Hacker News Reader

Comment by chrisjj | Hacker News Reader

> MIDI should be studied by anyone interested in protocol design

Absolutely. A shambles hopefully never repeated.

Probably won't be since MIDI 1.0 solved most problems good enough. Even MIDI 2.0 after decades is mostly just increased resolution. The only thing I'm aware of that's really missing is polyphonic aftertouch.

> Even MIDI 2.0 after decades is mostly just increased resolution

The biggest change is that 2.0 is duplex which allows devices to discover each other on the same MIDI network. No more in/out/thru, it's all just one USB (or ethernet, or wireless) connection from a device to every other device. It also allows property exchange, which means devices can exchange semantic information about their configuration and parameters with each other.

On top of that, individual voices can now carry pitch and control information, while devices can exchange tuning information. That's a big deal for microtonal and non-western music.

It would be hard to undersell how monumental MIDI 2.0 is as a change from MIDI 1.0, but somehow the MMA manages to do it by making it out to be "32 bit!" (which ironically, it isn't - you don't get 32 bits of resolution in your CCs with MIDI 2.0).

Where MIDI 1 can more or less express the pitch and dynamics/timbre changes within a single piece of western music, MIDI 2.0 can express almost everything you do within a DAW these days, short of actual audio i/o.

> you don't get 32 bits of resolution in your CCs with MIDI 2.0

The standard (section 7.4.6) defines 32 bit resolution for CCs.

Ah my bad (also not sure which document you're looking at, its in M2-104). I think I was confusing it with note on/off which only has 16 bits for velocity.

Poly aftertouch is in MIDI 1.0 -- it just isn't used much.

What MIDI 2.0 adds per-note pitch bend. That's a big deal, but you can also accomplish the same thing without all the complexity of MIDI 2.0 by using MPE, which is an extension to MIDI 1.0.

In a way, MPE is easier to deal with than per-note pitch bend anyways. MIDI 2.0 didn't bother to extend the 7-bit note number range, so you're still stuck with 128 notes. (This despite the existence of instruments like the Tonal Plexus H-Pi, which has 205 distinct pitches per octave, each with its own separate button!) That means in MIDI 2.0 you have to pick the nearest unused note and bend from there.

In MPE it's a bit easier because you can just grab an unused channel, pick the closest note to the one you want, and bend. It's a little more straightforward, and you know you'll never have to bend more than 50 cents either way.

So far, it looks like most of the expressive and microtonal instruments out there are using MPE. The only one I know of that uses MIDI 2.0 is the Lumatone.

MIDI 2.0 has attributes in note-on messages. Attribute type 0x3 gives you 7 bits for a note in semitones and 9 bits for a fractional part of a semitone, so a resolution of around 0.2 cents.

That's effectively just per-note pitch bend -- which is great to have, but it's still kind of problematic to only have 128 unique notes you can individually toggle on and off or otherwise control on keyboards that have more than 128 keys.

It's possible to work around the 7 bit limit, but it's awkward.

I was waiting for this comment :)

I'm sure nobody expected it to live this long, as the grumbling about its various shortcomings must've started in the late eighties at the latest. The curse of good enough strikes again ...

I've recently come into possession of a number of back issues of "1/1 - the Journal of the Just Intonation Network" from the 80's and 90's. It's interesting how much content was devoted just to the topic of figuring out how to use non-12-tone-equal-temperament scales on the MIDI instruments of the time.

Most of one issue (volume 3 number 4 from 1987) was taken up with an article on how the author was able to reverse-engineer and modify the firmware of the Ensoniq Mirage to support the kind of microtuning he wanted.

> the grumbling about its various shortcomings must've started in the late eighties at the latest.

Disproof: https://www.muzines.co.uk/articles/midi-what-s-wrong-with-it...

I want to study it and understand what you mean! Would you please like to describe the shortcomings / problems you see and how it could be done better? That would be so helpful to us! Thanks!

Anyone who posts that is an obvious idiot that you shouldn't listen to about anything.

Just someone who either has nothing to do with music or does but has an absolutely obnoxious personality that will point out some meaningless detail in the specification.

It is just mind blowing that my 1997 Korg Prophecy and 10 year old Waldorf Blofeld are all linked up to my brand new computer using the latest version of Abelton live with a midi interface that I don't even remember when I got it but its old too.

I once had the opportunity to use a genuine Steinway full-length grand piano which had been fitted with a MIDI input and actuators. I connected it to the output of a Kurzweil K2500 I believe it was, and I used the arpeggiator and sequencer on that to trigger the grand piano. The Steinway became a different instrument. It was not obvious! It was easy to make it sound like a mishmash.

This is why Synthesisers allow you to control the ADSR Envelope - the decay and sustain on a piano, makes its a bit of beast to sit comfortably in a mix or - as you've noted - fairly dissonant and unmusical when played as a keyboard as opposed to a discrete acoustic instrument with its own preferred range of operation in terms of pitch, timbre and dynamics.

The answer? Tailor the use-case to the instrument, as orchestras have done for thousands of years. A trumpet riff won't sound appropriate or evocative played on a xylophone anymore a Cello aria on a French Horn.

A masterclass on this is Wendy Carlos' Opus 'Switched On Bach', whose entire inspiration was to make "appealing music you could really listen to" using the then-new synthesiser technology.

https://en.wikipedia.org/wiki/Switched-On_Bach

Well, #1: Don't put out an Implementation Chart template with Yes and No the wrong way around.

If this doesn't sound serious, you had to be there.

Reply on news.ycombinator.com