Eighty-Five Teeth and Mars Still Drifted
09:14 — Forty-five hobbies in and I’m doing orbital mechanics with a calculator app because I don’t trust my own code yet. Mars orbits in 687 days. Earth in 365.25. Ratio is 1.881-ish. Need to factor that into gear teeth.
09:22 — The term is “accumulated secular error.” Every approximation drifts. If I use 85/45 teeth instead of the true ratio, Mars falls behind by a fraction of a degree per simulated year. In a decade of ticks, it’s wrong by a full week. Acceptable? No idea yet.
09:41 — First CAD sketch: a verge escapement. Two pallets catch a toothed escape wheel, release, catch. The 13th-century clockmakers who invented this didn’t have parametric modelling software, and honestly that might have been an advantage. I’ve been adjusting pallet angles for twenty minutes.
10:03 — Realization: layer line orientation matters. If the pallet face is parallel to the print layers, it’ll be rough. Rough means friction. Friction means the escapement catches twice per tick instead of once. Rotating the part 45 degrees in the slicer.
10:17 — Print started. Module 1.5 gears, which should be coarse enough not to strip. Finer teeth would let me pack more ratio into less space, but my Ender’s tolerances aren’t that good.
10:48 — Waiting. Read about the Antikythera mechanism. Greeks used 37 bronze gears to track the Moon and Sun. Largest gear had 223 teeth — encodes the Saros eclipse cycle. Lost in a shipwreck for 2,000 years. I’m basically doing the same thing in PLA.
11:12 — Print done. The escape wheel looks clean. The pallets… do not.
11:19 — Pallet faces are textured wrong despite the rotation. Reslicing with ironing enabled for top surfaces.
11:34 — Second print running. Coffee. Flipping through notes from the puzzle box project — same tolerance problems, same “hidden mechanism engineering,” same lesson about PLA shrinkage. That box had a pin that was supposed to drop into a slot at 0.3mm clearance. Took four prints to get it right. This escapement has moving parts that need to mesh under spring tension.
12:01 — Ironed pallets came out better. Glossy face where it matters.
12:14 — Assembly attempt. The escape wheel sits on a brass bushing I pulled from an old RC servo. Pallets mount on a vertical “verge” rod — hence the name. Whole thing should swing back and forth like a metronome, releasing one tooth per swing.
12:23 — It doesn’t swing. It jams. Pallet entry angle too steep.
12:40 — Third pallet design. Shallower entry, 7.5 degrees instead of 12.
13:15 — Print. Wait. Coffee is cold.
13:52 — New pallets installed. The escapement ticks. Once. Then jams.
14:04 — Problem identified: the escape wheel has slight wobble. The brass bushing isn’t perfectly concentric with the printed hub. When the high side comes around, it binds.
14:21 — Redesigning the hub with a press-fit bore instead of a friction fit. Adding 0.15mm interference. Should grip the bushing tighter, hold it straighter.
14:33 — Fourth print queued. This is the part where the orbital notification project spoiled me — software iteration takes seconds, not forty-minute print cycles. But I wanted something that ticks. Something physical. Something that encodes planetary motion in brass and plastic instead of sound.
15:08 — New escape wheel. Better concentricity. Reassembled.
15:14 — Three ticks in a row before it stops. Progress.
15:19 — The stop happens at the same tooth every time. Inspecting under magnification — there’s a tiny string of filament bridging the tooth gap. Invisible to naked eye. Scraping it off with an X-Acto knife.
15:26 — Seven ticks. Eight. Nine. It’s running.
15:27 — It stopped.
15:31 — The weight I’m using (fishing sinker, 28 grams) isn’t heavy enough to overcome static friction after the system settles. Need more mass or less friction. Or both.
15:44 — Heavier weight scavenged from a curtain rod. 85 grams. Overkill, probably.
15:51 — Sustained ticking. Twelve seconds of continuous motion before the weight hits the table. Each tick advances the escape wheel by one tooth. Fifteen teeth means fifteen ticks per revolution. If I gear this into a planet carrier at the right ratio, each tick is one sidereal day.
15:53 — Sidereal, not solar. A sidereal day is 23h 56m 4s. Solar is 24h flat. The difference is about 4 minutes, which accumulates to one full rotation per year. I’ll use sidereal. Stars don’t lie about where they are.
16:02 — Battery on laptop dying. Didn’t bring the charger downstairs. Ending session with a working escapement and no planet gears yet. Tomorrow: the Earth-Mars compound train. 85 teeth meshing with 45.
16:03 — The escapement is still ticking on the bench. I can hear it from here.