The Timbre
of Starlight

A journey in additive synthesis

The Milky Way Panorama, Photo Credit: ESO/S. Brunier

Gahlord Dewald :: 3/22/23 :: Mānoa, Hawai‘i


In 1925 Cecilia Payne published a method of deriving the chemical composition of stars by examining spectrographs stellar light. The Timbre of Starlight is a project which builds on discoveries by Payne and others to transform the data of spectrographs into sound texture. Some of this work is more sonification–creating consistent and accurate mappings of spectrographic data into distinct sounds. Some of this work is more musical–using field recordings and other methods to excite filters resonating with the sonic fingerprint of stellar light.

This is a work in progress. As it develops I'll be adding to and revising all of the sections. As the project spools up most of the detail will be in found in reviewing the project log. In particular the .plan section gives a heads up on the current top-of-mind tasks. Starlight Encoded discusses the technologies and techniques used to produce audio based on stellar spectra. The Aural Signatures section houses most of the results. There is a Glossary expand as I go.

Starlight Encoded

In order to render the spectral data of stars in a form that our ears can comprehend, I’ve created two pieces of technology.

Wireframe outline of a sound processor designed for adding sine waves
The elements of the sound computer used to generate audio for this project.

The Additive Sines Computer (TASC) is a collection of eurorack modules installed in a portable case. This machine was originally designed for analog and manual control over additive synthesis—a form of synthesis which is most often approached via digital "in-the-box" methods.

Though designed for manual control, TASC contains several monome crow modules which allow for digital control/computation using the Lua language. It is this capability that makes it possible to assign specific frequencies and amplitudes of component waves, discussed more below.

For audio generation and manipulation there five kinds of modules are used:

In addition to TASC, audio is gathered into a computer by way of an analog-digital converter (in my case this is most often an rme Babyface Pro for simple tests and a Merging Technologies Hapi for more involved things) and recorded to computer disk by Apple Logic Pro X. Tuner and test oscillator plugins within Logic Pro X also assist during the tuning and configuration of TASC prior to a recording session.

The signal flow of The Additive Sines Computer is as follows:

  1. An +5 to -5 control voltage (CV) is generated by a Pulp Logic Att-Off 1U tile module.
  2. The CV is through a Tendrils multcable into the inputs of two separate monome crow eurorack modules.
  3. The monome crow modules Each monome crow module can drive two LPFs and two VCAs—this results in four partials being available for Timbre of Starlight. evaluate the CV and send pitch CV to 2hp LPF eurorack modules and amplitude CV to channels on the Doepfer A-130-8 Octal Linear VCA eurorack module.
  4. The LPF eurorack modules are configured such that they are self resonating, these are the sources of our sine waves. The pitch CV is evaluated to set the frequency of the sine waves.
  5. Audio from the LPF modules are sent to channels of the Octal Linear VCA eurorack module. The amplitude CV sent to the Octal Linear VCA eurorack module in step 4 determine how much of each audio signal is allowed to pass through.
  6. The Octal Linear VCA eurorack module has a summing bus which sums the audio of all the partials. This summed audio signal is fed to a Whimsical Raps Rip (DIY) transformer output eurorack module.
  7. The ¼ inch output of the transformer output eurorack module is fed into an Audio/Digital converter for audio capture.

An additional Pulp Logic Att-Off 1U tile module is used during the tuning and measuring process to provide simple volume control.

The Additive Sines Computer is, for the purpose of this project, driven by a set scripts I collectively refer to as the cecilia-payne.lua scripts. Currently there are two separate scripts: one for setting frequencies for four component waves and one for setting the amplitudes for the four component waves. These scripts are loaded into the monome crow eurorack modules and receive input on two channels and send cv on four channels.

The script dealing with frequencies is named cecilia-payne-angstroms.lua. This script encodes the base frequencies and component frequencies of each element. Those frequencies are translated into control voltage and sent through the outputs. We can select which element's frequencies are being output by sending control voltage to the input of the crow. I am using a Pulp Logic Att-Off tile which can send a roughtly +5 to -5 control voltage and has a simple knob interface.

The script dealing with amplitudes is named cecilia-payne-oscillatorstrength.lua. This script encodes the relative amplitudes for each component wave. Those amplitudes are translated into control voltage and sent through the outputs. We can select which element's frequencies are being output by sending control voltage to the input of the crow. I am using a Pulp Logic Att-Off tile which can send a roughtly +5 to -5 control voltage and has a simple knob interface.

Both scripts contain a tuning mode for use in tuning the oscillators and for assessing the available dynamic range of the vcas. Additionally there are utilities for:

These additional tools are accessible via druid connection and require a laptop connection to read the outputs.

Aural Signatures

The end result of this project is an exploration of sound design, sonography, and musical resources. Can we hear and distinguish the composition of a star through this modeling effort?

In the Elements: Spectral Timbre sound examples, four persistent spectrographic lines for each of the alpha process elements have been selected (based on a criteria of greatest "Intensity" with ties broken by greatest transition probability). Frequency in ångstroms, amplitude as transition probability (Aki(108s-1)) normalized to 5v. The entire series has been lowered one octave to make some of the higher overtones easier to hear.

Some methods for learning to hear the differences between these sounds include:

Current results of the Timbre of Starlight project

Here are some descriptions of the timbre of the sounds created so far. These are provided to help you listen to this information as data or to hear the characteristics within them.

These are my own subjective listening notes. You might have your own way of hearing and understanding these sounds, and that's great.

  1. "Tuning"a sine waveTimbre Waveform: Tuning Sine is the four oscillators set to the same pitch. It fades in and out due to phase cancellation. The slower the swells in volume are the more in tune the oscillators are. A perfectly in tune and in phase collection of oscillators would be loud and without volume swells. Perfectly in tune but out of phase oscillators would have regular volume swells, slower being more in tune. Including it here is a way of helping confirm the accuracy of the equipment.
  2. "Oxygen"similar to a sine wave but much more raggedTimbre Waveform: Oxygen contains a fair amount of space between a lower tone which wavers ever so slightly and some tightly spaced middle tones. The tight spacing of these middle tones is responsible for the "buzzing" tone in the sound. There is also an extremely quiet high tone which may be difficult/impossible to hear. However this high tone adds a very subtle wavering texture to all of the sounds which are directly perceptible.
  3. "Neon"similar to a sine wave but with a smaller modulation wavering the lineTimbre Waveform: Neon also has a large gap between the lowest tone and middle and upper tones. However, since the middle and upper tones not tight, as in "Oxygen," the result sounds less buzzy.
  4. "Magnesium"ragged and chaoticTimbre Waveform: Magnesium features some tones closer together in the lower range, not as tightly spaced as in "Oxygen" but close enough to create a little bit of buzz. A higher pitch can be heard above these two. The relationships between all of the tones creates a very subtle pulsing effect.
  5. "Silicon"a wave that, though a little complicated, has a discernable and regular patternTimbre Waveform: Silicon features tones that are all spread apart from one another. This makes it one of the least buzzy of the timbres. It is also one of four of the alpha process elements which has a subtone--the lowest tone is not the loudest.
  6. "Sulfur"Sine-like but with clear, smooth animation of louder and softer Timbre Waveform: Sulfur has a noticeable, quick pulsing effect. It's the only element of the alpha process series that has two subtones--and those two tones below the most prominent tone are responsible the pulsing. The relationships between the various tones are close enough to being simple multiples of one another that the overall loudness of Sulfur is greater than most of the other elements. It has a "bright" sound. Sounds prettier than it smells.
  7. "Argon"Sine-like but a little wavering/raggedTimbre Waveform: Argon includes a very tight buzzing in the upper partials. The relationship between the lowest tone and the upper buzzing tones sounds more "harmonic" or "in tune" than "Oxygen" in which the tones are more tightly spaced. The wide distance between the low tone and the upper tones is like "Neon" but the buzzing in the upper tones of "Argon" makes it possible to distinguish between them.
  8. "Calcium"Tight alternating loud/soft waves that also have clear loud/soft animation.Timbre Waveform: Calcium has a noticeable buzzing as well a noticeable pulsing, all in the upper registers. The relationship between the buzzing tones is give the buzz a sort of growl or ticking sensation, where the buzzes can be heard distinctly--to fast to count out loud but slow enough to perceive. The distance between the lowest tone and the upper tones is not very great but neither is it tight enough to create even more buzzing.
  9. "Titanium"Uniform and clear loud/soft animation.Timbre Waveform: Titanium features a slow, shallow pulsing effect. It also has a subtone, the lowest pitch is not the loudest. The relationships between the audible tones are tight, but not so tight as to create prominent buzzing. The long shallow pulsing makes "Titanium" distinguishable from "Sulfur" which has a quicker pulse.
  10. "Chromium"A little ragged and with clear animation of loud/soft over time.Timbre Waveform: Chromium has tightly spaced tones resulting in a clangorous timbre that has prominent pulsing. It has a subtone--the lowest note is not the loudest, it hangs below the most noticeable pulsings sounds. Where "Calcium" similarly has tightly spaced tones and has an upper register wavering tone, "Chromium" is distinguishable by having no upper register wavering tone.
  11. "Iron"rough looking patternTimbre Waveform: Iron I has a beating/buzzing texture that pulses in and out of the upper part of the sound. The distance between the lowest tone and the beating texture in the upper part is not as large as in "Argon" but also not as tight as it is in "Oxygen" either. The pulsing of the beating texture is unique among the alpha process element timbres, at least in this frequency range.
  12. "Tuning (Again)"a sine waveTimbre Waveform: Tuning is a check to see if the tuning drifted while gathing these audio sample. Since the waveforms of The Additive Sine Computer are constant (they don't reset from zero but rather continue on eternally) the pattern may be slightly different as the four oscillators go in and out of phase.

Project Log

If you would like occasional updates about this project and other things I'm working on, I encourage you to join my newsletter, Sound|Community|Culture.

.plandata{select a star abundance table}, research{what is the metallic abundance of the Sun}, testing{confirm/encode all frequencies/amplitudes}, documentation{write the basic overview of cecilia-payne.lua}

|09:03:2023"Stellar Light Study #5" has been released by Germany's re:natura collective. This piece uses the elements in approximately the same proportion as our Sun. The filters are excited by a field recording of a beach here in Honolulu.

|16:12:2022Spoke at Bogliasco FoundationThe Additive Sines ComputerPhoto: Sondra Lapage re: my creative practice and had great conversation about Timbre of Starlight in particular.

|28:11:2022Documenting the math used to convert Hz to CV.

|05:09:2022New recordings that make use of implementation of dynamic range controls on cecilia-payne-angstrom.lua, testing process to determine lowest practical voltage to open the VCAs. Changed audio streaming on Creative Music Player to use an older standard as I couldn't figure out how to get HLS to play nice with Firefox.

|03:08:2022Wireframe for The Additive Sine Computer complete, opening of the "Encoding Starlight" section.

|10:05:2022A twitter thread that gives some biographical details about Cecilia Payne and contains a link to a a review of What Stars Are Made Of.

|03:04:2022Replacement parts addedThe Additive Sines ComputerPhoto: Leilehua Lanzilotti , auxiliary VCA sidecar populated (may prove unnecessary or a limitation in the long run, additional VCA capabilities have now been added to the additive sound computer).

|12:03:2022Starting to wrap my head around how to read an abundance table, courtesy Swinburne University of Technology, Australia.

|18:02:2022Making preparations to send the module I fried back for repair. While the system will be disassembled I'm taking advantage of the downtime by reorganizing the layout and swapping some parts. Namely, moving a collection of 8 VCAs onboard (though this project can only accommodate 4 partials, the system is cable of 8 independent sine waves), improving the summing capabilities, and adjusting the layout for more sensible cable management. I always tell myself that a modular system is done but it's never truly done. I've also cooked up an idea for a 1u Tile "sidecar" that I think will work.

|16:02:2022Got hardware working well enough to record new samples for Elements: Spectral Timbre. One of the jacks is definitely fried on the synth, likely have to order a replacement module. Ordered new VCA modules built on the 1280 chip to improve consistency between VCAs as well better testing control. Project budget totally blown, haha. Added wave imagery and timbre descriptions, fleshed out Aural Signatures a little bit.

|15:02:2022Added S and Cr to data tables and code. Then promptly fried the input of a critical module. :'( Sure sounded rad for a moment.

|14:02:2022Hardware tuning process complete & documented. First light on the sounds recorded. Opened the Aural Signatures section. Ran into system distortion until the maximum amplitude was brought down to 3.8v, might try to find where that's coming from and resolve it so I have more dynamic range. Though the compression might be nice for hearing upper partials better. Gathered wave imagery to accompany but ran out of time, will post later.

|13:02:2022New data assignment explored. Completed cecilia-payne-angstrom-b2.lua and cecilia-payne-oscillatorstrength-b2.lua

Frequency vs Probability for 8 alpha process elements
eight of the alpha process elements, frequency vs probability, Aki(108s-1) normalized to 5v for highest probability in a set.

|12:02:2022This is all really about time isn't it?

|11:02:2022Geoff posts an intro and some science context. It answers a question that's been nagging me "Hey what about our sun? How do we get a continuous spectrum?" Answer: we don't get a continuous spectrum Researched wave-particle duality & "aether" (while trying to understand similarity/differences between light "waves" and sound waves).

|07:02:2022Minor setback upon realization that using kj/mol for amplitude does not increase the available data, as it maps fairly simply to the frequencies chosen.

A graph of frequencies vs kj/mol which shows a pretty clear and simple relationship between the two.
Noticing that frequency and difference between energy levels are different ways of saying the same thing.

"A measurement of any one of the entities frequency, wavenumber, or wavelength (in vacuum) is an equally accurate determination of the others since the speed of light is exactly defined."A. Kramida, Yu. Ralchenko, J. Reader, and NIST ASD Team (2021), Atomic Spectra Database (version 5.9), National Institute of Standards and Technology, accessed 07 Feb 2022

Switching amplitude to something in the probability column of the NIST data. More research required.

|06:02:2022Repurposing my Sleipnir Lua script to drive the project. Prototype 1A prototype of the Timbre of Starlight SystemPrototype 1: modular synth controlled via Druid/Lua/Crow. complete: four frequencies drive analog sine waves and voltage controlled amplifiers to generate a timbre to represent oxygen.

|04:02:2022With arrival of Winterbloom’s uncannily appropriately named Hydrogen module, the analog additive synthesis computer assembly is complete.

|21:01:2022Cecilia Payne’s Stellar Atmospheres.

|20:01:2022First table of data prepared containing frequency selections,

|10:01:2022Research continues: Locating scientific papers which outline the abundance ratios of various stars. Mild comprehension.

|06:01:2022Research continues: What elements are in stars? What is the alpha process? What does the energy level of the NIST table mean?

|04:01:2022Research continues: Density, Translational Motion, Rotational Motion, Wavelength, Abundance Ratios

|03:01:2022@_TheGeoff launches his version, complete with a redshift and blueshift slider. StellarSynth is web-playable and includes background science information.

|27:12:2021Research begins: Bohr’s Postulates, Balmer, Lyman, Paschen. What elements are in a star? What does it mean that angular momentum is quantized?

|26:12:2021An initial image crossed the transom, a question pushed forward in the wake and condensed a further question.

@_TheGeoff sparks the idea:

“The new #JWST launched yesterday, is powerful enough to see the thin fringe of a planet's atmosphere as it passes in front of a star. We know the spectrum of the star, so we can see which colours the planet's atmosphere ‘subtracts’. We can see its chemistry. e.g. maybe life. Fun music analogy: the same note on a flute, piano and violin sound different. We're getting an oscilloscope trace of a whole orchestra (light spectrum) and are able to pick out the different instruments (atoms/molecules).“

Geoff: Fun music analogy: the same note on a flute, piano and violin sound different. We're getting an oscilloscope trace of a whole orchestra (light spectrum) and are able to pick out the different instruments (atoms/molecules). Gahlord: I suppose you could do a timbre map of each in some way, converting light frequencies to music frequencies, (assuming more than one light frequency is involved) then pitch down to human range. Use additive synthesis to sum it together. Probably noisy haha.


Please note that I am an experimental musician, not an astrophysicist. This glossary may well contain errors and oversimplifications.

alpha process
the order in which stars create elements beyond carbon. Helium plus the product of the previous step yields the next element in the process. The order is:(C + He →︎) O + He →︎ Ne + He →︎ Mg + He →︎ Si + He →︎ S + He →︎ Ar + He →︎ Ca + He →︎ Ti + He →︎ Cr + He →︎ Fe
oscillator strength
an alternative measure of transition probability
phase cancellation
sound waves of the same frequency and amplitude will cancel each other out to the degree that their amplitude peaks and troughs align
the quality of a sound, resulting from the mix of fundamental and the overtones it contains, which makes it distinct and recognizable
transition probability
the probability that an electron transitions between shells in an atom (and thereby releases or absorbs photons)