The No.1 Website for Pro Audio
Free software measures MIDI latency and jitter.
Old 14th September 2015
  #151
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by Don Solaris View Post
Nice! Any chance to repeat the test at 140bpm, 1/16ths, 1minute audio?
I recorded around 3 mins and got this..

MIDI note on jitter 0.028ms (max 0.11ms)

Attached Thumbnails
Free software measures MIDI latency and jitter.-16ths-140bpm-live-9-um880.png  
Old 14th September 2015
  #152
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by Don Solaris;11332303But below 1ms in your result kinda shocked me. Seems like the mac OS is taking full control of the USB port and doesn't let other tasks to interrupt it at all.

[url
https://www.gearslutz.com/board/11229657-post9.html[/url]
OSX Core audio has a feature called MIDI time stamping. From what I've read MTS requires the MIDI interface drivers to support it, firmware within the MIDI interface to support it, and also for the DAW to support it by passing time stamps ahead to the interface.
It is not solely OS dependant.

Last edited by Jauqq; 14th September 2015 at 05:33 PM..
Old 14th September 2015
  #153
Quote:
Originally Posted by Jauqq View Post
I recorded around 3 mins and got this..

MIDI note on jitter 0.028ms (max 0.11ms)

Still better than a PC with E-MU 1820m.
Old 14th September 2015
  #154
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by Don Solaris View Post
Still better than a PC with E-MU 1820m.
Its 28 microseconds RMS with a max of 0.11ms.

The UM880 does this from all of its 8 MIDI out ports without latency being increased from port to port.

You send 8 MIDI tracks to each of the 8 MIDI out ports and they all send data together......
If using Live 9 that is ....
Old 14th September 2015
  #155
Lives for gear
 

@ Jauqq

Nice work!

Incidentally, Ableton use MIDI to audio cables to test latency in Live. They mention it in the manual somewhere.

Live 9's a real CPU hog for me ATM though.
Old 14th September 2015
  #156
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by DanRand View Post
@ Jauqq

Nice work!

Incidentally, Ableton use MIDI to audio cables to test latency in Live. They mention it in the manual somewhere.

Live 9's a real CPU hog for me ATM though.
The info about Live's MIDI timing is here:
Ableton Live timing

"In general, timestamps are an extremely reliable mechanism for dealing with MIDI event timing. But timestamps are only applicable to data within the computer itself. MIDI data outside of the computer can make no use of this information, and so timing information coming from or going to external hardware is processed by the hardware as soon as it arrives, rather than according to a schedule.

Even within the computer, the accuracy of timestamps can vary widely, depending on the quality of the MIDI hardware, errors in driver programming, etc. Live must assume that any timestamps attached to incoming MIDI events are accurate, and that outgoing events will be dealt with appropriately by any external hardware. But both situations are impossible for Live to verify.
"

"OS X:
Interface C: In all tests, the maximum jitter was +/- 1 ms, with most events occurring with no jitter.
"

They never mention what "interface C" is......
Maybe I've found it ?




But I'm more of a linear sequencer guy so wish Logic X would perform the same.
Old 14th September 2015
  #157
Lives for gear
 
Jauqq's Avatar
 

MIDI Clock test:

Live 9, Edirol UM880, MIDI clock @ 120bpm for 124 bars, three tests.

1) MIDI clock jitter 0.03ms RMS (max 0.28ms)
2) MIDI clock jitter 0.029ms RMS (max 0.39ms)
3) MIDI clock jitter 0.03ms RMS (max 0.42ms)

No push/pull cycle with drops in the jitter wave.
Note the histograms which show little or no jitter, with MIDI clock being very tight.

Again another good performance with the Edirol UM880 + Live 9 combo, however its MIDI clock output does have slightly higher max figures than its MIDI messages note output.
Attached Thumbnails
Free software measures MIDI latency and jitter.-1-live-9.png   Free software measures MIDI latency and jitter.-2-live-9.png   Free software measures MIDI latency and jitter.-3-live-9.png  

Last edited by Jauqq; 14th September 2015 at 09:14 PM..
Old 14th September 2015
  #158
Lives for gear
 
Jauqq's Avatar
 

Same interface as before, Edirol UM880 but this time testing MIDI clock in Logic X 10.2.

Logic X 10.2, Edirol UM880, MIDI clock @ 120bpm for 124 bars, three tests.

1) MIDI clock jitter 0.4ms RMS (max 0.87ms)
2) MIDI clock jitter 0.39ms RMS (max 1.35ms)
3) MIDI clock jitter 0.35ms RMS (max 1.03ms)


This shows Logic X 10.2 has higher jitter for MIDI clock (as well as MIDI messages) than Live 9 even with the same MIDI interface.

Logic X also display as push/ pull cycle with large shifts in timing (the above three tests jitter waveforms over 124 bars shown in the attachment)

Also histograms for the above tests shown. Note the jitter !!
Attached Thumbnails
Free software measures MIDI latency and jitter.-logic-x-midi-clock-jitter-waves.jpg   Free software measures MIDI latency and jitter.-1-logic-x.png   Free software measures MIDI latency and jitter.-2-logic-x.png   Free software measures MIDI latency and jitter.-3-logic-x.png  
Old 14th September 2015
  #159
Lives for gear
 

Glad to see you're repeating each test.

I did run the Logic 9 + MIDI Express 128 + MIDI clock test for over 20 minutes though! I couldn't test all 20 mins*, but found a long drift over a range of 5.5ms+ over that 20 mins, which is sort of twice as bad as your 10.2 + UM880 results.

*Jitter usually increases the more data samples you take, but then reaches a plateau where the min, max & average latencies hardly move. I think Roland used this fact to provide drift correction on the TR-8 and SBX-1 - they just take an average over a longer period of time. Unfortunately, this mean the ability to change tempo quickly becomes a problem, because it takes them extra time to re-adjust the average.

Roland would make a killing if they re-released the UM880 without any performance changes (unless it gets better ).
Old 14th September 2015
  #160
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by DanRand View Post
Glad to see you're repeating each test.

I did run the Logic 9 + MIDI Express 128 + MIDI clock test for over 20 minutes though! I couldn't test all 20 mins*, but found a long drift over a range of 5.5ms+ over that 20 mins, which is sort of twice as bad as your 10.2 + UM880 results.

*Jitter usually increases the more data samples you take
Ive been testing around 124 bars @ 120bpm as this seems to be the max time that MLA can check. Maybe the next update will be able to test longer files.


Quote:
Originally Posted by DanRand View Post

Roland would make a killing if they re-released the UM880 without any performance changes (unless it gets better ).
Would also be helpful if all the DAW developers would write DAW's that take advantage of the UM880, in the way the Ableton guys have.

Ive also found the UM880 to have great timing with Numerology (0.05ms RMS/ max 0.5ms. No push/pull cycle) and Renoise (0.05ms RMS/ max 0.8ms. No push/ pull cycle)

Of the sequencers I've tested (Live, Numerology, Renoise and Logic) its only Logic X that has poor MIDI timing.

EDIT: LOGIC X SENDS MIDI TIMESTAMPS. SEE FURTHER POSTS IN THIS THREAD

Last edited by Jauqq; 12th October 2015 at 01:16 AM..
Old 15th September 2015
  #161
RDP
Gear Maniac
Cable is probably beyond my means....

Are you using Live on Windows or OSX?

RDP
Old 16th September 2015
  #162
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by RDP View Post
Cable is probably beyond my means....

Are you using Live on Windows or OSX?

RDP
Im using Live 9 on OSX 10.9.5.

What is being demonstrated with the Live 9 and Edirol UM880 results can only be down to MIDI time stamping, and in this case the time stamping that is part of OSX Core Audio.

I remember reading many years ago in the Apple developer notes that the only USB MIDI interfaces with firmware to support OSX Core Audio MIDI time stamping were the MOTU and Edirol ones. However I've also tested a MOTU MIDI Express 128 with MLA and it gives very different results to the UM880, which may be down to older hardware and clocks within the MOTU unit. (i.e. higher jitter, and always a push/pull cycle with all DAW/ sequencers)

The evidence indicates that the possible difference in results between Live 9 and Logic X with the Edirol UM880 may be because Logic X is not time stamping its external MIDI output even though its part of Apple OSX !!!!??????
Old 17th September 2015
  #163
RDP
Gear Maniac
If I remember correctly Logic use the emagic timestamping, from before the Core Audio days. I have a feeling if you try an Unitor or AMT device you will get your accurate timings. Not sure what version of OSX was last compatible with those devices though....

UPDATE: No, ignore what I wrote: https://www.muffwiggler.com/forum/vi...1c405059006732

RDP

Last edited by RDP; 17th September 2015 at 05:52 AM..
Old 17th September 2015
  #164
RDP
Gear Maniac
You note the MOTU device has MLA, and bad timing. Wasn't the requirement for it to support MTS?

RDP
Old 17th September 2015
  #165
Lives for gear
 
Jauqq's Avatar
 

When I first began checking MIDI timing by recording MIDI to audio I was only counting cycle to cycle peaks in an audio editor. This got tiring very quickly, and I was only checking through a small number of bars.

But using the MLA software, this has automated that process, allowed for checking of a number of minutes of data, and also allowed to see how jitter shifts over time and if it has a push/pull cycle etc
(MLA is the fantastic software that Mmmmqac has written to do this. A big thanks to him for this software )

AMT (Active MIDI transmission) was Emagics protocol and relied on firmware within the AMT8 and Unitor8Mk2 units to function. This ceased with the change from OS9 to OSX. The last Emagic Logic to support this was version 6 on OS9.

AMT is not present in OSX, as Apple Core Audio MIDI time stamping within OSX replaced it.

However for MIDI time stamping to work, the MIDI interface has to have firmware to support it. As I noted the only ones which "supposedly" support it are the MOTU and Edirol units.

But from the two the Edirol has superior timing, like having 8 Atari ST MIDI outs !!!! (only with certain sequencers) while the MOTU unit has average jitter 0.3ms, as per MOTU advertising, but larger peak jitter timings and also has a push/pull cycle with all sequencers. I was not able to see this with my earlier cycle to cycle counts, but the MLA utility has shown this to the the case.

Why two interfaces with MIDI time stamping firmware behave very differently, I don't know......

Not all MIDI time stamping interfaces and firmware are equal !!!
Old 17th September 2015
  #166
I am about to add the batch-mode and CSV file output to allow exporting results to Excel. A few minutes ago I successfully tested a few improvements:
  • Increased capacity to 2800 notes and 16,800 clocks, which is five minutes of 16ths at 140 BPM.
  • Display the loaded file's minutes and seconds in the status-bar as a sanity-check
  • Toughened the MIDI decoder against un-even rise and fall times causing missed notes
  • Ignore truncated MIDI bytes at beginning and end of recording that caused framing errors.
The CSV export will be easy, but the batch mode will be interesting.
Old 17th September 2015
  #167
Stellar!!!!
Old 23rd September 2015
  #168
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by Jauqq View Post

The evidence indicates that the possible difference in results between Live 9 and Logic X with the Edirol UM880 may be because Logic X is not time stamping its external MIDI output even though its part of Apple OSX !!!!??????
Confirmed!!!!

Using MIDI Monitor it's possible to inspect outgoing MIDI streams from different apps. Within MIDI monitor preferences, if the "expert" mode is switched on and the DAW being checked is not sending timestamps there will be a "0' in its time column.

Live 9 and Numerology all give time stamps.

Logic X a big column of "0".... Logic X "Nil Pois" !!!!

If Logic X were to support MIDI timestamps, then together with a confirmed time stamping firmware supporting USB MIDI interface such as the Edirol, there would be Atari ST tight MIDI output (as with Live 9, Numerology etc)

Instead Logic X is sending MIDI data as soon as its generated, giving rise to the push/pull cycles, and jitter drops that have been recorded by MLA.

So Apples flagship DAW does not incorporate a feature that is part of it's own operating system !! :face palm:


EDIT: LOGIC X IS SENDING MIDI TIMESTAMPS VIA THE EXTERNAL INSTRUMENT PLUG-IN. UPDATED INFO FURTHER ON IN THIS THREAD
Attached Thumbnails
Free software measures MIDI latency and jitter.-logic-x-nil-pois.png  

Last edited by Jauqq; 12th October 2015 at 01:17 AM.. Reason: Updated information
Old 25th September 2015
  #169
Quote:
Originally Posted by Mmmmqac View Post
I am about to add the batch-mode and CSV file output to allow exporting results to Excel. A few minutes ago I successfully tested a few improvements:
  • Increased capacity to 2800 notes and 16,800 clocks, which is five minutes of 16ths at 140 BPM.
  • Display the loaded file's minutes and seconds in the status-bar as a sanity-check
  • Toughened the MIDI decoder against un-even rise and fall times causing missed notes
  • Ignore truncated MIDI bytes at beginning and end of recording that caused framing errors.
The CSV export will be easy, but the batch mode will be interesting.
I have the batch process working now, and exporting to Excel. Any volunteers for beta testing?
Old 27th September 2015
  #170
Lives for gear
 
Jauqq's Avatar
 

Roland SBX1 sync box MIDI clock tests.

The SBX1 is set to USB as the clock source, and the “advanced” USB driver is being used. MIDI clock from MIDI port 1 being tested.

MacBook Pro, OSX 10.9.5

Live 9 DAW. In Live 9 MIDI sync preferences, SBX1 output has “sync” on.
4 tests @ 120bpm, for 124 bars.

1) MIDI clock jitter 0.038ms RMS (max jitter 0.59ms)
2) MIDI clock jitter 0.038ms RMS (max jitter 0.21ms)
3) MIDI clock jitter 0.039ms RMS (max jitter 0.61ms)
4) MIDI clock jitter 0..038ms RMS (max jitter 0.42ms)

Very similar results to these Live 9 + UM880 results!!!

No push/pull cycle MIDI jitter wave.
Attached Thumbnails
Free software measures MIDI latency and jitter.-1-live-9-sbx1.png   Free software measures MIDI latency and jitter.-2-live-9-sbx1.png   Free software measures MIDI latency and jitter.-3-live-9-sbx1.png   Free software measures MIDI latency and jitter.-4-live-9-sbx1.png  

Last edited by Jauqq; 27th September 2015 at 03:05 AM..
Old 27th September 2015
  #171
Lives for gear
 
Jauqq's Avatar
 

The SBX1 is set to USB as the clock source, and the “advanced” driver is being used. MIDI clock from MIDI port 1 being tested.

MacBook Pro, OSX 10.9.5

Logic X DAW. In Logic X, project setting, synchronisation, MIDI, MIDI clock set to transmit to “Destination 1” being “1 SBX-1 SBX-1”
4 tests @ 120bpm, for 124 bars.

1) MIDI clock jitter 0.42ms RMS (max jitter 1.12ms)
2) MIDI clock jitter 0.40ms RMS (max jitter 1.18ms)
3) MIDI clock jitter 0.43ms RMS (max jitter 1.09ms)
4) MIDI clock jitter 0.43ms RMS (max jitter 1.68ms)

Very similar results these Logic X + UM880 MIDI clock tests….

What caught my attention was that the SBX1 allows it to be set to either “generic” or “advanced” USB mode… the exact same as the UM880.

Its already been established that Logic X does not time stamp its MIDI output, but that Live 9 does.

From the evidence it seems the SBX1 has the same time stamping firmware as the Edirol (Roland) USB MIDI interface range. With a DAW that sends time stamps, the SBX1 in USB slave mode will have decent sync, as good as an Atari MIDI clock.

With a DAW that does not time stamp its MIDI output, the SBX1 will give higher jitter values, large shifts in timing, and push/pull cycles.

The jitter waves for the above four Logic X tests are in the attachment.

Best to stick to Innerclock + Expert Sleepers for the tightest MIDI clock from a DAW.
Attached Thumbnails
Free software measures MIDI latency and jitter.-logic-x-sbx1-jitter-waves.jpg   Free software measures MIDI latency and jitter.-1-logic-x-sbx1.png   Free software measures MIDI latency and jitter.-2-logic-x-sbx1.png   Free software measures MIDI latency and jitter.-3-logic-x-sbx1.png   Free software measures MIDI latency and jitter.-4-logic-x-sbx1.png  

Old 27th September 2015
  #172
Lives for gear
 
Jauqq's Avatar
 

Roland SBX1 sync box

Roland SBX1 internal sync, 120bpm, 124 bars.

MIDI clock out.

1) 0.15ms jitter RMS (max 0.48ms max)
2) 0.24ms jitter RMS (max 0.56ms max)

jitter waves for the above tests shown, both displaying push/pull cycle.

SBX1 average jitter is tighter under external sync from a MIDI time stamping DAW, then under its own internal sync
Attached Thumbnails
Free software measures MIDI latency and jitter.-sbx1-internal-sync-jitter-waves.jpg   Free software measures MIDI latency and jitter.-1-sbx1-internal-sync.png   Free software measures MIDI latency and jitter.-2-sbx1-internal-sync.png  

Last edited by Jauqq; 27th September 2015 at 06:09 PM..
Old 4th October 2015
  #173
Lives for gear
 

Quote:
Originally Posted by Jauqq View Post
Roland SBX1 sync box

Roland SBX1 internal sync, 120bpm, 124 bars.

MIDI clock out.

1) 0.15ms jitter RMS (max 0.48ms max)
2) 0.24ms jitter RMS (max 0.56ms max)

jitter waves for the above tests shown, both displaying push/pull cycle.

SBX1 average jitter is tighter under external sync from a MIDI time stamping DAW, then under its own internal sync
Jauqq, Thanks for your tests, they have been revealing.

I'm wondering if you have observed or measured how significantly (multiple streams of) cc's affect jitter of the clock, MIDI note timing and/or cc data?
Old 7th October 2015
  #174
Gear Head
 
NGC 5139's Avatar
 

@ Jauqq

Thanks for your testing with the UM-880!

I hope there is a linear sequencer software for Mac like Cubase 8 or Digital Performer 9 which makes use of MTS in Apple Core Audio in the just released OS X 10.11 (El Capitan) and also shows the same excellent results with the UM-880 like your test setup with 10.9.5 and Live 9. That would be a future proof setup for me...

If that becomes a reality, I switch back to Mac from PC...

And shame on you Apple, that your own flagship sequencer software Logic X doesn’t support MTS!
Old 7th October 2015
  #175
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by NGC 5139 View Post
@ Jauqq

Thanks for your testing with the UM-880!
You'll also get the same excellent timing with the other Edirol units which support time stamping. So the UM550, UM-ONE mk2 etc along with a time stamping DAW/ sequencer.

Quote:
Originally Posted by NGC 5139 View Post
@ Jauqq
I hope there is a linear sequencer software for Mac like Cubase 8 or Digital Performer 9 which makes use of MTS in Apple Core Audio in the just released OS X 10.11 (El Capitan) and also shows the same excellent results with the UM-880 like your test setup with 10.9.5 and Live 9. That would be a future proof setup for me...

If that becomes a reality, I switch back to Mac from PC...

And shame on you Apple, that your own flagship sequencer software Logic X doesn’t support MTS!
It is incredible that Logic X does not support MIDI time stamps

Two other DAW developers that do include time stamps have informed me its down to 64 bit math, and also its not rocket science...

Perhaps the Logic Developers feel that their "sending MIDI as soon as its generated" method is ok. Obviously they're not wiring MIDI leads to audio jacks and then looking at the waveforms or else they'd see its not an ideal method and instead see the "MIDI meh" Logic X is sending.

Ive been submitting feature requests/ bug reports about this to Apple for some time. If there is any one else that would like Logic X to include time stamps, perhaps take a few mins to submit feedback to them?
The squeaky wheel and all that....
Logic Pro feedback page

EDIT: LOGIC X IS SENDING MIDI TIMESTAMPS VIA THE EXTERNAL INSTRUMENT PLUG-IN. UPDATED INFO FURTHER ON IN THIS THREAD

Last edited by Jauqq; 11th October 2015 at 03:55 PM.. Reason: Updated information
Old 7th October 2015
  #176
Gear Head
 
NGC 5139's Avatar
 

Quote:
Originally Posted by Jauqq View Post
Ive been submitting feature requests/ bug reports about this to Apple for some time. If there is any one else that would like Logic X to include time stamps, perhaps take a few mins to submit feedback to them?
The squeaky wheel and all that....
Logic Pro feedback page
Good idea, just left a feature request regarding the MIDI Time Stamping (MTS), I hope it's more then just a drop in the ocean...

Jauqq, you mentioned, that there are two other sequencer softwares for Mac which support MTS and would give the same rock solid midi timing with Edirol midi interfaces. Care to share? And did you test them too?
Old 7th October 2015
  #177
Quote:
Originally Posted by Jauqq View Post
Confirmed!!!!

Using MIDI Monitor it's possible to inspect outgoing MIDI streams from different apps. Within MIDI monitor preferences, if the "expert" mode is switched on and the DAW being checked is not sending timestamps there will be a "0' in its time column.

Live 9 and Numerology all give time stamps.

………...
I'll test this with Bitwig and Logic 9 asap.
thanks!
Old 7th October 2015
  #178
Old 8th October 2015
  #179
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by NGC 5139 View Post

Jauqq, you mentioned, that there are two other sequencer softwares for Mac which support MTS and would give the same rock solid midi timing with Edirol midi interfaces. Care to share? And did you test them too?
Sure, but they're more sequencers than true DAW's. Its Numerology and Renoise. Testing them with the Edirol UM880 I found Numerology was 0.05ms RMS/ max 0.5ms. No push/pull cycle, and Renoise was 0.05ms RMS/ max 0.8ms. No push/ pull cycle. A higher max jitter for Renoise, so that could do with a little work to tighten it up.
Old 8th October 2015
  #180
Lives for gear
 
Jauqq's Avatar
 

Quote:
Originally Posted by Reptil View Post
I'll test this with Bitwig and Logic 9 asap.
thanks!
Just tested Logic 9 with "MIDI monitor" and guess what.....?
Another big column of zero's !!!

So Logic 9 also not sending MIDI time stamps


Im going to test Logic 7 at some point, but my guess is that will also not be sending MIDI time stamps.


EDIT: LOGIC X DOES SEND MIDI TIMESTAMPS! INFO BELOW!

Last edited by Jauqq; 12th October 2015 at 02:37 AM.. Reason: Updated information
Topic:
Post Reply

Welcome to the Gearslutz 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…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump