ORIGINAL: dewdman42

ORIGINAL: pianodano
I guess what I was trying to say is I can spend several hours tweaking a part to sound so good (to my ears). Freeze it and it can play ok for a few measures and suddenly for no apparent reason, start stuttering ever so slightly. After working on the part for so long, you would have heard the part so many times you know every nuance and easily recognize when something is amiss. It never seems to regain the timing again once the rendered tracked gets fouled up. As I said yesterday, I quit trying here and just put it on tape now. It's one of those time is of the essence things if you know what I mean.

Yea, a frozen track is "frozen". It will never change after that. The question is why the "stuttering" you mention is getting frozen in. When you say "stuttering" can you describe that a little better..you mean that it sounds like a guitarist with bad sense of rhythm?

Yep - a couple beers too many

ORIGINAL: dewdman42

I'm really most suspicious that RealGuitar has bugs in how its handling midi timestamps, but there might be a way to work around it, try the fast and real time bounces. According to Noel, there are some issues with multi-processor machines sometimes. He says the work around is to do a real time bounce instead of fast bounce. I would try without even using "freeze" per say. Just hit record on the audio track and let it play until the end, capturing the audio performance on a track from your midi track...just like you would be doing if you were sending it to tape, but send it to an audio track instead. See if that works. I think that is more or less Noel's suggested work around for these kinds of plugin bugs.

Are you saying the 1.5 version is giving you the timing problems, the 2L version, or both?

2L. I don't ever remember freezing a track in 1.5

ORIGINAL: dewdman42

Disk streaming is another area where its possible that if it can't keep up with the streaming there might be some timing glitches potentially I guess. If there is a way to turn off disk streaming, try to do so to see if that helps.


I am pretty convinced at this point that there is nothing wrong with Sonar, but apparently there are plenty of ways for plugin developers to screw things up.

Someone else reported issues with TTS, so its not unique to only RealGuitar..but perhaps some common traps that some plugin devs might fall into. On KVR they pointed out that the original SDK for DXi had a flawed example project that did not correctly check timestamps and its possible that some plugins out there were based on that flawed example. And guess what, the nature of that particular flaw is/was that if the audio buffer size is increased the jitter gets worse. Sound familiar? TTS is a prime candidate for this possibility since its a darn old DXi based synth from way back when DXi's first came out. It was probably created in japan by Roland and nobody's looked at it since, but I could be wrong.

Complicated, I know that much. And so far I've only talked a little about RealGuitar.

What a LONG read. I've been wanting to post this for several hours, but did not want to till I read the thread. I just wanted to say that I argued with cake tech guys a couple of years ago about midi timing (no it wann't Noel! Glad he's here). I had some midi percussion I was layering with another midi track. I will say that the project was loaded! CPU was very high, audio buffers was very high 300+. It's the only way I could get the project to play. This was a AMD2400 and Sonar 5, & Layla 24 midi. I only noticed the drift, because there were percussion hits that were supposed to be at the same time. What I got was a flanged sound because of waveform phase cancellations. The cake tech guy told me I shouldn’t zoom in that far and that I could not hear what I was hearing! This timing was way more than a few ms. I did 2 things to help. I changed from a Layla 24 midi port to a Emagic Unitor 8. This helped some, but what helped the most was to get a dual core computer. This faster computer allowed me to lower the audio buffers. I don't exactly know why lowering the audio buffers would change the timing accuracy of midi, but it did. I’ll be interested in learning more about this.

Oh Yea, if you knew what keyboards were in my avatar you would know I'm older than in that photo!

Max Arwood

Spelling error again, If I had to choose spelling talent or music talent..........

ORIGINAL: MArwood

What a LONG read. I've been wanting to post this for several hours, but did not want to till I read the thread. I just wanted to say that I argued with cake tech guys a couple of years ago about midi timing (no it wann't Noel! Glad he's here). I had some midi percussion I was layering with another midi track. I will say that the project was loaded! CPU was very high, audio buffers was very high 300+. It's the only way I could get the project to play. This was a AMD2400 and Sonar 5, & Layla 24 midi. I only noticed the drift, because there were percussion hits that were supposed to be at the same time. What I got was a flanged sound because of waveform phase cancellations. The cake tech guy told me I shouldn’t zoom in that far and that I could not here what I was hearing! This timing was way more than a few ms. I did 2 things to help. I changed from a Layla 24 midi port to a Emagic Unitor 8. This helped some, but what helped the most was to get a dual core computer. This faster computer allowed me to lower the audio buffers. I don't exactly know why lowering the audio buffers would change the timing accuracy of midi, but it did. I’ll be interested in learning more about this.

Oh Yea, if you knew what keyboards were in my avatar you would know I'm older than in that photo!

Max Arwood

Thanks for your input. I don't even know who started this thread, but I'm glad they did. Re your avatar, I see the B but I can't tell exactly what other others are. The one you're playing with your left hand really looks like a RMI (I remember the white tabs in the middle on mine) but I've never seen one in a wood case. Is your Hammond black ? I restored mine last year and did it in black. It is a very early 56 and I'm older than that.

Did you buy your new machine from one of the specialty type computer guys. I really hate to throw another $2500 to 3 grand into this stuff now because I have always had some timing issues but never before as bad as they are now. But I will if it's really the answer to solving timing problems. In desperation, I spent most of this year getting the tape recorder rebuilt, buying and setting up the new analog console and procurring vintage outboard compressors, reverbs and such so I could try to get away from these troubles. I just couldn't deal with it any more. The hard part for me is I love my sample libraries and I thought to just use the application mainly for edits and the libraries.

So, were you seeing recorded midi (and sometimes audio) events as late as 40 to 50 ms? Please tell more about the problems you were having in the past. My 1st machine for Sonar 3 was a AMD and I had no luck at all with it. The P4 I replaced it with would handle 30 or more audio and untold midi in 4, a bit less in 5 but in six I don't know what has happened. Task mgr cpu performace shows spikes at 98% running 9 audio and 24 midi tracks.

Dewdman, what's wrong with the metronome ? I haven't read the links yet. Is it late? eratic ?

Danny

RT:
I know this sounds weird but, I had better midi sync in WDM mode. I changed computers during S5. It was so much better I have not even checked S6 or S7. It was a rubber-band-ish "jitter" timing. Some notes ahead of where they were supposed to be and others behind. At first I was kinda freaked out that some notes were recorded before they were played . Then I just figured that it was buffer stuff.

Danny:
Avatar - I think it is a MS20 on top, Korg string ensemble, Unknown synth, Crumar T1, Unknown Piano, and a clav of to the right. I wished that was a B, but the T1 was great through a Leslie, and wow was that T1 light! Now I play a K2500 and a Motif XS. Either one of these keyboards blow away all that whole stack!

I had a DAW computer built. Jim Roseberry did it for me. I spent about 2500.00, but you could get the same one for less than 1/2 that price now. I would probably get a quad core now. The thing I really like about my computer is that is has no fans in it. Well, it does have and Aux fan I can turn on when I need to. I have it in an enclosure, but I it's so quiet I leave the door open most of the time. I like to be able to put in flash drives and CD's with out messing with the door. When serious recording starts I shut the enclosure.

Now I'm really curious. I still have 5 installed, I'm gonna check and see if 5 is different than 7.


Max Arwood

ORIGINAL: Jim Wright

ORIGINAL: Nick P

I just hope the Cakewalk guys take notice of this thread and test and improve (if needed) the MIDI timing of Sonar and Project 5. If Ableton detected and fixed this issue, then based on logic, and my ears, I will assume that it is an issue with Cakewalk sequencers as well.

Nick - In this post, and your first post of the thread, you seem to be implying that Ableton is doing a better job with MIDI timing than Sonar and Project 5.

Do you have any factual basis for this opinion? Have you compared Ableton and Cakewalk products, head-to-head, on the same PC, with the same MIDI interface? (I haven't done that comparison -- I have an old copy of Live Lite, but that's it.)

Just curious - and I wouldn't want to slam either Sonar or Live without being fairly sure of the facts.

- Jim

Jim - I wouldn't use the word "better". And I'm definitely not making any assumptions here. It could be that Cakewalk has done extensive MIDI tweaking which we don't know about. We've yet in this thread to hear from any of the Cakewalk guys. What I'm saying is that it looks (sounds) like this is an issue much more general than applied to any one software package. It very well may be an issue relating to the whole way computers deal with MIDI input and output as opposed to hardware sequencers (which of course have small computers within them, but not with Microsoft operating systems!). Possibly Ableton is just a bit ahead of the curve in recognizing and dealing with the issue.

After all of these posts and all of the interest in this thread (which amazes me), I just hope someone from Cakewalk will come in and let us know what is company's policy and philosophy on this issue.

As regards testing, with all due respect to the many scientific methods expanded upon in this thread, I still think I will stick with the "ears" test if I ever get around to it: Record some unquantized MIDI playing on a weighted 88 note controller to Sonar, my MPC2000, and maybe even dust of the old Roland MC-500 MKII and see how they all sound (feel).

Original: Nick P

Jim - I wouldn't use the word "better". And I'm definitely not making any assumptions here. It could be that Cakewalk has done extensive MIDI tweaking which we don't know about. We've yet in this thread to hear from any of the Cakewalk guys.

Actually, we have: Noel has posted a number of times.

What I'm saying is that it looks (sounds) like this is an issue much more general than applied to any one software package. It very well may be an issue relating to the whole way computers deal with MIDI input and output as opposed to hardware sequencers (which of course have small computers within them, but not with Microsoft operating systems!).

It is a more general issue. Some companies have dealt with it by creating integrated hardware/software packages. MOTU did this with Digital Performer and their inhouse MIDI interfaces (the combination has very low jitter, although I'm not sure about their jitter figures). Digidesign used a similar approach with their "MIDI I/O" box (10 ports, $595 list; timing is claimed to be "accurate to far less than one millisecond").

In tracking down the Digidesign reference, I came across people complaining bitterly about MIDI performance with PTLE on a Mac. So, it's not just a Windows issues (although I suspect that CoreAudio generally does a better job of handling MIDI I/O than XP... I won't talk about Vista).

Note that Windows does have the necessary APIs to support high-resolution time-stamped MIDI (e.g. DirectMusic Core). However - an awful lot of software still uses the old "MM" APIs, which do not directly support time-stamped MIDI I/O. Further, the DirectMusic Core APIs probably have issues (The earthvegaconnection guy reported long-standing bugs with how DirectMusic handled SysEx, and there may also be issues with high-res system clock stability). Also, DirectMusic Core is not being developed any further, and I don't know what (if anything) Microsoft plans to replace it with. DAW users are a very small market....

It may be that the industry (music software industry) needs to develop their own standard for high-resolution MIDI ports, and not depend on Microsoft (or Apple) to provide it. Consider ASIO, VST and the like. Consider the MFX and DXi standards that Cakewalk pioneered. Of course, any such standard would need to be cross-platform (that's been a huge advantage for ASIO, VST, Rewire over the years). There is an open-source effort underway to do just that ("PortAudio", which includes "PortMIDI"), but it hasn't gotten much traction in the commercial software world. (See papers at http://www.portaudio.com/docs/, including the ACMC2003 paper that has lots of detail on Windows-specific issues).

Possibly Ableton is just a bit ahead of the curve in recognizing and dealing with the issue.

Now, here I think you are making an assumption. Unless you've tested Ableton and Cakewalk products head-to-head, you've got no basis for that statement. It seems equally likely that Ableton's MIDI performance was somewhat lacking until recently, and they're only now catching up to their competition (e.g. Sonar). Of course, I don't know either way.

I'm looking forward to seeing Ableton's promised "MIDI Engine Fact Sheet", which "documents MIDI timing tests on both Windows and Mac platforms using various MIDI interfaces and describes exactly what users can expect in terms of MIDI timing accuracy." Once that's available, we'll have a better idea what's what. (I wish that Cakewalk would publish a similar document).

Further, Cakewalk has doing things for years, like the 'Windows Audio Professional Roundtable' events at NAMM shows, to promote quality improvements for MIDI and audio on Windows. Cakewalk has also pushed hard on Microsoft, to get them to turn a 'consumer multimedia' operating system into something that professionals could use. I'm not aware of Ableton doing anything like that.

- Jim

Edit: I just rechecked Nick's post. He wrote:

>> I still think I will stick with the "ears" test if I ever get around to it

Okaaayyyy - so Nick haven't tested anything himself. He hasn't even done a listening test. He's just going on hearsay, and buying Ableton's marketing. That's not cool.

To be fair - Live 7 is not shipping yet, so probably Nick can't do a listening test - unless he's a beta tester, and presumably under NDA. Without any hard evidence, any assertion that Live has better MIDI timing -- is just hot air.

ORIGINAL: Jim Wright

Now, here I think you are making an assumption.

Edit: I just rechecked Nick's post. He wrote:

>> I still think I will stick with the "ears" test if I ever get around to it

Okaaayyyy - so Nick haven't tested anything himself. He hasn't even done a listening test. He's just going on hearsay, and buying Ableton's marketing. That's not cool.

To be fair - Live 7 is not shipping yet, so probably Nick can't do a listening test - unless he's a beta tester, and presumably under NDA. Without any hard evidence, any assertion that Live has better MIDI timing -- is just hot air.

Jim - without wanting to stir up any conflict, you are waay off base in interpreting the intent of my comments here. Note that I used the word possibly. That means I'm not making an assumption.

As regards testing, I consider close to 25 years of working with all manner of MIDI hardware and now software sequencers and drum machines enough of a "test", at least as applies to the ears. While it's true I haven't done the waveform interrogation thing using an audio editor, I sure know what my playing comes back like via using different manner of sequencers, hardware and software. And to my ears, hardware boxes like the MPC do a better job of accurately recording unquantized MIDI and playing it back than do software solutions. While certainly subjective, I trust my ears as well.

ORIGINAL: RTGraham
Any suggestions?

The only things I can suggest for now is to record hi-res midi on a good hardware sequencer, or record live audio from an external synth, or only use midi for instruments with a slow attack, and/or curse at Windows. :)

I suppose the most logical place to start would be the MMA (Midi Manufacturers Association) http://www.midi.org

If you click on 'About MIDI', the bottom of the page lists companies that contributed to the document: Yamaha Corp US, Korg USA, Cakewalk, Edirol, Evolution, BitHeadz, MadWaves, PreSonus, Steinberg, E-Mu Systems, Berklee Media, Keyboard Magazine, Electronic Musician.

It's a pity that we have computers 1000x more powerful than they were when midi was first introduced and the only solution to the problem seems to be to return to the MPU-401 Smart Mode concept again that was required on 4.77mh PCs.

RTGraham, dstrenz --

The MMA is a good place to send suggestions and concerns about MIDI timing. Tom White (head of the MMA) will make sure they are circulated (and not "put in a circular file").

If a group of Cakewalk users want to collaborate on some kind of "open letter" or other document, I can forward it to the MMA if you like (assuming I agree with it ). Of course, you can just send it directly to Tom White.

I was on the MMA Tech Board for four years, and chaired the MMA Transport Layer Working Group (which is why I did the "temporal fidelity" work, some of which is published here.) During that time (through mid-2001), I was trying to get the MMA to agree on, and endorse, standards for timing accuracy. It was not possible to get a consensus on the issue. It would be inappropriate for me to comment further on what happened back then.

Many things have changed since then (6 years -- in "tech time', that's a different era!). A new effort might be more successful. However, one fundamental issue is unchanged: to generally improve MIDI performance on PCs (or Macs) - tends to require involvement/agreement by the OS vendor (e.g. Microsoft, Apple). We (DAW users) are a very small market, without much clout. A number of companies have found that it's much more practical to just develop their own solution, and have done so. Of course, for various reasons, such solutions tend to work only with products from the same company.

Note that one big thing has changed since the late 90's. There are now OS-level APIs that handle timestamped MIDI, on both Windows (DirectMusic Core) and Mac (Core Audio). Companies that make MIDI interface hardware can write drivers for the newer APIs; companies that make sequencers can support the newer APIs as well as the older non-timestamping APIs. Then, customers have to "connect the dots" and pick the right pieces (sequence and MIDI interface) to get good performance. If you have a great sequencer and use it with a lousy MIDI interface - you'll get lousy timing performance. End of story.

Note also: a truly low-jitter MIDI interface may cost a bit more. Customers have flocked, in droves, to the cheapest MIDI interfaces they could find -- and that has affected the market (why should a company make a better MIDI interface, if they won't sell enough to even cover development costs?). Dirt-cheap USB MIDI interfaces can't do better than 2 milliseconds of jitter, at best - and may do a lot worse. (That's on XP; on Windows 2K, they would probably perform worse. On Vista? I have no idea). The low-jitter MIDI interfaces I'm familiar with - have an embedded 'micro sequencer' to fire off MIDI events at just the right instant. That costs more than a dirt-simple USB MIDI box, that just fires off MIDI events whenever they pop off the USB bus.

- Jim

Supposedly Cubase has DirectMusic support. Its the only one I know of. Steinberg was probably more motivated to try it because Cubase has had notoriously bad midi timing for years.

The reason Microsoft is dropping DirectMusic has nothing to do with this issue. DirectMusic is this huge mammoth of a development platform for developing midi music for games, which nobody uses now. MS has provided a new mammoth development platform for that which is all audio based, no midi. At some point they have written off the DirectMusic mammoth and there are very very few people utilizing this one little small corner of it to drive their midi sequencers more accurately. So it will be dropped.

The best thing to do related to nudging Microsoft is to plead with them to include the midi queuing capabilities into the basic OS, rather than relying on DirectMusic and trying to support that whole mammoth. MS needs the equivalent of CoreMidi.

Strangely though, if you look around the web, you will find VERY few forum posts, etc.. of people asking about or complaining about midi timing at this level. No wonder MS is not paying any attention to the issue. My feeling is that technical people feel its good enough now. Fine musicians don't feel its good enough now, but they don't have enough technical understanding to get into the kinds of discussions we've gotten into in this thread. So get out there, visit the WDM development forums. Speak your opinion. Go to the Microsoft related usenet groups. Write emails to that department if you can find the email address. Etc..

But, I have to say again, 2ms is actually quite good. I doubt its going to be better than 1ms anytime in the next 10 years without expensive hardware timestamping we have discussed. Most of you cannot hear 2ms. If you hear horrible timing happening, then either your setup could be optimized with a better midi device/driver or there is something else going on. Remember also that if you want to make your computer more responsive to realtime timing, there will be a performance penalty in other areas such as audio.

BTW - in case anyone is interested. I have it from a very reliable MOTU source that the windows driver for the MOTU midi interfaces DOES NOT support hardware midi timestamping as of right now. Apparently this feature has only been supported by MacOS applications that specifically support it. FWIW. It appears to me, that basically hardware midi timestamping is not an option on windows at this time.

In my book that makes the parallel port interfaces king of the hill on windows.

ok, back to previously interrupted discussion...

No USB 2.0 will not get below 1ms. That constraint is imposed by the windows MM timer. But a USB2 interface, if its well implemented, may help ensure that you're getting pretty consistently within the 1-2ms window.

ORIGINAL: Jim Wright
Just to reinforce what you said earlier - 2 milliseconds max jitter should be fine for most people. Jitter at that level is really pretty subtle.

Yep. I think anyone who feels they need <1ms midi resolution with accuracy should invest in a hardware sequencer. 1-2ms is fine for most of us. Over about 5ms is intolerable, for me anyway.

If you're hearing serious jitter - then, most likely, something is seriously wrong in your system configuration./

I think anything under 2ms should just be ignored for the time being. Anything over 2ms can probably be fixed by most people.

The longer discrepancies could be one of several things I think we've identified through this thread:

- Could be system problem in general, though this seems to be more and more rare IMHO.

- Could be a midi interface is being used that does not have the greatest device driver.

- Could be the buggy Sonar metronome (try turning it off, if that fixes the problem, then contact Cakewalk and wail!)

- Could be that the soft instrument plugin is not using midi timestamps correctly. IMHO this is most likely the fault of the plugin in some way, but there is no way for end-users to know for sure whether its the plugin mishandling the midi timestamps; or if Sonar is somehow not feeding it with correct timestamps. However, I tend to think that if Sonar were messing this up, there would be a lot more people kicking and screaming. I believe in this situation, the plugin is probably to blame. If raising the audio buffer size causes it to get worse, this is likely a problem. Try different modes of rendering audio, including real-time mixing to audio, real time freeze, fast freeze, etc. Try with different audio buffer sizes. Eventually you should be able to get it to work right. Then contact the plugin developer and wail!

- Could be that the soft inst is using disk streaming and can't keep up with stuff. I would think that fast bounce or freeze would not be susceptible to disk streaming starvation, but in any case, if your plugin uses disk streaming, try turning it off or all of the different audio rendering approaches mentioned above. Try a larger audio buffer size. Then contact your plugin developer and wail!

- Could be some other bug in the soft inst that mis calculates timing. This could happen due to any number of buggy reasons including threading, Multiple CPU issues, etc.. If you have multiple CPU's, try turning off MP support to see what happens, etc.. If you can identify a focused problem, then contact your plugin developer and wail.

I can't think of any other things that have been identified recently.

ORIGINAL: RTGraham

Without getting into my credentials, I'll just say that I'm fairly certain I hear, and feel, the 2-millisecond jitter.

That's why I said most people. Not all people. There are a very very few golden ear'd people that might hear it. Most people will not. and I would argue that even the golden eared people may only hear it when they are using their hands to play a midi instrument and hearing the jittering latency in some way. People listening to the track will not hear it for all intensive purposes. Anyway, haven't we barked up this tree long enough?? We keep going round and round in circles. Sub 2ms is not solvable on Windows right now. Write your senator and get over it. ;-)

To a certain extent, I have probably lowered my standards and forced myself to adapt to the way it feels and sounds, so that I can manage to get through MIDI sessions without losing my mind, but I would love to have it feel better. I actually was already thinking about getting a hardware sequencer, based in large part on this discussion

if you truly are that sensitive, then I would definitely consider it. Though note that someone else showed that their hardware sequencer was also jittering around a fair bit in the 1-2ms range. So its hard to say. I think the Akai MPC is well liked because it has its own groove templates that give a famous MPC swing to the tracks done with it. But hey...maybe they are tighter. I'm not aware of any critical tests that have been done to compare.

Like I said, if you are that sensitive, its certainly worth a try to get one of those and try it. Pick up an MPC-4000, a Roland groove box, or a Yamaha QY700 if you want more than 92ppqn. I've thought about getting one also, just because I'm paranoid about it. But in the end, I think its overkill, $500 I'd rather save for something else. Still I wouldn't mind trying one a while to see if I can feel a difference or not.

record the tracks on the hardware box. Once its completely ready for soft instrument rendering, then import the midi file to sonar and finish it there. The question is, will the timing differences be worth the extra cost and hassle factor? I know I probably won't feel satisfied until I try it myself to find out.

Even though we're a niche market and a small percentage of Microsoft's market share, we still have a right to request that manufacturers create the tools that we need to operate at the best of our ability - and it seems that a software sequencer like SONAR, with the timing integrity of a traditional hardware sequencer but the graphical editing capabilities we've all gotten used to using, would be one such tool.

Yea, I think most of us would. Keep begging Microsoft.

I would consider going to a Pro Tools TDM system, even with it's bloated price tag, to take advantage of what people have described in the way of Digi's MIDI I/O interface's timing stability - but I just don't like working in Pro Tools as much.

Did someone say TDM midi is more stable? Is that measured? I would not expect it to be any more stable. ProTools has always been about audio and midi as an afterthought. I really doubt its a single bit more stable then everything we've been talking about.