of Microprocessors

MUSIC SYNTHESIS SOFTWARE 641
Often, each level is served by a distinct program or set of subroutines
and in fact the programming languages used may vary from level to level.
Overall a music synthesis software system for programmed performance
operates as an interpretive system that extracts information from the userprepared
score and then acts on the information. The higher levels do the
information extraction and organization, while the lower levels do the actual
execution.
The lowest level synthesis routines, which will be called "Levell"
routines, operate at the sample rate of the final sound. This means that a
single "loop" through the synthesis routine generates only one sample of the
final sound. Basically, the routine accepts sound parameters as input and
generates sound samples as output. The output samples may be mixed with
the output of other Level 1 routines or the same routine with different
parameters. Details of the routine, of course, depend heavily on the synthesis
method chosen. If waveform table lookup is the primary scheme utilized,
then each loop involves the computation of a new lookup address based on a
frequency parameter, actual table lookup based on a waveshape parameter,
amplitude adjustment based on an amplitude parameter, and output based
on a location parameter.
Usually, a Level 1 synthesis routine will generate a "block" of samples
before returning to the next higher level. This saves most of the considerable
overhead associated with call and return linkage and the passing of parameters.
In the hietarchical system being outlined, the parameters of the
sound being synthesized are assumed to be constant throughout the block.
Often, a sample buffer is used to hold one or more blocks of samples. As
samples are computed, they are algebraically added to corresponding samples
in the buffer in a manner analogous to the audio bus of an analog voice
modular system. If stereo or quad is being synthesized, there will be a
distinct buffer for each channel of sound.
In some cases, there may be a "Level 0" routine that processes sound
samples from the buffer before playing them through the DAC or writing
them on a storage device. The usual function ofsuch a routine is the addition
of reverberation or a chorus effect to the sound. Parameters from the upper
levels are often required to control the effects produced, particularly when
they are used in a piece for dramatic contrast.
Level 2 routines operate at what can be called the "envelope sample
rate." This is the rate at which the fundamental parameters needed by the
Level 1 routines are updated and is equal to the audio sample rate divided by
their block size. Rates of 50 sis to 2,000 sis are useful, which corresponds to
update periods of 0.5 msec to 20 msec. Level 2 routines accept envelope
parameters from the higher levels and supply sound parameters to the Levell
routines. Envelope parameter descriptions vary widely, however, and a typical
system may use different Level 2 routines simultaneously according to
envelope type and description.