The Disc That Knew a Star but Forgot How to Sing

The disc plays nothing.

Three hours of setup, two hours of exposure time, forty minutes of stacking, and the music box comb plucks silence. Not wrong notes—no notes. The pins don’t reach.

I drove out to my usual spot east of the airfield, same place I shot the star-drift compass cards last week. Clear sky, -19°C, Sirius blazing low in the southeast like it owed me something. The diffraction grating was taped to the lens hood with electrical tape because I couldn’t find my filter adapter, and the whole assembly looked like a prototype built by someone who’d lost an argument with their parts drawer.

Which, to be fair.

Where it went wrong

The imaging worked. Sirius spread into a bright smear with dark notches where the hydrogen Balmer lines punch through—H-alpha at 656 nm, H-beta at 486 nm, the fingerprint of a star 8.6 light-years away eating specific wavelengths. The spectrum stacked cleanly. The calibration against H-alpha looked reasonable. I had seven strong lines mapped to MIDI notes, compressed into a C major scale because I wanted something that sounded intentional instead of like a fire alarm.

The CAD model built without complaint. Pins at the right angular positions, heights set to 1.0 mm based on a tutorial I found on a music box forum. Exported to STL. Sliced. Printed overnight in matte black PLA while I slept the sleep of the competent.

The disc doesn’t play because I got the pin height wrong by 0.3 mm.

Music box tolerances are brutal—I knew this in the abstract the way you know turbulence exists before you fly through it. The comb teeth sit at a precise height, and the pins need to lift them just enough to pluck, then release cleanly. Too short and the tooth never engages. Too tall and the pin drags, mutes the note, or jams the mechanism entirely. The tutorial said 1.0 mm. My comb wants 1.3 mm. Different manufacturer, different geometry, and I didn’t measure.

I spent twenty minutes holding the disc up to a desk lamp, squinting at the pin shadows, convinced the problem was the note spacing. It wasn’t. The spacing is fine. The star’s hydrogen lines are correctly positioned at 47°, 112°, 156°, 189°, 224°, 271°, and 318° around the disc. Sirius would play a slow, stubborn arpeggio if the pins could reach.

They can’t.

What I learned by failing

Pin height tolerance on FDM prints is worse than I assumed. The layer lines run perpendicular to the pin axis when you print the disc flat, which means each pin has a slightly different effective height depending on where the layer boundaries land. The variance across my seven pins is probably ±0.15 mm. Add that to my 0.3 mm baseline error and some pins are 0.45 mm too short. The comb doesn’t even twitch.

The fix is obvious: measure the comb first, model second. Or print the pins horizontally so the layers run along the pin length instead of across it. Or switch to resin, which would solve the layer problem and probably create three new problems I haven’t imagined yet.

I could reprint tonight. Adjust the pin height to 1.3 mm, re-slice, wake up tomorrow with a disc that might actually work. But I’m tired, and the coffee is cold, and there’s something instructive about holding an object that contains a star’s spectrum and produces no sound at all. The data is all there. The mapping is correct. The fabrication failed at the last millimetre.

The APRS skytrace sculptures taught me that a printed object is a confession—every wobble visible, every hesitation preserved. This disc confesses a different sin: I trusted a tutorial instead of a calliper.

Sitting with it

Sirius is still up there, burning through 25 solar luminosities and not caring whether my music box works. The spectrum data is saved. The wavelength-to-pitch mapping is saved. Tomorrow I’ll measure the comb properly, adjust the model, and try again.

But tonight the disc sits on my desk, seven silent pins pointed at the ceiling, and I’m not sure whether to call it a prototype or a paperweight. Maybe both. Maybe the difference is just whether I reprint it.

The sky gave me the data. The printer gave me a disc. The comb gave me a lesson about assumptions. That’s not nothing—it’s just not music. Not yet.