The problem
Every underwater vehicle project dies in the same place: the seal. At 100 m depth, seawater pushes on every square centimetre of the hull with roughly 11 bar of pressure, and it only needs one path in.
Commercial housings rated for that depth start around €800. The Lamprey budget for the entire hull was €150. So we built our own and tested it to destruction.
Design constraints
- Internal volume: 1.2 L — enough for the flight controller, ESCs, battery and a Raspberry Pi Zero.
- Target depth: 100 m operational, 150 m survival.
- Penetrations: 6 wet-mate passthroughs (thrusters ×4, tether, sensor bus).
- Cost ceiling: €150 including end caps and connectors.
Material choice
We compared three candidate tubes:
| Material | OD × wall | Crush estimate | Cost/m |
|---|---|---|---|
| PVC pressure pipe (PN25) | 110 × 8.1 mm | ~28 bar | €14 |
| Cast acrylic | 100 × 5 mm | ~18 bar | €52 |
| Aluminium 6060 | 100 × 5 mm | >60 bar | €38 |
Acrylic is the classic choice because you can see your electronics (and your leaks). But cast acrylic crazes under cyclic load, and our budget said PVC. We kept a small acrylic viewport in the end cap as a compromise.
End caps and O-rings
The end caps are machined POM (Delrin) with double radial O-rings (NBR 70, 3 mm cord). Two lessons the first flooded test taught us:
- Surface finish matters more than tolerance. A 0.8 Ra groove face sealed; a sawn face did not, even with grease.
- Radial seals forgive assembly misalignment. Face seals do not. Use radial.
Potting the passthroughs
Every cable passthrough is a potential leak. We drilled the POM caps, threaded M10 cable glands, then back-filled each gland with slow-cure epoxy. The trick is to strip the cable jacket inside the potting — otherwise water wicks between the jacket and the conductors and walks straight past your epoxy.
The test rig
No pressure chamber? Use the real thing. We sank the sealed housing, loaded with paper towels and a €4 leak sensor, on a weighted downline:
- 25 m for 30 min — dry.
- 60 m for 60 min — dry.
- 110 m for 60 min — dry.
- Winched to ~140 m equivalent using a hand pump chamber built from a fire extinguisher — first drip at 13.5 bar, at a passthrough we had potted with fast-cure epoxy as a control.
What we’d change
Slow-cure epoxy everywhere, one fewer penetration by multiplexing the sensor bus, and a vacuum port so every assembly can be vacuum-tested on the bench before it ever touches water.
Total spend: €138. The ocean is not impressed by budgets, but it does respect O-rings.