SouthPing · the temperature probe
A waterproof DS18B20 on a three-wire lead has to get from the sea — on a LAND probe, from a pocket of air or snow — into a sealed hull, and end in a connector the carrier board accepts. Three wires in, three pins out, and one of the hull register's thirteen holes (thirteen holes and one seam, hull spec §7) in between.
Companion to buoy-parts-book.html and battery-pack-wiring.html. Pin order from the carrier brief §5, which keeps the rev. 3 order spec's keying decision (decision 9) and supersedes that spec's cross-plug matrix; the gland from the hull spec §7 row 14a.
The probe's lead reaches the dry cavity through a PG7 cable gland in the sensor-pocket roof. The bore is unconditional in sphere200g_p123, so every hull will carry it — none has been printed yet (HULL-21) — and the penetration exists whether or not the gland is fitted. The gland is chosen and not fitted, and the joint has never been pressure-tested.
That is one of three blocking rows on the hull (HULL-09, HULL-10 and HULL-20), and it is the reason this page ends on a gate rather than on a build. Everything below is buildable on the bench today; none of it goes in the water until HULL-09 closes.
A DS18B20 can run on two wires — parasitic power, where the sensor steals charge from the data line between transactions and stores it in an internal capacitor. It is a real mode and it halves the conductors through a gland, which is exactly the constraint here.
This design uses three anyway, on its own 3V3 feed. Parasitic mode needs a strong pull-up switched onto the bus during temperature conversion, is fussier about cable length and capacitance, and turns a sealed, unrecoverable probe's only sea-temperature reading into a mode with more ways to go marginal. The third wire buys margin on the one measurement that cannot be repaired at sea, at the cost of one more conductor through a gland already rated for the cable.
J8 is the only fully-populated 3-position port on the board — the thermistor and leak ports both leave their centre pin vacant, which is what makes a wrong plug harmless there. J8 cannot do that, because it needs all three. So it is keyed a different way: by pin order.
With DATA and 3V3 on the outer pins, an NTC harness — which is loaded on its outer positions — plugged into J8 lands across DATA and 3V3, and sees 3.3 V at both ends (DATA through R7 and pulled up by R3, 3V3 through R6): zero static current [C], nothing damaged. A leak harness in the same position puts ~0 V [C] across the hull pads: no electrolysis, no standing drain. Reorder those three pins for tidiness and both of those become live faults.
Every other cross-plug on the board is either impossible or harmless. This one was not: a 2-position solar housing pushed into J8 lands both contacts, and would put up to ≈ 8.5 V open-circuit in the cold [C] on IO4 (rated 3.6 V [D]) or straight onto the 3V3 rail. Nothing about pin order fixes that, because the port has no dead pin to absorb an offset.
The carrier brief closed it on 2026-10-03 by how the solar cable is built: its charger end is soldered, its one XH-2 plug mates at the combiner, and a plug pulled off the combiner carries no panel voltage — so the loose live housing the hazard needs does not exist. Keep it that way: no connector at the charger end. As built, though, every panel lead ends in an XH plug into a socket on the combiner's pad pair, on all three combiner boards, where the combiner spec §4 has the leads soldered; a panel plug that parted would be live. Whether to accept those plugs or solder the leads is open as FW-29, and this page does not decide it.
Step 2 is the one that is expensive to get wrong: crimp before threading and the housing will not pass the gland, and the only way back is to cut the crimps off.
The probe arrives with three stripped wires — red, yellow and black — and the near-universal convention for these is red = VDD, yellow = DATA, black = GND. Cheap waterproof DS18B20s are known to ship with the mapping shuffled. On 2026-10-15 board 1's probe was rung out by diode test — red probe on black, black probe on yellow 587 mV, on red 675 mV, and open both ways with the probes swapped [M] — which puts black on GND; a temporary lead (yellow → 1, black → 2, red → 3) then read a ROM ID and 26.94 °C on J8 [M]. Red and yellow are the convention, and the search agrees. Board 2's probe answered first time on the same order, but its wire test was not recorded; the other two probes are not rung out (HULL-28).
Reversing red and black destroys the sensor, and there is no spare per probe once all three deploy. Ring out each unrung probe with a meter against the sensor's own datasheet pinout before its first power-up, and write what you find on the harness. Everything in the drawing above assumes the convention holds for them.
| Wire | Net | J8 pin | What it does |
|---|---|---|---|
| yellow | OW_DATA_J | 1 | The whole bus, after R7 (100 Ω). 1-Wire carries clock and data on one line; the sensor never drives it high, only pulls it low against R3 |
| black | GND | 2 | Return. On the centre pin, which is the position an offset-inserted 2-pin housing is most likely to land on |
| red | OW_3V3 | 3 | Supply, from the always-on rail through R6 (100 Ω) — the reason there are three wires rather than two |
R3, the 4.7 kΩ pull-up, and R6 and R7, the 100 Ω resistors the port leaves through since rev. 4 (ADR 0017, decision 12), live on the carrier and not in the harness — so the harness is three plain wires and nothing else, and a spare can be made anywhere. The 100 Ω is what lets a flooded probe cost a thermometer, not the station; it saves the rail, not the pack. A flooded cable is a 33 mA load [C] on the always-on rail that nothing can switch off, about three weeks of a full pack [C] (FW-21).
Part of SouthPing — southping.org. This page is licensed CC BY 4.0: reuse it, adapt it, publish it, with credit to SouthPing and a link back. Source code in the project is Apache-2.0. Manufacturers' datasheets and product photographs remain theirs and are not covered by this grant.