSPs for a perfect timestamp, instead of be done for windows.

[garyb]

yeah, midi's not that tight anyway.
for most music, it's not an issue, but some kinds of electronic music using stacked low freq instruments might have some comb filtering...

anyway, recording the instruments as audio is the correct thing to do, imo, in any case. i'd only leave instruments as midi for as long as it takes to solidify the arrangement, then for mixing, i usually prefer to have recorded all the instruments as audio. this leaves more resources available for mixing and usually makes the tracks sound more integrated.

[dawman]

I have tried 3 different DAWs for MIDI and all 3 have been disappointing compared to a hardware sequencer.
As GaryB said though its not really important for most folks.
But if you are using lighting consoles faders, you want the recorded movement stored and recalled with accuracy. Akai MPC's, and most hardware sequencers will do this effortlessly.

But this is on 300mm motorized faders where 960+ ppq is desirable.
Cubase jitters badly, Reaper is smoother with slight jitter, the Akai MPC is one solid quantized motion which actually outperforms a human hand. This is a preferred method as now the Lighting assistants have absolutely no reasons to miss their spotlight cues.......

[xite knight]

+1 on the MPC

When midi-sequencing internal sounds, my MPC-4000 will - under ideal circumstances in my test setup - spit out audio events on the grid +/- 1 or 2 samples. That's an occasional deviation off the grid by 2 samples divided by 44100 samples per sec = approx. 0.00005 sec deviation off the grid.

That's a tight spot!

I agree on this accuracy not being required by the majority, but If anyone wants it even tighter than that, it may be possible with an analog sequencer but that would contradict with the midi theme of this thread

[lagoausente]

mmm, the Akai MPC is too much extra $$ to spend for just record midi tracks, it seems...
I have not tested it yet, but... if record a Native sampler audio output will have the same jitter as on the midi tracks? Have you tested that?
I have tested with the STS, recording it´s output, but since is on DSP the STS output is solid, no midi jitter, no variations, but I guess what will happen with a Native one...

[jksuperstar]

To reduce jitter between computer and hardware, avoid using the computer to generate the MIDI Clock. Always use hardware. The system clock is subject to interrupt latencies, so even if the clock itself is tight, it takes a while for the system to react to it. Hrdware going into a computer has latencies, but that jitter isn't cummulative over time. I have Ableton locked to an external clock (Yamaha QY-70, but should become the XITE soon), and it's rock solid. The display will show +/- .10 bpm, but it's far better than Ableton thinking it's rock solid, but the actual clock or note output being off (which if I monitor Ableton's clock by a hardware device, the best I'll see is +/- 3-5bpm,).

Also, 1 single midi note-on takes 1ms to send (.001sec). So sending chords over MIDI, no matter how tight the clock is, could take 3ms, or ~64 samples @ 44.1kHz. Nothing you can do, really.

[xite knight]

To Lagoausente:

My post was more like a praise of the MPC accuracy than anything else.

But as a comment on your initial post, I think personally there's at least 4 main reasons for the inaccuracy that you experience:
1. As you mentioned, the sequencer software is depending on the clock of non-dedicated native processing.
2. The midi accuracy capability of modern Windows versions is not suitable for my nervous system.
3. The midi accuracy of a physical midi port controlled by a Windows midi driver? See my 2nd bullet point.
4. The physical mid spec/bandwith itself which is almost 30 years old

Regarding your question about Cubase sequencing a native sampler, no I haven't tested that.

But I guess you would get better timing out of that when compared to Cubase sequencing a sampler via the physical midi port due to my 3. and 4. bullet points.

As a final public disclaimer, this is an attempt to help Lagoausente on the basis of my humble experience. It is neither an inquisition against native DSP/sw sequencers nor an invitiation to start a flame war between one faith against the other. Why? See my 2nd bullet point.

[Bud Weiser]

I have tried 3 different DAWs for MIDI and all 3 have been disappointing compared to a hardware sequencer.

Which hardware sequencer ? QX-1 ?
If yes, I agree,- it´s the tightest.

Cubase jitters badly, Reaper is smoother with slight jitter

Well, I have solid MIDI timing w/ Cubase SX3, Midex-8, driver 1.92 and USB /ASUS P5WD2 Premium

I also have solid MIDI timing w/ Reaper 3.x, MusicQuest 8portSE and Earthvegaconnections DirectMusic driver for WinXP and conneted to the LTP port / ASUS P4T-E

I´m using System Timestamp in both cases.
System Time Stamp uses QPT (Query Performance Timer) clock of the MoBo.
So it´s neccessary to find MoBos w/ this clock being stable to use System Time Stamp.

Other MIDI interfaces like Unitor-8(mkII), AMT and AFAIK also these of RME cards use TGT clock which is less accurate because there´s no System Time Stamp.
That doesn´t mean these interfaces don´t work or are crap.
There are also users having stable MIDI timing w/ Cubase and AMT type interfaces or the RME MIDI ports.
So I think it highly depends on the system configuration and tweaks.

Cannot speak for Win 7

I only use the SCOPE cards MIDI ports for SCOPE devices which is perfect,- so my machine using a SCOPE card has a 2nd MIDI interface connected for the additional native software like VST and Reason, works like a charme.

Bud

[Andy_F]

Ive heard people say older setups like Cubase on the Atari is better than modern Cubase on WinXP for midi timing.
Makes sense as the OS has to handle much more now.

[lagoausente]

To reduce jitter between computer and hardware, avoid using the computer to generate the MIDI Clock. Always use hardware. The system clock is subject to interrupt latencies, so even if the clock itself is tight, it takes a while for the system to react to it. Hrdware going into a computer has latencies, but that jitter isn't cummulative over time. I have Ableton locked to an external clock (Yamaha QY-70, but should become the XITE soon), and it's rock solid. The display will show +/- .10 bpm, but it's far better than Ableton thinking it's rock solid, but the actual clock or note output being off (which if I monitor Ableton's clock by a hardware device, the best I'll see is +/- 3-5bpm,).

Also, 1 single midi note-on takes 1ms to send (.001sec). So sending chords over MIDI, no matter how tight the clock is, could take 3ms, or ~64 samples @ 44.1kHz. Nothing you can do, really.

64 samples is not a problem on their own. Problem is if on time is 64 samples and next time is 240 samples, or just randomly.
I didn´t understand well your setup. If you sync Ableton to the Yamaha Midi clock, do that means that the position of the event will be told by the Yamaha? Are you sure of that? I don´t know in depth how the midi clock works, but if your setup works, it seems a solution. Can you detalied a little more how the Yamaha midi clock helps Ableton on midi tracks?
I´ll check recording the sampler audio output as soon as posible and compare the audio events with the midi events.
Regarding xite knight points:
Point 4 seems not the problem itself, if you record STS audio output and measure it, it´s rock solid, no variations.
jksuperstar point seems interesting.
I´ll try to do the Native test this weekend.

[Bud Weiser]

Ive heard people say older setups like Cubase on the Atari is better than modern Cubase on WinXP for midi timing.
Makes sense as the OS has to handle much more now.

MIDItiming on Atari is the best up to now, but also on Atari, Notator´s MIDItiming was better than Cubase.
Beside my DAWs, I still have Atari´s running Cubase and Notator.

I find Notator SL w/ Log3 and Unitor2 still good as a MIDI front end for hardware devices as well as Scope instruments.
But I don´t use that setup much since years, it´s just there, I rarely sell stuff and it´s so pretty nostalgic

Bud

[jksuperstar]

First, if all of the MIDI and Audio is done completely NATIVE within a DAW, then there is no problem. Any jitter is absorbed by the audio driver's buffers, since the audio hardware runs on it's own clock, which is rock solid for this discussion (we don't need to talk about jitter in audio samples).

However, the problems start showing up when you sync a computer (either OSX or Windows, or Linux for that matter) with hardware. If the computer is the clock, the time it takes for the system to respond to an interrupt for the timer, process the timer, and then schedule a MIDI output can vary from one clock to the next. So if you connect a hardware device to the computer, and it locks to the computers clock, you'll see the clock jumping all over the place (+/- 5 bpm @ 120bpm).

If I press play on Ableton, it sends midi data out, and audio, and there is some noticeable jitter when I use hardware effects such as delay locked to that MIDI clock.

In my case, I use the QY-70 sending clock not only into the DAW, but also other hardware (and other computers sometimes). I set up Ableton to lock to an external clock (the QY-70) via MIDI input, and the jitter Ableton reports is +/- .1 bpm (at 120bpm). The other hardware I have connected (nord G2), doesn't report any jitter. The QY-70 has a very high PPQN clock, 480 if I remember correctly. Some DAW sequencers say they are accurate to 960, but the jitter in the system clock far outweighs this theoretical number in my experience.

When I hit play on the QY-70, Ableton automatically starts playing. My delays are now locked solid, and percussive elements from both hardware and ableton feel like they are very tightly synchronized.

Also, to reduce jitter, if you have a multi-channel MIDI interface, dedicate one port to clock if you can. This reduces the chance of jitter on the serial MIDI line.

[spacef]

>>>>>
MIDItiming on Atari is the best up to now, but also on Atari, Notator´s MIDItiming was better than Cubase.
>>>>>

I used to think the same but when I recorded the atari songs to cubase (SX2), i didn't think the same anymore.

I also recorded songs from the EPS16+ sequencer into cubase... the timing is awful (you can talk about random quantize, lol) , but back in the days, it was just fine and it wasn't a problem at all when you listen to the song. I had a techno listening session witha friend, tracks from 1990-1993.. I felt they are 200 times more lively than what you hear nowadays... in that time, nobody gave a damn of those few samples unprecisions, as no-one really knew about them (and anyway, what would you have done? cut tapes? impossible) and it is probably what makes the interest of those tracks nowadays. i have lots of friends in trance that make sample precise music, it is kind of boring after a while. i don't listen this style of music mainly because of this....

my 2 cents on the cubase vs atari timing thing,
it is not a "jitter and latency is good" point of view at all

[astroman]

to that I subscribe, Mehdi - you just made my day... eh night

cheers, Tom

[dawman]

MIDI Schmiddy........
Great info there Bud..

The reason I like hardware still is simple. I dont need to run 76 MAC 600 washlights, 18 MAC 2000 Profiles, 12 MAC 300 washlights and 4 MAC 500 profile spots from 300mm faders, but the stages I work on have this stuff, and they still prefer a hardware SMF Recorder, or in some cases Reaper. It's 960+ ppq is incredibly stable for whatever reason, and since these guys don't give a hoot about the scientific data, just the results, it does make one wonder.

But the main reason I still prefer hardware is because there's nothing lovlier, fatter and thicker sounding than digitally controlled, Analog synths and consoles.
My SE-1's are incredible when I can insert the fader movement, and then go back to quantize it and align its start and stop points.

I really long for a giant knob for live work though. I hate having to sacrifice improvisational skills so some clowns can dance to my filter sweeps. But wtf, Im making bread.

I am working on getting this new controller though. I think it will handle any imperfect movements pretty well.

http://createdigitalmusic.com/2010/10/a ... -hardware/

[t_tangent]

I really long for a giant knob for live work though. I hate having to sacrifice improvisational skills so some clowns can dance to my filter sweeps. But wtf, Im making bread.

I am working on getting this new controller though. I think it will handle any imperfect movements pretty well.

http://createdigitalmusic.com/2010/10/a ... -hardware/

Haha....loved the link...cheers

[Bud Weiser]

I really long for a giant knob for live work though. I hate having to sacrifice improvisational skills so some clowns can dance to my filter sweeps. But wtf, Im making bread.

I am working on getting this new controller though. I think it will handle any imperfect movements pretty well.

http://createdigitalmusic.com/2010/10/a ... -hardware/

Haha....loved the link...cheers

He, he, he,- me too !!!

Bud

[lagoausente]

To reduce jitter between computer and hardware, avoid using the computer to generate the MIDI Clock. Always use hardware. The system clock is subject to interrupt latencies, so even if the clock itself is tight, it takes a while for the system to react to it. Hrdware going into a computer has latencies, but that jitter isn't cummulative over time. I have Ableton locked to an external clock (Yamaha QY-70, but should become the XITE soon), and it's rock solid. The display will show +/- .10 bpm, but it's far better than Ableton thinking it's rock solid, but the actual clock or note output being off (which if I monitor Ableton's clock by a hardware device, the best I'll see is +/- 3-5bpm,).

Also, 1 single midi note-on takes 1ms to send (.001sec). So sending chords over MIDI, no matter how tight the clock is, could take 3ms, or ~64 samples @ 44.1kHz. Nothing you can do, really.

I have just bought a Qy70 on ebay. For what I have reading 480 ppq is good enough, but I have a doubt. If sync Cubase to QY70 midi clock, do that mean the driver will take the hardware clock instead of the Mobo? If so, that should mean that I should have same accuracy recording teh midi track on Cubase than recording on the own QY70..., or I´m misunderstanding the issue? What I ´mean is that is Cubase is synced externally, midi track event positions will be determined by the QY70 clock? .., but still, windows will have delays to manage that incoming clock or not?
Or just the best way is simply use the QY70 to record the midi tracks?