16.10 Practice and Psychology
265
Csikszentmihalyi, M. (1996).
Creativity: Flow and the Psychology of Discovery
and Invention
. Harper Perennial.
Ghiselin, B (ed.) (1952).
The Creative Process
. Regents of University of Cali-
fornia.
Hofstadter, D. R. (1979).
G¨
odel, Escher, Bach: An Eternal Golden Braid
. Basic
Books.
Sommerville, I. (2004).
Software Engineering
, 7th ed. Addison Wesley.
Tarkovsky, Andrey A. (1988).
Sculpting in Time: Reflections on the Cinema
.
University of Texas Press.
Online Resources
Sven Carlsson’s site:
<
http://filmsound.org
>
.
Dan Stowell, “Designing Sound in Supercollider” Wiki book entry:
<
http://en
.wikibooks.org/wiki/Designing Sound in SuperCollider
>
.
17
Techni
q
ue 1
Summation
SECTION 17.1
Additive Synthesis
With this technique we work in the frequency domain and build sounds piece
by piece. Here we revisit as a synthesis method what we have known infor-
mally all along, that almost any arbitrary function of time can be expressed in
terms of simpler functions,
1
and through the work of Bernoulli, D’Alembert,
Euler, Fourier, and Gauss we arrive at some special cases for harmonic peri-
odic sounds which are the sum of sinusoidals. In Fourier’s harmonic theory any
periodic waveform, which only need be defined over the interval 0–2
π
, is the
sum of a trigonometric series
f
(
θ
) =
1
2
a
0
+
∞
k
=0
a
k
cos(
kθ
) +
b
k
sin(
kθ
)
(17.1)
in which
θ
is 2
πωt
+
φ
, where
φ
is the initial phase of a sinusoid. The coeffi-
cients of this expression are sinusoidal and cosinusoidal components in a simple
integer series; where the sound is harmonic they are all multiples of the lowest,
fundamental frequency. This describes a static, steady state spectrum (and as a
mathematical feature, one that is assumed to be infinite in duration). Calculat-
ing one period is enough, since all the others are the same. But if we reevaluate
the equation with a different set of coefficients every few milliseconds we can get
dynamic, evolving sounds with attacks and endings. This is called
discrete time
Fourier synthesis
, and we can use a
discrete time Fourier transform
(DTFT)
to analyse an existing sound into a set of coefficients and then use those to
resynthesise the sound. If we replay analysis data to an array of oscillators we
recover the original sound, but it leaves little room for actual design unless we
can manipulate and transform the intermediate parameters.
It’s one of the oldest digital methods. Max Mathews and Jean Claude Ris-
set did much of the groundwork for additive synthesis in the 1950s and 1960s.
Since the number of harmonics in most real sounds is very large, to do practical
additive synthesis we must employ
data reduction
, to boil the sound down to
its most important features. We look for envelopes that can be used with more
1. In fact not every function works, because of convergence problems.
Do'stlaringiz bilan baham: |