Conference PaperPDF Available

Parthenope: A Robotic Musical Siren


Abstract and Figures

Parthenope is a robotic musical siren developed to produce unique timbres and sonic gestures. Parthenope uses perforated spinning disks through which air is directed to produce sound. Computer-control of disk speed and air flow in conjunction with a variety of nozzles allow pitches to be precisely produced at different volumes. The instrument is controlled via Open Sound Control (OSC) messages sent over an ethernet connection and can interface with common DAWs and physical controllers. Parthenope is capable of microtonal tuning, portamenti, rapid and precise articula-tion (and thus complex rhythms), and distinct timbres that result from its aerophonic character. It occupies a unique place among robotic musical instruments.
Content may be subject to copyright.
Parthenope: A Robotic Musical Siren
Michael Sidler
Worcester Polytechnic
100 Institute Road
Worcester, MA 01609
Matthew Bisson
Worcester Polytechnic
100 Institute Road
Worcester, MA 01609
Jordan Grotz
Worcester Polytechnic
100 Institute Road
Worcester, MA 01609
Scott Barton
Worcester Polytechnic
100 Institute Road
Worcester, MA 01609
Parthenope is a robotic musical siren developed to produce
unique timbres and sonic gestures. Parthenope uses perfo-
rated spinning disks through which air is directed to pro-
duce sound. Computer-control of disk speed and air flow
in conjunction with a variety of nozzles allow pitches to be
precisely produced at different volumes. The instrument
is controlled via Open Sound Control (OSC) messages sent
over an ethernet connection and can interface with common
DAWs and physical controllers. Parthenope is capable of
microtonal tuning, portamenti, rapid and precise articula-
tion (and thus complex rhythms), and distinct timbres that
result from its aerophonic character. It occupies a unique
place among robotic musical instruments.
Author Keywords
Musical Robots, Aerophones, Sirens
CCS Concepts
Applied computing Sound and music computing;
Computer systems organization Robotics;
Sirens are often thought of as ”noisemakers” that have lim-
ited utility as musical instruments [4]. Part of the reason
for this attitude is due to the fact that many sirens are
unidimensional in terms of pitch and rhythm and are hard
to control. When considering instruments to add to the
robotic ensemble of WPI’s Music, Perception and Robotics
Lab (, we were intrigued by the possibility of
harnessing the unwieldy utterances of the siren through
computer control of electromechanical actuators. There are
few robotic instruments that have ventured down this path,
thus an opportunity to explore new territory in musical ex-
pression was presented.
Licensed under a Creative Commons Attribution
4.0 International License (CC BY 4.0). Copyright
remains with the author(s).
NIME’20, July 21-25, 2020, Royal Birmingham Conservatoire Birming-
ham, UK.
Figure 1: Parthenope
A siren has a simple design: a pitch is created by passing
air through concentric holes arranged radially on a spinning
disk. The frequency produced is equal to the number of
holes multiplied by the rotational speed of the disk. A disk
with 44 holes spinning at 10 rotations per second will result
in a frequency of 440 Hz, producing the pitch A4 [4].
f[Hz ] = n[holes]v[rps] (1)
Sirens are most often recognized for their use as alarms
(e.g. air raid sirens). They are recognized by their sweep-
ing portamenti and piercing timbre. As a result, their use
in musical pieces is often for textural or representational ef-
fects. They helped illustrate the clamour and intensity of an
industrializing world in the early 20th century in works such
as Edgard Var`ese’s Ionisation or the Ballet ecanique by
George Anthiel. The use of the siren is also heard in the
popular domain: the song Ridin’ the Storm Out by REO
Speedwagon opens with the sweeping sound of a siren that
commands the attention of listeners.
Variations of the siren’s design have been made to try to
achieve the functionality of more traditional musical instru-
ments, which can produce a larger range of pitches and mu-
sical gestures. A number of siren machines have been built
that feature a motor that spins a perforated disk. None
of them are fully autonomous though in that each requires
some degree of human involvement.
The Helmholtz double siren, as built by Rene Bakker in
2008, has two spinning disks that are capable of sweeping
portamenti [1]. The tone is changed by altering a secondary
disk, which limits the flow of air. The timbre produced
is reminiscent of an accordion or harmonica, and can be
changed by moving a case over the disk. The instrument
is limited by its complex design, which makes it harder to
reproduce musical gestures.
The Loman Siren Organ was patented by Abraham Lo-
man in 1915 and interpreted by Rene Bakker in 2008 [2].
This instrument has a series of six disks that can produce
two tones in semitone increments over the span of an octave.
The disks are encased, displayed in the familiar visage of an
organ, which affects the timbre produced by the instrument.
It is played manually with an octave wide keyboard. The
range can be changed by pulling a stop that increases the
speed of the disks.
In 2014, a group from Stanford created the ”Siren Or-
gan”, which operates by actuating values to allow air to flow
through the three disks on the instrument [3]. The speed
of the disk and the flow of air is human controlled. How-
ever, this control relies on human input, limiting the range
of musical expression. Precision is difficult when physically
moving a valve or potentiometer, echoing the issues of mu-
sical gesture reproduction with the Helmholtz double siren.
The above examples suggested that there was musical pos-
sibility in robotic sirens that had yet to be discovered. As
roboticists, many of the machines that we work with create
noise as a by-product. The fundamental idea of Parthenope
was to create what is typically seen as a noisemaker into a
musical voice through computer control of electro-mechanical
actuation. Such an approach could surpass the limitations
of the previous works mentioned.
These motivations resulted in a number of specific de-
sign objectives for Parthenope. We sought to create an in-
strument that could play continuous gestures and discrete
notes in ways that humans could not. The disk’s rotations
were to be controlled to ensure speed of intonation and ac-
curacy of pitch. The timbre produced should be variable,
compelling, and in combination with a wide dynamic range,
could be either the primary or supporting voice in an ensem-
ble. We sought to minimize system latency so that the ma-
chine could be used in real-time interactions. We wanted to
make our instrument accessible to less-technical musicians,
thus it was to be played through Digital Audio Workstations
(DAWs) via the MIDI protocol.
3.1 Disk Design
Figure 2: Diagram of one of Parthenope’s disks
The design of Parthenope’s disks dictates the structure of
the rest of the instrument. The cast acrylic disks have been
laser cut to create four concentric sets of holes. A note is
produced by passing air through any one set of holes.
The number of holes in each of the four concentric sets af-
fect the way that Parthenope generates sound. The control
algorithm prioritizes the smallest change in angular velocity
when choosing what holes to send air through. This helps
to minimize the portamento and time the instrument takes
to settle on the correct pitch.
To determine the range and speed optimizations of a par-
ticular disk, the following calculations are used: First, the
number of holes in each set are chosen (e.g. ranging from
inner most to outer most set: 11, 12, 14, 15). Next, a start-
ing note is picked (e.g. C4). The disk speed in rotations per
second is calculated for the innermost hole set using equa-
tion 1. The note C4 has a frequency of 261.626 Hz, which
when divided by 11 holes gives a speed of 23.784 rps. This
disk speed is used to find the frequency that each of the
other hole sets would play. For example, the second hole
set would have a frequency of 12 23.784 = 285.408H z.
This note is closest to the pitch D4 (293.665 Hz). In prac-
tice, the motor speed would be changed to account for this
difference. The nearest notes of the remaining two hole sets
are calculated according to this process (in this example,
E4 on the third hole set and an F4 on the fourth hole set).
A spreadsheet aids with this calculation process.
A disk with 11, 12, 14, and 15 holes in each set will have
a range of five semitones in the pattern W W 1/2 between
the four notes. In Parthenope’s design, the inner hole set of
the second disk will have a fixed offset from the inner hole
set of the first disk.
This offset is given as an integer number of semitones
and is referred to as the disk semitone offset. The offset in
speeds of the two disks can be used to optimize various con-
figurations. The disk semitone offset can be set to seven to
produce a configuration for Parthenope that is optimized to
play a diatonic major scale. (Figure 3). This optimization
only exists over a one octave range as there are only eight
sets of holes. Any notes played outside of this range will
require the motor to make a greater speed change, resulting
in a pitch glide up or down.
This procedure can be followed to create any pattern that
the composer desires. A range of eight octaves is obtained
with two disks, 48 semitones apart, each with a hole pat-
tern of 1,2,4, and eight holes (Figure 4). A consequence of
this configuration is larger pitch glides to non-optimized fre-
quencies. This will cause a decrease in accuracy as tempo
increases. Additionally, the intonation time will increase,
as the instrument has a greater speed change to overcome.
A disk semitone offset of six could be chosen to alter this
configuration to have a three and a half octave range and
a tighter grouping of notes (Figure 5). The desirability of
such behavior will depend on the music being played.
3.2 Motor and Motor Controller
In order to precisely control and change the rotational ve-
locity of the disks, a high power motor must be used. For
this task we chose the RCRunning 3650 brushless motor
and an ODrive v3.4 24V motor controller. A brushless mo-
tor was chosen for its quiet operation and high power. The
ODrive controller was chosen for its implementation of field
oriented control (FOC). FOC is a method for driving the
three phases of a brushless motor using sinusoidal waves in-
stead of square waves. This prevents the undesirable high
pitched whining noise that is customary of standard brush-
less motor controllers.
3.3 Air Supply
An Eagle Silent Series 2-HP 20-Gallon air compressor is
used to supply 30 PSI air through the instrument to the Air-
TAC 4V210-08 solenoid valves. These are five port, two way
pilot operated valves, with one of the channels closed off.
This prevents air from flowing unless one of the solenoids
is purposefully actuated. The solenoids’ valves direct the
release of air through nozzles. These nozzles are of different
sizes to produce different amounts of airflow and therefore
Figure 3: Optimized frequencies of a disk with 11, 12, 14, and 15 holes and a semitone offset of 7
Figure 4: Optimized frequencies of a disk with 1, 2, 4, and 8 holes and a semitone offset of 48
Figure 5: Optimized frequencies of a disk with 11, 12, 14, and 15 holes and a semitone offset of 6
different volumes. Diameters of 0.2mm, 0.6mm and 1.0mm
were chosen as they produced a wide dynamic range. Three
nozzles (one of each diameter) correspond to a hole set on
a disk. With four sets of holes on each disk, one rotational
speed can produce four different pitches at three different
velocity levels each (see figure 2).
3.4 Instrument Control
Parthenope is controlled by an Adafruit Grand Central M4
Express. This is an ARM Cortex M4F development board
chosen for its large pinout, floating point processing sup-
port, and configurable serial ports. The microcontroller can
interface with a computer or any other device over ether-
net through the Adafruit Ethernet FeatherWing. This is an
ethernet interface built around the WizNet WIZ5500 chip.
Parthenope accepts OSC messages over the ethernet inter-
face, which can be used to play notes and adjust instru-
ment parameters. The primary way to play Parthenope is
through a DAW such as Ableton Live. A Max for Live
patch was created to translate the MIDI information from
the Ableton Live Set into OSC messages that are sent to
the microcontroller. The patch also includes buttons that
can start/stop the disks.
3.5 Aesthetics and Visualizations
The enclosure for Parthenope was inspired by the analog
synthesizers of the 1960’s and 70’s. The two have a tim-
bral connection that we sought to reference in Parthenope’s
physical design. We envisioned Parthenope as a product of
modern technology that is infused with the spirit of early
experimental electronic music.
The casing of the instrument is made from a translucent
black acrylic that allows for a glimpse at the internal mech-
anisms when lit internally with signal LEDs. The case also
displays analog voltage and current indicators to visualize
the power of the instrument. The clear spinning disks are
prominently displayed above red modules that couple each
system together. We find the design simple and engaging.
Our technical evaluation of Parthenope focused on intona-
tion time, pitch accuracy, timbral envelope, dynamic range,
and system latency. One of the ways that Parthenope pro-
duces different pitches is by changing the rotational speed
of the motor, which takes time. We measured the time it
took for the instrument to accurately produce all intervals
within a chromatic scale relative to a tonic (in this case
the pitch C5), both ascending (e.g. 1 >2, 1 >3, ... 1 >
12) and descending (12 >11, 12 >10, ... 12 >1). Af-
ter a note message was sent, a timer was started and the
estimated pitch produced by the instrument was measured
using Max’s fzero object (which uses wavelet transforms).
When the produced frequency reached a threshold +/- 5%
of the target frequency, the pitch was considered accurate
and the timer was stopped. The experiment was repeated
twice for a total of three blocks. The resultant values were
averaged, which can be seen in Figure 6.
Figure 6: Intonation time of ascending and descend-
ing pitch intervals in the chromatic scale. The y-
axis represents time (msec) and the x-axis indicates
pitch interval size (semitones).
Intonation times tended to rise with interval size (more
with ascending than descending intervals), though this re-
lationship was variable. Intonation times were generally in
the 100-250 msec range, which, in practice, would affect the
amount of pitch ”glide” produced at the start of a note.
To assess pitch accuracy, we collected reported frequen-
cies for each pitch produced once they reached the specified
threshold. The absolute values of these readings were av-
eraged to give a measure of pitch accuracy. The pitches
produced were quite accurate with most values deviating
from ideals by 1% or less.
The timbral envelope of a tone produced by Parthenope
is characterized by a percussive attack (produced by the
solenoid valves) with energy at 200-400 Hz and 700-800 Hz,
which lasts roughly 300 msec. The lower partials of the
sound develop quickly, starting about 3 msec after the per-
cussive beginning, and are followed by the higher partials
emerging by 15 msec. The spectrum is strongly harmonic
with surprisingly little energy at the fundamental, with em-
phasis on harmonics 4-7 and considerable range that ex-
tends into the 7 kHz region. These details can be seen in
Figure 7.
Figure 7: A Spectrogram of Parthenope playing the
pitch C5 (log frequency scale).
To test the dynamic range of Parthenope, we placed a
LotFancy Sound Level Meter 1ft from the front face of the
instrument. With the instrument powered off in a quiet
room the sound level was measured at 54.4dBA. With the
motors powered on, the sound level was 59.3dBA. Next, the
instrument played a two octave C major scale starting at
C3 at each of its three volume levels. The dBA reading
was taken for each note at each volume level. The read-
ings for each volume level were averaged. The quietest was
80.5dBA, the middle was 95.7dBA, and the loudest was
The main source of latency between receiving a message
and sounding the appropriate note is the solenoid valves.
This latency was defined as the time difference between
the illumination of an LED, representing the power to the
solenoid, and a thin piece of paper being moved by the air
released from the nozzle. We recorded the solenoid actuat-
ing with a camera at 240fps. From the footage, an average
of ten trials gave us a consistent six frames between the
two events. This results in a solenoid actuation latency of
25msec +/- 2.4msec. The other potential source of latency
is the microcontroller processing the message, however, this
will be an order of magnitude smaller than the solenoid la-
tency. This is due to the 120MHz clock speed and floating
point processor core of the microcontroller.
The instrument occupies a unique musical niche as a con-
sequence of its physical design, means of sound production,
and computer control. The quality of the sound produced
is a cross between a sawtooth oscillator on an analog synth
and acoustic brass. There is a rawness and a roughness to
it that when sustained, can be grating (particularly when
played in isolation). Parthenope finds a more comfortable
home when articulating rapidly, which mollifies its unruly
qualities. At high speeds, the jaggedness of the timbre pro-
vides a sharp attack, helping it cut through noisy or busy
sonic environments. Thanks to the minimal activation time
of the solenoids, Parthenope is able to play quite rapidly (it
is comfortable at 50 msec intervals).
Parthenope excels in producing portamenti. By modify-
ing the speed of the disks while air is flowing, continuous
pitch changes are created. This allows the instrument to
produce a range of gestures from long sweeps to abbreviated
synth-like pitch glides that are typically found in electronic
dance music. Such variability in combination with the ca-
pacity to produce discrete notes is unique in the world of
musical robots. Another strength is in articulatory varia-
tion. Contrasts can be created between legato phrases with
portameti only to transition to aggressive staccatissimo at
the change of a note.
In addition to the pitches it produces, Parthnope’s phys-
ical design results in sonic idiosyncracies, originating from
the mechanical sounds of solenoids and hum of spinning
disks. The sound of the disks’ rapid rotation can intensify
musical gestures or provide a background texture that can
help congeal disparate sound sources.
Parthenope’s configurable design allows it to support a
potentially unlimited number of disks through a simple re-
configuration of constants in its code. Each disk can be
altered to contain different numbers of holes, which will in
turn affect the amount of portamento relative to a partic-
ular tuning system. The instrument is thus an excellent
candidate for microtonal explorations.
When composing with Parthenope, the instrument most
often occupies the lead voice because of its piercing sound.
However, its timbre allows for contrast and combination
with other instruments to voice unique phrases. The third
author composed a piece for Parthenope and (human-played)
vibraphone that explored such sonic interactions. The sharp
attack of the instrument allows for clear rhythmic articula-
tions, which enables it to assume a percussive role if desired.
Parthenope is successful as it allowed us to harness the en-
ergy of a noisy siren and transform it into a voice for mu-
sical expression. The controls afforded by the instrument
and the idiosyncratic sounds that it produces inspire our
musical creativity in new ways.
Despite this success, there are still areas for improvement.
We would like to expand the number of playing modes.
Currently the instrument can only play monophonic parts,
but new control algorithms could be written to allow each
disk to play independently, allowing Parthenope to produce
polyphonic passages. Adding more disks to Parthenope
could increase the range of the instrument, add to the num-
ber of notes that could be played in the polyphonic mode,
or even be used to create auditory effects by placing disk
modules in different locations relative to the composer or
listeners. Parthenope could also be augmented with alter-
native controllers, allowing it to be played as a continuous
instrument by a human, similar to a theremin. In addi-
tion to the above technical improvements, we hope that
time will allow more musicians to explore Parthenope and
discover creative ways to incorporate its unique sound into
new musical works.
[1] R. Bakker. Helmholtz double-siren., July
[2] R. Bakker. Loman siren organ., July
[3] R. Collecchia, D. Somen, and K. McElroy. The siren
organ. In Proceedings of the International Conference
on New Interfaces for Musical Expression, pages
391–394, London, United Kingdom, June 2014.
[4] B. Hopkin. Musical Instrument Design. See Sharp
Press, Tucson, Arizona, 1996.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
In this work we describe a MEMS instrument that resonates at audible frequencies, and with which music can be made. The sounds are generated by mechanical resonators and capacitive displacement sensors. Damping by air scales unfavourably for generating audible frequencies with small devices. Therefore a vacuum of 1.5 mbar is used to increase the quality factor and consequently the duration of the sounds to around 0.25 s. The instrument will be demonstrated during the MME 2010 conference opening, in a musical composition especially made for the occasion.
Helmholtz double-siren
  • R Bakker
R. Bakker. Helmholtz double-siren., July 2008.