The Knight That Learned to Bend Starlight

19:42 — Mask came off the print bed thirty minutes ago and it’s still warm. Black PLA, 1.2mm thick, the silhouette of a chess knight cut clean through the center. The edge resolution looks good. Layer lines run parallel to the bed so they shouldn’t scatter the starlight.

19:58 — Adapter ring friction-fits perfectly over the 80mm refractor’s dew shield. I should probably design a proper bayonet mount eventually. Tonight the friction fit will hold.

20:14 — First test frame. Vega, 15 seconds at ISO 800. The spikes are there. Asymmetric. Chaotic. The knight’s mane throws light in three directions the rook never would.

20:16 — Zooming in. The L-shape of the knight’s head produces overlapping spike families that don’t reinforce each other. This is Fraunhofer diffraction—the spike pattern is the Fourier transform of the mask’s edge geometry. Sharp edges produce clean spikes; the knight’s curves produce something halfway between a spike and a halo.

20:31 — Swapped to the rook mask. Same star, same exposure, same everything. The spikes are completely different. Four clean rays from the crenellations. Regular. Predictable. The knight was jazz; this is a metronome.

20:33 — The difference is immediately obvious. Two masks, two personalities. The rook looks like it belongs on a Hubble press release. The knight looks like it belongs in a dream someone forgot to finish.

20:47 — Mounted the pawn mask. Squat, rounded silhouette. The exposure looks almost normal—no defined spikes at all, just a soft halo around the star. Curves don’t generate diffraction structure the way edges do. The pawn is optically boring. This is useful information.

21:02 — Cold enough that my phone battery dropped 40% in twenty minutes. Brought it inside my jacket between exposures.

21:15 — Back to the knight for a longer exposure. Thirty seconds this time. The spikes are sharper but there’s field rotation creeping in—on an alt-az mount the camera rotates relative to the sky during tracking. The spike pattern smears. Should have remembered this from the Knight’s Tour mosaic when I was obsessing over coordinate systems. Equatorial mounts eliminate field rotation. My alt-az doesn’t.

21:18 — Solution: shorter exposures, more frames, stack later. 10 seconds each. The diffraction signature should align if I register on the spike pattern rather than the star centroids.

21:34 — Twenty frames in the can. Knight, rook, bishop, pawn. Four masks, four stars, eighty exposures total. Fingers are freezing. Coming inside.

22:17 — Stacking in DeepSkyStacker. The knight frames combined cleanly. Spikes crisp now that the rotation has been averaged out. The asymmetry is even more pronounced at high signal-to-noise. One spike dominates—the horse’s nose throwing a hard shadow that becomes a hard line of light.

22:28 — Side-by-side comparison on the monitor. Four stars, four geometries. The king mask I printed but didn’t test yet is sitting on the desk. Tall, narrow silhouette. I’m betting it throws vertical-dominant spikes.

22:31 — Each piece changes the sky differently. That was the hypothesis and it’s confirmed. The knight produces something no other piece can replicate.

22:45 — Stray thought: John Herschel did this in the 1830s with triangular apertures. He wasn’t trying to make art—he was calibrating optical theory. I’m going backward through history without a plan.

22:52 — The bishop mask has a crack. Noticed it when I set it down. Ran a test exposure anyway and the crack shows up as a faint secondary spike bisecting the mitre’s signature. 0.3mm fracture propagating from a layer-adhesion failure. Should reprint tomorrow with higher bed temp.

23:01 — Sitting in the dark looking at the knight frame full-screen. Vega is 25 light-years away. The diffraction pattern is entirely local—it exists because light bent around a chess piece I printed this afternoon. The star didn’t change. The geometry was mine.