HARMONIC SHIFT OSCILLATOR

HARMONIC SHIFT
O S C I L L AT O R

SCINOMRAH
YCNEUQERF

COARSE LEVEL

FINE STRIDE

FREQ STRIDE LEVEL

M O D U L AT I O N

FM HS HL

V/OCT OUT _ OUT \

NEW SYSTEMS INSTRUMENTS

For the Generation of Harmonic and Inharmonic Spectra

Manual Revision 1.0.2

NEW SYSTEMS INSTRUMENTS

SPECIFICATIONS

Size 12HP
Depth 25mm
Power Consumption +12V 80mA
−12V 75mA
Tuning Range 7Hz–50kHz
Tuning Accuracy ~7 octaves
Output Range ~+9dBu–~+12dBu,
−8V–+8V peak
Input Impedance 20kΩ (FM and HS),
100kΩ (others)
Output Impedance 150Ω
Output Drive 2kΩ (min), 20kΩ+ (ideal)

INSTALLATION

Before installing the module, make sure the power is off. Attach the power
cable to the module and
to the bus. Double check the alignment of the red
stripe (or the brown wire for a multicolor cable)
with the markings on the
module and the bus. The red stripe should correspond with −12V, as is
standard in Eurorack. Check the documentation of your bus and power solution
if you are unsure.
Screw the module to the rails of the case using the
provided screws. (M2.5 and M3 size screws are
provided.)

New Systems Instruments modules all have keyed headers and properly
wired cables. But please
remember to double check the other side of the cable
for proper installation with the bus. Addi-
tionally, if using a different power
cable, note that not every company wires modular power cables
such that the
red stripe will align properly with a keyed header. While our modules are
reverse po-
larity protected as much as is practical, it is still possible that
you could damage the module, your
power supply, or another module by
installing the power cable improperly.

Lastly, please fully screw down the module before powering on your case.
The electronics are po-
tentially sensitive to shorts, and if the module is not
properly attached to a case, there is a risk of
contact with conductive or
flammable matter.

1
BASIS

The Harmonic Shift Oscillator creates a sound from abstract parameters

describing the character-
istics of that sound. Specifically, given angular
frequency ω , harmonic level L, and harmonic stride
S , this
produces two waveforms according to the following equations:

∞ ∞
∑ Ln sin((nS + 1)ωt) ∑ Ln cos((nS + 1)ωt)
​ ​

n=0 n=0

These are the real and imaginary components of the complex waveform:

∞
∑ Ln ei(nS+1)ωt
​

n=0

EXPLANATION

All waveforms can be constructed by adding together a set of simpler

waveforms at various ampli-
tudes and frequencies. The relationship between
frequency and amplitude for a given complex
waveform is known as that
waveform’s spectrum. Generally this spectrum is given in terms
of sine
waves, which our ear hears as pure tones.

When the spectrum only has nonzero amplitudes at frequencies that are
integer multiples of a fun-
damental frequency, that spectrum is said to be
harmonic. Otherwise, the spectrum is inharmonic.

The formulae above each have two parts: the part expressing the frequencies
of the components,
and the part expressing the amplitude of each of those
components. The Harmonic Shift Oscilla-
tor will produce sounds where the nth
harmonic is nS + 1 times the fundamental frequency, ω .
When S
is an integer, nS + 1 will also be an integer, and you’ll get a
harmonic spectra. Otherwise,
you’ll get an inharmonic spectra. The other
part of these equations is the amplitude of each of
these waves: Ln .
Spectra sound “brighter” or “darker” depending on how
much high frequency
content is present. Maximum brightness is achieved when
L is 1, and so every harmonic has the
same volume. The analog limitations
mean this is not quite possible in practice, but very bright
sounds are still
achievable. Lowering L to 0.5 would produce a spectrum with harmonics at
relative
amplitudes 1 for the first, 0.5 for the second, 0.25 and 0.125 for
the third and fourth, etc.

By controlling L, you directly control how bright or dark the sound is.
By controlling S , you con-
trol the character of the spectrum. And lastly,
by controlling ω , you control the fundamental
frequency.

2
HSO Spectrum at S = 1.3 and L = 0.707

3
INTERFACE

HARMONIC SHIFT
O S C I L L AT O R
1. Coarse Frequency – Control the
frequency, ω ,
ranging from about 7Hz to 50kHz.

1 3 2. Fine Frequency – Control the frequency, ω , rang-

ing about
a fifth above or below the frequency deter-

SCINOMRAH
YCNEUQERF

COARSE LEVEL mined by the Coarse knob.

3. Harmonic Level – Control L, the brightness of the

2 4 spectrum,
ranging from 0 to 1.

FINE STRIDE 4. Harmonic Stride – Control S , the spacing be-

tween each
harmonic, ranging from 0 to about 4.
5 7
6 5. Frequency Modulation Attenuator – Attenuator
FREQ STRIDE LEVEL
for #8, the
FM input.
M O D U L ATI O N

FM HS HL 6. Harmonic Stride Modulation Atten­uator – Atten-

8 10 uator for #9, the
HS input.
9
V/OCT OUT _ OUT \
7. Harmonic Level Modulation Attenuator – Attenu-
11 12
13 ator for #10, the
HL input.

NEW SYSTEMS INSTRUMENTS

8. Frequency Modulation – Modulate ω , the funda-
mental
frequency. Typical range of ± 8 octaves, or
1.6 V/octave.

9. Harmonic Stride Modulation – Modulate S , the spacing between

each harmonic. Typical range
of about ± 4.

10. Harmonic Level Modulation – Modulate L, the brightness of the

spectrum. Typical range of ±
1.

11. V/Octave Input – Control ω , the fundamental frequency,

using a V/Octave scale.

12. \  Output – Outputs one phase of the waveform. This is

the cosine, real, or +90° component.

13. _ Output – Outputs the other phase of the waveform.

This is the sine, imaginary, or 0°
component.

4
USING THE HARMONIC SHIFT OSCILLATOR

The Harmonic Shift Oscillator offers a huge field of possible waveforms,

and the best way to get to
know it is to use your ears and walk through the
wavescape by turning the knobs. Unlike many
other methods of inharmonic
synthesis, the Harmonic Shift Oscillator is very intuitive, and won’t
require much
theoretical knowledge to get good outcomes.

HSO waveforms for various values of S and L.

That being said, if all those possibilities are overwhelming, here are some
tips to get started.

First, turn the LEVEL knob somewhere between halfway and 4

o’clock; and plug one of the outputs
to your mixer, or however else you
get audio out of your modular. To get harmonic sounds, you
need to set S
to an integer. S is set to 1 when the STRIDE knob is about
halfway, or vertical. To set S
, slowly turn the STRIDE knob,
listening for the beat frequencies. When they slow down (they won’t
ever
completely stop), you’ll know you’ve reached your target.
After that, they’ll start speeding up
again. The STRIDE knob
is very sensitive, so take your time! When you’ve reached 1, the output
will
sound like a rich, “normal” waveform. Then slowly increase
S to 2, which is found when the
STRIDE knob is just before 3
o’clock. It will sound hollow, or flimsy. Continue to set S to 4,
which is
found just before the end. It should sound really sparse, and almost
digital.

In addition to these, you can get a harmonic sound with S at 0.5 (at
about 9 o’clock), which will
sound like it’s an octave lower, and
at 3, which sounds a little strange and is more difficult to find
than 2 or
4.

Next, try experimenting with inharmonic frequencies, setting the

STRIDE knob somewhere in be-
tween the harmonic values. When you turn
the STRIDE knob a bunch, the harmonics will move but
the
fundamental frequency will not. Because of this difference in motion,
humans tend to pull these
two sounds apart and perceive them separately. To
perceptually stick them back together, move the
fundamental frequency,
especially with a V/Octave controller.

All the parameters of the Harmonic Shift Oscillator can be modulated.

Modulation of harmonic
stride can produce nice percussive tones, while
modulation of harmonic level adds dynamism. Fur-
ther, the outputs of the
Harmonic Shift Oscillator function well as inputs to its various modulation
capabilites. In this way, incredibly rich and varying soundscapes can be
created from simple
controls.

5
HARMONIC AND INHARMONIC SPECTRA, TUNING, AND BEAT
FREQUENCIES

Constructive and Destructive Interference

Any two waves interact with each other to create constructive and
destructive interference. That is,
when the crests of one wave line up with
the crests of another wave, the overall waveform gets
louder. Alternately,
when the crests of one wave line up with the troughs of another wave, the
over-
all waveform gets quieter. The distance between the crests of two
different waves is the relative phase
of those waves. When two waves
of two different frequencies interact with each other, that relative
phase
changes at a constant rate, as the slower wave keeps falling further behind
the faster one.
This continual change in phase produces a cyclic change in
amplitude, known as a beat frequency.
This frequency will always be
equal to half the difference between the frequencies of the two waves
being
considered. If we ignore the sign of this waveform, focusing just on the
change in amplitude
itself, this is just the difference in frequencies. We can
express it like this:

sin(a) + sin(b) = 2 cos((a − b)/2) sin((a + b)/2)

Beat Frequencies

6
Note that these are two different ways of perceiving the same phenomenon,
not two different phe-
nomena. Sometimes we perceive things one way or the
other, but more often it’s a mixture of both.
Thus two very close
waveforms might be perceived as a single wave at the average frequency,
(a +
b)/2, with a slow beat frequency at a − b (twice the frequency of
of the cosine modulator, since
we’re just hearing the absolute value of
the amplitude). On the other hand, once these waves have
been sufficiently
separated, we’ll notice two dissonant frequencies as well as the beat
frequency.

When the two waves being considered are waves in a harmonic

series, we can rewrite the equation
like this:

sin(nωt) + sin((n + J)ωt) = 2 cos(J(ω/2)t) sin((2n + J)(ω/2)t)

Where n and J are two integers.

As we can see, both the beat frequency and average frequency of a harmonic
series are themselves
frequencies in a harmonic series of half the frequency.
Thus, both ways of perceiving these two
waveforms give rise to waveforms
within the same basic harmonic series.

When the series is inharmonic, that is no longer the case. Beat frequencies
other than those present
within the spectrum can be perceived. But also, the
overall perceived frequency can be difficult for
the human ear to determine.
This difficulty is what enables some inharmonic sounds to be used
outside of a
harmonic context (for example, purely rhythmically). Other inharmonic
sounds—most
of those produced by the Harmonic Shift
Oscillator—will still be perceived as tonal, but the rela-
tionship of
that tone to the fundamental frequency will be complicated. These sounds
should be
carefully tuned by ear.

PHASE AND THE TWO OUTPUTS

The Harmonic Shift Oscillator includes two different outputs in orthogonal

phase with each other.
The phase of the \  output trails the
_ output by one quarter turn, or 90°. Thus, while these two
outputs have similar spectra, they have peaks and troughs which occur at
different times.

The perception of phase in humans is complex. While we do not seem to be

able to directly detect
absolute phase at all, phase differences between our
two ears play an active role in our perception
of the spatial characteristics
of a sound. Further, these phase differences have other effects on the
sound
that are directly audible. First, the different placement of peaks
and troughs cause these two
sounds to saturate differently at different times,
resulting in frequency content that varies with time,
but which is different
for each output. Second, the interaction of each of these outputs with anoth-
er
signal will produce beat frequencies which are out of phase with each other by
one eighth turn.
Lastly, because the phase of each individual component is
orthogonal, combining the two wave-
forms does not produce a comb filter effect,
but merely a new waveform at an intermediate phase.
However, the
differentiated application of the two waveforms to a multidimensional delay
environ-
ment, such as natural or artificial reverberation, results in a more
complex comb filter pattern than
would the application of either signal
alone.

While these two outputs are useful for spatialization, generally that is
not best achieved by simply
assigning one to each channel of a stereo output.
Further, there are other textural uses of this phase
difference beyond the
needs of spatialization.

7
THE SPECTRA OF NATURE

It is not the intention of the Harmonic Shift Oscillator to mimic any

natural sound. Nevertheless, it
has certain characteristics that align with
natural acoustic phenomena.

Were there no other factors, a single dimensional resonator would produce a

perfectly harmonic
waveform proportional to its length, mass, and elasticity.
But in nature, there are always other fac-
tors. Columns of air have endcaps and
bodies that have different elasticity, mass, and volume than
their contents.
Strings have anchor points which themselves are pulled and prodded while the
string vibrates in space. As a result of this, many acoustic instruments
produce slightly inharmonic
sounds, where the space between harmonics is
stretched or shrunk to just less than or just more
than 1, or just less than
or just more than 2.

Since it’s not really possible to dial in an exactly

integral harmonic stride, the Harmonic Shift Oscil-
lator generally ends up in
this barely inharmonic state. Further, just how inharmonic is readily tun-
able.
This makes the Harmonic Shift Oscillator sound “alive” in a way
that most oscillators don’t
(and to mimic this liveliness, other
oscillators must be doubled or tripled).
A two dimensional resonator—a drum
head, for exam-
ple—has vibratory modes proportional to the zeros of
Bessel functions of the first kind. Although these res-
onators can’t be
exactly modeled as a sum of frequencies
proportional to Sn + 1, we can
get remarkably close.
Adjusting to place the fundamental frequency at 1, the
first 5 modes of a two dimensional resonator would be at:
1, 2.295, 3.598,
4.903, 6.209. Tuning S to exactly match
the first harmonic, we get the
first five modes for the
Harmonic Shift Oscillator as 1, 2.295, 3.591, 4.886,
The Frequencies in a Two Dimensional 6.182, which is a difference of +0, +0, +4, +6, and +8
Resonator and Those
Producible by the cents,
respectively.
HSO
As far as the brightness of natural spectra, generally this
brightness
changes as a function of changing resonance, as in speech, but also as a
function of over-
all energy of oscillation, as in a plucked or hammmered
string. While the first can be modeled with
a filter, the second is better modeled by modulation of L, where an exponential envelope results in
greater rates of decay of the higher than the lower harmonics, while keeping
the overall exponen-
tial character of that decay for each harmonic.

THE HARMONIC SHIFT OSCILLATOR AND CONVENTIONAL ANALOG

SYNTHESIS

Conventional analog synthesis—“subtractive

synthesis”—begins with four waveforms: sawtooth,
square, triangle, and sine. Leaving aside the sine
wave, we can express these waveforms according to
their spectra:

∞  
∑ sin((n + 1)ωt)/(n + 1) (sawtooth wave)
​

n=0

∞  
∑ sin((2n + 1)ωt)/(2n + 1) (square wave)
​

n=0

8
 ∞
 
∑ cos((2n + 1)ωt)/(2n + 1)2
​

n=0 (triangle wave)

It is fairly straightforward to reproduce

these spectra
with the Harmonic Shift Oscillator. First, we can see
that the
saw contains all harmonic frequencies, while the
triangle and square have only
odd frequencies. We can
produce this with the Harmonic Shift Oscillator by
set-
ting S to 1 or 2, respectively. The levels of these har-
monics can only
be approximated, however. The first
five harmonics of a Sawtooth wave
(1/(n + 1)) are 1,
0.5, 0.333, 0.25, 0.2. Matching the first and third
har-
Harmonic Levels of a Sawtooth wave and monics (L = 1/ 3 = 0.577), the harmonic shift oscilla-
​

an Approximation by the
Harmonic Shift tor gives 1,
0.577, 0.333, 0.192, 0.111. Whichever har-
Oscillator monic you choose to match, the
Harmonic Shift Oscilla-
tor has a slightly greater amplitude of prior harmonics,
and slightly lower amplitude of subsequent harmonics. This is similar for the
spectra of the triangle
wave, which has the first five harmonics
(1/(n + 1)2 ): 1, 0.111, 0.04, 0.020, 0.012. Again, match-
ing the third
harmonic (L = 0.2), we get 1, 0.2, 0.04, 0.008, 0.002.

Typical waveforms vs. phase-corrected HSO variants

While these spectra are different enough to give the Harmonic Shift
Oscillator a consistently dis-
tinct sound, in the practice of subtractive
synthesis these traditional waves are rarely used entirely
on their own.
Instead, the sound source is shaped by mixing together different waveforms,
and
then passing the result through a chain of filters. In some cases, by
using the Harmonic Shift Oscil-
lator one can forego this whole process and
arrive at the desired spectrum directly. But, filters offer
a substantially
different and interesting method of shaping sounds from that provided by the
Har-
monic Shift Oscillator. Fortunately we don’t have to choose. The
Harmonic Shift Oscillator is in-
tended to complement rather than replace
subtractive methods. It works well with filters and the
other tools of
subtractive synthesis.