JI tuning on MIDI

Mark Nowitzky said:

>Yeah, the first time I played JI chords using MIDI, they sounded
>out-of-tune to me. Then my opinion would shift. I wondered if JI was an
>"acquired taste". But then I figured, well then, equal temperament is an
>"acquired BAD taste".

I have my doubts about the ability of a MIDI instrument, since it is limited
to 100 cents per semitone, to produce the same JI tuning that
tuning-flexible acoustic instruments (strings, voices, etc.) do. Every time
I try to adjust my M-1 to what I want to hear as an "in tune" triad, I find
nothing but frustration. When a colleague asked me to select my "preferred
tunings" while he played intervals in one cent increments, I was unable to
do so in a consistent manner; yet I can tell when my singers are "perfectly
in tune" and when they are even slightly "out of tune."

Comments anyone?

Jerry

>I have my doubts about the ability of a MIDI instrument, since it is
>limited to 100 cents per semitone, to produce the same JI tuning that
>tuning-flexible acoustic instruments (strings, voices, etc.) do. Every time
>I try to adjust my M-1 to what I want to hear as an "in tune" triad, I find
>nothing but frustration. When a colleague asked me to select my "preferred
>tunings" while he played intervals in one cent increments, I was unable to
>do so in a consistent manner; yet I can tell when my singers are "perfectly
>in tune" and when they are even slightly "out of tune."

For the vast majority of applications, cent-accurate tuning should be quite
satisfactory. The important thing to remember is that the accuracy is up to
your synth, not to MIDI. MIDI can send as much tuning data as you would
like with sysex, and more than most synths can reproduce with pitch bends.
Instruments do exist that are capable of greater than 1 cent accuracy
(usually, the accuracy is in hertz, so log-accuracy will vary over the range
of the instrument).

-Carl

To my post:

>>I have my doubts about the ability of a MIDI instrument, since it is
>>limited to 100 cents per semitone, to produce the same JI tuning that
>>tuning-flexible acoustic instruments (strings, voices, etc.) do. Every time
>>I try to adjust my M-1 to what I want to hear as an "in tune" triad, I find
>>nothing but frustration. When a colleague asked me to select my "preferred
>>tunings" while he played intervals in one cent increments, I was unable to
>>do so in a consistent manner; yet I can tell when my singers are "perfectly
>>in tune" and when they are even slightly "out of tune."

Carl Lumma responded:
>
> For the vast majority of applications, cent-accurate tuning should be quite
> satisfactory. The important thing to remember is that the accuracy is up to
> your synth, not to MIDI. MIDI can send as much tuning data as you would
> like with sysex, and more than most synths can reproduce with pitch bends.
> Instruments do exist that are capable of greater than 1 cent accuracy
> (usually, the accuracy is in hertz, so log-accuracy will vary over the range
> of the instrument).

Good news, indeed. Thanks, Carl. The bad news, I would imagine, is that such
instruments are likely wildly expensive. Wouldn't you know that my needs
would fall outside "the vast majority of applications."

In regard to super-sensitive synths, I'm almost afraid to ask........what
synth? And even more reluctant to get into "log-accuracy." The phrase "over
my head" comes to mind. (Not to mention, over my budget.)

May I assume that your input here, Carl, is to improve my faulty concept
(which it does) of the limitations of MIDI and not to stifle my complaints
about the inadequacy of 100 cent increments to match what the ear can hear?
That, after all, was the point of my post in response to Mark's reference to
his synth not being able to produce satisfying tunings.

All of this makes one appreciate the "miracle" of human perception of pitch
relations. Thank goodness, we don't have to teach kids math in order for
them to sing in tune.

Jerry

[Gerald Eskelin wrote...]
>In regard to super-sensitive synths, I'm almost afraid to ask........what
>synth?

There are others here far more knowledgeable on this subject than I. But I gather that there are two parts to the question:

1. How much tuning resolution can the synth address -- how much does it even attempt to produce?

2. How closely does the sound played actually match the note the synth is trying to play.

To answer the first part, you've got to decide how you want to send the tuning information to the synth -- pitch bend or sysex? Pitch bend works on all synths, and the resolution is potentially very great, but most synths ignore at least some of the pitch bend resolution. I don't know which ones are good at this. Usually, working with pitch bends is awkward -- to my knowledge, all the musicians who have successfully used pitch bends for tuning have controlled their synth from a PC, using software they wrote.

The far more popular and easier option is sysex. To find out about the sysex tuning capability of a synth, I always consult the microtonal synthesis web site...

http://home.att.net/~microtonal/

The two most flexible designs are the Kurzweil K2000/K2500/K2600, and the Emu Audity 2000 / Proteus 2000. The Kurzweil units accept tuning data in 1-cent increments, and the Emu units accept data in 1.56-cent increments.

Then there is Kyma, from Symbolic Sound. It requires a PC, and costs about $3300 for an entry unit. It allows arbitrary precision of tuning data.

Answering the second part is more tricky. There was a discussion here a while back about wavetable synths delivering poor and inconsistent accuracy. I wasn't involved in this discussion, so I can't tell you much about it. But to my knowledge, the most accurate synth as far as producing what it's trying to produce (that's the second part), is again Kyma. Supposedly, it's never off by more than 0.0026Hz.

>And even more reluctant to get into "log-accuracy." The phrase "over
>my head" comes to mind.

Well, cents are a logarithmic measure of frequency. You know, to add two ratios (like 5/4 and 5/3) you multiply them (25/12). But to add to intervals in cents (386 and 884), you add them (1270). Deal is, if a synth is off by 1 cent, that could be 0.064 hertz at A=110, or it could be 1.017 hertz at A=1760. So if you're mis-tuning a just ratio way down low in the frequency range by 1 cent, you'll have to listen much longer to hear a beat than you would mistuning by 1 cent in the upper range. If a synth was off by a certain number of hertz, that would be many cents down low, but few cents up high. Most synths are accurate in cents as far as what they're trying to do (first part), and in hertz as far as what they're actually doing (second part). Follow?

>May I assume that your input here, Carl, is to improve my faulty concept
>(which it does) of the limitations of MIDI and not to stifle my complaints
>about the inadequacy of 100 cent increments to match what the ear can hear?

I'm not sure. Keep in mind that 100-cent increments is 12tET. I took you to mean 1-cent increments. When tuning notes of the same timbre, in the most sensitive range of the ear, differences of less than a cent in a harmony can be heard, as piano tuners prove every day. But to hear mistuning at this level requires a trained ear, and careful listening. In a musical situation, such small mistunings come across as a _subtle_ change in timbre. Nevertheless, it is one which piano lovers and audiophiles can hear -- it has been said that a good tuning "brings the piano to life". I would encourage you to discover exactly what the accuracy of the synth is question was, and to qualify your statements by answering questions like,

What type of music was I listening to?
What limit was the harmony?
What timbre was being used?
How did the mis-tuning sound to me?
Was this a blind test?

-Carl