subtractive

feature synthesis

synth secrets
espite the title of this article, I won’t be
PART 1: WHAT’S IN A SOUND?
In the first part

D revealing any actual secrets during the

course of this new series. But before you
turn the page in disgust... what I will be
doing is taking a look at the basic principles of that
most common form of sound synthesis, subtractive
pleasing sound if their lengths were related by
simple integers (ie. whole numbers). For example,
if one string was half the length of the other (a 1:2
relationship) the result sounded quite nice. If the
relationship was 2:3, that sounded pleasant too.
of this new
series exploring
the world of
synthesis (these principles are well known, and Pythagoras was blown away by his discovery,
therefore hardly secret), and, later on in the series, and placed numerology at the heart of his subtractive
how these principles are applied on specific
synthesizers. The aim is that if you have a synth
philosophy. Unfortunately, he and his followers
then went off the rails a bit and tried to determine
synthesis,
that works on subtractive principles, and you know similar numerical relationships for the periods and Gordon Reid
how to get sounds out of it that you like, but don’t orbits of the five known planets, the sun and the
understand why or how it makes the sounds it moon, thus giving rise to the mythical ‘music of the
goes right back
does, this series should fill in some of the blanks spheres’. If they had only looked at the very small to basics. What
(they’re the ‘secrets’, you see). OK, maybe we instead of the very large (discovering Quantum
should have called the series Why Your Synth Does Mechanics in the process) they would have been are waveforms
What It Does When You Twiddle That Knob Or Slide much more successful.
That Slider... but, let’s face it, that’s not exactly a But why did Pythagoras’s strings have integer
and harmonics,
catchy name. So Synth Secrets it is. First things relationships? Why weren’t similar, pleasing sounds where do they
first, then: what is subtractive synthesis? generated by two strings when one of them was
The name ‘subtractive synthesis’ is derived from 1.21346706544 times the length of the other? come from, and
the method itself, wherein you attenuate or remove
harmonics from harmonically rich waveforms to Let’s Get Plucking how does the
create new sounds. You can do this in a static To start answering this, let’s consider a stretched theory relate
fashion to create simple tones, or you can use the string that is fixed at both ends, but free to vibrate
facilities offered by the filters, envelope generators, along its length. Figure 1 shows such a string at rest. to what we
and modulators in your synthesizer to make
dynamic sounds that change as time passes. But...
Now imagine that we gently pluck the string
exactly halfway between the ends. As you might
actually hear?
you may be lost already. What actually are imagine, this causes it to vibrate in the way shown
harmonics? What are waveforms? Where do they in Figure 2.
come from? This month, I’m going right back to This is an example of a ‘standing wave’. It does
basics, and answering just those three questions. not run up and down the string like waves on the
The stuff about VCFs, EGs and LFOs will come later. surface of the sea, but vibrates up and down. If the
vibration (or ‘oscillation’) is as simple as that shown
It’s All Greek To Me in Figure 2, a point at the centre of the string moves
To answer these fundamental questions, we have in a simple, repeating pattern called a sine wave
to jump into the Sound On Sound time machine (it’s (see Figure 3). We call this pattern the oscillation’s
tucked away somewhere behind the photocopier) ‘waveform’, and the frequency with which the
and head off into the dim recesses of the past. Back waveform completes one ‘cycle’ is called the
before physical modelling, before samplers, before ‘fundamental’ frequency of the string.
analogue polysynths, even before monosynths… The fundamental mode is, however, not the only
Actually, we’re in serious Dr Who territory here, way in which the string can vibrate — although
because we need to head back 2,500 years and because it is fixed at both ends, the number of ways
Figure 1: A simple
reacquaint ourselves with an Ionian chap by the and the speeds with which it can do so are severely
stretched string.
name of Pythagoras. Pythagoras was perhaps the
world’s first pure mathematician, yet we know
relatively little about him or his achievements
(much of what we do know about him may be no
more than legend — in contrast to what every
schoolboy knows, the Babylonians discovered Figure 2: A vibrating
Pythagoras’s theorem about 1,000 years before simple stretched string.
Pythagoras was born).
One of the lesser-known discoveries attributed
to Pythagoras was that plucking two similar
strings stretched to the same tension gave a

180 SOUND ON SOUND • may 1999

 constrained. Imagine placing your finger in
Displacement the exact centre of the string (but so that
the string can still vibrate along its entire
length) and plucking it on one side or the
Time other. You can see from Figure 4 that a
standing wave with half the wavelength of
the original looks perfectly feasible.
Likewise, if you place your finger
one-third of the way along the string, a
Figure 3: A sine wave. standing wave of one-third the
wavelength of the original should be possible (Figure
5) and so on.
Indeed, these standing waves can exist at all the
integer divisions of the wave shown in Figure 2, and
we call these the ‘harmonics’ of the fundamental
frequency.
If you study the maths of standing waves, you’ll
discover that you can represent such a wave as two

Figure 4: A standing
wave with one-half
the wavelength of
the fundamental.

Figure 5: A standing
wave with one-third ‘running’ waves moving in opposite directions along the
the wavelength of string (no, don’t ask why, or we’ll be here until page
the fundamental.
304). Knowing this, however, leads us to a simple
conclusion: if you halve the wavelength, the frequency
of the ‘running’ waves required will double. Similarly, if
you divide the wavelength by a factor of three, you
triple the frequency; quarter the wavelength and you
multiply the frequency by four, and so on… Only whole
“The name numbers will work because, if you tried to introduce a
non-integer change in the frequency, the string would
‘subtractive need to be somewhere other than at the zero position at
one of its ends (ie. part of the way through an complete
synthesis’ is cycle), and this isn’t possible because, of course, the
ends are fixed.
derived from the Anyway, we’ve now answered our first question by
identifying the harmonics which can be produced by a
method itself, simple oscillator: they are the permissible modes of
vibration. Of course, this analysis doesn’t only apply to a
wherein you vibrating string. Consider the air in an enclosed space
such as a cubic room. Forgetting for a moment any
attenuate or complicating factors such as furniture, the air can
vibrate anywhere in the room except at the walls, floor,
remove harmonics and ceilings. In other words, the vibrations in the room
are constrained in the same way as those on a string.
from harmonically This is why regular rooms have ‘resonances’ — they are
the harmonic frequencies of the room itself. And this is
rich waveforms why cathedral organs work — pipes are also simple
harmonic oscillators.
to create new In all but some esoteric cases, the first harmonic
(the fundamental, called f) is the pitch that you’ll
sounds”
▲

perceive when you listen to the sound of the plucked

 subtractive
feature synthesis
▲

string. The second harmonic (also called the first this is simply a shorthand way of saying, “this
Amplitude
‘overtone’) is half the wavelength of the setting generates a particular set of harmonics
fundamental and therefore twice the frequency. In with amplitudes of x, y and z…”
isolation we would perceive this as a tone exactly
one octave above the fundamental. Subtractive Synthesis Time
The third harmonic has a frequency of 3f So let’s apply these ideas to a synthesizer.
(which is the perfect fifth, one and a half octaves Look at the waveform in Figure 6. You would
above the fundamental) and the fourth harmonic, never get this from a plucked string, but you’ll
with a frequency of 4f, defines the second octave find an approximation to it on almost every Figure 6: The simple
above the fundamental. The next three harmonics synthesizer ever built. It’s an ideal ‘sawtooth’ sawtooth wave.
then lie within the next octave, and the eighth wave, so named because of its shape.
harmonic defines the third octave above the This waveform has a simple harmonic
fundamental. And so it goes on… relationship, expressed as follows:
This is the information we need to understand
Pythagoras’s observation. The shorter of the two Amplitude
strings in the 1:2 relationship is producing a
fundamental at the same frequency as the second
harmonic of the longer one. It’s exactly one octave The harmonic
spectrum of a Figure 7: The harmonic
higher. In the example of the 2:3 strings, the third sawtooth spectrum of a simple
harmonic of the longer string is at the same waveform sawtooth wave.
frequency as the second harmonic of the longer
one. In other words, the harmonic structures of
the two strings are closely related to one another,
and we hear this as musically ‘pleasing’.
Frequency
The Nature Of A Sound
Now consider this: when you pluck a string, you
Amplitude
don’t hear the sound of a single harmonic. The
conditions for creating such a pure tone are — in
the real world — almost impossibly precise, so Truncated harmonic
spectrum
any naturally occurring tone is likely to be a
composite of many harmonics present in differing Figure 8: A sawtooth
amounts. At any given moment it is this spectrum truncated to
combination that determines the waveform of the its first five harmonics.
sound and, because of the number of harmonics
present, this waveform will be much more
convoluted than the simple sine wave shown in Frequency
Figure 3. You only have to look at a sample of a
guitar or a human voice in a waveform editor to Every harmonic is present, and the amplitude of the
see how complex a real waveform can be. nth harmonic is 1/n times that of the fundamental.
This would make analysis of sound — or its OK, so it doesn’t look so simple when written in
resynthesis — almost impossibly difficult, had it English but, believe me, there are far nastier ones
not been for a French mathematician named Jean than this. Anyway, Figure 7 shows the first 10
Baptiste Joseph Fourier. Another guy with a harmonics in a sawtooth wave, and you
colourful life, Fourier was in turns a teacher, a can see how they taper off at higher and Amplitude
secret policeman, a political prisoner, governor of higher frequencies.
Egypt, Prefect of Isère and Rhône, and a friend of But what happens if you truncate this
Napoleon. Despite this, he still found time to series of harmonics? Let’s say you
determine that any periodic motion, no matter remove all but the first five of them (for
Time
how complex, could be broken down into its which you need a device called a ‘filter’).
harmonic components. This procedure is called Figure 8 shows this spectrum, and Figure
Fourier Analysis in his honour. Furthermore, 9 shows the waveform to which it
Fourier analysis also shows that, given a set of corresponds.
harmonics, you can derive a unique waveform. As you can see, the new waveform
Hold on a second… the waveform defines the looks different from the sawtooth wave.
harmonics, and the harmonics determine the It sounds different too. But the only
waveform? Clearly, harmonics and the waveforms difference between them is that you have
are just two ways of expressing the same thing. truncated the harmonic series of the sawtooth so Figure 9: The waveform
This is a key point: the natures of musical tones that only the first handful of harmonics remain. In created by the first five
are defined by the numbers and amplitudes of the other words, you have used a ‘filter’ to ‘subtract’ harmonics of the 1/n
series.
harmonics contained within them, and any given harmonics, thereby creating a new waveform, and
set of harmonics gives us a given waveform. So thus a new sound.
when we look at the oscillators on a synth and see Welcome to the world of subtractive
things such as ‘square’ waves or ‘sawtooth’ waves, synthesis! SOS

182 SOUND ON SOUND • may 1999