Red Means No Fix
At first blush, you’d think that GPS on a scooter designed to be underwater is a dumb idea. Totally unnecessary, and if necessary, something that can be absolutely under-engineered. Then you realize that you need to get a fix somewhere besides the beach — say at the descent buoy. Then the only antenna you’ve got is a postage stamp that wants clear sky and about a foot and a half of freeboard it does not have.
Act I: Simplicity Is King, Until It Isn’t
The first build put the GPS board right in the main NAV tube, same housing as everything else. One tube, no extra penetrations, no extra seals to fail. At the time that felt like the obviously correct call — every penetration is a future leak, so don’t add one you don’t need. Furthermore, you only really need GPS at the surface, so there’s no sense in having anything external for GPS. No penetrations, no extra complexity. Great.
The GPS chip antenna doesn’t care about my design philosophy. It wants a clear view of the sky and it wants to be something like 12–18 inches above the water’s surface. Apparently RF waves in the 1000MHz range like to bounce off water, and the antenna isn’t great at differentiating between direct hit waves and bounced waves, so the time difference of arrival needed for good GPS signal just doesn’t work. Below about 12-18in, or through any meaningful thickness of water, it sees nothing. “Simplicity is king” turned into me treading water at the buoy, holding a 52-pound Cuda 650 a foot and a half over my head, trying to get a fix before my arms gave out. I’m in decent shape and that was still absurd. It is not a fix process anyone should have to repeat every dive.
Act II: Wire It Out, Rubber-Band It On
Iteration two kept the GPS board where it was and ran a cable out to an external antenna instead — straight off the board’s u.FL connector, up a short PVC mast, mounted on the side of the scooter under a couple of rubber bands. Potted the junctions so it wouldn’t flood.
When it worked, it worked well. The antenna only needed to clear the water, not the whole scooter, so a fix was finally achievable without a shoulder workout. Holding the “mast” aloft was easy. But “under rubber bands” turned out to be a worse idea than it sounded on the bench. Getting the mast out from under the band was fine. Getting it back in — underwater, in gloves, while getting beaten up by surface chop, one-handed while also holding a scooter — was close to impossible. Worse, the u.FL connector was a real weak point. I added real strain relief at the u.FL connector, which helped for a while, but u.FL connectors are only rated for something like a few dozen mate/unmate cycles before the connection degrades, and every time I opened the case (pulling GPS apart) was wear the unit wasn’t prepared for. Eventually, no strain relief could keep GPS intact, no matter how high the antenna was.
Act III: Move the Whole Board
Iteration three gave up on the mast, with all it’s rubber band complexity, but kept GPS outside the main tube. I moved the entire GPS board into its own GoPro housing. Easy to prototype, and it solved the original problem cleanly: now I only had to get the small housing out of the water for a fix, not hoist the whole scooter. No rubber bands - I just tied on a bolt snap like a proper technical diver and clipped it to the nose of the scooter. Done and done.
It also traded up to a different set of problems. Same tiny antenna, so the fundamental sensitivity issue never went away — I’d just made the housing I had to lift smaller. Two new penetrations and a whole extra housing meant two new places to flood. And the one nobody warns you about: the only fix indicator was an LED on the GPS board itself, inside the housing, which I could not see well enough to read at the surface with a mask on and sun glare on the dome. So the actual workflow became: bring the housing up, can’t read the light, lower the housing toward my face to look at it, which puts the antenna back under more than 0.00 mm of water and drops the fix the instant I need to confirm it. A closed loop with no good exit.
Act IV: Stop Guessing
After a dive where the housing was leaking and I couldn’t read the fix light and I couldn’t hold the thing high enough anyway, I stopped patching this iteration and redesigned the whole chain at once:
- A real active antenna. Not the postage stamp — a proper multi-constellation (GPS/GLONASS/Galileo/BeiDou) active antenna. More sensitivity, less fuss about exact angle and clearance.
- A milled Delrin housing instead of the GoPro shell. Cheaper, and I control the geometry instead of working around a housing designed for a camera.
- A cheap microcontroller inside, dedicated to one job: read the GPS fix quality and drive a single LED. Red for no fix, yellow for a poor fix, green once it’s held a good fix long enough that the main NAV unit has actually logged it.
- Printed clips on the mast and the scooter body, replacing the rubber bands, sized to be worked one-handed in dry gloves.
The antenna and housing fix the signal problem. The LED fixes a completely different problem — not whether the GPS can get a fix, but whether I can tell that it has, without surfacing fully or lowering the antenna to check.
Where It Stands
The fix process for this weekend is: unclip the mast, hold it up, watch the LED. Red, wait. Yellow, wait a bit more. Green, clip it back in, go. No judging antenna angle, no squinting at a dome, no deciding whether the dive computer agrees with the mast.
It’s unflown as a complete system — this write-up is going out ahead of the dive that will actually test it — but the individual pieces (active antenna sensitivity, Delrin housing fabrication, clip-in mounting) are all things I’ve already proven out elsewhere on this project. What’s new is putting them together with a status light that answers the one question that actually mattered: not “is there a fix somewhere in this circuit,” but “can I see from here that there is.”
Four iterations, with this most recent one happening over the course of a single week. Not because any single one was hard, but because I didn’t really understand the problem well enough to produce a working solution. I think at this point I’ve experienced the pain, and this answer should solve all of it.