ORIGINAL: brundlefly

Interesting. I guess the 1010's MIDI ports are right on the PCI card?

Yup, that's right.

My 1820m's ports are on the dock, which probably slows things down a little. But I don't understand why my MIDI loopback test yields a round-trip of only 2.6ms, while the Tester says 5.7ms. I'd be curious what results you get from a basic MIDI loopback test, using Sonar. If the tester is right, you should have re-recorded MIDI events come in no more than 2 ticks behind the source MIDI at 125BPM (.5ms/tick at 960PPQ). That would be pretty impressive.

I don't think it would matter that the ports are on the dock, unless it's doing some additional processing there or something.

I did a seemingly simple loop back test in the Sonah, as well as Reaper, because I didn't believe it. Disabled recording latency compensation. I just put a quarter note in a measure and looped it, sent that to midi out, created a second track and recorded midi in. Used 44.1Khz.

Sometimes I'd count the samples between the midi events: 80, 80, 66, 52, 45, 83, 74, 68. Avg is 1.55ms round-trip (at 44.1)
Another time it'd be twice as much, avg of 3ms round-trip
Sometimes it would be 30 samples or less between events
Sometimes the recorded MIDI events would occur before the outputted events!
Sometimes the recorded MIDI events are late by 4 samples, or exactly on.

Well maybe not so simple of a loop back test. I have more questions now than answers.

Sometimes I'd count the samples between the midi events: 80, 80, 66, 52, 45, 83, 74, 68. Avg is 1.55ms round-trip (at 44.1)
Another time it'd be twice as much, avg of 3ms round-trip
Sometimes it would be 30 samples or less between events
Sometimes the recorded MIDI events would occur before the outputted events!
Sometimes the recorded MIDI events are late by 4 samples, or exactly on.

I don't understand how you could get MIDI events at these tick intervals. 125BPM is approx. 21.7 samples per tick at 44.1kHz and 960PPQ, and 100BPM is 25.9. You should only be able to have MIDI events on whole-number multiples of these values. What tempo did you use for testing?

Yeah, me neither. You probably doubt these results, heck I wouldn't believe me unless I saw them with my own eyes. I tried tempos of 1000bpm down to 10bpm, and 120bpm; it didn't seem to matter. Neither did increasing the tick resolution beyond 960. Again I was testing in both programs, mostly in Reaper as it is easier to measure the sample differences, and it has dummy audio, where the sound card is not used (results were unchanged with that). Recording delay compensation didn't seem to affect MIDI timing, nor did ASIO buffer sizes from 64 samples up to 2096 samples. I'm not running the MIDI through anything weird; I've used multiple cables of different lengths. I guess Imust have some sort of super computer.

Again I was testing in both programs, mostly in Reaper

Okay, I missed that. Maybe Reaper doesn't enforce tick boundaries on MIDI until you quantize it? I'm not familiar with it. It's not that I doubt what you're seeing; just trying to understand what the software is really doing. Obviously having recorded MIDI placed ahead of the sent event isn't right. Makes me wonder if Reaper is doing some sort of MIDI latency compensation. As you mentioned, audio latency compensation should not affect MIDI.

I thought this would be a good place to get help to find out how to get my midi timing better. Here is a link to a screen capture I did. I thought the Roland TD-8 would have great timing since it is an electronic drum module.

Link
http://img408.imageshack.us/img408/4140/td8ye2.jpg

This screen shot is the same mini track being recorded several times to show the variations. The midi box is a Emagic Unitor 8. Computer is a AMD X2-4800.
Settings:
Audio Latency 16ms @ 24/44 64bit double prec., read and write cache disabled, dithering - none, WDM, trigger and freewheel, use multiprocessor, record latency 136 (set from other thread - THANKS!!)
Midi 128 - tried 256,512,1024 all about the same.

Soft synth - EZ drummer - timed out a little late on all the clicks. Have not measured yet but it looks close. Will count samples tonight.

Thanks for any help!

Max Arwood

We should probably move this to its own thread, but at first glance this looks like only +/- 1 tick of variation - not unusual at all. Probably due to the jitter of the interface as discussed ealrier in this thread, as opposed to a timing problem with the TD-8. Also it appears that audio is a little over- compensated, assuming the MIDI events were at 2:01:000. This is a little unusual, as the response time of outboard sound modules together with outbound MIDI latency usually makes the audio a little late.

brundlefly: Thanks for looking. I hope you can help. This amount of variation can cause phase problems.

My problems thread.

http://forum.cakewalk.com/tm.asp?m=1327188

Max Arwood

Ok, I'm on page 5 of this thread. and, just wanted to push it to the top again as in my view, this is the most important thread that has ever been on this forum.

Looking back in hindsight, thru the years of my searching for information on these timing issues and just giving up and thinking I was crazy because no one else had the same timing issues. But , then again I'm coming to the realization that, really, it's kind of hysterical to think that human beings could actually assume they could create some kind of hardware or software that could take from the air and/or heart and duplicate it into 1s and 0s and then recreate it.

But then, hmm, tape decks did this. A simple piece of oxide passing along a head sending electrical signal thru all it's thingys , making music, not just sounds. There it was, it had feeling.
WTF is wrong with computers? How can a 15ips tape machine be faster than souped up computer??

I've been with Cakewalk since Dos.
numerous tape decks before that.
After 20 years of trying, I am getting to the point of realization that one basically cannot make "music" with computers.
Or, is it Cakewalk?
Pro-tools must be the standard for a reason. Do they have these kind of issues?

EDIT:
Ok, I thunk about it a bit after my little emotional outburst.
Computers can and are a great recording medium and will record audio quite well.
I guess trying to use it as an instrument via anything midi is where things get screwy.
amazing that midi hasn't gotten any better.

CAKEWALK, Please go hire Roger Lynn to sort all this out. Or, buy his code or whatever it was that he used in his
Lynndrum back in the 80s. I'm guessing that the Akai boxes mentioned in this thread are similiar.

Tim

I feel your pain timidi. And I hear how Sonar (and P5, and probably every other computer-based MIDI recording "solution") tends to butcher the timing of a MIDI performance unless you quantize it.

From trying to keep up with the massive amount of information in this thread, the main answer seems to be that, at least with PC-based recording, Windows interferes with the accurate timing of incoming and outgoing MIDI due to its own processes taking place at the same time.

Truthfully, I've experienced MIDI inaccuracies even since the days of hardware sequencers, so I can't in all frankness say "bring back the Linn/MPC/Roland Microcomposer. But it does seem that these hardware boxes were/are better at playing back an accurate rendering of a non-quantized MIDI performance. And, just as importantly, it seems like with all of the advances in computing, that this would be something that would automatically be improved. Instead, it's worse. I used to think it had to do with clock resolution. Obviously that's not the case. Whatever the reason, it sucks. I never think of Sonar when I think of something to record accurate unquantized MIDI performances on. Is there such a solution in software?

Well, I guess to answer my own question, considering what prompted me to start this thread, it would currently be Ableton Live.

I never think of Sonar when I think of something to record accurate unquantized MIDI performances on. Is there such a solution in software?

Well, I guess to answer my own question, considering what prompted me to start this thread, it would currently be Ableton Live.

If there's one thing that everyone should be able to agree on in this thread, it's that MIDI jitter is inherent in the technology and the limitation of hardware performance, and cannot be completely resolved through software alone with the existing MIDI standard. Ableton has no chance of being any better than SONAR.

I also want to reiterate that while I contributed to this thread, and found it all very interesting from a technical standpoint, I still maintain that the vast majority of musicians do not have sufficient personal timing accuracy or consistency to justify concern about a tick or two of MIDI jitter (.5 to 1 ms at typical tempos). If you can hear timing issues in a MIDI performance recorded by a properly functioning SONAR installation with a good interface, then the problem likely lies somewhere between your brain and your fingers.

I firmly believe that when you are actually performing, you hear the performance differently than when you listen back to the recorded performance. Your brain is in the groove, even when your fingers aren't.

Try recording the same live performance two or three times, and I predict you will find pervasive timing variations of individual notes from the grid and intervals between notes that are an order of magnitude or more greater than the MIDI jitter in your system.

The delay issues that Timidi is having are a different issue altogether. The cumulative delays he is dealing with due to inadequate audio latency compensation are a real problem.