[sdiy] MIDI isn't musical : Flame bait? (long, verbose, and rambling)

[Jonathan Lippard]

At 12:36 PM 1/16/02 -0800, Don Tillman wrote:

>Is it possible that MIDI was designed by folks who don't actually play
>keyboard?

Here's the whole issue with MIDI and its shortcomings.  Let's look at the 
world in which MIDI was first developed:

-No synths had more than 8 notes of polyphony unless you went for a massive 
custom job.  I'm not aware of any single keyboards that would let control 
more than that at once, anyway.

-The vast majority had less.  The DX-7 didn't come out until after the 
protocol was designed, and yes I'm excluding string machines and 
string-machines-on-crack like the Polymoog and Korg PS series for the sake 
of argument.

-Very few synths were velocity sensitive and even fewer had aftertouch 
capability.

MIDI really was just designed in order to send basic control information 
produced by a keyboard, no more, and no less (and just this information was 
a big deal at the time IMO, since most synths required the use of pedals to 
control parameters like filter cutoff while you were playing 
two-handed).  It let you control more than one sound source at once.

It wasn't designed to instantly capture every nuance of playing for any 
single instrument, most of which seem hard to capture at all--they wanted 
to capture the qualities that were easiest and most directly applicable to 
performance (dare I say, only the obvious parameters).  Any of the 
parameters on top of Note On/Off, Velocity, Aftertouch, and Pitch/Mod 
Wheels is really icing on the cake.  But they added multiple registered 
continuous controllers, as well as the ability to define your own 
controllers.  Clearly they were thinking about ways you could add more 
nuance to a sound you were playing, and the responsibility for lack of 
usage of these controllers lies squarely with the manufacturers of 
MIDI-capable equpiment.  Might I also point out that in a 8-bit at best 
world, 14-bit capability was extremely good? Let's not hold the original 
design issues in the light of today, where a 32-bit parallel bus is 
standard in out little hunks of silicon, and criticize them.  Let's 
criticize them for one or two things they should've given thought to back 
then (see below).

Sure, the designers of MIDI obviously weren't performance virtuosos who 
might pay more attention to more subtle aspects of acoustic 
instruments--they were just interested in what was easily achieved 
regarding performance IMO, since the primary goal was to standardize the 
way two synths talked to each other.

The shortcomings are pretty damn obvious: a slow serial stream with 20% of 
the bandwidth unusuable by default, despite the fact that proprietary 
parallel interfaces were a practical option at the time.  Better Oberheim's 
System, Palm's parallel buss, or even DCB than MIDI, perhaps.  What was 
Dave Smith thinking?

Back then it didn't make much difference.  Today it does, especially as the 
cost of implementing something like poly AT or throwing in a ribbon 
controller, an XY joystick, and some extra pedals/wheels is lower than it 
used to be.  MIDI clogs just too damn easily, even with low polyphony, and 
since lots of boards now have 32 notes or more of polyphony, it's a royal 
pain.  Let's not even mention SysEx ("It's a way you can waste *all* the 
bandwidth on one message!")  MIDI is obviously in many ways outdated.  Only 
a handful of performance parameters immediately at your control with a Note 
on or Aftertouch message.  Only 16 channels.  Coarse resolution when it's 
obvious that now we can easily deal with much higher resolutions.  Speeding 
up the MIDI clock would solve a congestion problem but not other issues...

And for some things that makes MIDI problematic or impractical to 
use.  Which is fine with me: if I'm going to play a mindblowing solo over 
something, I feel that's best done in realtime anyways.  I feel the example 
of a Hendrix or, say, a Charlie Parker solo is a bit specious at best.  You 
couldn't replicate any Hendrix solo with a trumpet or an acoustic piano, so 
why slag MIDI for the same limitation? Of course you can't replicate the 
nuance of a Hammond organ: MIDI is an abstraction from the performance 
qualities that are different for each type of instrument, and making MIDI 
emulate the nuance of a Hammond is highly likely to cause serious problems 
if you want to emulate the nuance of another instrument. Of course, half 
the idiosyncrasy of any solo is timing--that's really more an issue of how 
high the resolution is on your sequencer clock, pulses per quarter note and 
all that.  There are ways around issues with alternate tuning, as well, 
it's just that too few synths support that and it might come down to having 
to record your MIDI tracks to final form one at a time.  But that's what 
hard disk recording is for, isn't it?

This also seems to me like slagging keyboards as controllers in general, 
but with any type of controller there are a limit to how many performance 
parameters you've got.  No one's pointed out that no two stringed, blown, 
or keyboard instruments have the exact same set.  Harpsichords didn't have 
much in the way of dynamic volume: clavichords remain unparalleled in that 
you get velocity, aftertouch, *and* vibrato if you wiggle the damn 
key.  I'd like to see a bit more of that on future controller 
keyboards.  I'm more interested in first seeing the bandwidth and protocol 
issues resolved first.

Does any of this make sense, or is it all incoherent babbling?

I guess my final point is this--I complain about some things MIDI doesn't 
do or does poorly, but there's a lot of utility contained therein, and for 
many things I'm glad I've got it around.  But it's not everything.  No 
protocol will be.  And every protocol has ways around that. ;)

-Jonathan