of Microprocessors

MUSIC SYNTHESIS SOFTWARE 691
harmonic. The subgroup starts with the keyletter H (harmonic) immediately
followed by a number defining the harmonic number. Commas then separate
the parameters in the subgroup. The first parameter after the harmonic
number is amplitude (in percent) and the second is phase in units of 211"/100
(thus, 50 specifies 180°,25=90°, etc.). The last parameter in a subgroup is
followed with a semicolon if any more subgroups or groups follow; otherwise,
the end of the statement is assumed. If the phase parameter is omitted,
a random-number generator determines the phase, and, if the amplitude
parameter is omitted, a 100% amplitude is assumed. Omitted harmonics, of
course, have zero amplitude. The amplitude percentages in a waveform
specification simply indicate relative harmonic amplitudes; there is no restriction
on their sum, since the waveform will be scaled after it is computed.
DW, SW, and RWare similarly used to specify groups of parameters for the
other three waveforms. If a waveform is not defined for an envelope phase,
the waveform of the previous segment will be used, which saves space in the
waveform table area of memory. The attack segment must have a waveform
specified.
PRCUS Statement
The PRCUS statement, which is detailed in Fig. 18-22, defines the
sound of a percussion instrument. The ID number and duration and
amplitude parameters closely parallel those of the voice statement. Since a
two-phase envelope is used, only the AD and RD parameters are needed.
The use of filtered noise in the percussion sound model creates some
interesting problems in specifying the amplitude of the filtered noise. The
worst problem is that the parameters specified for the filter will greatly
influence the perceived as well as actual amplitude of the filtered noise.
Although the interpreter could look at the filter bandwidth and adjust the
amplitude parameter, the correction will be inaccurate if the bandwidth
changes during the sound. For simplicity, the noise amplitude parameter
should simply be passed through with any necessary correction supplied by
the user. Another problem with filtered noise is the fact that its peak
amplitude cannot be predicted with certainty even if uniformly distributed
noise samples are used. Nevertheless, this will seldom cause a problem unless
the noise filter bandwidth is quite small and the percussive voice amplitude
is a sizable portion of the DAC's range. If a problem is encountered, it would
be easy to incorporate a test in the percussion sound generation routine to
clip excessively large filtered noise samples.
The noise filter mode parameter is different because it uses mnemonics
for its value. While convenient for the user, mode numbers could be used
just as well to simplify the interpreter program. Parameter transitions are
specified by giving two numbers separated by commas, the first of which is
the initial value of the parameter and the second is the final value. The