From: Lord of Wolves [email protected] Reply-To: [email protected] >To: [email protected] >Subject: [VST-Mac] USB MIDI (was Re: Pluggo) >Date: Fri, 26 Jan 2001 00:35:10 -0800 > >At 9:16 AM +0000 1/25/01, NYMo [email protected] wrote:

> >Hi Orren, I've got my G4/533 Mp on the way too...what will you use for a midi interface?? You know, that's a really good question. :) I've had a B&W G3 for a few years now (got the first rev. 1s off the >line) and I've been using 2 MIDIsports. But I really must say that I >don't like them, I find them flakey, and I don't trust midiman's >drivers (yes, even the brandest spankingest newest ones, that I have, >that supposedly solve everything). Since they are working (mostly) I >don't *have* to run out and get a new MIDI interface, but I really am >considering it. > >Steinberg's Midex looks good, but I heard they are holding up >releasing mac drivers until we get a new congress in the US in 2003. >;) I also am looking seriously at the Emagic interfaces. VERY >seriously (I'm a Soundiver user as well). I'd personally only go >with the MOTU USB MIDI interfaces with Digital Performer, but a lot >of people do get them to work with Cubase, and well.

Now, I know that this comes up every so often, that the Mac gets a slagging >for its Midi timing, and I am not going to join the debate as to whether >the >old Atari was more stable. > >BTW, VST5 has a much better timing engine in it. I have a test song which >we >use in-house to check the 'apparent' timing of various devices by driving >them to the absolute theoretical port-speed limit & its playback is the >closest I've ever heard to accurate. There isn't a PC we've tested yet that >can play it, but actually that's cheating a bit because the PC's absolute >port speed is slower than the Mac. > >However, you have always to bear in mind exactly _what_ you are asking the >machine to do. > >Let's take your example... >You send a stream of single notes to one sound module (this is, of course, >taking no account whatsoever of how long it takes the module to respond to >the message, which can be anything from 7 - 30 ms depending on >model/manufacturer.) > >Midi is a serial transmission, every note-on takes 1 millisecond to >transmit. So does every note-off - something most people forget. (I'm not >interested in Running Status here, by the time you're running a complex >track, it becomes less important) > >If you have your 5 notes all happening at once, you've got a 5ms lag from >the first to the last - if your notes are all exactly one 16th or 32nd long >- you're doubling that lag by having the note-offs also occur exactly on >the >beat. >Add your next 4 tracks & you've generated a potential 18ms spread between >the first & last note on - and that's if the receiving module can process >at >ANYTHING LIKE the speed you can send it down the serial port, most of them >are nowhere close to keeping up - and we haven't even mentioned how good >the >Midi translator box is at keeping the stream going. > >It's always possible that the tone module's rolling DVA (Dynamic Voice >Allocation) isn't getting enough time to release the oldest note in time >for >the newest note to be triggered, so you could actually be running out of >polyphony even with apparently only half the maximum number of notes >running. Any given sound will have a minimum gate-time that the system will >hold onto it before it can let it go to perform another task. Drum sounds >are notorious for this, as they can have a fixed gate-time of up to 500 ms, >even though the sound itself may be shorter. > >It is perfectly possible, by a little judicious juggling (nice alliteration >there G ;-) to get a whole lot more than what you have there going in >"apparent" time. >OK, so I work for Yamaha, so that gives me plenty of time to get paid to >piss around with this kind of stuff, but without even changing the position >of a single note-on, you could randomise your note lengths, so they're >spread, change the order of the tracks on screen - Cubase reads down the >page, so put your most time-sensitive stuff at the top of the list & it'll >get sent out first. > > Get rid of all the info in the Inspector panel & put it separately in a >set-up bar at the head of the song. That'll save the bloody info being sent >out every time you cross a box. (Sorry, my pet hate, the Inspector - the >most oft-ignored source of timing problems) > > Get any controller information off the beats - throw it around at random >or >just move it all 20 ticks early. It takes most tone modules between 20 - 50 >ms to respond to a CC anyway, to stop zipper-noise. > > Get any slow-attack sounds off the beat - slide them early to keep up >apparent timing. > > If the tone module can do it, set each channel's voice-allocation to >exactly or only slightly above the actual number of simultaneous voices >you're using for that channel, so you don't get note-stealing from other >channels. > >Try borrowing a Yamaha MU 50, or preferably an MU100 or 128 - NO, I'm NOT >trying to flog them, just eliminate one source of delay - run it using the >"to Host" port directly into the Mac instead of using your translator box & >see how much faster the whole thing responds. The MUs have quite a large >buffer compared to most single/dual port translator boxes & are running on >a >much faster chip-set. (OMS thinks they're MTPs. running at 1MHz) > >Just to keep the balance & so you don't think I'm trying to flog Yamaha >gear... for a laugh you could borrow an EX5 & try the same with that - it >has a response time measured in weeks!

I have tried this out on various combinations of track/module/channel >assignments. >The first time I noticed this effect....the 4 x piano 16th arpeggios were >all >on *different* VST midi tracks (vst midi tracks 1 to 4 inclusive)...but >all >assigned to the *same* midi channel.... and controlling one channel on a >Roland JV3080 module...whilst the 32s were on a *single* midi track >(5)...and >assigned to a *different* midi channel on the same Roland 3080. Assigning >*different* midi channels for each 16 part did not steady up the 32 part. >The four piano 16th parts play steadily enough to be usable on their own. >When de-muting the 32s part, the piano parts get a fraction sloppier (but >still usable), but the 32s are all over the place...sounding like >randomized >drunken triplets. On muting the piano parts one by one...the 32s tighten up >, >until when playing alone..they are steady as programmed. In this test, only >5 >midi tracks are playing at maximum. >I then entered the same note data, but as 32nd piano parts, and a 64th >percussion, and halved the overall tempo....but no change or improvement. >I have re-inputted all the data several times from scratch...done the same >experiment...but no change. > >Now here is the *weird* bit: >I can copy and paste the same 16ths data...to 4 more midi tracks (6 to 9 >inclusive)...and the same midi channel on the 3080 as the original piano >parts...mute the original 16th parts on midi tracks 1-4, and play the 4 new >piano parts simultaneously with the 32s.....and now the timing sounds OK, >*including* the 32 percussion parts!!!! > >Yesterday, I did the same experiment on a different computer with the 16th >parts on an Ensoniq MR Rack, and the 32s playing back on a Kurzweil >K2500...same effect. However today, I tried the same experiment ... the >16th >parts went to a Roland JD990 module and the 32nd percussion parts to an >Alesis DM5 drum module...and everything was fine! >So as of yet I cant tell if its the midi timing in VST which is the >culprit, >or the midi timing resolution in the synthesizer hardware itself. I am >still >a tad baffled as I haven't found a *completely* repeating trend as of yet, >except using the Roland 3080. > >One other thing while I am in here. I recall a thread awhile back dealing >with latency on the Sonorus studi/o card. Any clues as to how to set this >up >to get the theoretical minimum (of 6 mS or so?) without either (a)crashing >the computer, or (b) getting loud'n'nasty distorted splatty farting sounds >each time I hit stop during playback? (I've pretty much tried everything so >far without any joy). > >Thanx, >Nik > >TAT wrote: > > > > > >Hi list, > > >Has anyone come across this internal midi timing issue? Here are the > > >symptoms: (you could try this yourselves, see > > >if the same problems manifest...) > > >Here is a way of observing this effect: > > >There are 4 midi tracks up, each with a pattern of 16th-note > > >arpeggios, playing the same sound (piano) on one midi > > >channel of a Roland external module. > > > > Could you clarify this a bit please? Are you addressing 4 tracks of > > arpeggios simultanesouly to the the Roland module? > > Are you also assigning the 32s to the same module? > > > > Some Roland stuff has atrocious MIDI interfaces. Which module?