Continuous Order Polygonalwaveform Synthesis
Continuous Order Polygonalwaveform Synthesis
and further enhances the possibility of emergent musical Continuous Order Polygonal Waveform Synthesis
behaviour.
532 Proceedings of the International Computer Music Conference 2016 Proceedings of the International Computer Music Conference 2016 533
Most of the challenges when porting synth engines into
a usable device, both virtual and hardware, are rooted in
robustness and edge behavior. We highlight several aspects
that need to be taken into account here:
4.1 (Anti)-Aliasing
One general artifact of digital synthesis is aliasing, see
[14]. This specifically holds true for most of the wave-
forms generated here: they often contain discontinuities
in the slope, which in turn result in high frequency content
Figure 2: Projection of a square polygon (order n = 4) from the two-dimensional x/y plane into the time-domain. that is beyond the typical audio Nyquist limit of 22.05 kHz.
To alleviate this, we propose generating the waveforms at
Figure 5: Spectrogram over phase modulation frequency, four times the final audio sampling rate, lowpass filtering
2.2 Projection be changed quasi-continuously in real-time. There is a fmod ∈ [2, 200] Hz with order n = 3, f0 = to the Nyquist limit using a 128th-order FIR, then deci-
hard lower limit of 2, corresponding to the polygon col- 100 Hz. Visual artifacts due to non-continuities in mating by a factor of 4. Such 4× oversampling dropped
The projection from the complex plane onto the time do-
lapsing into a line, which, depending on the phase offset sweep. the aliasing below the noise level in our tests.
main is done by simply taking the real (or imaginary) part
and projection, results in zero or infinite amplitude. For
of the polygon P :
n → ∞, the waveform approaches a pure sine wave. Fig-
P (φ, n, T, Φ) = p(φ, n, T ) · ej(φ+Φ) (3) ure 4 shows the spectrogram of a logarithmic sweep over Depending on the speed of a continuous phase modula-
the orders n ∈ ]2, 11] with a constant fundamental pitch of tion, both slowly evolving shapes or harsh, metallic sounds
x = { P (φ, n, T, Φ) } (4)
f0 = 100 Hz. may be generated.
y = { P (φ, n, T, Φ) } (5)
Although Fig. 2 only depicts the extraction of the y com- 3.3 Teeth T
ponent, it should be noted that the only difference to the x The parameter T , named for its visual effect on the poly-
component is a phase shift by 90◦ . The additional phase gon, allows the over-extension of the polygon’s vertices.
(a) n = 2.17, T = 0.17, fLP = 835 Hz, R = 0.94
offset Φ rotates the polygon in the complex plane and al-
lows for phase modulation of the time domain signal.
n h1 h2 h3 h4 h5 h6
Figure 6: Spectrogram over the parameter Teeth 4.2 Lookup Table vs. Phase Accumulation
2.001 1f 3f 3f 5f 5f 7f T ∈ [0, 0.5] with order n = 3, f0 = 100 Hz.
3 2f 4f 5f 7f 8f 10 f Digital waveforms can be either generated as a lookup ta-
4 3f 5f 7f 9f 11 f 13 f Increasing T can make the polygon exceed the unit cir- ble or on the fly, or in a mixed approach. Lookup tables
5 4f 6f 9f 11 f 14 f 16 f cle and consequently overdrive the output amplitude. For might generally be faster and can be pre-antialiased, but in
6 5f 7f 11 f 13 f 17 f 19 f lower values, this will only amplify present harmonics as this case require a sophisticated layout or interpolation to
shown in Figure 6, which shows the sonic effects of sweep- accommodate the various lengths of the waveforms due to
Table 1: Ratios of harmonic overtones to the fundamental f ing the parameter T ∈ [0, 0.5]. Depending on the em- the required cycles c to close a non-integer shape. In the
with increasing order. For non-integer orders, over- ployed limiting technique, higher values will drive the os- prototype we chose to evaluate Equation (5) with a con-
tones are continuously interpolated. tinuously varying angle φ, accepting an increase of high
cillator into saturation, allowing fine-grain control of addi-
tional harmonic partials. frequency noise when rotating non-integer orders.
Figure 3: Projections of polygons P (φ, n, T, Φ) from the At lower orders, strong harmonics form at specific ratios
2D space into the time domain, Φ = 0. 4.3 Amplitude Limiting
as noted in Table 1. They split and drift upwards with in-
creasing order, until only the fundamental is left and the 4. IMPLEMENTATION As mentioned in Section 3.3, non-zero values of T result
waveform is recognized as a pure sine wave. in amplitudes that can exceed the unit circle. To keep the
A monophonic version of the proposed synthesis method
3. EVALUATION waveforms in arbitrary but strict amplitude limits, clipping
was implemented in Max/MSP [13] to explore the physical
3.2 Phase offset Φ or compression must be applied to the output signals. A
In this section we discuss the three synthesis parameters interaction with the available parameters. Figure 7 shows
simple hard clipper with a variable input attenuator is ap-
order n, phase φ and teeth T and their influence on the Modulating the phase of a polygon P by adjusting Φ is the GUI, with both the polygon and the time-domain signal
plied to the oversampled signal in our implementation to
sonic properties of the waveforms. non-trivial for non-integer orders, as discussed in Section 2.1. drawn in real-time. Figure 8 shows two visually and son-
keep the audio signals within [-1,+1] limits.
For closed loop polygons, phase modulation results in in- ically interesting polygons, their time-domain representa-
3.1 Order n teresting spectral behavior as shown in Figure 5, where tion and their settings. 1 4.4 Filtering
As the order n ∈ ]2, ∞] of the polygon is specifically the fundamental and even overtones are bend up while odd 1Please find a small selection of audio samples at https:// A traditional sculpting lowpass filter as known from sub-
not bound to be an integer, the shape of the polygon may overtones are bend down. ccrma.stanford.edu/˜chohner/polygon_samples.zip tractive synthesis is employed to further shape the pro-
534
535 Proceedings of the International Computer Music Conference 2016 Proceedings of the International Computer Music Conference 2016 536
Most of the challenges when porting synth engines into
a usable device, both virtual and hardware, are rooted in
robustness and edge behavior. We highlight several aspects
that need to be taken into account here:
4.1 (Anti)-Aliasing
One general artifact of digital synthesis is aliasing, see
[14]. This specifically holds true for most of the wave-
forms generated here: they often contain discontinuities
in the slope, which in turn result in high frequency content
Figure 2: Projection of a square polygon (order n = 4) from the two-dimensional x/y plane into the time-domain. that is beyond the typical audio Nyquist limit of 22.05 kHz.
To alleviate this, we propose generating the waveforms at
Figure 5: Spectrogram over phase modulation frequency, four times the final audio sampling rate, lowpass filtering
2.2 Projection be changed quasi-continuously in real-time. There is a fmod ∈ [2, 200] Hz with order n = 3, f0 = to the Nyquist limit using a 128th-order FIR, then deci-
hard lower limit of 2, corresponding to the polygon col- 100 Hz. Visual artifacts due to non-continuities in mating by a factor of 4. Such 4× oversampling dropped
The projection from the complex plane onto the time do-
lapsing into a line, which, depending on the phase offset sweep. the aliasing below the noise level in our tests.
main is done by simply taking the real (or imaginary) part
and projection, results in zero or infinite amplitude. For
of the polygon P :
n → ∞, the waveform approaches a pure sine wave. Fig-
P (φ, n, T, Φ) = p(φ, n, T ) · ej(φ+Φ) (3) ure 4 shows the spectrogram of a logarithmic sweep over Depending on the speed of a continuous phase modula-
the orders n ∈ ]2, 11] with a constant fundamental pitch of tion, both slowly evolving shapes or harsh, metallic sounds
x = { P (φ, n, T, Φ) } (4)
f0 = 100 Hz. may be generated.
y = { P (φ, n, T, Φ) } (5)
Although Fig. 2 only depicts the extraction of the y com- 3.3 Teeth T
ponent, it should be noted that the only difference to the x The parameter T , named for its visual effect on the poly-
component is a phase shift by 90◦ . The additional phase gon, allows the over-extension of the polygon’s vertices.
(a) n = 2.17, T = 0.17, fLP = 835 Hz, R = 0.94
offset Φ rotates the polygon in the complex plane and al-
lows for phase modulation of the time domain signal.
n h1 h2 h3 h4 h5 h6
Figure 6: Spectrogram over the parameter Teeth 4.2 Lookup Table vs. Phase Accumulation
2.001 1f 3f 3f 5f 5f 7f T ∈ [0, 0.5] with order n = 3, f0 = 100 Hz.
3 2f 4f 5f 7f 8f 10 f Digital waveforms can be either generated as a lookup ta-
4 3f 5f 7f 9f 11 f 13 f Increasing T can make the polygon exceed the unit cir- ble or on the fly, or in a mixed approach. Lookup tables
5 4f 6f 9f 11 f 14 f 16 f cle and consequently overdrive the output amplitude. For might generally be faster and can be pre-antialiased, but in
6 5f 7f 11 f 13 f 17 f 19 f lower values, this will only amplify present harmonics as this case require a sophisticated layout or interpolation to
shown in Figure 6, which shows the sonic effects of sweep- accommodate the various lengths of the waveforms due to
Table 1: Ratios of harmonic overtones to the fundamental f ing the parameter T ∈ [0, 0.5]. Depending on the em- the required cycles c to close a non-integer shape. In the
with increasing order. For non-integer orders, over- ployed limiting technique, higher values will drive the os- prototype we chose to evaluate Equation (5) with a con-
tones are continuously interpolated. tinuously varying angle φ, accepting an increase of high
cillator into saturation, allowing fine-grain control of addi-
tional harmonic partials. frequency noise when rotating non-integer orders.
Figure 3: Projections of polygons P (φ, n, T, Φ) from the At lower orders, strong harmonics form at specific ratios
2D space into the time domain, Φ = 0. 4.3 Amplitude Limiting
as noted in Table 1. They split and drift upwards with in-
creasing order, until only the fundamental is left and the 4. IMPLEMENTATION As mentioned in Section 3.3, non-zero values of T result
waveform is recognized as a pure sine wave. in amplitudes that can exceed the unit circle. To keep the
A monophonic version of the proposed synthesis method
3. EVALUATION waveforms in arbitrary but strict amplitude limits, clipping
was implemented in Max/MSP [13] to explore the physical
3.2 Phase offset Φ or compression must be applied to the output signals. A
In this section we discuss the three synthesis parameters interaction with the available parameters. Figure 7 shows
simple hard clipper with a variable input attenuator is ap-
order n, phase φ and teeth T and their influence on the Modulating the phase of a polygon P by adjusting Φ is the GUI, with both the polygon and the time-domain signal
plied to the oversampled signal in our implementation to
sonic properties of the waveforms. non-trivial for non-integer orders, as discussed in Section 2.1. drawn in real-time. Figure 8 shows two visually and son-
keep the audio signals within [-1,+1] limits.
For closed loop polygons, phase modulation results in in- ically interesting polygons, their time-domain representa-
3.1 Order n teresting spectral behavior as shown in Figure 5, where tion and their settings. 1 4.4 Filtering
As the order n ∈ ]2, ∞] of the polygon is specifically the fundamental and even overtones are bend up while odd 1Please find a small selection of audio samples at https:// A traditional sculpting lowpass filter as known from sub-
not bound to be an integer, the shape of the polygon may overtones are bend down. ccrma.stanford.edu/˜chohner/polygon_samples.zip tractive synthesis is employed to further shape the pro-
534 Proceedings of the International Computer Music Conference 2016 Proceedings of the International Computer Music Conference 2016 535
Textual and Sonic Feedback Loops: Simultaneous conversations as a
collaborative process in cmetq
duced waveforms. As expected, this introduces rounded [4] D. Gerhard and others, “Audio visualization in phase ABSTRACT character of language and the quality of sonic material, de-
vertices in the polar domain and overshoot depending on space,” in Bridges: Mathematical Connections in Art, fine the complexion of cmetq. The creation of cmetq was
the resonance setting. Music and Science, 1999, pp. 137–144. cmetq is an concert length work for baritone voice with dominated by two conversations and supplemented with
live processing, fixed electronics and video projection. The several side channels of conversation. This paper will illus-
Any polygon with a geometric centroid different from 0
[5] J. Whitney, Digital Harmony: On the Complementarity text was created to highlight notions of etiquette associ- trate the two main conversations, the development of text
additionally introduces a dc offset when projected into the
of Music and Visual Art. Byte Books, 1980. ated with the emergence of the telephone in the 19th cen- and musical material, and the contributions of each process
time domain. While this can be done deliberately to pro-
tury and social media/mobile telephony in the 21st century. to the realization of cmetq.
duce modulation signals, it should be avoided in the audio [6] B. Alves, “Digital Harmony of Sound and Light,” The text was collaboratively realized and began as a se- We begin with a look at the conceptual framework for
domain. A DC blocker [15] is implemented at the output Computer Music Journal, vol. 29, no. 4, pp. 45–54, ries of tweets between the authors that were collected and the cmetq project, discussing the relationship between our
of the synth. 2005. edited. This paper will articulate the motivation that influ- subject, communication etiquette, and the resulting com-
4.5 Phase Modulation enced the formal design as well as the unique workflow for position. The following section investigates how the feed-
[7] L. Vieira-Barbosa. (2013 (accessed Dec 1,
composing cmetq. This collective development of a con- back loop of conversation serves as a model for the col-
A simple phase modulation scheme based on an LFO was 2015)) Polygonal sine animation. [Online]. Avail-
cert length work results in a synergy, exploiting the unique laborative strategies. The conversation that generated the
implemented, which can be toggled between sinusoidal and able: http://1ucasvb.tumblr.com/post/42881722643/
assets of the constituent collaborators. text occurred predominantly on social media, erupting in
linear waveforms. Very slow LFO frequencies allow for the-familiar-trigonometric-functions-can-be
Tags: Composition Systems and Techniques, Collabora- short bursts and moving across material in a nonconsecu-
ever changing soundscapes, whereas modulation frequen- [8] D. Chapman and M. Grierson, “N-gon Waves–Audio tive Work tive fluid manner. In contrast, the creation of each song fol-
cies in the audible range allow for FM-esque sounds [16]. Applications of the Geometry of Regular Polygons in lows a sequence from text to complete song, occurring as
the Time Domain,” 2014. 1. INTRODUCTION a series of recordings. Finally the composer reports about
5. CONCLUSIONS a compositional translation tool and how it evolved as a re-
[9] H. S. M. Coxeter, Regular polytopes, 3rd ed. New cmetq is a concert length stage work for baritone voice with sult of this particular work.
The proposed continuous-order polygonal-waveform syn- York: Dover Publications, 1973. fixed electronics and live processing. The work collects
thesis is able to generate a wide variety of timbres, ranging statements around notions of communication etiquette as it
from more traditional waveforms such as square and trian- [10] J. Sampath. (2016, Feb.) DIN Is Noise. [Online]. relates to the 19th century telephone and 21st century cell
gle to harsh digital sounds. The unusual control parameters Available: http://dinisnoise.org/ phone and social media. The title of the work cmetq (pro- 2. CONCEPTUAL FRAMEWORK
give a new approach to modulation and our test implemen- nounced c m e t q) is a compression of the words commu- Once the world is technologized, we can not
tation shows that interacting with it is quite rewarding. [11] M. Le Brun, “Digital waveshaping synthesis,” Journal
of the Audio Engineering Society, vol. 27, no. 4, pp. nication etiquette. This compression is a nod to the Hun- go back. - Nicholas Carr [2]
Future work should include the effect of adding addi- garian Notation system and the title, collaborative method-
250–266, 1979.
tional voices which can be synced, detuned and cross mod- ology and dramatic scope of HPSCHD by John Cage and cmetq is designed to motivate the listener to consider the
ulated. LFO tracking and more sophisticated filter topolo- [12] User: Raskolnikov. (2011, May) Paramet- Lejaren Hiller. This compression reflects the intermingling unique position of technologies in our daily life. This is
gies would also open further sonic sculpting capabilities. ric equation for regular n-gon. [Online]. Avail- of automated and intuitive processes that were used to cre- inspired by others who’ve asked similar questions, such as
Because of its immediate visual appeal, an implementation able: http://math.stackexchange.com/questions/41940/ ate cmetq. Youngblood [3] in Expanded Cinema.
with a larger display surface should be employed, which is-there-an-equation-to-describe-regular-polygons The work focuses on the language of etiquette surround-
could happen both in the virtual as well as the real world. ing communication technology. This paper documents the What happens to our definition of ”family”
[13] M. Puckette, D. Zicarelli et al., “Max/Msp,” Cycling, collaborative methods used to generate both the text and when the intermedia network brings the be-
Acknowledgments vol. 74, pp. 1990–2006, 1990. the musical composition. The text of cmetq was culled havior of the world into our home, and when
from a conversation between the authors, conducted over we can be anywhere in the world in a few
The authors would like to thank CCRMA for all the oppor- [14] T. Stilson and J. Smith, “Alias-free digital synthesis
social media. The compositional and melodic material was hours?
tunities during their research stay and Jens Ahrens for the of classic analog waveforms,” in Proc. International
great support. created via an exchange between the performer and com-
Computer Music Conference, 1996. The relationship to technologies is manifest in the behav-
poser using sound recordings. We supplemented these ex-
iors that arise socially. With the growth of population and
[15] J. O. Smith, Introduction to Digital Filters with changes with telephone, video conferencing, email, and in-
the increasing range of inter-human communication op-
6. REFERENCES Audio Applications. W3K Publishing, 2007, person conversation. McLuhan reminds us ”The medium
tions, the social aspect of the human experience expands.
https://ccrma.stanford.edu/˜jos/filters/. is the message” [1] and cmetq emphasizes this influence of
[1] M. D. Waller and E. F. F. Chladni, Chladni figures: A In the guise of etiquette, we collectively agree to a set of
the medium on the message. The working process was de-
study in symmetry. G. Bell, 1961. [16] J. M. Chowning, “The synthesis of complex audio interpersonal rules. We vote on these rules through our
signed to reveal the effects and artifacts of the medium. In
spectra by means of frequency modulation,” Journal actions, endorsing with adherence to convention and chal-
compiling the final work these effects, such as brevity, the
[2] D. Smalley, “Spectromorphology: explaining sound- of the Audio Engineering Society, vol. 21, no. 7, pp. lenging by ignoring conventions of behavior. Cmetq is not
shapes,” Organised sound, vol. 2, no. 02, pp. 107–126, 526–534, 1973. a lexicon of the rules of etiquette, but rather a collection of
Aug. 1997. c
Copyright: 2016 Christopher Jette et al. This is an open-access article observations that are drawn from considering the conver-
distributed under the terms of the Creative Commons Attribution License sation around etiquette in both the present and historical
[3] J. A. Lissajous, Mémoire sur l’étude optique des mou- 3.0 Unported, which permits unrestricted use, distribution, and reproduc- realms. The authors are asking a question about the evolu-
vements vibratoires, 1857. tion in any medium, provided the original author and source are credited. tion, or lack thereof, in social conventions.
537
536 Proceedings of the International Computer Music Conference 2016 Proceedings of the International Computer Music Conference 2016 538