SUBTRACTIVE SYNTHESIS OF VIOLIN SOUND

G.Ranjith kumar D.Sampath

D.No: 5-89, Devi Nagar, Vijayawada,
Vijayawada, Andhra Pradesh
Andhra Pradesh, India. India.
E-Mail: [email protected] E-Mail: [email protected]

Abstract-This paper will deal with subtractive but in principle it could even be white noise ,
synthesis is a sound synthesis technique, which is “spectrally rich".Because of the
became popular of its analog synthesizers. need for pitch perception,a source is
The principle behind this technique is to needed, so that it contains harmonic
approximate the spectrum of the desired frequencies. Removing the excessive non-
sound by subtracting all the harmonic frequencies from the white noise
“nondesired" components from a can be very difficult. This problem is avoided
“spectrally rich" sound source. It dicusses by choosing periodical waveforms like saw
about source, filter and amplifier. Source tooth, square and pulse wave. The filter can
should be a sawtooth to avoid excessive non- be very basic, starting from simple lowpass
linear harmonic frequencies. The filter filter, or more complex structures, trying to
approximates the response of a violin’s body model the amplitude response of the
to the forces in the bridge. In source side instrument, if the technique is used to
discusses about band limited sawtooth and emulate some existing musical instrument.
vibrato effect. In filter side it discusses about The envelope shaping is intended to give
second order equlizer filter. The last step is to additional time varying characteristics to the
envelop shaping which will discuss “ADSR” generated sound. This may follow a typical
pattern. The final output is the product of the ADSR pattern (“attack",“delay", “sustain",
filter output and envelope shaping function. “release"), as shown in figure 2.For a basic
description of subtractive synthesis, see [1],
1.INTRODUCTION [2] and [3].

The process of subtractive synthesis can be 2. BASICS OF VIOLIN PHYSICS

split in three cascading blocks as shown if
figure 1. First we have to know about the violin
physics, to justify the choice of source and
filter. Then, different techniques and their
application to the synthesis of the violin
sound can be understood in an easier way.
Figure 1: Block diagram for generic
subtractive synthesis
The main parts are the body and the bridge,
Source: generator of the “spectrally rich" and the main accessory is the bow. When
signal the strings are bowed, there is motion of the
Filter: responsible for subtracting the strings, relative to the bow, depending on
“undesired frequency components" the friction force resulting from the bow
Amplifier: signal amplitude shaping in time pressure over the strings:
domain
1. In the beginning, the friction is
The source should be selected according enough to displace the string with
to the type of sound to be synthesized , the bow.

1
 3. At some point the friction is big
2. After restorative forces become big enough to drag the string again and the
enough, the string begins to slip back cycle repeats.
And the restorative forces to diminish.

Figure 2: Generic ADSR envelop shaping

Figure 3: Violin response to the sawtooth forces in the bridge

The sawtooth has a line spectrum containing

As a result of the above described both odd and even harmonics. Some
procedure, the contact point of the string considerations have to be made about it:
oscillates, producing forces over the bridge
that have magnitude like sawtooth waves.
The bridge transmits the forces into the 1. The digital waveform has to be band
body, which then has an amplitude response limited, in order to have an unaliased
like the one shown in figure 3. Spectrum.
2. The line spectrum itself would not yield
3.SOURCE a realistic sound, requiring some spectral
widening around the peaks. This can be
The brief description of the violin physics in accomplished by linear combining slightly
section 2 suggests both source and Filter detuned saw tooth signals and/or introducing
structure in order to apply the subtractive some jitter in the fundamental frequency
synthesis technique to the generation of a (vibrato effect).
violin sound. The source should be a
sawtooth, and the filter, something 3.1. Generation of band limited sawtooth
approximating the response of the violin signals
body to the forces in the bridge.

2
The main problem in generating a source to sampling frequency of the system (Nyquist
be filtered, apart of the required “spectral criterion). There are two ways (at least) to
richness", is that it has to be a band limited generate the bandlimited sawtoothwave:
signal, due to the restrictions imposed by the

Figure 4: Bandlimited sawtooth wave from BLIT method

Figure 5: Spectrum of the bandlimited sawtooth wave (BLIT method)

1. By integrating a train of band limited 3.1.2. Softening sharp discontinuities

pulses. It is described as an alternative technique, in
2. By softening sharp discontinuities which the minimum phase band limited step is
with smooth functions as the minimum stored in a lookup table, in order to be mixed
phase band limited step (minBlep). with a “naive" (sharp edges) sawtooth and thus
soften the abrupt transitions, effectively limiting
3.1.1. Bandlimited impulses the bandwidth of the resulting signal.A minimum
The exact and non-bandlimited saw tooth can be phase band limited step is shown in figure 6.
obtained by integrating a train of impulses in
time, where essentially there is an impulse every 3.2. Linear combination of sawtooth signals
time the sawtooth has a sharp change. The most
immediate approach to construct a band limited The detuning should be controlled, so that it
sawtooth would be integrating a train of produces some audible effect, but it doesn't
bandlimited impulses. A bandlimited sawtooth is exceed the frequency width of one
shown in figure 4, and its alias free spectrum, in semitone, in order to avoid the overlapping
figure 5. of the tone perception (being able to

3
distinguish one note from the immediate
semitone neighbors.
(1) (1)
The general form of the Nst combined
signals would be:

Figure 6: A minimum phase band limited step

Figure 7: Introduction to vibrato effect

Where the stk(n) represent the detuned

sawtooth waves. 4. FILTER

3.3 Introduction of vibrato effect In order to approximate the frequency response

of the violin body to the forces in the bridge (fig
The well known vibrato effect of the violin 3), a filter bank is used. Each filter is a “Regalia-
sound was emulated by introducing some Mitra" second order equalizer filter, and is based
pitch jitter, through the addition of a low on an all pass filter, three adders and two gains.
frequency and small amplitude sinusoid to This structure is advantageous, since the
the fundamental frequency of each sawtooth frequency, bandwidth and gain parameters are
signal, as illustrated in figure 7. easy to change.
Each filter has a transfer function: eqution no (3)
(2)

4
 Figure 8: Filter bank approximation to modified violin body response

The coefficients for the filter design, as given 5.ENVELOPE SHAPING

by Regalia and Mitra are computed as
follows: The envelope principle can be expanded to
encapsulate the filter and the oscillator. An
envelope used to control the cut-off point of the
filter with respect to time will alter the quantity
of upper harmonics present. For example, using
an envelope with a gradual rising attack, the
sound will appear to brighten as the cut-off point
rises in line with the envelope. The same
(4) principle can be applied to the oscillator. This
time the pitch of the fundamental frequency
varies in line with the (5) envelope shape.
Depending on the complexity of the synthesizer
there may be three envelopes, one for each of
Pitch, Timbre and Volume, or a single, universal
where is the normalized 3 dB bandwidth, envelope which can be assigned to any or all of
K the gain and w0 the normalized central these.
frequency of the filter
For the modeling of the violin body • ATTACK - Immediately upon pressing
response, a filter bank comprising three of a key, the envelope "opens" from zero
the above mentioned equalization filters to full. This is controlled by Attack
was used, with center frequencies 300Hz, time. An Attack time of zero means the
700Hz and 3000Hz. The 3dB bandwidths envelope goes from zero to full
were 20, 60 and 500 Hz,and the gains 11, instantly (ie sharp attack). Increasing
20 and 40. The first two filters in the filter the Attack time means that this will
bank are narrower than the third one .The happen more slowly.
true response of the violin body (to the • DECAY - While the key is still pressed,
sawtooth forces in the bridge, not the body the envelope then drops from full to the
impulse response) is shown in figure 4,and Sustain level. This is controlled by
the approximated version (by the filter bank) Decay time. A Decay time of zero
is shown in fig 8 . means the envelope goes from full to
sustain instantly. Increasing the Decay

5
 time means that this will happen more remains at the Sustain level. This is set
slowly. by Sustain level.
• SUSTAIN - While the key is still being
pressed, the continuing envelope just

• RELEASE - When the key is released,

the envelope then drops from the
Sustain level to zero. This is controlled
by Release time. A Release time of zero
means that this will happen instantly.
Increasing the Release time means that
this will happen more slowly.

6.CONCLUSIONS
In this case, this will try to reproduce the
The subtractive synthesis technique was
characteristic attack and release pattern of
successfully applied to the generation of a
the violin sound. This patterns depend both
violin sound A vibrato effect based on a
on the interpreter and the the music
slow modulation of the period was included
style.The final sound is computed by
to yield a more realistic sound More realism
multiplying the output of the filter with the
could be achieved by fine tuning the source
envelope shaping function in the time
and filter blocks.In particular, it could be
domain:
considered:
1. Add more equalizers to the filter bank,
v(t) = F(s(t))e(t) (5)
thus making the approximation to the violin
response more precise
where v(t) is the final violin sound, s(t) is the
2. Increasing the number of sawtooth waves
source output, F(.) denotes the composite
to be combined to produce a spectrally
action of the filter bank and e(t) is the
richer source. The combining parameters
envelope shaping function Our
(amplitudes and detuning) can still be
implementation of the ADSR envelope
optimized
neglected the decay phase out of simplicity
and considering it not to be essential for the
7.REFERENCE
sound synthesis. Furthermore,
we employed one of the simplest possible
[1] T. Yahaya Abdullah, Subtractive
approaches, by fitting rising and decaying
synthesis Available on the Internet at
exponentials, according to the attack and
http://tyala.freeyellow.com
release time limits
[2] David Marshall, Basic sound synthesis
Available on the Internet at
http://www.fortunecity.com

[3] Tim Collins, Subtractive synthesis

Available on the Internet at
http://www.eee.bham.ac.uk