Hi Nick,
1. Re: USB MIDI interfaces
USB MIDI interfaces
cannot perform quite as well as PCI or parallel-port interfaces (assuming all interfaces have decent drivers, etc.). Why? USB MIDI adds at least 1-2 milliseconds of jitter on top of the "baseline" jitter resulting from the use of the 1 millisecond Windows "MM" timer used for timestamping MIDI data.
See
http://forum.cakewalk.com/fb.asp?m=1191358 for more detail about where jitter comes from.
See
http://forum.cakewalk.com/fb.asp?m=1186189 and
http://forum.cakewalk.com/fb.asp?m=1186298 for some recent non-scientific timing tests I ran, on several MIDI interfaces I own. It's very clear that my USB MIDI products (Edirol UM550 interface, M-Audio Axiom 61 with built-in interface) have about 2 milliseconds more jitter than my EMU 1820M PCI card (with 2 MIDI ports). So, for timing-critical parts, I will use the MIDI-DIN outputs of my Axiom 61, connected to the MIDI-DIN ports on my EMU PCI card.
Use your ears. If you are happy with the results you get from your USB MIDI controller (or interface) -- great! Don't let us mess with your mind!
If you're concerned - try a different kind of interface
(PCI or parallel-port recommended; Firewire should also work better than USB MIDI, but I haven't tested that).
2. >> "Also, if USB is not the best choice for accurate MIDI recording, how does that square with the hundreds of USB-MIDI controllers on the market."
Most people don't notice the difference. USB MIDI is dirt-cheap, compared to other alternatives. Microsoft changed Windows in ways that made it harder to do parallel-port or PCI MIDI interfaces. Hooking up a USB cable is easier (and less risky) than installing a new PCI card. Modern motherboards have fewer PCI slots..... there are lots of reasons why the market has gone towards USB MIDI. Market realities dictate that many companies have to use USB MIDI, regardless of whether it's the "best" solution or not.
It may be worth noting, again, that the MMA has endorsed Firewire MIDI (I edited that spec, actually;it's linked from the MMA website) - but the MMA has never endorsed USB MIDI. Since lots of MMA members make USB MIDI products, this interesting fact doesn't tend to get much press. In fairness, I should also note that current USB MIDI interfaces perform a lot better than they did circa 2000 (7+ milliseconds of jitter, always, with occasional spikes well above that -- puh-leeeze!). The MMA might well decide to endorse USB MIDI today, because it's improved a lot, and is usable. My two cents; YMMV. 3. Re: "What is Ableton talking about?"
As dstrenz said - you'd have to ask them. My guess is that they had some sloppy code in older versions of Live, decided to clean it up, and decided to turn a bugfix into a marketing advantage. It's also possible (barely) that Live 7 includes a kernel-level event engine with < 1 millisecond jitter
(see my first linked post above for a ramble about kernel-level stuff ..... but also note that performance with USB MIDI interfaces would not be improved much/at all by a kernel-level engine, because the 1-2 millisecond jitter of USB MIDI would "swamp out" any benefits of a kernel-level engine -- because jitter is additive.) Their marketing claims really don't tell us anything about whether Live 7 is better or worse than Sonar 7 wr..t. MIDI timing. We just don't know. We'll have to wait for Live 7 to ship, and for someone to do some credible testing.
I really do like the idea of companies publishing a "MIDI Engine FactSheet". Here's what Ableton has promised to do, for Live 7:
MIDI Engine Fact Sheet
The MIDI Engine Fact Sheet 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. Coming Soon.
I wish Cakewalk would publish a similar document.
- Jim, MIDI curmudgeon