The interferometer needed a mount for its 520 nm laser diode module (§M2). Somebody on MakerWorld had already designed one, close but bored for a different diode than my copper-head module, so I imported it into Onshape, cut a new bore through it with a Boolean subtract, re-exported, and printed it. The module seats snugly in a mount I only had to design a small fraction of.
That is the second printer's whole job. §M4 was the resin machine, the precision end of this project. The other machine on the desk is a Bambu Lab A1 mini (the mini, not the full A1, and without the AMS multicolor add-on), build volume 180 × 180 × 180 mm [24]. It prints one color at a time, and in this project that color is either black or white, in plain PLA that ranges from official Bambu spools down to unbranded rolls off Amazon. For fixtures and structural parts, I honestly couldn't tell you which roll a given part came from. Generic PLA has been fine.
The role of FDM in a precision project
The two printers split the work by what each is bad at. Resin prints are precise but small, brittle in thin sections, and every print costs a wash-and-cure cycle (§M4c). FDM prints are fast, cheap, tough, and big, but the surface is ribbed and the dimensional accuracy is a step below what the Mars 5 Ultra's 18 µm pixels allow. So the standing rule from the project's ground-truth table is resin for precision parts and anything with fine features, FDM for bodies, fixtures, and jigs, and never FDM for flexure hinges, because layer adhesion under cyclic load is a fatigue failure waiting to happen [project CLAUDE.md §2].
In practice that makes the A1 mini the shop's fixture factory. It runs unattended, a part is in hand the same evening, and if the bracket is 2 mm off you print another one and throw the first in the failed-prints box without feeling anything.
Don't start from zero
MakerWorld already showed up in §M4c, as the source of both the EVA and the clearance test models. For FDM work it became the default starting point, with an escalation ladder depending on how wrong the existing model is for my needs.
- Print as-is. Someone already solved the problem. Credit them, print, done.
- Rescale in the slicer. When the shape is right and only the size is wrong, the slicer's scale options fix it in ten seconds without touching CAD.
- Export to CAD and edit. When the geometry itself needs to change, the model goes into Onshape (browser-based CAD). For editing downloaded mesh models, Boolean cut-and-add (union, subtract, intersect) got me further than any other CAD operation. You build a clean solid (a cylinder, a box) and use it as a knife or a patch on the imported mesh, instead of trying to rebuild someone else's geometry feature by feature.
- Model from scratch. Several parts are fully custom Onshape designs, sliced in Bambu Studio on the desktop (never the web slicer) and sent to the A1 mini [builder, 2026-08-07].
The laser-diode mount
That mount from the opening is the clearest example of tier 3. In Onshape I sketched a cylinder at my module's actual diameter and Boolean-subtracted it through the existing mount body, which is the whole edit.
The module is 7.9 mm in diameter by 18 mm long according to its vendor listing, whose string is literally "7.9X18 Green 5mW 1PC" [builder, 2026-08-07].
The FDM parts ledger
Here is the ledger. These are the parts this project's files and articles reference, with the machine that made each one and the attribution it rests on. Everything on it is FDM except the one row in bold, which is resin.
| Part | Role | Made on |
|---|---|---|
| Laser-diode interferometer mount (modified) | holds the 520 nm module in the Michelson (§M2) | A1 mini ✓ (this chapter) |
| Custom rail base with printed V-grooves | the linear guide's structural body (§M2, §M9) | A1 mini [builder, 2026-08-07] |
| Slider carriage (carries the bearing bolts) | the moving element the alumina rides on (§M2) | A1 mini [builder, 2026-08-07] |
| Preload bracket + spring/screw seat | applies the calibrated 0.099 N/tick preload (§M2) | Mars 5 Ultra (resin), part of the flexure body [builder, 2026-08-07] |
| Optics fixtures (beamsplitter / mirror / photodiode holders) | the Michelson's skeleton (§M2) | A1 mini [builder, 2026-08-07] |
| Tilt board for the friction-coupon test | µ_s measurement rig (§M8 plan) | A1 mini, per the blanket rule [builder, 2026-08-07] |
| Everything else (jigs, stands, cable guides) | the blanket rule, which is that if it isn't the resin flexure, its foot, or its preload bracket, it printed on the FDM machine | [builder, 2026-08-07] |
Two notes on that table. First, the precision-critical FDM parts (rail, carriage) work because their precision lives in the right places. The bearing surfaces are the bolts and bearings themselves, and the printed V-grooves only need to be straight and consistent, which FDM does fine. Second, everything on this list is replaceable in an evening, which changed how I design. Parts stopped being precious.
What the A1 mini adds up to
The A1 mini is the least glamorous machine in the project and probably the one that earned the least attention for the most work. It got zero calibration drama (§M4a was a resin story, and the Bambu ran well out of the box), it feeds on the cheapest filament I can find, and it quietly built the skeleton that the precise parts bolt into. Between MakerWorld's base models, slicer scaling, and Onshape Booleans, most "I need a thing to hold a thing" problems in this project got solved in under an hour of actual work.
The part of the project that did not run well out of the box comes next. §M6 is the drive electronics, starting with a published circuit board whose files turned out to be broken.