[sdiy] MIDI isn't musical : Flame bait?

[Helin Mikko.A (NET/Tampere)]

In LTB protocol the notes are sent to driver/interface in advance, with
a time stamp of output time. That's to avoid computer latency, which can
be as much or even larger than the MIDI transmit delay (on MIDI 1.0
Physical Layer). LTB is not very useful on 1 channel MIDI output
devices, but it improves timing on multichannel devices, which usually
use cable select command (F5h + channel byte, 2 byte msg) before sending
data to another output port.

Everybody knows that MIDI 1.0 physical layer is somehow restriced, but
that is not the end of world. As I wrote earlier there are now many
devices which send MIDI over USB. There are some keyboards (like PC-300
Roland and Midiman Oxygen - with not the greatest action, but I guess
there will soon be better ones for Fatar etc.). Also companies like
Clavia (with Electro) and Akai (Z series) have USB on instrument but so
far it supports only sending and editing patches. Roland/Edirol SC8850,
SC8820 and SK-500 can though send and receive MIDI information via USB.
On the other hand the 'MIDI over FireWire' spec doesn't seem to allow
more than 3X speed over MIDI 1.0 phys. layer, if both ends can negotiate
the speed.

Another thing is that the internal latency for digital samplers and
synths can be much more than the MIDI latency. The Keyboard magazine
measured latencies couple of years ago, and there were substantial
latencies seen on the measurements (couple of ms for individual notes).
These days virtual synths on computers can have very small latency. I
have been able to play at 1.5 ms latency with my computer and the action
feels surprising fast, almost too fast, the sound comes out of speakers
earlier than the key has stopped, which is somehow annoying. Usually 12
ms soft (ASIO) latency + MIDI latency is enough, sometimes I use 6 ms
latency setting for the ASIO drivers. 

-Mikko
> -----Original Message-----
> From: ext Glen [mailto:mclilith at ezwv.com]
> Sent: 17. January 2002 9:02
> To: Ian Fritz; jdm at synthcom.com; synth-diy at dropmix.xs4all.nl
> Subject: Re: [sdiy] MIDI isn't musical : Flame bait?
> 
> 
> At 10:00 AM 1/16/02 , Ian Fritz wrote:
> 
> >Yes, that agrees with some of the studies I've been reading 
> about. Test on 
> >timing simultaneity, for example, show that the best that 
> the most gifted 
> >people can resolve is around 10 ms.  People here who claim 
> being able to 
> >resolve 1 - 2 ms don't appear credible according to these 
> results. Also the 
> >fastest most people can tap a finger is about 9 Hz. 
> 
> Even if that last line is true, please remember that one 
> isn't limited to using
> a single finger on a polyphonic keyboard. The peak "event 
> rate" could be much
> higher than 9 events per second, even with "normal" people 
> playing the keys, if
> we allow them to use all ten fingers.
> 
> 
> Lastly, I just want to mention that Steinberg has an MP3 demo 
> file of their new
> LTB MIDI interface, which offers greatly increased MIDI 
> timing accuracy, at
> least when used with CuBase. If you play that demo, you will 
> hear a brief
> passage played through a "standard" midi interface, and then 
> the same passage
> will then be played using the "improved" accuracy of the 
> Steinberg LTD MIDI
> interface.
> 
> There certainly is a noticeable improvement in timing 
> accuracy, when the
> passage is played through the Steinberg interface. However, I 
> have to say that
> I'm not really sure if I actually like the sound of so much 
> accuracy! Maybe I'm
> weird, but I think that the drums in that demo sound a bit 
> better when played
> on the standard MIDI interface, timing errors and all! If 
> anyone else downloads
> the demo, I'd like to know what you think of my opinion.
> 
> 
> Later,
> Glen
>