MIDI SYNC: the ultimate SYNC thread

There has been positive/negative talk for the past year about how the MPC 2500 syncs to DAW and other gear via Midi Clock and Midi Time Code. Some say the earlier Akai operating systems are better, others say the JJ OS is better, while others have no clue how sync works.

Well, I sat down this morning and started a very comprehensive eval on MIDI sync using the latest JJ OS 3.04, and All the previous Akai OS's for the MPC 2500. I also tested the Akai MPC 3000.

Items used:
Akai MPC 3000
Akai MPC 2500
Motu Midisport 8x8 USB MIDI interface
Various operating systems
MIDI sequence (120 tempo)
Nuendo 3 as MTC/Midi Clock Master/TEMPO 120

The 4 bar TEST MIDI sequence I used was made up of drum hits. I intentionally played very bad and even used grid edit to randomly throw in extra off beat hits. The first 2 bars of the sequence are blank to allow time for devices to sync. I loaded this MIDI sequence into both MPCs using various operating systems. I recorded these sequences directly into Nuendo as MIDI tracks. I imported the TEST MIDI sequence directly into Nuendo. I used Nuendo's MIDI event list which allows you to view exactly where the drum hits along a time line..down to .oo1 of a second.

MIDI EVENTS LIST(timeline in seconds):
Imported TEST SEQUENCE
1. 4.016
2. 4.036
3. 4.057
4. 4.099
5. 4.115
6. 4.182
7. 4.214
8. 4.255
9. 4.344
10. 4.401

MPC 3000:
Midi Clock Slave
1. 4.021
2. 4.043
3. 4.064
4. 4.106
5. 4.120
6. 4.190
7. 4.222
8. 4.262
9. 4.349
10. 4.409

Midi Time Code Slave
1. 4.024
2. 4.045
3. 4.064
4. 4.107
5. 4.124
6. 4.192
7. 4.223
8. 4.263
9. 4.353
10. 4.410

MPC 2500 (JJ OS 3.04)
Midi Clock Slave
1. 4.011
2. 4.029
3. 4.050
4. 4.092
5. 4.113
6. 4.173
7. 4.218
8. 4.261
9. 4.341
10. 4.405

Midi Time Code Slave
1. 4.019
2. 4.042
3. 4.063
4. 4.107
5. 4.126
6. 4.188
7. 4.221
8. 4.260
9. 4.350
10. 4.411


MPC 2500 (Akai OS 1.22)
Midi Clock Slave
1. 4.009
2. 4.028
3. 4.048
4. 4.092
5. 4.112
6. 4.175
7. 4.218
8. 4.257
9. 4.343
10. 4.403

*Midi Time Code Slave (MPC waiting for Time Code)
1. 4.025
2. 4.046
3. 4.065
4. 4.110
5. 4.127
6. 4.194
7. 4.224
8. 4.265
9. 4.353
10. 4.410


MPC 2500 (Akai OS 1.20)
Midi Clock Slave
1. 4.007
2. 4.029
3. 4.049
4. 4.090
5. 4.113
6. 4.177
7. 4.216
8. 4.259
9. 4.341
10. 4.402

Midi Time Code Slave
1. 4.023
2. 4.043
3. 4.066
4. 4.107
5. 4.120
6. 4.191
7. 4.219
8. 4.262
9. 4.351
10. 4.408


With these resuts..generally, slaving the MPC 2500 to Midi Clock will have MIDI notes showing up early when recorded into my DAW. However, when slaving the MPC 3000 to Midi Clock, events show up later.

What do you guys think about the MTC results?

Remember that this test is with a sequence that is played very badly with no timing correct. I will post the results of my test with heavily quantize tracks and report back.

Please feel free to drop comments...

Did you repeat the recordings to see if timings vary, or is this from a single recording?

Did you repeat the recordings to see if timings vary, or is this from a single recording?

Yup...and they vary with each recording. However, each new recording on a particular OS stay within a few thousandths of a sec of each other.

Interesting test. I would suggest that a PC with a USB midi interface is probably not the greatest way to purely test the MPCs slaving abilities, but obviously it's a 'real world' test. (I've never felt confident with PC midi timing despite owning many different PCs and midi interfaces. Somehow they never seem as tight as my old Atari ST *sniff*)

I'd be really interested to see how the timing varied if you used the other MPC as the recording device instead of the PC. Obviously that test becomes less about interaction with the DAW and more about the pure clocking.

Interesting also that your results are so good using midi clock. I've always had measurably better results using MTC but there doesn't seem like much difference in your tests. What is encouraging is that there doesn't seem like much, if any, drift between individual notes. It's just that entire sequences are early or late.

Did you try enabling or disabling Nuendo's "use system timestamp" setting? This is often cited as a cure for 'early' midi recordings.

I do have MTC/MIDI clock follows time/position checked/activated. Not really sure how much this settings come into play but I can/will post the results of not using those settings.

As far as the USB inteface not being ideal for the test, I do have an earlier post doing a similar test using an RME PCI audio/MIDI card. The results were practically the same.

I did this test because of all the talk about the MPC being unstable when used with a DAW. I can do a test using just both MPCs but it will most likely not be of any use to a majority here.

jahrome wrote:I do have MTC/MIDI clock follows time/position checked/activated. Not really sure how much this settings come into play but I can/will post the results of not using those settings.

That's not the setting I mean. Try going to Device Setup / Midi Port Setup (I'm on Cubase 4 so it may be worded differently). Then enable or disable "Use System Timestamp for Windows Midi Ports". It can have a dramatic effect on midi record timing. NOTE: It's not applicable if you are using direct x emulated ports.

I did this test because of all the talk about the MPC being unstable when used with a DAW. I can do a test using just both MPCs but it will most likely not be of any use to a majority here.

Yeah I'm sure you're right. Probably not worth it.

Question for you two. I use an M-Audio 1814 firewire with ProTools Mpowered. I see you used a dedicated midi interface. Is that necessary? Would you suggest having a separate audio and midi interface for trying to sync the MPC to DAW? I know this isn't directly about your results Jahrome, but just a question.

Also I mentioned before that the type of sync I'm looking for is running the MPC audio into PT, while PT is rewiring with Reason3, having all synced up. Also want to add a Motif rack to the system. I've had problems making this work for long audio recording out of the MPC. Tempo always drifts off, where Reason and the MPC lose sync. I've been recording "clips" of eight bars or so and then duplicating and manipulating in PT. May not be proper but its been working for me. One.

Rozzer wrote:
Did you try enabling or disabling Nuendo's "use system timestamp" setting? This is often cited as a cure for 'early' midi recordings.

I found that setting. For the tests, System timestamp was activated.

Prophacee wrote:
Question for you two. I use an M-Audio 1814 firewire with ProTools Mpowered. I see you used a dedicated midi interface. Is that necessary? Would you suggest having a separate audio and midi interface for trying to sync the MPC to DAW? I know this isn't directly about your results Jahrome, but just a question.

Also I mentioned before that the type of sync I'm looking for is running the MPC audio into PT, while PT is rewiring with Reason3, having all synced up. Also want to add a Motif rack to the system. I've had problems making this work for long audio recording out of the MPC. Tempo always drifts off, where Reason and the MPC lose sync. I've been recording "clips" of eight bars or so and then duplicating and manipulating in PT. May not be proper but its been working for me. One.

Previously, I only used the MIDI interface of my RME sound card. But with the MIDI gear I have, I needed additional MIDI I/O so I just purchased the M-Audio 8x8. So far I am happy. I last purchased Steinberg's discontinued 8x8 MIDI interface and it was a piece of crap (another story). With that said, I don't think you need a seperate MIDI interface unless you need additional MIDI I/O.

As far as your issues...I would first see if I could get Pro Tools to sync to the MPC. Then I would try to get Reason and the MPC to sync properly. If both of those work, then the problem could be with Rewire. Also, once Reason is Rewired...Pro Tools should be the master...and you should be trying to sync the MPC with it versus Reason.

Nice one Jah! Thats some serious MPC dedication.

impressive Oh well, Jahrome great job this is very informative this should be a bigger priority for AKAI to look in to.

jahrome wrote:I did this test because of all the talk about the MPC being unstable when used with a DAW.

Just to throw that in... if it's all about stability here, wouldn't the variation between different recording runs be more meaningful then? I mean, it can still be rock solid if it's a few ticks early or late, but it stays that way... I realize that this is a lot more work though.

I plan to do some more test and post the info. Hopefully, it will be of some use.

I tried recording MIDI from the MPC 2500 into the 3000...the 3000 almost had a heart attack It couldn't handle the amount of MIDI notes being recorded at that tempo of the project. I will have to probably slow the test down and repeat it again.

If you look closely at the MPC 3000 MTC results, and compare them with the different MPC 2500 operating system results...MTC using Akai OS 1.22 (although it is broken) is appears to be closer to the MPC 3000 than the other operating systems.

Also, using Akai OS 1.20, timing seems to be shifted forward and timing shifted later for OS 1.22.

Just to throw that in... if it's all about stability here, wouldn't the variation between different recording runs be more meaningful then? I mean, it can still be rock solid if it's a few ticks early or late, but it stays that way... I realize that this is a lot more work though.

Exactly what I was trying to say in earlier threads. But there were a few that said they have problems with sync messing up their grooves. Another point I want to highlight is that these tests are showing differences down to .oo1 seconds. I would think that if an MPC plays a hit .003 seconds off from time to time...it will make for a more natural sounding drum pattern...than heavily quantized tracks???

But again....this is what I will test tonight.

jahrome wrote:Exactly what I was trying to say in earlier threads. But there were a few that said they have problems with sync messing up their grooves. Another point I want to highlight is that these tests are showing differences down to .oo1 seconds. I would think that if an MPC plays a hit .003 seconds off from time to time...it will make for a more natural sounding drum pattern...than heavily quantized tracks???

Yes of course...that's what it's about for most people I think.

jahrome wrote:If you look closely at the MPC 3000 MTC results, and compare them with the different MPC 2500 operating system results...MTC using Akai OS 1.22 (although it is broken) is appears to be closer to the MPC 3000 than the other operating systems.

Also, using Akai OS 1.20, timing seems to be shifted forward and timing shifted later for OS 1.22.

And remember, you're talking about differences as low as 1-2 digits of accuracy... that's not significant unless you prove that it is reproducible over like 1000 runs. I would not state that one is shifted later than the other from that.

For me, Akai OS 1.20 and 1.22 have exactly the same timing seeing these results.

Here is my final data. I did the same test as above...but this time, I quantized the tracks 1/16 with 50% swing. However, this time I used the MPC 3000 as the Master and slaved the MPC 2500 to it. I recorded MIDI from the MPC 2500 into the MPC 3000. Here are the results:

Original Test Track
1. 4.000
2. 3.375
3. 4.500
4. 4.750
5. 5.375
6. 5.500
7. 5.750
8. 6.375
9. 6.500
10. 6.875

Now, it is very obvious that this track is heavily quantize. As I have done above, I used a MIDI event list editor in Nuendo to display where along the time line the MIDI notes hit.


MPC 2500 JJ OS 3.04
MTC Slave
1. 4.000
2. 3.380
3. 4.505
4. 4.755
5. 5.380
6. 5.505
7. 5.755
8. 6.380
9. 6.510
10. 6.880


MPC 2500 Akai OS 1.20
MTC Slave
1. 4.005
2. 3.380
3. 4.510
4. 4.760
5. 5.380
6. 5.510
7. 5.755
8. 6.380
9. 6.510
10. 6.885


MPC 2500 Akai OS 1.22
MTC Slave (although it doesn't work properly)
1. 4.010
2. 3.385
3. 4.510
4. 4.760
5. 5.385
6. 5.510
7. 5.760
8. 6.385
9. 6.510
10. 6.885


Now casually looking at these results, its obvious that these identical tracks produced similar results...all three tracks appear to be heavily quantized as the original test track. However..one of these tracks really stand out above the other two when you compare its timing to the original

Please feel free to comment....

thanks jah... all i get from this damn sync thread though is that sync is poop... what gives? does time stamping work? thats suppose to create sample accurate sync, no? if so, anyone suggest any boxes that notably use such? i was told motu is the way to go... i dunno... lol