Hey guys. I am trying to use Sonar to play a midi piano track and record the audio back into Sonar at the same time. I don't get the feeling like the midi is playing back 100% in time let alone alone with sample accuracy. What can I do to get rock solid sample accurate midi playback especially while recording (if that even makes a difference). Also my computer is just about as fast as possible so I know that is not an issue. I am using an m-audio midisport usb 4x4 as my midi interface. Please help. Thanks.

"Sample accuracy" is reserved for totally ITB virtual instruments. The "samples" in question are audio samples, so we are talking about timing that is perfectly integrated with your DAW system clock (e.g. 44.1, 48, 96kHz) etc.

MIDI is an old, slow, 1980's hardware protocol for connecting digital musical instruments (Musical Instrument Data Interface). MIDI clock is very slow - nothing even close to a digital audio clock. All MIDI devices have latency - usually very sloppy latency that can't be fixed with an offset.

What keyboard are you using? It will have some latency and slop. What piano are you triggering? If external hardware, it will have some latency and slop. Is it perhaps a GM soundset on a PC soundcard? If so, that must be considered to be an external MIDI device with a *lot* of latency and slop.

To get the "sample accuracy" that we love you need to be running a virtual instrument - a software sampler or virtual piano for example. The only respect in which this will be "sample accurate" is with reference to the midi notes in your DAWs sequencer.

With conventional external midi hardware, if you - for example - cloned a piano midi track and sent it to two identical midi pianos - you would hear a lot of flanging and phasing. That's because of the latency and slop. MIDI is a *serial* protocol - so two identical notes to two identical midi machines have to be sent one after the other.

Within a DAW, although we are using MIDI to communicate with the outside world, the notes are processed internal at system audio clock speed - way faster, and *parallel*. That means if you clone a piano track and send it to another instance of the same piano, the two tracks will be perfectly in phase, and they will simply sum to a 6dB boost. That is the perfect "sample accuracy" of which we speak.

You will still have possible problems with MIDI latency and slop with respect to your keyboard playing and the notes recorded. With good fast USB drivers, this will be well within the tolerance of a human player and not likely to cause musical problems.

If you are recording an external digital piano, I wouldn't expect the recorded midi notes to line up perfectly with the audio notes recorded. MIDI hardware just isn't that good.

But you could perhaps make use of software quantisation options to align your midi notes with the recorded audio notes. Perhaps use the audio to create a groove template, and then quantise the midi to that groove template.

But if you are then sending the midi notes back out of the box and trigger the external piano - the latency and slop will undo your efforts.

What exactly are you trying to do? There are many tools for doing most things you can think of.

Well actually I just tried reducing the latency in my apogee control panel to 4 ms the stream buffer to 2 ms and the io buffer size in sonar from 1024 to 128. This seemed to solve most of the problem. I am using a yamaha cp300 as my external keyboard I am wondering if I would be better off (soundwise too) with a vst instrument like synthogy or something.

Hey, here´s a radical idea: Record the piano as an audiofile (Yes, it is actually possible )! Then you edit the audiofile. It will be rock solid. And sample accurate.

(...And yes, I am serious)

I agree. Forget MIDI if you need really tight, repeatable timing. I believe a lot of the ultra-tight quantised '80's stuff was actually analog sequencers. Or internal sequencers of drum machines and samplers, with the audio recorded without the use of MIDI.

I've tested several MIDI interfaces for both jitter of MIDI events and alignment with the audio stream, and the best that I have found so far is a the Fireface 800/400. The jitter of MIDI events with it is under 1ms; it's the only interface I've been able to get my hands on that compares in timing to the MPC4000. The FF400 is 2x2, whereas the FF800 is only 1x1.

I've compared it to the Digidesign MIDI I/O, MOTU MTP/AV, Edirol UM3-EX, and a couple of oddball interfaces.

-synthoid

Sample accurrate isn't gonna happen. Even with ITB plugs, IME.

However, the key to my WIndows setup's timing getting usable (and up to the standards of my old Macs) was going BACK to using a parallel interface. I bought a MOTU MTPAV for $60. After going through all kinds of interfaces--USB and PCI, it's provided the solidest timing I've found on the Windows platform.

If you're using an external sound module for the piano sound, you'll do much better with a software synth plug-in. Once everything is internal within Sonar the timing will be perfect.

--Ethan

Thanks a lot guys your input was all very helpful

Hey Synthoid how have you tested your midi interfaces? In my case the best timing I found was using a Lynx One with Sonar. I played a steady midi pulse out of sonar and back in to another track. I then compared the two tracks. You get interesting results in that tempo and timer resolution give different results. Generally I got 2-3ms variation between events at worst using this test which is probably a worst case since you are mixing together variations from playback and recording rather than just one. Other sequencers I have found to be far worse on the pc. Reaper is terrible using this test. Cubase can be bad or average depending on interface settings. My Amiga and atari are much better with the Amiga doing the best on this test. I could play and record a steady pulse with almost no variation when using a resolution of .89ms. Amazing.

In regard to getting good timing, You should set your tempo and resolution so that each clock tick is about 1ms. You'll have to do the math for your situation. Anything less is just wasting cpu time since midi note/off messages will take about 1ms to travel anyways. You can circumvent that on playback with multiple ports.

Diegel - it sounds like you are trying to play a vsti and record the output simultaneously. I would suggest you don't. Record the midi first then listen and edit and finally render to wave. The biggest issue you will have will not be midi but the latency and jitter when playing a vsti. Midi has little to do with this. Whether hardware or software the latency introduced by an instrument is really a function of the audio engine and how it handles the incoming midi.

Jon

No, it won't be. I can't believe how much we've come to accept in the way of shit timing and latency and what not over the last ten years. MIDI's timing was flawed then...it's MUCH more flawed now. We even allow for poor "autocompensation" for plug in latency, which now makes the timing of digital audio almost as sloppy-albeit less random.

I don't even MIDI anything I have anough fingers to play anymore. Only strings and (scratch) drums get any MIDI recording here. I have a really expensive, dedicated C4/VSL string sequencer...that's it. Want an organ? EP? Piano? Stereo into the converters...or better yet--the mics in front of the instrument.

Get an

Logic 8 coupled with Fireface400s/800s as MIDI interfaces is giving me resolution that is just under 1ms, just as you report for the Amiga. I used an MPC4000 for years to do my sequencing, and I tried several times to switch to Logic, but prior to Logic 8, and with the MIDI interfaces I had tried previously, the timing was awful (resolution in the 10s of ms), so I stuck with the MPC.

I've measured the timing in a couple different ways. I starting measuring MIDI jitter by sending the MIDI events from my sequencers into a simple (hardware) sound module that outputs a sharp transient, and recording that into an Alesis HD24. I then compare the distance between transients to the distance between MIDI events in the sequencer. I also like the simple MIDI loopback test you mention (loop the MIDI output back into the sequencer, record it on a second track, and compare to the original track). As you say, there are a lot of variables being smeared together by this test, but it is a quick test to perform, and most MIDI interfaces I have tried do very poorly in it. With the FFs and Logic 8, both of these tests give me latency and jitter in the neighborhood of 1ms. As a nice side benefit, the FFs do a good job aligning the MIDI events with the audio stream, since the single device is handling both jobs.

-synthoid

Hey you might make me want to switch to a mac Oh I wish the market could support an os/computer that would be ideal for music and media. Beos was my last hope but that went down the drain. Maybe the guys at haiku developing a freeware beos will actually make it happen (they are doing amazing things if you haven't looked). The Amiga still does it for me if only there was development but that is not going to happen either. Still for midi it is hard to beat and there is a ton of creative software you won't find on any platform other than the Atari. The mac is interesting but the os footprint seems so large and I look at my amiga taking up about 6mb for the os and think why. I'll look at the mac again if you are getting such good results. I gather from your comments that the performance is highly dependent on the interface used.

Hey Synthoid by the way, do you have any experience with the firefaces on windows? I wonder if the good performance you are getting can be had on XP.

Jon

Like pan said, Atari or (what was Sonars predecessor that ran over DOS, that had pretty tight midi timing).

RME's interfaces are solid midi devices also.

Rich

Happily, I do not have any experience making music under Windows! That's why there's still a spring in my step and a smile on my face.

-synthoid

I get sub-1ms accuracy using Nuendo and Midex 8 with Win XP.

It all comes down to the class of driver model that the hardware maker chose.

There is an excellent explanation below:

8 Port SE Drivers

Windows is capable of the same MIDI timing as OS X, regardless of the apparent need by some Mac users to feel superior. Which although I don't really uinderstand, is still pretty amusing most of the time.

Yeah, I think it is very much dependent on the interface and its drivers.

I'm not sure it is very meaningful to compare the OS footprint of a modern computer to that of the Amiga. I mean, I am no fan of bloated operating systems, but the fact is that 6mb was a pretty big fraction of the memory that the amiga could address. I think the real difference when it comes to MIDI timing is that on those old computers you could write device drivers that would talk directly to the serial and parallel ports with very little overhead. MIDI is basically an RS-232 protocol. It's not nearly so natural a match for USB; that's the main reason that MIDI timing has gone all to sh*t.

On top of Logic and all the commercial software for it, there is a ton of good open-source music software available for the Mac, since it's a Unix machine. That comes in very handy for me, as I like to write my own utilities to support my sound design and synthesis work.

I am sure that someone who is as comfortable with Windows as I am with Unix would find the Windows environment to be very productive as well. I'm not religious about computers, I just want to take advantage of what I already know and work with a minimum of sysadmin hassle. In my case, the only sensible choice is to use Macs. I like Linux also, but the commercial audio software for it doesn't compare to what's available for Windows or OS X.

-synthoid

Yeah I know. Just a personal thing with me. I like things neat tidy and well ordered. Anyone who has tried to find there way through windows registry will know what I mean. Sometimes I look at OSX and think it is just out of control but you know it comes down to does it work and work reliably over time and maybe OSX does and that's what counts.

I'll have to look into the midex interface. I also wonder if it's related to the app as well. Would the midex be just as good with Sonar as with Nuendo? Not sure.

Someone else also mentioned the dos versions of Sonar. There actually are none technically as Cakewalk for windows was a different program altogether from Sonar. Cakewalk for DOS was what he was referring to and the timing there was not any better in in my experience but that was maybe even more dependent of interface and driver as each interface had to have a new driver written for it. The timing was as I remember with my MPU401 in the 2-3ms window range.

Jon

Jon

Maybe your computer is not set up properly so you have background services and disk indexing etc kicking in? MIDI timing in Sonar has been rock solid here for years.

--Ethan

That's the convenient eternal Windows excuse, isn't it?

Is it possible to get accurate midi as in keyboards or hardware sequencer from PC or MAC? (besides ATARI)

I tired alot of things but can't get midi to be accurate.

O'm using Windows 7 64bit Pro OS.
My Computer is custom build (4th DAW computer I built in past 2 years I hope I will stay with this one since I did alot of research and transitioning) :
ASUS P6T ('regular' not deluxe or se),
i7-920(not overclocked, core parking, HT and other stuff turned off according to audio optimization guides)
Kingston 6GB DDR3 (1066MHZ i believe)
3 HardDrives, 1 external 1TB HD usb.

Sound Card Focusrite Saffire Pro 40 (also, I had many audio interfaces before: Presonus Firestudio, M-Audio 1010LT, MOTU 828MkII, E-Mu, even tried onboard audio and asio4all).
I also did latency test utility and it showed that audio is ok.

I get MIDI latency too, but I 'compensate' for it by shifting each midi track according to buffer size/latency.

This happens on all my midi devices: Motif ES( both on usb and edirol um-1sxmidi ports), Fantom-G (through USB or Midisport 1x1)(Fantom-G bit better than the rest of midi devices in terms of MIDI timing), Tr-Rack through edirol um-1x. I have also tried Saffire's MIDI (as well as midi on the previous audio cards I had).

I had Triton Extreme before, much better timing than any midi device I did more than 10 tracks going to Triton Extreme without noticeable MIDI timing shifting.

I have also tried installing windows XP, and win7 32bit before but MIDI was the same. an Reinstalling Windows.

I tried many different things for the past year or so trying to solve this MIDI problem.

I still have an old parallel port midi interface. Is parallel more stable than usb?

Getting accurate timing of any type on any modern machine is a challenge that nearly always results in a compromise, regardless of the OS.

Sorry if this has been already mentioned here ?

I have a MIDI setup I have been using now for 2 decades now, I use a MIDI patchbay MX-8. You can use other brands as long as you maintane a simple as possible straight MIDI in/out and not go through several MIDI thru's on the way to your sound modules. If you keep it simple like having only one master MIDI controller and or one master DAW sequencer, you then should have better MIDI timing results. Keep the MIDI cables as simple straight connection to the Master MIDI source(sequencer/DAW) as possible. Not going through multiple MIDI thru's. So that each MIDI sound module is connected directly with the master MIDI controller device, all being connected at the same time.

This works great when you are both playing back MIDI tracks and recording new MIDI tracks, it really does cut down on MIDI timing errors and latency.

For example, what I do with my MIDI patchbay is have my computer DAW and master keyboard MIDI outs both merged going into the MIDI patchbay MIDI ins then going back out to the multiple MIDI outs on the MIDI patcbay and then connected to all my outboard sound modules MIDI ins at once.(with the MX-8 MIDI patchbay it does not effect the MIDI latency or timing when merging the 2 master MIDI ins to multiple outs) For me this lowers the MIDI timing errors because each sound module is getting a direct link from the master controller or MIDI sequencer.

You can use this setup also with a master drum machine and a master keyboard controller, having both MIDI outs on the 2 MIDI master controllers merged into all your MIDI ins on all your sound modules makes it easy to both record and play MIDI at the same time. You can also use other MIDI setups as well with a MIDI patchbay, such as a drum machine/DAW/sequencer, multiple master keyboards etc.

MIDI patchbays eliminate all that MIDI daisy chaining ins/outs to thru's. MIDI patchbays are also great for changing your master controller or master sequencer setup instantly without having to unplug and plug in again and again. If you have more than one MIDI master controller and or MIDI sound module, then you definately need a MIDI patchbay !!!

You can also do seperate MIDI out interfaces on your master controller/DAW/sequencer if you want to send more than just one bank of 16 MIDI channels. Sending on seperate MIDI outs and seperate MIDI interfaces is best for MIDI accuracy. It lightens the load and spreads out the data on multiple seperate 16 channel MIDI data serial streams, by doing this on multiple MIDI banks you will get better accurate MIDI playback and recording.

Lets say If you have all 16 MIDI channels/tracks being played back and being recorded at the same time , using only one MIDI out interface, then you probably will get a major slow down (MIDI timing latency errors) using just one MIDI bank. So if your MIDI songs are complicated with allot of MIDI data streaming ,then it is best to use multiple MIDI banks.

As a general rule, use 8 MIDI channels on playback at once per MIDI output bank. There is no set rules for this just as long as you keep the data stream as simple as possible on each MIDI bank. You can use all 16 MIDI channels on each bank as long as it does not overload the MIDI stream on that particular bank and or cause timing errors. You can hear it if the MIDI stream is over loaded.

You can Get lets say another MIDI output interface and output on that seperate bank for when you are recording allot of MIDI drum tracks, and or for allot of MIDI notes playing back at once, especially for allot of MIDI controller information being recorded and played back, all at once.

You just may end up needing 64 MIDI channels ? (4 seperate MIDI banks) I am not saying you should use all of 64 MIDI channels, but rather just use as many MIDI out banks that are needed, and use only the MIDI channels that are necessary for the MIDI accurate playback of your song, thus keeping your MIDI serial timing errors down to a bare minimum for a rock solid MIDI performance.

So keep your MIDI data strean as simple as possible, and use as few MIDI channels as possible for each bank. I have used all 16 channels on one MIDI bank before and it does work. But if the MIDI serial data is too much, then you may get permanently stuck on MIDI notes, and many MIDI timing errors, and some MIDI tracks playing slower and or faster than others at random. So everytime you play back the same sequence of MIDI tracks it sounds different each time. Random MIDI track timing differences. This has happened to me before when I use just one MIDI bank in a fairly complicated song.

ANother words your MIDI sequencer sounds like really drunk musicians all playing together, which may be a good thing !?!


But then you don't have to worry about all of this MIDI timing stuff, rather just use your DAW to convert your MIDI tracks to audio tracks as soon as you can. Record your external MIDI sound modules into your DAW or multitrack recorder by playing your MIDI tracks 1 at a time on your MIDI sequencer, this way it will not overload the MIDI data stream and timing errors will not be recorded. So after this you can tweak all your New MIDI recorded audio tracks with your DAW, for sample accurate playback and rock solid timing, everytime.

MIDI is great for writing your music initially and getting your ideas down fast to a MIDI sequencer, but then you should not completely rely on MIDI throughout your whole song. But many artist and people do rely on MIDI. So It isn't all that bad, MIDI is especially good for getting your music printed out in music Notation.

You should when ever possible record your MIDI instruments live using the audio line outs on your keyboard or sound modules. It teaches you how to play and practice your musical style and teaches you to remember how to play every part of your song live. This ultimately prepares you for band practice, or for up and coming gigs, and or for live studio recording.

That link is dead, Wayback Machine gave me this:

Hi all,

I know this is an old thead, but I was wondering if getting a dedicated midi interface would help with the midi latency issue. Basically I already have an RME 802 but if I mess with the buffer settings, this ultimately messes up the midi sync to my hardware JV2080. If I got a dedicated midi interface to USB, would this solve the issue and be excluded from the RME's sample/buffer settings?

Please let me know. I am new to hardware gear and would love to integrate the sounds into my midi sequences.

Thanks.

Expert Sleepers devices is the only way I know to have 100 percent rock solid timing, although others at GS have sworn by MOTU midi devices.

My fear is if I buy the MOTU or any other Midi interface, it would not eliminate the latency I am experiencing. I am just wondering how did people in the 90s do this. especial folk doing 90s dance and techno with these hardware sound modules and still managed to achieve proper timing in their DAW or sequencer.

Please let me know if you have any ideas.