ï~~methods address the sound in bulk; there is no
within the sound.
3. ARCSYN ADDITIVE OSCILLATO
3.1. Oscillator Architecture
ARCsyn's additive oscillator has three contro
inputs - pitch, dynamic level, and spectral m
The underlying synthesis architecture is a
reflection of these dimensions. There is a datc
steady-state spectral information, with ever
tempered note having a "slice" that contai
dynamic and spectral morphing information. Eac
of the "slice" consists of N steady-state partials,
is defined by computational requirements.
motion
OR
1 signal
orphing.
direct
changes your coordinates within the database; if the
samples are done correctly the formant remains
stationary. For electronic timbres, the spectral data can
be designed to change (or maintain) character throughout
the tessitura of the instrument.
4. CONTROL SYSTEM: MODELED MOTION
4.1. Random Individual Partial Modulation
_/
abase of The SPEAR software package was used to look at Cello
y well- tones in all registers of the instrument. Vibrato is clearly
ns both evident in most tracks, being subtle in the lower partials
h corner and dramatic in the highest. Aside from this, one can
where N generally state that there is a fair amount of random
amplitude and frequency motion in the partial tracks.
To model this, all partials in Arcsyn are individually
- side 2 modulated by a stochastic control signal. The user
specifies the frequency of the update and the magnitude
of modulation; it is applied individually (and
side I asynchronously) to both the frequency and amplitude of
all partials. Introducing this movement greatly improves
the musical quality of the tone; it ranges from a subtle.............. "life" to granular mayhem.
4.2. Transient Modulation
forte layer
piano layer,,,
A A# B C C# D
Figure 1: Pictorial representation of the database.
The tone at any given time is determined by the value of
the three control signals which specify an exact location
within in the database. Interpolation is performed
between the eight nearest spectral data locations, as well
as the previous partial values. Note that this
interpolation is best done in the logarithmic domain for
musical character.
3.2. "Acoustically-Correct" Vibrato and Glissandi
One substantial benefit of using the spectral database
architecture is that it intrinsically models acoustic vibrato
and glissandos very well. This is a central problem in
conventional sample based synthesis. Wakefield [7]
writes: "Although the player creates mainly frequency
modulation, it results in changes in amplitude that are
crucial to the perception of vibrato..... The moving
harmonics are boosted and depressed according to the
resonances of the instrument body."
A time-based sampler shifts pitch by speeding up or
slowing down the playback speed. All partials are
shifted up or down while maintaining the same
amplitude, effectively shifting the instrument resonances.
Maintaining a stationary formant requires that one
"switch samples" as pitch changes. The Arcsyn
oscillator does this well - for acoustic instrument
emulation, the formant is encoded in the spectral
information for each note. Changing pitch simply
Partials are not perfectly harmonic during the attack
portion of a sound, but gradually find their way to
integer-multiple values as the attack portion gives way to
the steady-state. This phenomenon is more pronounced
at higher dynamic levels. Arcsyn models this by
increasing the above random frequency modulation
during note onset. This achieved cleanly by introducing
a new sort of control signal - the derivative of the
dynamic control signal. If this signal is changing
quickly, partials will be pulled farther from their
harmonic tracks. The user can control the duration and
amount of the inharmonicity, as well as a set amount
from each new note-on command. Note that this could
also be used to modulate the noise content of the
bandwidth-enhanced partials.
4.3. Partial Envelopes
The dynamic control signal is fed into the Arcsyn
oscillator and specifies the position between the pp and ff
layers. This signal can be driven by a continuous
performance interface for sustained tones, or from any
form of a control envelope for struck or percussive
sounds.
Every partial has its own attack and release value; they
track the dynamic control signal. When the dynamic
signal is greater than the partial's current level, it slews
up based on the attack setting. When the input dynamic
is less than the partial's current dynamic, it slews down
based on the release value. The overall temporal
envelope is defined by the combination of the dynamic
479
0