Pitch bends again

I've looked at how sibelius does pitch bends again, and I cannot find
an error in what I've done. I've checked against my own ear and I've
definitely used a 19-tone scale that is only marginally off EDO
(limitations with Roland SC88 do not permit me from using the other
pitch bend number ). I don't know where people get 4096 pitch bends per
semitone: sibelius is quite clear on this, the range is 0 to 127, with
64 being the centre: no pitch bend. I do set the total range to *1
semitone*, which by a simple test proved that when the range is set to 1
controller 100,0, 101,0 6,1 in decimal, then 127 represents a bend up
of one semitone and 0 represents a bend down of 1 semitone. I set my
'zero' on A, as that is the tuned pitch on the real recorders that
could play these pieces. If anyone is interested I can post an 'ear
test' Sibelius file which I put together when composing these works.
There are slight deviations caused by the facts that:

1 if a bend of 0 bends pitch down 1 semitone, then the semitone is
divided into 64 parts.
2. if a bend of 127 bends a pitch up by one semitone, then the range 65
to 127 must divide a semitone into 63.5 parts: my spreadsheet
computations insist on this. - I even typed out a big table of all 128
bends and computed the cent values given 0 cents as a bend of 64 as
above.
3. The granularity of these bends is of the order of 1.5 cents, so the
maximum deviation is half this:

bend:
k <---------------1.5625¢------------------> k+1
|----------------------------------------------|
^ max deviation = 0.78¢ - movement either way
will make the representation more accurate, not less.

Mark

--- In MakeMicroMusic@yahoogroups.com, Mark Gould <mark.gould@a...> wrote:
> I've looked at how sibelius does pitch bends again, and I cannot find
> an error in what I've done. I've checked against my own ear and I've
> definitely used a 19-tone scale that is only marginally off EDO
> (limitations with Roland SC88 do not permit me from using the other
> pitch bend number ). I don't know where people get 4096 pitch bends per
> semitone: sibelius is quite clear on this, the range is 0 to 127, with
> 64 being the centre: no pitch bend.

I have no way of knowing what you are hearing; all I have is the midi
file, and how that is heard depends on how the pitch bends are
interpreted. In any case, I rendered the midi files using Audio
Compositor set to a bend sensitivity of 8192 pitch bends to a
semitone, converted to ogg, and uploaded it. The ogg files and
discussion of your scale on the web page should now be correct; it
might be a good idea for you to take a listen and see what you think.

http://66.98.148.43/~xenharmo/coll.htm

I have to say this has all been highly educational.

--- In MakeMicroMusic@yahoogroups.com, Mark Gould <mark.gould@a...>
wrote:
> I've looked at how sibelius does pitch bends again, and I cannot
find
> an error in what I've done. I've checked against my own ear and
I've
> definitely used a 19-tone scale that is only marginally off EDO
> (limitations with Roland SC88 do not permit me from using the other
> pitch bend number ). I don't know where people get 4096 pitch bends
per
> semitone: sibelius is quite clear on this, the range is 0 to 127,
with
> 64 being the centre: no pitch bend.

***As I understand it, the first bend number goes from 0 to 127.
This is the "fine" number. The "coarse" number is the *second*
number after the comma. So in other words, ~B127,127 is the largest
bend you can have. Since there are, essentially then, 128 bends for
both coarse and fine (including 0) you get 128 x 128 or 16384.

Now since the bend range is over *two whole tones* or 4 semi-tones,
you have to divide by 4...

16384/4 = 4096.

best,

J. Pehrson