After the KAWS figure I wanted a print that would test everything at once, so I picked a hard one, a fully articulated EVA Unit-01 from Neon Genesis Evangelion, by the MakerWorld creator M Armory [17]. Articulated means every major joint actually moves, which means dozens of separate parts that have to fit each other after printing. Fine detail in every direction, and this time the parts had a functional job, not just an aesthetic one. That is also why an anime figure earns a chapter in a motor project. The clearance numbers this print forced me to measure are the numbers the dome-tip foot, the friction coupons, and eventually the flexure legs get printed to (§M9).
I went in confident, with the calibrated exposure from §M4a and the orientation and support discipline from §M4b. And the individual parts came out beautifully.
Then I tried to assemble them.
Tolerance and clearance
Joints that were supposed to snap together didn't. Pins wouldn't seat, and sockets gripped too hard or not at all. The parts were each printed well. They just didn't fit each other. That's how I learned the difference between two things I had been treating as one.
- Tolerance is how accurately a single printed part matches its intended dimensions.
- Clearance is the gap deliberately designed between two mating parts so they can move or slide together.
A perfectly calibrated printer still needs the right clearance in the model, because two parts printed at exactly nominal size will jam if the designer left zero gap between them. And clearance interacts with exposure. This connects straight back to §M4a, since any over-exposure bleeds light sideways and eats the designed gap. Your clearance numbers are only meaningful at your calibrated exposure. Mine below are for the Mars 5 Ultra at 2.6 s with the Anycubic gray resin [15], and they'd need re-testing on anything else.
Measuring clearances
Instead of trial-and-erroring the EVA joints, I printed two purpose-made test models from MakerWorld.
- "Tolerance Test" by kvnper [18], round and square pegs into holes with clearances stepped from 0.05 to 0.25 mm.
- "Clearance tolerance test" by 3D Maker Noob [19], a compact test of how tight a gap the printer can resolve between parts.
Two numbers came out of that stretch of work. The 0.15 mm for separate parts falls inside the range kvnper's model steps through, so it traces to a printed test [18]. The 0.30 mm for print-in-place sits above that model's top step of 0.25 mm, which means it did not come from that test, and this chapter can't point you at the run that did produce it. I design to it and I trust it, and it is the weaker of the two on provenance. Both are for my printer, my resin, and my exposure:
| Situation | Clearance that worked | What it means in practice |
|---|---|---|
| Separate parts assembled after printing (pins, sockets, joints) | 0.15 mm | Smooth fit, parts seat fully, joints move without slop |
| Print-in-place parts (printed already assembled) | 0.30 mm | Parts release from each other instead of curing into one piece |
Print-in-place needs double the gap because the mating surfaces are being printed a fraction of a millimeter apart in the same vat, and any bleed-over fuses them permanently. Separate parts get to be cleaned and cured before they ever meet, so they can run tighter.
Armed with those two numbers I went back to the joint-critical parts of the EVA (the shin, the foot, and the forearm). I rescaled them in the slicer rather than remodeling them, rearranging the supports for each attempt [builder, 2026-08-07], and the fits went from fighting to smooth. Scaling is a blunt way to buy clearance and I want to be clear about that. A percentage in the slicer moves every dimension of a part at once, and it only opens a joint if one side gets scaled and the mating side doesn't. The two numbers told me the gap I was aiming for. The slicer was how I got there by eye, and this chapter doesn't have the scale factors to give you.
Washing and curing without the appliance
Resin prints come off the plate wet with uncured resin, and the proper appliance for that is a wash-and-cure station. I didn't have one. The DIY workflow that did the job has four steps.
- Dirty bath. Parts go into a tub of 99% isopropyl alcohol straight off the build plate, still attached, for the first wash. This bath sacrifices itself, collecting most of the uncured resin and getting murky over time.
- Clean bath. From there into a second tub of clean 99% IPA for 5–10 minutes. Two-stage washing means the final rinse is always in nearly-fresh alcohol, so parts come out without a sticky film, and the clean bath stays usable far longer.
- Dry, then de-support. Removing supports after washing (and before curing) keeps the dimple damage down while the resin is still slightly soft.
- Post-cure. Each part goes under the custom UV curing station for 5–10 minutes per side, flipping so every face gets direct UV. Resin only hardens where light reaches it, so shadowed faces stay soft if you skip the flip. The curing station itself is a modified nail-polish curing stand with a story of its own (a short circuit, a heatsink, and a lesson in how fast UV LED modules cook themselves). That build is §M4d.
Results, including the failures
With correct clearances and the full workflow, the model assembled into an articulated EVA Unit-01 that stands and poses.
It did not go cleanly the first time, or the fifth. I collected a pile of misprints along the way, mostly from two causes I could name afterward: clearances I hadn't tested yet, and support placements that broke my own §M4b rules. Every one of them taught me something specific, which is the only reason to keep a failure pile.
The pipeline, quantified
Three chapters in, the resin pipeline is complete and quantified: exposure calibrated per resin (2.6 s, §M4a), orientation and supports systematized (§M4b), clearances settled for both assembly styles (0.15 / 0.30 mm, this chapter), and a repeatable wash-and-cure routine built from tubs and a hacked nail-polish stand. This is the exact toolchain the actuator work in §M9 depends on. The EVA was practice with a scoreboard.
And credit where it's due. M Armory's articulated EVA is excellent work [17], and the two free test models [18, 19] saved me from guessing at clearances one ruined joint at a time.
Next comes the little UV station that post-cured everything here, and what went wrong building it. That's §M4d.