Important caveat for Sibelius microtones

Well, more Sibelius microtone obsessing follows.

However, there is a *very* important caveat here, and that is that if
you use the quartertone high *AND* the quartertone LOW symbols in
Sibelius in the quartertone playback plug-in you run into trouble.

I was mentioning this obliquely before. As soon as you hit a
quartertone LOW symbol, one is thrown into a different scale, where
the numbers are different to get it to work right. One actually has
to *add* higher numbers to get a *SMALLER* microtone with the
quartertone LOW symbols.

Trust me. I've been obsessing about this all evening.

THE SOLUTION???

Well, fortunately there *is* one!

The solution is to use ONLY quartertone high symbols for the plug-in.

Then you can change the pitch bend numbers with impunity and they
always come out right.

Just *one* quartertone LOW symbol and it totally changes the scale!

(And by "scale" of course, I mean "pitch bend" numbers!)

If one stays with only the quartertone HIGH symbols, it works like
the manual says it should, which would be the following:

normal = B0,64

On the raising side:

^ = B0,69.5
> = B0,74.6
] = B0,80

And, on the lowering side:

v = B0,58.7
< = B0,53.4
[ = B0,48

Now that will work *perfectly* as long as you don't sneak a
quartertone LOW accidental in there someplace. Then the scale
changes.

And, you can't add the quartertone low symbol in later, or it ADDS
the quartertone lowering to the pitch, since the quartertones are
brought in so integral to the system.

Since there are so many problems with these "scales" when trying to
use more than "just quartertones" which, by the way, would be
*exceedingly* easy to do in Sibelius (but who, in our extended circle
of crazies would want to do *just* that...) I've decided that it
really *is* best to put ALL the microtonal accidentals in as TEXT
items.

So, that means that, basically, I'll be using the Sims SIM-bols,
after all!

And, probably that's for the best, since some players are used to
them in this system.

Oh... one further comment is that, if one wishes to use these
microtones with SHARPS and FLATS, one puts them in LATER. The pitch
bend is added to the sharping and flatting, so it works perfectly.

Just so long as both quartertone sharps and quartertone flats are NOT
used.

Great, if that's all you're doing, but for anything more complex..

BEWARE!!!!

jp

Finale is definitely harder to use than Sibelius, but Finale's
flexibility (with both built-in options and third-party plug-ins,
esp. those from Tobias Giesen) makes it (IMHO) the more powerful
program. There was a contest last year on an engraving list among
several notation programs, and despite being rigged against Finale
and some other serious problems, the contest was won hands down by a
Finale user. Let's hope this contest gets repeated in some form where
the abilities of the individual engravers are more balanced.

There are two curent methods of getting microtones in Finale.

The first, requiring a sythesizer with a preset full keyboard
retuning, uses the "KeyMap" option ín the "non-standard key
signatures" menu within the key signature tool. This works really
well for 19-equal, 31-equal, 53-equal etc, anytime sharps and
flats'll have fixed step values.

The 2nd option is to put in pitch bends for each note, attached
as "note expressions". The bad news is that pitch bends are
associated with channels, so polyphonic music has got to be carefully
managed to deal with this limitation, and that pitchbends are
notorious for latency problems, and every sythesizer or soundcard
seems to have a different implement of the midi pitch bend
instruction. The good news is that Finale's pitch bend message uses
more units than Sibelius, with tone at rest = 0, and allowing bend
messages from -8192 to 8191. Other synths or soundcards may react
differently, but on my generic soundblaster knock-off soundcard, 8191
will raise a note by 200 cents. That means 4096 parts to the
semitone, 2048 to the quartertone. It sounds like my card is picking
up messages of at least 4 units. So, for me, Finale is more accurate
than Sibelius by about 8-to-1.

Misc.: If you don't care about attaching a symbol to the pitchbend
playback message, you can also use the "Edit Continuous Data" menu in
the Midi Tool. Here is a graphic editor for the pitch bend message.
Also, the third-party "TGTools" includes an excellent tool for gliss
playback.

That's my Finale vs. Sibelius rap. If I didn't give a hoot about
midi playback, I would probably go to Graphire, which is purely
graphics, and now priced to compete with Sib. and Fin... but that's
another story, and would mean having to produce a score AND a
midifile. I'll wait until I can spare an extra half-life for that.

--- In tuning@y..., "alternativetuning" <alternativetuning@y...>
wrote:

/tuning/topicId_35620.html#35626

That means 4096 parts to the semitone, 2048 to the quartertone. It
sounds like my card is picking up messages of at least 4 units. So,
for me, Finale is more accurate than Sibelius by about 8-to-1.
>

***It seems as though this might not be quite an accurate statement.
One user on the Sibelius list offered the following:

"Finale "more precise"? Hmmm - I would think that implementing a MIDI
command should be the same on both applications. If it's possible to
increase the number of divisions of the octave, I'll bet you could do
it equally well in either program - all it would take is a MIDI
command.

As you've mentioned, a lot depends on the capabilities of the sound
card or synth that the computer is driving - you always need to check
and see which MIDI messages are supported and which are not. In that
aspect, Sibelius should be no better or worse than Finale - it's
essentially sending "dumb" commands to another device that may or may
not know how to handle them.

If you can find the MIDI command message to split a semitone into a
greater or lesser number of parts, try sending that command through
Sib and playing with the other pitch-bend commands and see what you
get.

Jim Stanley"

And, regarding the MIDI implementation in Version 2 of Sibelius:

"While MIDI data are just byte streams that any program can send,
there still needs to be a user "command" interface to construct the
messages, so there could be differences between MIDI applications in
that regard. For example, Sib 1.x provides only a few MIDI commands,
while Sib 2 now gives access to almost all MIDI messages and
controllers.

--N" (Norm Beck)

And from the Quicktime site that Manuel Op de Coul mentioned:

"Using these defaults it is possible to divide the semitone into 32
units. Using a pitch bend range to +/- 1 semitone you can 64 units
per semitones."

***It looks like that is the MIDI command that I need to learn to use
if I want to use the "refined" scale with Sibelius.

Now how to go about that... ??

Anyway, the "coarse" scale works well enough for the moment.

Does anyone really think the 3 cent error is a "big thing" when
considering the current state of "sound card technology."

They don't seem like "precision musical instruments" at this point,
do they?? Of course, a *real* player would probably put in more than
3 cents in vibrato or error as well... :)

jp

Joseph wrote:
>***It looks like that is the MIDI command that I need to learn to use
>if I want to use the "refined" scale with Sibelius.

The command is simply MIDI pitch bend. But the point is if
they let you have full control over the pitch bend parameter.
With anything less than 4096 steps / 100 cents that's not
the case. If so, there's no excuse for these lame Sibelius
or Quicktime programmers.

>Does anyone really think the 3 cent error is a "big thing" when
>considering the current state of "sound card technology."

Certainly.

>They don't seem like "precision musical instruments" at this point,
>do they??

You can very easily check the precision of your sound card's
pitch bend implementation with Scala by finding the smallest
cents increase where you start to hear beating when you play
two tones. I bet that is less than 3 cents.

Manuel

> That means 4096 parts to the semitone, 2048 to the quartertone. It
> sounds like my card is picking up messages of at least 4 units.

You can tell the resolution of your card if you set up two notes, on two
separate tracks, and set one of the pitch bends to 1024, 512, 256, 128, 64,
etc, and see where they fuse. Even if it's only a little out, you should
hear a slow phase. Just so you know how it's rounding to the nearest bit,
when you find the point where it fuses, try adding or subtracting one from
that number and twice that number. It's rare (or at least usually
expensive) that sound cards or synths actually handle the full
implimentation of every bit of a pitch bend. Most only look at the most
significant byte.

Or you can look at the specs of your card. It might show what the bit
resolution of pitch bends is.

Marc

On 3/13/02 11:11 AM, "alternativetuning" <alternativetuning@yahoo.com>
wrote:

> The first, requiring a sythesizer with a preset full keyboard
> retuning, uses the "KeyMap" option ín the "non-standard key
> signatures" menu within the key signature tool. This works really
> well for 19-equal, 31-equal, 53-equal etc, anytime sharps and
> flats'll have fixed step values.

Actually yeah I like this. I use it quite a bit. It's nice to be able to
take a modified QuickTime Musical Instruments extension and tweak the sheet
music into producing the right note.

Marc

--- In tuning@y..., manuel.op.de.coul@e... wrote:

/tuning/topicId_35620.html#35643

> Joseph wrote:
> >***It looks like that is the MIDI command that I need to learn to
use if I want to use the "refined" scale with Sibelius.
>
> The command is simply MIDI pitch bend. But the point is if
> they let you have full control over the pitch bend parameter.
> With anything less than 4096 steps / 100 cents that's not
> the case. If so, there's no excuse for these lame Sibelius
> or Quicktime programmers.
>

***Thanks, Manuel! Well, this is one of those cases where it's
SUPPOSED to do that, according to the manuel. That would be
the "refined scale." However, it doesn't seem to be working; at
least, *I* couldn't get it to work and there have been complaints by
others.

> >Does anyone really think the 3 cent error is a "big thing" when
> >considering the current state of "sound card technology."
>
> Certainly.
>
> >They don't seem like "precision musical instruments" at this point,
> >do they??
>
> You can very easily check the precision of your sound card's
> pitch bend implementation with Scala by finding the smallest
> cents increase where you start to hear beating when you play
> two tones. I bet that is less than 3 cents.
>

***Hmm.. You're probably right. Well, I'll have to
keep "experimenting" and "bugging" the Sibelius people...

Joseph

Hi Joseph,

If programmers had any notion at all that someone
might want a better pitch resolution than 32 values per
semitone, they would surely have added it in.

It is a one line code change; most of the work is in
the user interface.

In c:

ipitch_bend & 0x7F, ipitch_bend >> 7,

to unpack the pitch bend valuee (range 0 to 16383) into two 7 bit
values to send to midi.

In making pitch bend that coarse, they are following the
example of pitch bend wheels on many keyboards, as they
often only send the coarse part of the pitch bend
parameter as part of the work to reduce the number
of pitch bend messages sent in the stream of data.

The problem is, if one had the finest possible resolution,
then as you move the pitch bend wheel to change the
pitch by a semitone, you need to send pitch bend values
for every possible position (as user could stop abruptly
at any point).

Sending 8192 pitch bend messages whenever you change
the pitch by a tone using the wheel is clearly
not very practical, even nowadays with really
fast machines - maybe in a few years it will be.

On my machine it does maybe half a dozen to a dozen
midi events in two milliseconds. So to relay
8192 events would take it over a second,
and the p.c. would be completely occupied with
doing that - it would stop responding to anything
else as sending midi is done at nearly the maximum
priority possible; only a few system tasks will have
higher priority.

They do other things to reduce it too, - maybe
if you move the pitch bend wheel fast the resolution
will drop by a factor of two or more; so it would
be no surprise if ones pitch bend wheel has a pitch
bend resolution of more like 6 cents most of the time.

It would be possible to do the opposite and increase
the resolution when one moves the wheel more slowly
and so recover some of the extra pitch bend resolution,
though only for slow movements of the wheel. So
it would be practical to build a wheel that had
8192 pitch bend messages for a tone, but most of
the time if you moved it fast would send a message
only for every 256 points or so in the range,
gradually increasing resolution to every data
point if you moved it slowly enough. Maybe
this is done - I don't know much about it.

Anyone know?

We are probably a bit unusual in wanting such exact
pitch bends - of most interest if one wants an instant
pitch bend jump rather than a glide.

Anyway if inserting a single instant pitch bend jump using
a sequencer or score editor, then the stream of data
aspect of it isn't a consideration at all, and so one should
use the maximum precision available.

Robert

On 3/14/02 5:04 PM, "Robert Walker" <robertwalker@ntlworld.com> wrote:

> It would be possible to do the opposite and increase
> the resolution when one moves the wheel more slowly
> and so recover some of the extra pitch bend resolution,
> though only for slow movements of the wheel. So
> it would be practical to build a wheel that had
> 8192 pitch bend messages for a tone, but most of
> the time if you moved it fast would send a message
> only for every 256 points or so in the range,
> gradually increasing resolution to every data
> point if you moved it slowly enough. Maybe
> this is done - I don't know much about it.
>
> Anyone know?

It's a two-byte message, with a Most Significant Byte and a Least
Significant Byte. Most software and most synthesizers ignore the LSB, and
process it as one-byte bends. Same with recording. If you had an itchy
pitch wheel, it would generate tons of data every time you breathed.

On the software side, Performer has a "Thin Continuous Data" command which
can round every significant jump up so many values and discard everything in
between. I think most sequencers have something like this built in.

Marc

Hi Marc,

I've just had an idea about it, just this minute.

Perhaps one could deal with it by keeping track of
the time from the last message sent.
E.g. if it is more than 2 milliseconds since the
last pitch bend message, it is okay to send
another one, and send it for whatever position
the pitch bend wheel is at at the time.
Also send it if the wheel stops moving for more
than 2 ms, for the most recent value reached
before it stopped.

That way you get 8192 units per tone,
at most one pitch bend message every 2 millisceonds
which is very reasonable and should satisfy everyone,
and you get the maximum precision because when
the pitch bend wheel stops, then you record the
exact position it is stopped at.

Surely someone must have done this?

One might find the wheel sends a continuous
stream of data even when you aren't touching it
as the sensors vary their readings, but you could
deal with that too by making it so that it
only sends a message if there has been a change of
pitch of at least, say, half a cent in the
hundredth of a second (or whatever values
are appropriate).

If not for a pitch bend wheel, how about for a
string with the pitch varied by changing the
position of ones finger on the string?

BTW when one really notices this is if one
changes the pitch bend range to, say
an octave so that one can control huge
pitch bend glides from the pitch bend wheel.
Then you find the midi pitch bend wheel
using just the MSB has a very poor pitch
resolution indeed, but you do get your huge
pitch glides.

Robert

--- In tuning@y..., "Robert Walker" <robertwalker@n...> wrote:

/tuning/topicId_35620.html#35652

>
> The problem is, if one had the finest possible resolution,
> then as you move the pitch bend wheel to change the
> pitch by a semitone, you need to send pitch bend values
> for every possible position (as user could stop abruptly
> at any point).
>

***Hi Robert!

I appreciate your explaining this...

Joseph

On 3/15/02 9:47 AM, "jpehrson2" <jpehrson@rcn.com> wrote:

>> The problem is, if one had the finest possible resolution,
>> then as you move the pitch bend wheel to change the
>> pitch by a semitone, you need to send pitch bend values
>> for every possible position (as user could stop abruptly
>> at any point).
>>
>
> ***Hi Robert!
>
> I appreciate your explaining this...

Another plug for Digital Performer... Of late you can have LINES of
continuous data. Instead of actual points. This, for volume fades,
expression swells, smooth pans. I'm pretty sure it works with pb as well.