A NIRCam frame minus its own transpose, read across the detector. Silence would mean a perfectly symmetric field. What sounds is the asymmetry, and nothing else. Choir and timpani.
Discrete notes instead of continuous sound: one scale degree per frame, repeats merged, so a note sounds only when the data moves. This is the edge-activity face of the cube, on marimba, in D Dorian. Rendered identically but driven by brightness, the same three minutes plays a third as many notes. The choice of which feature drives pitch turns out to matter far more than the choice of scale, and this is what that sounds like.
Brightness bands read through real bell partials — inharmonic, the way a bell actually rings — where the contrast of a band chooses the size of the bell that sounds it. Twelve distinct pitches in three and a half minutes: the sparsest and most physical thing in the collection. Shell changes ring rather than step.
The repaired mapping, alone. Pitch here is driven by edge activity — a channel that survives having the cube shuffled — rather than by brightness, which does not. It is the first melody this instrument has produced that a control cannot reproduce, and the whole of Series 2 is arranged around getting to it. Whole-tone tuning, so the scale samples pitch as evenly as twelve tones allow and the data’s own movement is quantised as little as possible.
The same cube minus itself, shifted eight slices along the wavelength axis. The residual is how fast the cube changes with wavelength: this is the only piece in the collection that sonifies a rate of change rather than a value. Oboe and bassoon, and long stretches where the cube is changing so smoothly that nothing sounds at all. A note on the silences: this piece runs quiet for 1m37s in total, the longest stretch 56 seconds. Nothing is wrong with the file. The silence is the recording, and here it carries a specific meaning. These pieces are subtractions: two files put on one shared scale and taken away from each other, with agreement rendered as silence. Where you hear nothing, the two files agreed. Nothing is wrong with the download.
A spectral cube minus a three-by-three median filter of itself. A median filter eats single-pixel spikes, so what is left over is the detector talking about itself: hot pixels, cosmic ray hits, bad samples. Each note is one defect. Music box and timpani. Half of it is silence, because most of the detector is behaving. A note on the silences: this piece runs quiet for 55 seconds in total, the longest stretch 55 seconds. Nothing is wrong with the file. The silence is the recording, and here it carries a specific meaning. These pieces are subtractions: two files put on one shared scale and taken away from each other, with agreement rendered as silence. Where you hear nothing, the two files agreed. Nothing is wrong with the download.
One piece, one frame, four million pixels — and an accident we kept. The traversal was asked for two minutes of audio from a NIRCam frame 2048 pixels square. A frame that size cannot be walked at sixty frames a second, so the recorder kept running while the traversal crawled, about twelve times slower than real time, until the process was stopped from outside. What you hear is a two-minute piece stretched by its own arithmetic to twenty-five. Nothing was time-stretched afterwards. The clock was simply outrun, and the sound of that is a very slow just-intoned drone moving through a field it will never finish crossing.
Twelve subtractions. Nothing here is a rendering of data. Two files are put on one shared scale and taken away from each other, and the residual is sonified so that agreement is silence and only disagreement sounds. The sign splits between two voices, one panned to each side. A cube minus a blur of itself is its own fine structure. A cube minus a median of itself is the detector’s defects — hot pixels, cosmic ray hits, single bad samples. A cube minus a surrogate that kept its Fourier amplitudes and threw away its phase is whatever shape survives that operation. A field minus its own mirror image is its asymmetry. Of every instrument in this collection this is the only one where not hearing something is evidence. Roughly a quarter of this programme is silence, and the silence is the recording. A note on the silences: this piece runs quiet for 10m36s in total, the longest stretch 2m02s. Nothing is wrong with the file. The silence is the recording, and here it carries a specific meaning. These pieces are subtractions: two files put on one shared scale and taken away from each other, with agreement rendered as silence. Where you hear nothing, the two files agreed. Nothing is wrong with the download. This programme contains material from the DESI Early Data Release (DESI Collaboration et al. 2024), used under CC BY 4.0. Required citation and the full acknowledgement: https://radio.enthropysystems.com/pages/attribution
The programme where the instrument is caught, and then repaired. Twice here a piece is followed by its ghost: the same render, made the same way, from the same cube after every value in it was shuffled into a random position. They are very hard to tell apart. That is the finding rather than a lapse of attention — measured, the feature driving pitch is statistically indistinguishable between the real cube and the shuffle. Then the mapping is changed. Pitch is driven by edge activity, a channel that does survive shuffling, instead of by brightness, which does not. The ghost stops matching: the real cube’s line covers twice the pitch range of its control. Near the end, two recordings with the view deliberately frozen, so you can hear how little the bands move on their own. That is what established the accurate version of the finding: the melody was reporting that the view changed, and could not report what changed. This programme contains material from the DESI Early Data Release (DESI Collaboration et al. 2024), used under CC BY 4.0. Required citation and the full acknowledgement: https://radio.enthropysystems.com/pages/attribution
Thirteen readings of the same instruction, forty-two minutes long. Each view of a data cube is quantised into eight brightness bands. The piece steps through those bands while the view itself advances in depth, weighted so that depth carrying more structure takes more time. That instruction never changes here. What changes is the tuning it lands on: struck gamelan bronze, real bell partials where the contrast of a band picks the size of the bell, quarter tones, a raga over a fixed drone, a just-intoned overtone series. Two sources. A quasar host observed by JWST’s NIRSpec, and a slice of the DESI redshift survey. Listen for the terraces, and for how differently they arrive when the scale underneath them changes. The next programme is about what those terraces are actually reporting. This programme contains material from the DESI Early Data Release (DESI Collaboration et al. 2024), used under CC BY 4.0. Required citation and the full acknowledgement: https://radio.enthropysystems.com/pages/attribution