Quantcast
midi in PT 6.0 / OSX - Gearspace.com
The No.1 Website for Pro Audio
midi in PT 6.0 / OSX
Old 2nd April 2003
  #1
Lives for gear
 
C.Lambrechts's Avatar
 
🎧 15 years
midi in PT 6.0 / OSX

been fiddeling around / testing / experimenting with the midi in PT 6.0 today. In combination with the midi I/O digidesign claims to offer 'sample accurate' midi timing ..... had to see that myself and here's the result :


- it is not sample accurate .... but my guess is that's no longer digi's fault .... but the midi protocol.


I tested several synths today ... both soft and hardware.

One of my main concerns with midi timing in PT was that if one recorded the same midi track to audio several times .... the timing would never be the same ... In the past I've done tests and that showed differences up to like 80 samples of difference between the different audio tracks recorded from the same midi file with the same midi device.
Another big issue (for me at least) was that in time ... the delay would NOT be consistent.

For example ... I know that my K2500R has 240 samples of delay when I record a midi track to audio (including of course AD) . Now in theory this would mean that in time EVERY audio attack would start 240 samples later then the according midi note. Under OS 9 / PT 5.x this was far from the case. It could be anything from say 100 samples to 350 samples and different for almost every midi note.


This is where the real improvement for midi timing in PT6.0 is to be found. It is consistent all the way and sample accurate.

Say when recording a midi track in PT 6.0 the first audio attack starts 240 samples later then it's midi note .... then all following audio attacks on that track will be exactly 240 samples later then their midi notes.

This makes midi track offset finally a usefull feature. Depending on the device and the sound / sample one can easy determine it's delay/latency and set the offset right.


Recording the same midi track up to 10 times in PT6.0 / OSX showed that the maximum difference I found between the track was 7 / 8 samples. I'd say that's pretty tight. AND consistent over the whole track.


Old 2nd April 2003
  #2
Founder
 
Jules's Avatar
Thats good news for you dance music folks then ain't it!

Old 4th April 2003
  #3
Gear Maniac
 
🎧 15 years
Midi delay

That's great news to my ears Chris!

I still used Logic synced to Pro Tools 5.1 using MTC for all my Midi programming because of the random timing fluctuations in Pro Tools... drove me crazy!

Now I can ditch the IMO unintuitive Logic and solely use Pro Tools.
This will also free up my PC for standalone stuff like Reaktor/Battery, etc.

I'm running PTs 6.0.1 in OS 10.2.4 - lovin it! Haven't experimented much with midi yet - just setup all my devices and tested they all work properly. Mostly I've been testing out the new Beat Detective features... the Groove analysis seems to work extremely well when applying to audio.

Burt
Old 5th April 2003
  #4
Gear Maniac
 
🎧 15 years
Midi delay

Hi Chris

just done a quick test of my own.

OS 10.2.4
Pro Tools 6.0.1
Emagic amt8
Waldorf Midi synth
Pro-Midi cables

setup a percussive stab with 0 Attack setting.

Pencilled in a single midi note and recorded 4 passes to
seperate audio tracks.

Delay from note ON to audio zero crossing start point:

325 samples
358 samples
283 samples
336 samples

hmmm... amt8 obviously does not = sample accurate and consistent Midi delay in Pro Tools 6.

Maybe I should trade for a Digi Midi I/O?

Burt
Old 6th April 2003
  #5
Lives for gear
 
C.Lambrechts's Avatar
 
🎧 15 years
Re: Midi delay

Quote:
Originally posted by Burt

325 samples
358 samples
283 samples
336 samples

hmmm... amt8 obviously does not = sample accurate and consistent Midi delay in Pro Tools 6.

Maybe I should trade for a Digi Midi I/O?

Burt

with the midi I/O this would result in for example :


325
327
320
326
325

samples that is .....
Old 6th April 2003
  #6
Gear Maniac
 
🎧 15 years
Midi delay

those figures surely look a lot more manageable.

To be honest, I generally tighten all audio to sample accuracy
using Beat Detective, but it would be good to have Midi being as tight as possible before printing to audio....

anybody wanna buy an amt8?

Burt
Old 15th April 2003
  #7
Lives for gear
 
Renie's Avatar
 
🎧 15 years
Re: midi in PT 6.0 / OSX

Quote:
Originally posted by C.Lambrechts

- it is not sample accurate .... but my guess is that's no longer digi's fault .... but the midi protocol.


Recording the same midi track up to 10 times in PT6.0 / OSX showed that the maximum difference I found between the track was 7 / 8 samples. I'd say that's pretty tight. AND consistent over the whole track.


Hey Chris and Burt

Thanks for the reports. I was going to get the AMT8 but I'll get the midi I/O now.

Chris, Do you think the midi protocol will improve so that we get true sample accuracy?

thanks,
Old 15th April 2003
  #8
Lives for gear
 
C.Lambrechts's Avatar
 
🎧 15 years
Re: Re: midi in PT 6.0 / OSX

Quote:
Originally posted by Renie
Hey Chris and Burt

Thanks for the reports. I was going to get the AMT8 but I'll get the midi I/O now.

Chris, Do you think the midi protocol will improve so that we get true sample accuracy?

thanks,
not very likely ... the timing is allready sample accurate now. Once it leaves ProTools it is out of it's hands ... Once it leaves the computer it is and will allways be MIDI. What the midi I/O does is make sure that all midi information leaves on all ports with sample accurate timing. From there on it is dependend on the module / synth to reproduce the sound or whatever midi information it recieves and translate it it into for example a sound.

But it is a MAJOR improvement to what it was before. Logic's AMT (and maybe others too) allready had a similar way of doing that.
Old 16th April 2003
  #9
Gear Addict
 
🎧 15 years
I did this test today and had different results. Using a Roland TD10 right back into Protools I got variances of 40 samples over a 8 bar section in comparison to the grid. This is no where near the 10 sample results I used to get with an opcode studio3 using 2 serial ports.

I think the problem is still in the USB timing.

So far, I'm not a believer in MIDI/IO. Then again, "sample acurate timing" is still an "upcoming feature" according to the Digi web site.

I don't know why I didn't think of this....What results would we get if we bounced a midi track. Midi out to midi in? I'll test that in the morning and report back.
Old 17th April 2003
  #10
Gear Addict
 
🎧 15 years
Well...... I did that test today and got up to 42 sample variances over 20 bars (clicking 1/8th notes. What's worse, repeating the test with two simultanious notes output, the timing was even more on the back side. Looks to me like at this point MIDI/IO is just another USB midi interface. Lets hope new software improves this.

I wonder if the midi timing on the 002 is any better? MLAN?

For what it's worth, I realized today that throughput on a 192 at 48k is a wopping 104 samples. Up from 60samples with an 888. Yikes!

With latency like that, 40 sample drift on a midi click track will still be closer than the 52 sample delay between when audio was played, and when it was recorded. hmmmmmm
Old 17th April 2003
  #11
Lives for gear
 
🎧 15 years
Guys,

Excellent research on your part. Most people never scientifically quantify this issue to this degree.

I did much of the initial testing of AMT for Emagic back during the development phase of Logic 4. Sadly, Midi is a quite dated 32Kb serial protocol that has severe inherent timing limitations. Worse than most people know. It takes approximately .6ms just to get the 3 Bytes of a single Midi Note On down the pipeline. These tests you're running are about single note timing repeatability, which is a valid test in itself, but the ugly truth is, once you start playing music with all the notes interacting, it gets much uglier.

Running Status can lessen the time required for subsequent notes to be transmitted, as it assumes subsequent Midi messages are of the same type until told otherwise, allowing the first byte of subsequent notes (that define the type of Midi message) to be dropped. Yes, it's sad that the friggin' thing is so slow, dropping a single byte really helps.

Your tests seem to all be single notes. If you want to really weep, make it 10 notes, one for each finger. Without Running Status enabled, 10 Midi notes to the same instrument will have about a 7-8ms spread, best case (assuming no controllers). With Running Status, more like 5ms. That's as good as it gets for Midi.

As you can infer, even with AMT or like technology enabled, truly sample accurate is a BS claim for any external Midi use. The best I've seen AMT do is about +/- .3ms average deviation from pass to pass.

Burt, your 4 passes had a max deviation of 75 samples, which is a little less than +/- 1ms (about a 2ms total window), and the average would be about +/ .6-7ms. Chris's 7-8 samples is by far the best Midi performance I've heard of, but again, once multiple notes are cranking down the line, the inherent slop in Midi will far exceed that spec.

For reference, prior to AMT a few years back, the best Midi timing I ever tested on a Mac was about +/- 2.5ms. Many were +/- 3-5ms, believe it or not. The Atari was better than that by an order of magnitude, God rest it's soul. That also explained why people who wanted to groove back then used an MPC instead of a Mac. It really was better, using internal timing protocols instead of a Midi cable for the drums.

We've come a long way.


Regards,
Brian T
Old 17th April 2003
  #12
Lives for gear
 
🎧 15 years
Quote:
Originally posted by OKden

For what it's worth, I realized today that throughput on a 192 at 48k is a wopping 104 samples. Up from 60samples with an 888. Yikes!


Wow. That's a big jump. By the time you get routed, bussed and Channel Strip or whatever EQ/comp plugin going, what's the total A/D in to A/D out through the system to the singer?

Sounds like it will be pushing 3ms+ on a 192 at 44.1K, with decent plugins. Hmmm.

Why is it worse? Is that why it sounds better?


Regards,
Brian T
Old 21st April 2003
  #13
Gear Maniac
 
🎧 15 years
Midi

Thx for the wisdom Brian....

Chris - now perhaps you understand why I've used Beat Detective on all of my tracks since its introduction! Its the ideal solution to tighten up those slack midi triggered drums.

Burt
Old 21st April 2003
  #14
Lives for gear
 
C.Lambrechts's Avatar
 
🎧 15 years
Quote:
Originally posted by BrianT
Wow. That's a big jump. By the time you get routed, bussed and Channel Strip or whatever EQ/comp plugin going, what's the total A/D in to A/D out through the system to the singer?

Sounds like it will be pushing 3ms+ on a 192 at 44.1K, with decent plugins. Hmmm.

Why is it worse? Is that why it sounds better?


Regards,
Brian T

most ADDA convertors do like 3 ms to ADDA, Apogee / Prism .... all in the same ballfield there. I don't think one will see very much of improvement there in the near future. Seems like that's what good convertors need. and BTW ... this counts for ANY DAW out there ... not only PT


192 has 103 samples when doing DA - AD .... plugins can be compensated for. That means that your singer is going to hear the music on your disks an 'WOPPING' 52 samples late and base her performance on that. That is like 1. something MILLISENCONDS.

Don't get me wrong here ... any picosecond of delay is imprtant but please also realise the relative importance of say 1.3 MS.

Even after her track coming back into PT it will be less then 3MS late. Try nudging the track 3 MS back and listen to the performance. Quite frankly most people I know won't here the difference. BUT that doesn't mean that one shouldn't be aware of it ... I often nudge back overdubs .... because it gives me a better feeling ... not because in blind tests I'm able to tell the difference.

Also ... try looking at multiple overdubs from a same person on the same track .... Those are FAR MORE apart then 3 ms ... even if you're the best and tightest drummer in the world. So 3 ms ADDA is a 'relative' unimportant issue when recording to daw. BUT again ... don't shoot me for saying this because I too am a nudger in some / a lot of cases. I just felt like going into it a little bit because using words like 'WOPPING' can cause confusion as to how bad it realy is .... on the contrary ... it's pretty good.
Old 21st April 2003
  #15
Lives for gear
 
C.Lambrechts's Avatar
 
🎧 15 years
Re: Midi

Quote:
Originally posted by Burt
Thx for the wisdom Brian....

Chris - now perhaps you understand why I've used Beat Detective on all of my tracks since its introduction! Its the ideal solution to tighten up those slack midi triggered drums.

Burt
Old 22nd April 2003
  #16
Lives for gear
 
🎧 15 years
Chris,

Correct me if I'm wrong, but in this case the singer will hear their own voice 3ms or more late while singing, depending upon the specific plugin chain chosen. Plugin comp and nudging audio will not affect the realtime throughput latency issue.

3ms may seem pretty quick, but I've become convinced that every ms counts in this case, because of the comb filter effect in headphones from the delay, and it's potential effect on a singer's pitch from case to case.

Have you ever had a singer with a pitch problem and improved the situation by merely changing the type of headphone? In a similar way, I believe the throughput latency can really affect some people. Others seem impervious. Those pesky humans are so unpredictable, aren't they?

One reason I expressed surprise is that the empirically measured latency in my system, from A/D into a channel, through EQ and compression, bussed to the stereo bus and output from D/A is ~1.3ms, realtime. Less than half the total latency, from a system designed 6 or 7 years ago.

That's just not what I would have expected. I still don't understand why the 192 latency is about double that of an 888. Anybody?


Regards,
Brian T
Old 22nd April 2003
  #17
Lives for gear
 
C.Lambrechts's Avatar
 
🎧 15 years
Quote:
Originally posted by BrianT
Chris,

Correct me if I'm wrong, but in this case the singer will hear their own voice 3ms or more late while singing, depending upon the specific plugin chain chosen. Plugin comp and nudging audio will not affect the realtime throughput latency issue.

correct. If you want to have the plugin's allready on the vocal track ... that adds to the latency. Depending on what plugins you use that can be a bother. I never record with plugins on the track I record to.

Quote:
Originally posted by BrianT
I still don't understand why the 192 latency is about double that of an 888. Anybody?


Regards,
Brian T [/B]
Convertor quality. All the high end convertors I know have about the same latency ... Apogee / Prism ...
Old 22nd April 2003
  #18
Founder
 
Jules's Avatar
Yes.... high end converters work on the conversion longer and dont send it out untill it is done properly and given a time stamped, 'inspected by' code. Most singers find the extra wait for their vocals in headphones a worthwhile trade off for the quality boost. Sting for example uses yoga as a calming exercise whilst waiting for his and Steven Tyler of Aerosmith had MANY years of doing top quality cocaine. (for example)
Old 24th April 2003
  #19
Here for the gear
 
hkrabye's Avatar
 
🎧 15 years
Re: midi in PT 6.0 / OSX

It's interesting that I found this thread, because I have done exactly the same thing as the first post here explains.

I tested the Core-MIDI (which never leaves the Mac) by playing three percussion tracks using the Indigo plug-in, and then bounced the mix to my harddrive two times. I then loaded the audio files into ProTools with the identical start point for the two. When I play the audio files back with identical start times (all MIDI is muted), I hear that they were recorded with small differences in when the attacks actually came. There are variations in the timing (for the two bounced files) between 20 and 100 samples, and this delay is enough to create a phase problem. Example: If the attack of my kickdrum is not hitting with the attack exactly at the same point as my sharp bass guitar sound, the two sounds will partly phase each other out. (The peak of the low frequency wave for the kick drum will occur when the buttom of the low frequency for the bass gitar sound occurs, and the result is a reduced level.)

I will test the MIDI delay by using my MOTU microExpress MIDI interface soon. I guess I will have to add a couple of hundred samples in delay?

By the way, the MOTU microExpress is supporting MIDI time stamping, but how do I know that the time stamping is active when recording and playing back? There is no info in the MOTU user manual (only for OS9). Does someone on this list know?

Helge
πŸ“ Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 295 views: 72253
Avatar for anguswoodhead
anguswoodhead 26th March 2013
replies: 15929 views: 1528483
Avatar for Ragan
Ragan 11th January 2019
replies: 1296 views: 178247
Avatar for heraldo_jones
heraldo_jones 1st February 2016
Post Reply

Welcome to the Gearspace Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…

Forum Jump
Forum Jump