Musical sequence
Marc LeBrun
mlb at fxpt.com
Thu Jan 19 01:45:31 CET 2006
>=Jon Wild
> ...for many, many more details than you'd care to know about.
Here's far too many of those details, which I feel compelled to offer as a
public service.
Tuning systems are fun, but they seem often to become a sort of obsession,
divergent from sense.
If you want to pursue this subject, you should know that the musical
realities are subtler than one might at first assume:
>=David Wilson:
> What if we abandon the notion of octave, but keep the idea of equally
spaced tones,
> so that our scale can be described by...a ratio r between adjacent tones
on the scale.
There are doubtless other good examples, but one bit of computer-generated
music I'm familiar with that does this is John Chowning's 1978 "Stria".
This piece uses the golden ratio, not only for defining the scale tones
used in the composition, but also in the spectra of the synthesized sounds
themselves.
When these sounds are played at the specified scale tones the harmonics
tend to line up and reinforce eachother, producing a pleasant, though
exotic, kind of consonance.
Nonlinearities in physical vibrating systems--not only in the instruments
but also in the ear--cause all real sounds to have complicated spectra.
The "fundamental" frequencies of a composition, as defined by the formal
scale, are only one small aspect of the perceived musical result.
> Is there a measure of goodness of r that would rate r = 2^(1/12) high?
Yes, it's motivated by the happy coincidence that the value of 2^(7/12) =
1.4983... is close to 3/2. But before you start making lots of continued
fraction expansions of weird numbers, see below...
> Presumably this rating would be based on the closeness of tone ratios to
simple rationals which represent pleasing harmonies.
Yes, that's true to a first approximation, but, as the "Stria" example
indicates, there's a lot more to the story.
As others have mentioned, in musical contexts 2^(7/12) is indeed a
reasonably close approximation to the so-called "perfect fifth".
But note how numerically coarse this "good" approximation actually is!
Moreover, it has no particular bearing on 2^(k/12) for k#7. As mentioned
elsewhere, 2^(4/12) = cbrt(2) = 1.2599... is an OK approximation to 5/4,
but there's nothing particularly nice about 2^(1/12), and 2^(6/12) =
sqrt(2) was called "diablo in muisica".
On top of all this, dissonance is a non-monotonic function of frequency
separation. The nonlinear intermodulations referred to above produce
difference components. For very close pitches they are subaudible. As the
pitches diverge you get increasing roughness, dissonance and beating, which
then tapers off again at larger separations.
This intermodulation applies to *all* the spectral components of both tones.
That is the reason why close approximations to the "simple ratios" tend to
produce consonance--it's the reinforcement and blending, not because the
ear somehow mysteriously does number theory!
For example playing two tones with fundamentals in 3:5 ratio means the 5th
overtone of the former and the 3rd overtone of the latter will fall
together inside the "critical band".
But this says nothing about, say, the 7th overtone of one against the 4th
overtone of the other, etc.
Or chords of three or more tones.
Now it's true, very roughly, that the signal power in higher harmonics
tends to fall off. You could imagine weighting the strength of each
cross-interaction, to produce some kind of global measure.
But there are many very significant deviations from monotonic fall-off--for
example in initial transients (which the ear is critically sensitive to for
"source identification"), the formants that distinguish sung vowels, etc etc.
Further, the overtones of real vibrating systems are at best only integral
to a first approximation--and then only for linear vibrators--surfaces like
drums and bells have all kinds of crazy modes.
Even in linear sources the physics often tends to stretch the actual
overtones upward, which in turn causes humans to prefer "bright"
tunings. Tests on the conductor Pierre Boulez, for example, demonstrated
(much to his surprise) a definite preference for sharp octaves--so even
that most basic 2:1 ratio is a fudge. Historically, I believe the music
world legislated the A=440 Hz standard in part because of the stress from
continual gradual "pitch inflation".
> Given this rating, what would be the best r?
Any such "best" you might want to define might produce an interesting
sequence for the OEIS, but I hope I've demonstrated its relevance to actual
real music is at most tenuous.
Have fun, but don't take it too seriously. Beware, tuning systems tend
encourage certain species of mathematical crankiness. There is a glut of
microtonal crud out there whose creators believe *has* to be great, reality
notwithstanding, because it embodies their pet theories.
But honestly, the best such music seems to succeed in spite of, or aside
from, the exotic tunings.
Perhaps I'm too reactionary.
Just to be contrary, a few years ago, as an experiment, I had a guitar
refretted with 14 equal-tempered tones per octave (instead of the usual
12). This peculiar number was chosen expressly to maximize musical
apostasy (and practical joke potential). Surprisingly, it was not too
difficult to noodle on it in a reasonably melodic way. I believe this was
partly due to players' tendency to unconsciously "bend" notes a little,
even on normal instruments, to deal with slight mistunings, untrue
frettings, etc. Experiments with exact synthetic tones tended to sound
much worse.
I concluded that other aspects of musicality vastly overwhelm tuning
technicalities.
More information about the SeqFan
mailing list