Converting Astronomical Spectra into Sounds: a Tool for Teaching and Outreach?

Paul Francis, Department of Physics, Faculty of Science, The Australian National University, Canberra 0200, Australia.

Joint appointment with the Research School of Astronomy and Astrophysics, Mt Stromlo Observatory, The Australian National University.

Abstract

In this paper, I present a new way of presenting spectra to the public: convert the electromagnetic spectra into sound spectra and let people listen to them. I explore the strengths, weaknesses and limitations of this technique. For emission-line objects such as quasars and planetary nebulae, it can work superbly, producing sounds that are both beautiful and informative. For many other objects, however, the results are less beautiful, less useful or both. I discuss possible educational, media and outreach uses for this technique, and present a library of astronomical sounds.

Keywords

Multimedia – Radiation and Spectra – Public Outreach – High-School – College non-majors – Museum & Planetarium

1 INTRODUCTION.

Over 70% of observations with most research telescopes consist of some form of spectroscopy, and spectroscopy is fundamental to our understanding of almost all astronomical phenomena. Despite this, imaging completely dominates the public perception of astronomy. This is not surprising. It is almost impossible to get spectroscopy into a media news story, primarily due to the severe time and space constraints. By the time you’ve explained that light is made of waves, that we can measure the different waves, that they tell us about what things are made of, you’ve already filled the press release and exceeded the attention span.

The dominance of imaging is almost equally prominent in introductory astronomy text-books. I informally surveyed a few popular “ASTRO 101” textbooks, and found that while the text gives due weight to spectroscopy, the illustrations are biased 10:1 or more in favour of colour images. A similar bias is obvious in science centre and planetarium exhibits and shows.

None of this should be surprising. Color images are one of the greatest assets that we as astronomers enjoy in our quest to educate and inform. They are beautiful, and both easy and intuitive to interpret (at least to some level). Spectra, wiggly lines plotted on paper, are neither.

It occurred to me (and has doubtless occurred to many other people) that perhaps we are using the wrong sense to present spectra. Perhaps we should use our ears, not our eyes, to examine spectra? Our eyes give us excellent spatial information but only a little spectral information, while our ears give us the opposite. Thus our ears seem a better match to the capabilities of most spectrographs.

The use of sound in astronomy education and outreach is nothing new. In the reference section, I give links to web pages presenting a wide variety of sounds, including the big bang, solar and stellar oscillations and the winds on Titan. In these cases, the signals being converted to sound are acoustic pressure oscillations: they really are sound waves, simply needing their frequencies to be shifted into the audible range. This limits the applicability of the technique: we only have data on acoustic oscillations in a handful of astronomical objects, and to date the sounds produced, while fascinating and evocative, are far from beautiful.

In this paper, I try something different: I experiment with converting electromagnetic wave spectra into sounds. This is a slightly bigger conceptual leap: we are converting electromagnetic waves into sound waves. But is does allow us to listen to a vastly wider range of astronomical objects, and gives us a new way to present spectroscopy to students and to the general public.

In Section 2, I give technical details of how to convert spectra into sounds. The programs I use to do this conversion are available on my web page, listed in the resources section. In Section 3, I systematically investigate the conversion, and what sorts of spectral information it can and cannot usefully convey. The conclusions drawn in this section are backed up by sound-tracks, also available on my web page. In Section 4, I discuss the educational and outreach uses to which these sounds might be applied, giving examples. Finally, conclusions are drawn in Section 5.

2 HOW TO TURN SPECTRA INTO SOUNDS: STEP BY STEP GUIDE

Perl programs to do stages 1-6 of the conversion are available on my web page, listed in the resource section.

1. Input spectrum. You will need an observed or synthetic digital spectrum covering a wide range of wavelengths, which is flux calibrated. One of my programs will take a list of emission-lines or continuum parameters and synthesize the spectrum: another reads in a pre-existing ASCII spectrum.

2. Convert to frequency. Many spectra come as tables of flux per unit wavelength against wavelength, and will need to be converted to flux per unit frequency against frequency.

3. Reduce the frequency. Almost any astronomical spectrum will have a frequency far above the ear’s audible range (~ 20 – 12,000 Hz). The frequency will thus need to be reduced. As discussed below, for optical/UV spectra I use a conversion such that the

H-alpha emission line (656.3nm) comes out as the note middle-C (261.63 Hz), which seems a reasonable compromise applicable to a wide range of spectra. This corresponds to reducing the frequency by a factor of 1.75 trillion.

4. Choose the phases. A phase must be assigned to each frequency bin. I choose these randomly, which is appropriate for almost all astronomical sources. Non-random phases can produce dramatically different sounds (the Fourier transform of a Gaussian emission line with uniform phases is, for example, a Gaussian, so instead of a continuous tone, you would hear a bleep).

5. Do the Fourier Transform. This spectrum is then converted into a discretely sampled waveform by taking its Fourier Transform, typically using the Fast Fourier Transform (FFT) algorithm. I take the square root of the spectrum before doing the FFT, to convert it from a power spectrum into an amplitude spectrum. The sampling rate used must be high enough to do justice to the sound: I use the CD sampling rate of 44100Hz, which eases importation of the sounds into sound editing programs.

6. Output the discretely sampled waveform as an ASCII file. The first line should list the sampling rate, while subsequent lines consist of an increasing integer and the sampled flux value (in the range –1 to 1). The file name should have the suffix “.dat”. Here are the first few lines of an example, to demonstrate the format.

; Sample Rate 44100

0 0.0459122

1 0.0646971

2 0.083316

3 0.101704

4 0.119761

5 0.13734

7. Convert the ASCII file into a “.wav” file. To make this conversion I use the freeware SoX Sound exchange utility (http://sox.sourceforge.net/) written by Lance Norskog and Chris Bagwell, which is available for a wide range of operating systems. This “.wav” file can now be played on a computer, inserted into powerpoint presentations or web pages.

8. Combine sounds. It may be helpful to combine different sounds, change the volume, record voice-overs etc. I use the Apple Garage-band program to do this: you can simply drag and drop the “.wav” files into it and edit/combine them in many ways (as long as the sampling rate is 44100). Similar programs exist for other computer types. Care is needed in tacking together multiple sound clips to make long sounds, as a click is often distractingly audible at the join. The solution is to generate longer clips to begin with, or to stick them together so that there is no amplitude discontinuity.

9. Convert to a compressed format. The “.wav” files are large (1 MB for an 11 sec clip). Converting them to a compressed format dramatically reduces the size with little penalty in quality. I use the Apple iTunes program to convert the clips to mp3 format, which reduces the file size by a factor of around eight. Many similar format conversion programs exist.

3. CAPABILITIES OF THE HUMAN EAR

I’ve experimented extensively with converting different spectra into sounds. The capabilities of the human ear determine which sounds are useful and/or attractive and which are not. When I started this experiment I had a wide range of ideas of how sound conversion might be useful: the limitations of the ear rule out many of this applications.

In this section I describe what I learned. The points made here are best demonstrated by sound-tracks, which can be found on my web page (resource section).

2.1 Frequency Range.

In principle, the human ear can hear sounds over an extremely wide frequency range: 20 – 20,000 Hz. In practice, I find that most computer speakers have trouble producing audible sounds below around 100 Hz, and that frequencies above around 7,000 Hz are hard to hear and unpleasant to listen to. Best results seem to be achieved if you choose your frequency normalisation to place the main spectral features in the range 200 – 5000 Hz or thereabouts. Listen to the demonstration sound file on my web page and make your own mind up on this.

This is still a very wide range: few astronomical spectrographs cover more than a fraction of this frequency range. This makes it hard to find data to convert that truly match the ear’s capabilities. The problem is particularly acute in the radio and sub-mm, where spectroscopic bandpasses are very narrow. I find that one good place to find spectra with wide frequency coverage is combining optical and UV data, the latter typically from the IUE satellite.

2.2 Frequency Resolution.

While the ear covers a much wider frequency range than most spectrographs, its spectral resolution is much more limited. Most people can clearly distinguish notes 6% different in frequency (a semi-tone). Below this, sensitivity varies from person to person: 3% is often distinguishable, but 1% is marginal even for the most musically gifted. Try out your hearing using the demonstration sound file on my web page.

This limited spectral resolution means, for example, that the difference in sound across a galaxy rotation curve, or the Doppler shift of a binary star is too small to be heard, unless we artificially exaggerate it. A 3% frequency shift corresponds to a Doppler shift of 9,000 km/s.

2.3 Frequency Separation of Two Notes.

Even though the spectral resolution of the ear is only around 3%, two notes closer together than this will be very noticeable, as they will beat together. Indeed, if you play two notes less than around 30% different in frequency, instead of hearing two notes, the ear hears one note mid-way between them, plus strong beating. This is both a benefit (it allows you to detect the presence of closely spaced notes) and a curse (you will not hear the separate frequencies).

2.4 Dynamic Range.

The dynamic range of the human ear is a second place in which the ear outperforms many astronomical instruments. If I play one note, and then add a second note, well separated in frequency, this second note makes a perceptible difference even if it has less than 1% the power of the first note. The perceptibility limit seems to be around 0.1%.

The loudness range at which a given sound can be played is also very great. With typical computer speakers, you can decrease the power of a sound by 100,000 times and still hear something. In practice, however, ambient noise when the sounds are listened to usually decreases this dynamic range significantly.

2.5 Emission-line Width.

I now experimented with playing emission-lines with finite velocity widths, rather than pure sine-wave tones. I use Gaussian line profiles.

Line width makes a striking difference. Below roughly 100 km/s (full width at half maximum height) there is no obvious effect, but for larger velocity widths, the different frequencies within the line interfere with each other and cause the amplitude of the line to wander up and down in a somewhat random way. The greater the line width, the faster these amplitude wanderings. For widths of a few hundred km/s the effect on a multi emission-line spectrum sounds something like a peal of bells, with different lines coming in and out of prominence. For wider line widths still, each line sounds “grumbly” or “querulous”. These fluctuations are probably hard to explain to students, the public or the media, but they do make the sounds more interesting, and they enhance the audibility of different notes by bringing them to peak volume at different times.

2.6 Continuous Spectra.

Continuous spectra (such as power-law continua or black-body spectra) sound somewhat like noise. The character of the noise does vary somewhat as the shape of the spectrum changes. A low frequency black body, for example, sounds like waves on a beach. At slightly higher frequencies, it sounds like sitting in jet aircraft. Higher frequencies still sound like static on a radio or TV.

I generated the sounds corresponding to black-body spectra at a range of temperatures. The human ear can hear differences of around 30% in temperature, so one could, for example, classify stars into spectral types by ear.

2.7 Emission or Absorption Lines Combined with Continuum Emission.

The biggest disappointment for me was the extremely limited ability of the ear to pick up emission or absorption lines superimposed on a continuum spectrum. Even very strong emission and absorption lines make little or no perceptible difference. For example, it is hard to tell the difference between a 6000K black body and the spectrum of the Sun, or between a power-law continuum and the spectrum of a QSO. The power in the emission line needs to be comparable to that in the entire continuum to be detectable.

This rules out many possible applications of sound conversion, unless we artificially reduce the continuum strength, so as to bring out the lines. It means that the most interesting sounds come from sources with the highest equivalent width emission lines, such as planetary nebulae, HII regions and supernovae remnants. These are, interestingly enough, exactly the sources that are most photogenic!

3 POSSIBLE APPLICATIONS

3.1 Ambient Music and Image/Movie Sound-Tracks

Planetaria and science centers have no shortage of gorgeous images and animations to excite the public. In many applications, however, it would be nice to have a sound-track to go with these visuals. Ambient “spacy” music is often used, and can be very effective. Converting spectra into sounds, however, is more “authentic” and physically motivated. The use of sounds may also provide a way to make astronomy more accessible to people with impaired vision.

This will only work if the sounds are attractive. To test this, I obtained combined optical and UV spectra for a wide range of astronomical objects, and produced the sounds. These sound clips are available on my web page (resource section).

It turns out that many emission-line astronomical sources have optical/UV spectra that convert into rather attractive, mysterious shimmery sounds. The emission-line regions of quasars, supernova remnants, planetary nebulae and HII regions all sound quite nice, especially (to my ear) the first two, which are gorgeous.

Stars and other continuum sources can sound like waves on a beach. Cool stars have an aggressive “growly” sound (due to interference in the TiO bands), which could be quite effective used in moderation.

One drawback is the “sameness” of many of these sounds. For example, the spectra of different planetary nebulae, or of different locations within a given planetary nebula only sound subtly different. Different classes of nebulae (eg. planetary nebulae compared to supernova remnants) sound more distinct, while broad-line sources such as AGN, or spectra dominated by molecular bands (eg. comets) are quite different. But there is only a limited range of sounds, unless you start playing with the frequencies things are shifted to, or start using non-random phases.

3.2 Education

Perhaps the most obvious educational use of these sounds is teaching about spectroscopy. One could, for example, play the sound associated with a given nebulae, and then play one by one the sounds coming from each of the elements found in that nebulae, combining them to show how the nebular spectrum is built up, and how you can use the spectrum to work out what the nebula is made of. This approach, combined with more conventional approaches, might get the students interested, and engage those with a preference for auditory (as opposed to visual) learning styles. It might also provide an interesting connection to sound physics and the “real world” of guitars and amplifiers.

As an experiment, I put together a couple of “sound stories”. The first was the life of our Sun, starting out with the spectrum of the warm interstellar medium, then that of a star formation region, then the spectrum of the Sun, getting slowly louder due to main sequence stellar evolution. It then gets very load and changes to a red giant spectrum, before becoming a planetary nebulae and then a cooling faint white-dwarf spectrum. The second was a comet, starting off with a faint reflected sunlight spectrum as it enters the solar system, with molecular bands emerging as it swings past the sun, and then fading away again. I’ve tried both out on a variety of students, and their feedback is positive: they find them an interesting, different and memorable way to present these “stories”.

It might also be possible to use sound conversion as an interactive lab-type exercise, to learn about both spectra and sounds. The current programs are probably not simple enough for this purpose (they require installing and running the Perl FFT module, a Perl program and a command-line utility, SoX), but it should be straightforward to run these programs through a cgi web interface – I would be interested in working with anyone who would like to try.

3.2 Media

These sounds may provide a way to get certain astronomy news stories into the media, which could not be made publicly accessible (given the short time/space constraints of most media coverage) in any other way. Some examples might include:

1. Discovery of a new emission line, or new element in some object. Given the ear’s excellent sensitivity to even weak lines, you could play the signal from this source with and without the newly detected feature.

2. Spectral diagnostics. If, for example, you find a galaxy with an unusual spectrum, you could just play it to show how weird it is. Or you could compare the sound with models to show how you worked out what something is.

3. Certain instruments such as integral field spectrographs are very hard to explain o the media: playing the sounds detected at different points could make this possible.

Use of attractive sounds may also help stories get on the radio, rather than being restricted to the print media.

4. CONCLUSIONS

The most heartening thing I learned from this project was that the sounds of astronomical spectra can be both beautiful and informative. The technique of converting spectra to sounds is useful for pure emission-line spectra, and for picking up gross features of continuous spectra. It is not, alas, helpful for presenting absorption lines or emission-lines superimposed on a continuum. I hope, however, that readers will find this technique useful, and look forward to hearing of any interesting applications. To help with this, I present a library of sounds, as well as the programs I used to make them, on my web page.

Resources

My web page: http://www.mso.anu.edu.au/~pfrancis/Music/

Sounds on Titan.

http://www.esa.int/SPECIALS/Cassini-Huygens/SEM85Q71Y3E_0.html

Sounds of the Big Bang: acoustic peaks converted into sounds.

http://faculty.washington.edu/jcramer/BBSound.html

http://www.astro.virginia.edu/~dmw8f/index.php

Sounds of Pulsars: pulsing signal amplitudes converted to sounds.

http://www.jb.man.ac.uk/~pulsar/Education/Sounds/sounds.html

http://pulsar.princeton.edu/pulsar/multimedia.shtml

Sounds of the Sun and other stars: solar oscillations converted into sounds.

http://solar-center.stanford.edu/singing/singing.html

http://www.eso.org/outreach/press-rel/pr-2002/pr-10-02.html