MIDI notes recorded earlier (aka Plonk test)

Cubase
#1

Hello,

Some of us suffer from MIDI recorded earlier than it should. Its been a quite big discussion here on this forum on this topic and finally we´ve get to differnetiate 2 different problems:

  1. The amount the “earlyness” of the MIDI is directly related to the latency value. This is caused because Cubase records the note when the key is actually pressed, not when the VSTi plays the note. This leads to people trying to compensate by playing early and thus the MIDI gets recorded early too. Steinberg already said they are gonna introduce a switch on the next major update to make this behaviour change so hopefully people complaining about this will have their MIDI spot on.

  2. MIDI is recorder early no matter which latency value you are working at. This post is about this case, not the first one.


    What i´d like is to get results from many users that also suffer this issue so that we can try to figure out what might be the cause of this time difference. The test is pretty easy to do: you only need cubase + a synth which also can be used as a controller or if you dont have one of this, a microphone.

The idea is to record aduio and midi at the same time and then check what is the difference between both. So you setup a session with no VSTi whatsoever, just 1 midi and 1 audio track, set the correct inputs hit record and try to record on the beat following the click (thats optional but if you record this way you can see clearly if the midi was always before the click or not). If you are too lazy to setup the session, just download this:

http://dl.dropbox.com/u/54950124/MIDI%2BAUDIO%20tracks.cpr

And chose the apropiate inputs. If you dont have a synth/controller you can do as i did and record the sound of the keys bieng stroke with your fingers (i used AKAI pads since the sound had more attack than weighted keys).

If the difference is somehwere between 1-2 ms it should be fine. Greater values will start to get annoying.

And thats it. I am pretty sure there´s a combination of hardware + Cubase that causes this, cause , as said earlier, doesnt happen on every machine so you cant blame Cubase only. If we all give our specs maybe we get to something. Steinberg doesnt seem to want to get involved on investigating about this this so i really hope some of you make the test!


My results:

7-8 ms difference , pics here --> http://dl.dropbox.com/u/54950124/Cubase%20test_2.rar
(128 and 4096 are the buffer settings but you can see it makes no difference)
SPECS: M-audio Keystation Pro 88, TC KONNEKT 6, Intel Core Quad q8300@2,50Ghz, ASUS P5QL PRO, 8 GB RAM

I am specially interested in people who use a TC KONNEKT 6: please run the test and post here, even if the result was within 1-2ms range!

Thanks!

#2

This is caused because Cubase records the note when the key is actually pressed, not when the VSTi plays the note.

There’s NO WAY Cubase can know when a key is pressed on another machine.
It knows when it ARRIVED in the system. (System being the OTHER end of the midi cable to the keyboard)
And if anyone has to compensate for latency then their buffers are set too high or their system is acting up in some way. I play live an awful lot and midi is just as good as the audio for any musical purpose.
Yes, there will be a difference in when it arrives and the VSTi spitting out the sound but on modern systems this is very close to real life onstage live latency.

If you can hear it then something is going on that shouldn’t be but in my opinion there’s more to it than just Cubase. Although it would be senseless to just say there’s not a bug I wouldn’t put money on it because it’d have to be a pretty obscure and hard to find bug manifesting in only a few system setups.

Curiosity question, have you tried different midi leads?

#3

Conman please, lets try to make this thread as pracmatical as possible ok? Just results and specs. Please.

#4

OK (but don’t know what pracmatical means) but one suggestion after you get results that may involve the TC Helicon is to swap out the usb leads and also to reinstall it’s driver if it’s an involved installation where things can go wrong with picky drivers.
And turn on the TC Helicon after everything else and before launching Cubase.
I had to do a combination of these when my Octa-capture threw a wobbler the other week. No early notes but they were randomising. In the end just disconnecting the usb was my cure.

And sorry for not seeing the rule that said only results and specs as I lost the power of telepathy at birth. :mrgreen:

The idea is to record aduio and midi at the same time and then check what is the difference between both. So you setup a session with no VSTi whatsoever, just 1 midi and 1 audio track, set the correct inputs hit record and try to record on the beat following the click (thats optional but if you record this way you can see clearly if the midi was always before the click or not). If you are too lazy to setup the session, just download this:

You need to specify the quantise setting used as if someone has 8ths he will see the notes placed right on the beat and so will think he’s ok.

#5

The TC is a firewire interface and works well. I only have one FW port anyway. The Keystation Pro i already reinstalled several times in all of my USB ports.

Conman, could you do the test and see whats the offset? I am quite sure yours is a working computer setup so i´d like to know what the time difference is there.

#6

I would suggest running this test, of course, with quantise turned off, otherwise it would make no sense at all. I would also suggest keeping quantise out of this thread as it has nothing to do with the issue. I would also suggest keeping the latency of vstis out of this thread, because, as the OP mentioned in his OP, the test to be performed involves no VSTis.

I did the plonk test here on my system, and I noticed a time difference of about 1-2ms. I seem not to be experiencing the problem here.

I did the plonk test on several other systems where problem Nr.2 has occured, and it always showed a discrepancy of roughly 10 up to 20ms, which was annoying.

I also suggest adding another test for significance: if you are experiencing problem 2 and your plonk test results in significant discrepancies, please also do the “record click” test: mike and audio record your Cubase click, then compare audio to grid to see if audio is recorded late. (result for me: 1-2ms on any setup, my own (finely working) one and the other (problem 2) ones.)

#7

I get case 2 pretty much all the time. The tech support have the MIDI controller drivers, I guess they are trying to see if they cause it.

In the next update I hope they implement the solution to the latency + MIDI issue, and I pray it solves it.

#8

I meant only change the cable (above in bold). I’m pretty sure my computer is working aok as, like you did, would have noticed stray early notes. Then, like you, I would have needed to do somethng like the plunk test.
However, I might still try it for form’s sake.

Now forgive me for not taking this in previously but you have, as far as I can tell, The TC through FWire and a controller keyboard into usb. That’s two streams of continuous data, even if they are “idle” then there is still recognition data continually present concerning the unit identity and timing at least. This will be interleaving into whatever needs it. There could be a timing discrepancy along the line as two time pointers cancel each other out or combine to give a false reading. These factors could depend on the motherboard. ?
I don’t know if the Keystation has only usb but if it has midi can that go thru the TC or is that just audio?
If you can’t because the TC is for audio and the keystation is your midi I’d strongly recommend borrowing an audio/midi interface and feeding your keystation through that, or another borrowed midi keyboard. So you have one continuous data stream rather than two.
Hope that makes sense. You would hope that the solution was that there are two data streams causing the timing discrepancy then you could go to support and give them something more definite to go on.

#9

I would suggest running this test, of course, with quantise turned off, otherwise it would make no sense at all. I would also suggest keeping quantise out of this thread as it has nothing to do with the issue. I would also suggest keeping the latency of vstis out of this thread, because, as the OP mentioned in his OP, the test to be performed involves no VSTis.

It does need stating that quantise be turned off. Quantise MUST have rather a lot to do with the issue as that is what defines the final note positions. It is always present even when “off”. Read the manual on the matter to remind yourself of that.
“The test to be performed involves no VSTis” Oh, yeah? So what’s with the statement?

Please, if you don’t know much about the problem, refrain from bandying erronious statements while not contributing anything to solving the problem except basically stating “Cubase done it.”

#10

Would you mind posting the system specs of each one of them, if its not too much asking?

BTW the “record click” test showed here something around 2ms too. I guess the 1-2ms range could be considered normal when recording audio in Cubase

would you mind posting the complete specs here please?

TC audio interface doesnt have MIDI I/O. I´ll try to get a MIDI interface from a firend but, again, a this midi interface is connected to the computer thru USB, so we still have the 2 paths: FW-USB. I also have to mention that the early MIDI appears not only on the Keystation 88 Pro, but also on the AKAI LPD8 (and i´d bet my money that it will most likely appear on any MIDI device i ´d connect to this computer). This rules out faulty dirvers on the M-audio - in this matter, there´s also the fact that MIDI is recorded normally on other sequecers.

#11

Holy crap Conman, cant you read??

The text you quote about the VSTi is in the first paragraph of my OP that refers to case number 1. If you continue reading you´ll see how i expicitly explain that we are not supposed to talk about case number 1 on this thread, only case 2!

#12

That:

  1. MIDI is recorder early no matter which latency value you are working at. This post is about this case, not the first one.

If you can tell me that makes sense as a case at all I’ll even eat your hat.
That looks just like a get out clause to me. If it isn’t why mention the other suff in the first place?
This thread’s getting old really fast. You all seem so keen to keep the problem I’ll let you.
You can have another twelve pages where all help is rejected and where Steinberg silence on the matter tells you nothing other than the fault must be at the hands of very well educated programmers and not part-time laymen bumbling about doing “plunk tests”.
Plunk away. Some day you’ll learn to play the banjo. :mrgreen: Bye.

#13

Enjoy eating Cubasero’s hat, and after that, please STFU. We are talking about a problem (Nr.2) that has nothing to do with another problem (Nr.1), although they are sometimes confused, and you insist that we should rather talk about problem Nr.1, since problem Nr.2 doesn’t occur on your system (it has been posted several times, that problem Nr.2 occurs on several systems, whereas it doesn’t on others. It has also been said, that the nature of this potentially having a hardware cause doesn’t mean it’s not a Cubase problem).

So, I suggest doing what the OP said (plonk test, record click test) to further narrow down the problem, that many people have encountered (including me, although not on my present home system).

My system specs: Cubase 6.07, Intel Core i7-2600K, GIGABYTE Z68X-UD3H-B3, 16GB PC3-10600 DDR3 SDRAM UDIMM, M-Audio Fast Track Pro.

#14

So these are the specs of your correctly working (non-plonky?) computer right?

#15

there´s an important data missing: what do you use to input MIDI and how is it connected to the computer?

#16

Yes, it’s the working one, and my MIDI is of course connected through the M-Audio interface. I’ll check on the specs for the “MIDI early” ones as soon as I get back to them. But I do remember they were running with USB MIDI keyboards, which, as we have suspected earlier, might be part of the problem.

#17

Cool, thanks Thommok.

Anyone else is willing to do the plonk test? It takes just 5 minutes!

#18

This is probably rather related to the 1st problem, if one turns plugin delay-compensation off, the early midi notes problem isn’t as prominent as when delay-compensation is on.

#19

I just did the plonk test (number 2) with latency compensation turned off and i have the same 8 ms difference between audio/MIDI. If you experience a difference its probably because yu only have problem number 1 on your computer. So…no, they are not related.

In any case…Subpantelis could you do the plonk test as described on the first post please? Thanks

#20

Today i got a Firewire 410 from a friend. Did the tests (256-44K) and this is quite shocking:

MIDI from the AKAI LPD8 came in 5ms AFTER the audio plonk
MIDI from the Keystation88 (via USB) came in 14ms AFTER the audio plonk
MIDI from the same Firewire 410 came 2ms AFTER the audio plonk (it was the Keystaion MIDI output, with the 5-pin cable hooked to the MIDI IN of the FW410.

All of these with Timestamped on, which was slightly better.


Some conclusions:

· The audio plonks were ALWAYS closer to the beat (like with the previous plonk test, with the TC Konnekt, where MIDI appeared earlier), so its quite safe to assume there´s something going wrong with the MIDI timing.

· Looks like on the same computer we got MIDI early or late depending on the audio interface we connect. Its not only a time difference but it can change the way it goes: before or after the real plonk, the offset can be positive or negative. So seems like audio drivers play a role here?

· Different USB MIDI devices get a different offset. Some of them are closer to the plonks than others.

· The best results seems to be obtained when the MIDI data is travelling on the same bus (firewire in this case) that the audio interface.


The worst thing of this test for me is that, even if i buy a new MIDI interface to send my MIDI via 5-pin cable to it, it will still be connected by USB, cause my audio interface, the TC Konnekt, doesnt have MIDI connectors so the MIDI will still be travelling on USB and my audio on firewire, so i´ll most likely be suffering from a MIDI offset (early, late…who knows). Its just incredible that a plain MIDI recording can become such a pain in the ass. Specially these days where pluggin your USB MIDI controller and sending MIDI thru it is the most common thing.

Steinberg support guys, can you add something now that i posted these new test results with other audio hardware?