Food for thought

Retunable keyboards:
How about adding the Yamaha Portasound PSS-480 to the list of retunable
keyboards? Maybe not, but it _is_ possible to generate just pitches by
changing the frequency operators of the FM synthesis. Some interesting
melodies can be played after saving piano/flute sounds with frequency
retuned to 7, 8, 10, 11 and 13 in the five banks of user timbres. This gives
five 12TET scales separated by just intervals. If you consider 12TET to be
3-limit, the resulting 13-limit lattice is fun to explore.

MOS - Rothenberg - Balzano property:
Has anyone studied this from an information theory perspective? The more
unique intervals there are in a scale, the more "information" in would
contain.

Microtones on MTV:
The flute obstinato at the end of Eminem's "The real slim shady" is
microtonal.
9/4 2/1 3/2 4/3 21/16 7/6 9/8 1/1

Neural networks:
Do you know of any musical uses for Neural Networks? Maybe giving
scales/chords in input to a Kohonen Map would generate some pretty looking
patterns. What classification (of chords or scales) would emerge as the
result of unsupervised learning? Since NN's are good approximaters, maybe
this classification/property would be a generalisation of the concept of
"pitch class."

-Kami

>MOS - Rothenberg - Balzano property:
>Has anyone studied this from an information theory perspective?

Rothenberg has; he makes cryptic remarks about it in his papers, including
one spot where he criticizes previous efforts to apply information theory
to music:

"Much confusion has resulted from the application of standard statistical
'redundancy' measures to musical 'messages' without considering the
limitations introduced by the efficiency of the code being used."

>The more unique intervals there are in a scale, the more "information" in
>would contain.

One important feature of the Rothenberg model is that it postulates that
listeners construct reference frames to extract information from music.
So, we have three places to look: the music, the scale, and the reference
frame. Then, we must ask what we want. Do we want to maximize the amount
of information in any one of these? Random tone generation has far more
information than normal music, but most listeners would not prefer it.

-Carl