Additive Synthesis, Amplitude Modulation

and Frequency Modulation

Prof Eduardo R Miranda

Varèse-Gastprofessor
[email protected]

Electronic Music Studio TU Berlin

Institute of Communications Research
http://www.kgw.tu-berlin.de/
 Topics:

 Additive Synthesis

 Amplitude Modulation (and Ring Modulation)

 Frequency Modulation
Additive Synthesis

• The technique assumes that any periodic waveform can be modelled as a sum
sinusoids at various amplitude envelopes and time-varying frequencies.

• Works by summing up individually generated sinusoids in order to form a

specific sound.
Additive Synthesis

eg21
Additive Synthesis

eg24
• A very powerful and flexible technique.

• But it is difficult to control manually and is computationally expensive.

• Musical timbres: composed of dozens of time-varying partials.

• It requires dozens of oscillators, noise generators and envelopes to obtain

convincing simulations of acoustic sounds.

• The specification and control of the parameter values for these components
are difficult and time consuming.

• Alternative approach: tools to obtain the synthesis parameters automatically

from the analysis of the spectrum of sampled sounds.
Amplitude Modulation

• Modulation occurs when some aspect of an

audio signal (carrier) varies according to the
behaviour of another signal (modulator).

• AM = when a modulator drives the

amplitude of a carrier.

• Simple AM: uses only 2 sinewave

oscillators.

eg23
• Complex AM: may involve more than 2 signals; or signals other than
sinewaves may be employed as carriers and/or modulators.

• Two types of AM:

a) Classic AM
b) Ring Modulation
Classic AM

• The output from the modulator is added to an offset amplitude value.

• If there is no modulation, then the amplitude of the carrier will be equal

to the offset.

am = ac × mi
ac
mi =
am

eg22
• If the modulation index is equal to zero, then there is no modulation.

• If it is higher than zero then the carrier will take an envelope with a
sinusoidal variation.
• In classic simple AM, the spectrum of the output contains 3 partials: at
the frequency of the carrier + two sidebands, one below and one above
the carrier’s frequency value.

• Sidebands = subtract the frequency of the modulator from the carrier

and add the frequency of the modulator to the carrier.
• Amplitudes

- The carrier frequency remains unchanged

- The sidebands are calculated by multiplying the amplitude of the
carrier by half of the value of the modulation index, E.g. is mi = 1, the
sidebands will have 50% of the amplitude of the carrier.

am = ac × mi
amp _ sidebands = ac × (0.5 × mi )
Ring Modulation

• The amplitude of the carrier is determined entirely by the modulator signal.

• If there is no modulation, then there is no sound

eg23
• When both signals are sinewaves, the resulting spectrum contains energy
only at the sidebands.

• The energy of the modulator is split between the 2 sidebands.

• The frequency of the carrier is not present.

• RM distorts the pitch of the signal; original pitch is lost.

• The multiplication of 2 signals is also a form of RM.
• Both classic AM and RM can use signals other than sinusoids, applying the
same principles.

• Great care must be taken in order to avoid aliasing distortion (above 50% of
the sampling rate).
Frequency Modulation

• Modulation occurs when some aspect of an

audio signal (carrier) varies according to the
behaviour of another signal (modulator).

• FM = when a modulator drives the frequency of

a carrier.

• Vibrato effect, good example to illustrate the

principle of FM, with the difference
that vibrato uses sub-audio as the modulator
(below 20 Hz).

• Simple FM: uses only 2 sinewave oscillators.

eg25
Simple FM

• The output of the modulator is offset by a

constant, represented as fc.
• If the amplitude of the modulator is equal to
zero, then there is no modulation.

• In this case the output of the carrier will be a

simple sinewave at frequency fc.

• In the amplitude of the modulator is greater than

zero, then modulation occurs.

• The output from the carrier will be a signal

whose frequency deviates proportionally to the
amplitude of the modulator.

FM1
• The “amplitude of the modulator” is called
frequency deviation, and is represented as d.

• The parameters of the simple FM algorithm are:

Frequency deviation = d
Modulator frequency = fm
Carrier amplitude = ac
Offset carrier frequency = fc
• If fm is kept constant whilst increasing d, then the period of the carrier’s
output will increasingly expand and contract proportionally to d.

• If d is kept constant whilst increasing fm, then the rate of the deviation will
become faster.

FM2
The spectrum of simple FM sounds
• The spectrum is composed of the carrier frequency (fc) and a number of
partials (called sidebands) on either side of it, spaced at a distance equal to
the modulator frequency (fm).

• The sideband pairs are calculated as follows, where k is an integer, greater

than zero, which corresponds to the order to the partial counting from fc:

fc + k × fm
fc − k × fm
• The amplitude of the partials are
determined mostly by the frequency
deviation (d).

• If d = 0 then the power of the signal

resides entirely in the offset carrier
frequency (fc).

• Increasing the value of d produces

sidebands at the expense of the power in
fc.

• The greater the value of d, the greater

the number of generated partials and the
wider the distribution of power between
the sidebands
• Modulation index helps to control the
number of audible sidebands and their
respective amplitudes:

d
i= d = i × fm
fm
• As i increases from zero, the number of
audible partials also increases and the
energy of fc is distributed among them.

• The number of sideband pairs with

significant amplitude can generally be
predicted as i = 1.

• Example if i = 3 then there will be 4

pairs of sidebands surrounding fc.

FM3
Estimating the amplitude of the partials
• fc “may” often be the most prominent partial in an FM sound; in this case it
defines the pitch.

• The amplitudes of the partials are defined by a set of functions: Bessel

functions.

• They determine scaling factors for pairs of sidebands, according to their

position relative to fc.
Bessel functions
• ac usually defines the overall loudness of the sound

• The amplitudes of the partials are calculated by scaling ac according to

the Bessel functions.

• Example: B0(i) gives the scaling for fc, B1(i) for the first pair of sidebands
(k=1), B2(i) for the second pair (k=2), B3(i) for the third (k=3), and so on.
Bessel functions
• The vertical axis is the amplitude of scaling factor according to the value
of i (mod. index) represented by the horizontal axis.

Example:
if i = 0 then fc = max factor and all sidebands = 0

[B0(0) = 1, B1(0) = 0, B2(0) = 0, B3(0) = 0, etc. ]

BN (i )

d
i=
fm

N = sideband pair
Example:
if i = 1 then fc = 0.76, 1st pair of sidebands = 0.44, 2nd pair = 0.11, etc.

[B0(0) = 0.76, B1(0) = 0.44, B2(0) = 0, B3(0) = 0.11, B4(1) = 0.01, etc. ]
“Negative” amplitudes

• The Bessel functions indicate that sidebands may have either positive
or “negative” amplitude, depending on i.

• Example:
If i = 5, then 1st pair of sidebands will be = -0.33

• “Negative” amplitude does not exist: it only indicates that the

sidebands are out of phase.

• Can be represented by plotting them downwards.

“Negative” amplitudes

• In general, the phase of the partials do not produce an audible

effect…

• … Unless another partial of the same frequency happens to be

present.

• In this case the amplitudes will either add or subtract, depending on

their respective phases.
Negative frequencies & Nyquist distortion

• If fc is too low and/or the i is too high, then the modulation produce
sidebands that fall in the negative domain.

• As a rule, negative sidebands fold around the 0 Hz axis and mix with
the others.

• Reflected sidebands will reverse their phase.

 Negative frequencies

• Reflected sidebands will reverse their phase.

Example:

f c = 440Hz, f m = 440Hz, i = 3
Nyquist distortion

• Partials falling beyond the Nyquist limit also fold over, and reflect into
the lower portion of the spectrum.
Synthesising time-varying spectra

• Modulation index i is an effective parameter to control spectral

evolution.

• An envelope can be employed to time-vary i to produce interesting

spectral envelopes that are unique to FM.

• A partial may increase or

decrease its amplitude
according to the slope the
respective Bessel function.

• Linearly increasing I does

not necessarily increase
the amplitude of the high-
order sidebands linearly.

FM4
Frequency ratios & sound design

• FM is governed by two simple ratios between FM parameters:

d : f m = i (mod index)

f c : f m = frequency ratio

• Freq ration is useful for achieving variations in pitch whilst maintaining

the timbre virtually unchanged.

• If the freq ratio and the mod index if a simple FM instrument are
maintained constant, but fc is modified then the sounds will vary in
pitch, but the timbre remains unchanged.

FM5
• It is more convenient to think of in terms d : f m = i (mod index)
of freq ratios rather than in terms of
values for fc and fm.
f c : f m = frequency ratio

• It is clear to see that 220 : 440 are in ratio 1:2, but not so immediate for
465.96 : 931.92.

• As a rule of thumb, freq ratios should always be reduced to their

simplest form. For example, 4:2, 3:1.5 and 15:7.5 are all equivalent to
2:1
FM directives in terms of simple ratios

FM6

FM7

FM8

FM9

FM10

FM11
Composite FM

• Involves 2 or more carrier oscillators and/or 2 or more modulator

oscillators.

• Produces more sidebands, but the complexity of the calculations for

predict the spectrum also increases.

• Basic combinations:

a) Additive carriers with independent modulators

b) Additive carriers with one modulator
c) Single carrier with parallel modulators
d) Single carrier with serial modulators
e) Self-modulating carrier
Additive carriers with independent modulators

• Composed of 2 or more simple

FM instruments in parallel.

• The spectrum is the result of

the addition of the outputs from
each instrument.

FM12
Additive carriers with 1 modulator

• One modulator oscillator

modulates 2 or more oscillators.

• The spectrum is the result of

the addition of the outputs from
each carrier oscillator.
fc

FM13
Single carrier with parallel modulators

• Modulator is the result of 2 or more

sinewaves added together.

• The FM formula is expanded to

accommodate multiple modulator freq (fm) and
mod indices (i).

• In the case of 2 parallel modulator the

sideband pairs are calculated as follows:

f c − (k1 × f m1 ) + (k 2 × f m 2 )
f c − (k1 × f m1 ) − (k 2 × f m 2 )
f c + (k1 × f m1 ) + (k 2 × f m 2 )
f c + (k1 × f m1 ) − (k 2 × f m 2 )

FM14
 f c − (k1 × f m1 ) + (k 2 × f m 2 )
f c − (k1 × f m1 ) − (k 2 × f m 2 )
f c + (k1 × f m1 ) + (k 2 × f m 2 )
f c + (k1 × f m1 ) − (k 2 × f m 2 )

• Each of the partials produced by one modulator oscillator (k1 x fm1)

forges a “local carrier” for the other modulator oscillator (k2 x fm2) .

• The amplitude scaling factor result from the multiplication of the

respective Bessel functions: Bn(i1) x Bm(i2).
Example: (see Appendix I of Computer Sound Design Book)

FM15
Single carrier with serial modulators

• The modulating signals is a frequency

modulated signal.

• The sidebands are calculated using the same

method as for parallel modulators, but the
amplitude scaling factors is different:

• The order of the outermost modulator is used to

scale the modulations index of the next
modulator: Bn(i1) x Bm(n x i2).

• Note: no sidebands from Bm(i) are generated:

B0(i1) x B1(0 x i2) = 0.

FM16
Further reading:

• Three Modelling Approaches to Sound Design, by E R Miranda (PDF file

tutorial3.pdf)

• The Amsterdam Csound Catalogue:

http://www.music.buffalo.edu/hiller/accci/