One drawing per box: what lives inside it, how the pieces wire to each other, and every cable through the wall. Companion to the Electronics and Wiring Plan; entry numbers and gland sizes match Appendix A hole for hole. Every decision in the v10.3 guides is drawn in, the slip ring through the wing’s turning joint among them.
How to read a sheet. The heavy outline is the box wall. Blocks inside are the boards and parts that live in the box; colored lines are the wires between them. Numbered circles on the wall are the cable entries, numbered exactly as in Appendix A of the wiring plan, with the gland size under each and where the cable goes just below. Every entry is a nylon IP68 gland unless it is marked as an RJ45 coupler or an SMA bulkhead, and only cables that leave the cabin get a mix-31 ferrite, on the outside (Electronics and Wiring Plan sec 7.5). Entry positions on these sheets are drawn for readability; the real holes go where the drilling templates put them.
Printing. US letter, landscape. Each sheet prints as its drawing on one page and its notes on the next; the simplest boxes (A5, A13) fit drawing and notes on one page.
| Sheet | Box | Holes |
|---|---|---|
| 0 | The whole network, box to box | |
| A1 | Pi A box (aluminum, main island, lightning stud) | 9 + B (S and 7 share one split gland; 12, S, 7 to the hub box; 5 not drilled) |
| A1a | USB hub box (beside Pi A; the hub, science adapters, dAISy, Iridium A’s USB isolator and wing readers; box not picked) | 6 + one 5-port split gland + Iridium A’s cable (open) |
| A2 | Pi B box (aluminum, backup island, floats; Iridium B’s USB isolator inside) | 8 (7 a split gland, 4 not drilled) |
| A3 | Central node (release, two relay modules, the charging stop, the Ewellix steering drive) | 16 + 2a + V (one more if entry 15 is needed) |
| A4 | BMS box as bought, on the battery box’s aft wall (watchman; monitoring only; one gland per cable, no vent) | 4 |
| A5 | FSESC box: deleted, neither boat | none |
| A6 | Aft junction (plastic, passive, stern; nav lights arrive switched, no photocell; no network parts) | ~14 (measure the box) |
| A6a | Network switch box (aluminum, beside the aft junction, glued to the hull; the TSW202 switch and its supply only, on a DIN rail), with the network surge board near the deck (the five Ethernet surge protectors, not a box) | 11 |
| A7 | VESC A box (main propulsion) | 8 |
| A8 | VESC B box: Boat 2 only, not fitted on Boat 1 | 7 (Boat 2) |
| A9 | Battery box (main bank; one gland per signal cable) | 5 + V + relief stub + fuse, BMS box bolts |
| A10 | Emergency bank: a bought Dyness battery and the backup battery board beside it, no box to build | none |
| A11-A | Iridium A box (main island; its own USB cable to the hub box) | 3 |
| A11-B | Iridium B box (emergency island; its own USB cable to Pi B) | 3 |
| A12 | Detector mini-box (Playing With Fusion AS3935 board, 5 V) | 1 |
| A13 | Starlink enclosure (deck) | 2 + deck stub |
| A14 | Native devices, no drilling; the wing through its slip ring into the new wing box below the deck, where the wing's wires enter the hull | tee + drop; wing box: 5 |
| A15 | Release Node B (second release chain, stern) | 4 |
| A16 | Transom terminal box: DELETED (the release cables run straight to the stern) | none |
| A18 | Wet lab box (plastic, beside the wet lab; a house bus node that switches each wet lab part and reads its sensors) | 14, no vent |
| R1 | Release path map: which node releases what | no box |
What it shows. Two CAN rails, one Ethernet star, and one floating island. Steering and release channels ride the control bus; the busier monitoring traffic rides the house bus. Only the Pis touch both rails. Boat 1 steers with one actuator; there is no FSESC box.
Bus power. Each bus carries 12 V on its power pair, fed once from its own 5 A breaker at the main bus board and its own 24 to 12 V converter at the compute multiport.
Emergency island. Pi B, Iridium B and the lightning detector run on the emergency bank with no copper path to the main system. Every one of their data links crosses on an isolator, a transformer, or plain UTP (unshielded network cable), whose bulkhead jack’s metal touches no wire or shield. The one power bridge is the Orion-Tr, a charger isolated by its own transformer, which central node relay channel 10 runs 90 minutes a day.
Bus order as built. Control: power tee, 4-port multiport (Pi A, Pi B), VESC A, central node. The Electrak tee is Boat 2 only and unwired; Boat 1 steers with the Ewellix CAHB-22, a plain DC actuator driven from the central node, not a bus device. House: power tee, multiport, em-trak, VESC B (Boat 2 only; its tee is unwired on Boat 1), the wet lab box (A18), and Release Node B as the bus end. The aft junction is on no bus: nothing in it talks on CAN. The terminators (the 120 ohm end resistors) sit on the spare port of the end tees; HAT termination jumpers off.
Also shown. Release Node B (Sheet A15) at the stern. The central node’s release cables run straight to the stern, with no terminal box (A16 deleted, 2026-10-27). The USB hub box beside Pi A (A1a), the Orion-Tr on the backup battery board (A10), the wet lab box (A18) on its own house bus tee, and the wing on its slip ring, no radio, into the wing box below the deck (A14). The wing box sends one shielded 4-pair cable to the hub box, where Pi A reads the tip wind meter and the wing compass and talks to the trim actuator on the actuator’s own CAN link, kept off both buses. Each Iridium modem talks over its own USB cable through a USB isolator: Iridium A on a hub port in the USB hub box, Iridium B on a USB port inside the Pi B box (A11-A, A11-B).
What it is. The main brain. Glued to the hull below the waterline for cooling, at least 2 ft from the main battery and motor cables where space allows, for the IMUs (a guess at a safe distance); bond stud wired to the lightning copper. Entries go on any face but the glued back face, with drip loops (drawn on the bottom). Every line from off the island crosses an isolator or an optocoupler before it touches the CM5.
Power. The 10 A breaker at the bar (entry 1), then inside: the TVS, the INA226, the terminal strip, WD-A’s relay and the Mean Well DDR-60L-5 converter (5 V 12 A, 18 to 75 V in, isolated, on a short DIN rail at the far end of the plate from the IMUs; output set to 5.1 V, output negative tied to input negative), then the carrier. The relay is a 5 V 1-channel module with Pi A’s 24 V feed through its normally closed contact, ahead of the converter: coil off means Pi A on, so a dead Nano leaves Pi A running. WD-A cycles Pi A after five minutes of heartbeat silence. WD-A runs on its own Pololu D36V6F5, fed ahead of the relay, so the relay never cuts it. The hub’s 5 V lead takes off the converter’s output on a 3 A inline fuse and leaves by entry 12, so the hub restarts whenever WD-A cycles Pi A.
Entries. Ten, plus the bond stud: 1 to 4, 6 to 8, 11, 12 and S, in nine holes. Not drilled: 5 (Iridium A’s own USB cable goes to the USB hub box through its USB isolator, Sheet A1a, so no Iridium line and no isolator chip for it in this box), 9 and 10 (Pi B and Starlink run to the TSW202), 13 (no wing radio) and 14 (no RTL-SDR). One mix-31 ferrite, on the GPS A coax only. The CAN shield drain lands on the bond stud here and nowhere else. Entry S carries the hub’s USB lead to Pi A’s first USB port, with no USB isolator; entry 7 the wing CAN adapter’s own lead to the second; S and 7 share one Altelix PG25 4-port split gland (on hand; two ports plugged; Mike, 2026-10-14), so both USB plugs pass without cutting; the adapter’s lead goes never a hub port, so the wing’s control never depends on the hub. The kill line (entry 4, beside the VESC UART) and the alive line to Pi B (entry 6) leave WD-A through a HiLetgo PC817 optocoupler board. The Ethernet bulkhead jack’s outside face is greased and taped (Sheet A6a).
Inside. Two IMUs, A1 and A2 (BNO085), on the same I2C pair, A2 at the chip’s second address, one at each end of the plate, far from the converter, each with a forward arrow on the plate. The safe-mode wire runs from a WD-A output pin to header pin 37, the 15k in the wire and the 22k from pin 37 to ground, so the pin sees about 3 V, not 5 V; no box hole. Pi A’s motor kill pin, when assigned, must not use pin 37. No vent; desiccant, humidity breakout; boards conformal-coated. The 4 TB NVMe and the Hailo-8L go in the carrier’s two M.2 slots; bench-prove both. GPS A is a NEO-7M, position only. USB cables clamped within 2 in of their plugs, plug bodies staked. FTDI chips only, never CH340.
USB hub box. Now on its own sheet, A1a: the hub, the science adapters, the dAISy, Iridium A’s USB isolator and the wing’s two readers. Three leads link it to Pi A: the hub’s 5 V lead (entry 12), the hub’s USB lead (entry S) and the CAN adapter’s lead (entry 7).
What it is. A small box of its own beside Pi A (Electronics plan A1a), because Pi A’s box is full. It holds the USB hub and everything that plugs into it, plus the wing’s two readers. It is on no bus: its only links to the computer are two USB leads to Pi A.
Power. Two feeds. The hub runs off Pi A’s 5 V: a short 2-core lead from Pi A’s Mean Well, on a 3 A fuse inside the Pi A box, out by Pi A entry 12 and in by this box’s top PG7 gland, so the hub restarts whenever WD-A cycles Pi A. The box’s own 24 V feed (16/2 on its own 5 A breaker at the branch bar) feeds only the third YRDZXG 24 to 12 V converter, with no fuse on either side of it: the 5 A breaker is the only protection (Mike, 2026-10-14); the 12 V goes out with the SDI-12 line to the BaroVUE 10 and HygroVUE 10, never the 24 V bus.
The hub. StarTech USB210AIND, 10 ports, 5 to 24 V in on a terminal block. On M3 nylon standoffs (the kit on hand), not a DIN rail; zip ties round the hub body and across the plug grips hold every plug in; no screw-lock cables (Mike, 2026-10-19). Its ports: the SDI-12 adapter, the SBE 45’s FTDI USB-RS232-WE-1800-BT-0.0 cable (RS-232, Mike 2026-10-14; the 0 V version, no power out on its wires), the DO sensor’s FTDI cable, the Waveshare wing reader, the dAISy, the SBMS0 lead, the MOTU M2 and Iridium A’s USB isolator; two spare. The DO sensor’s and SBE 45’s FTDI cables run on to the wet lab box (Sheet A18), their bare ends joining the sensors’ cables on terminals. Its own USB lead is detachable: clamp and stake it at both ends.
Leads that keep their plugs (Mike, 2026-10-14) go through one Altelix PG25 5-port split gland in the top wall: the hub’s lead and the CAN adapter’s lead to Pi A, the SBMS0 lead from the BMS box, the AIS antenna coax with its PL-259 on, and the MOTU M2 lead if the MOTU sits outside (else that port takes its plug). The cable lays into a slit seal from the side, so no USB plug is ever cut off. Lines with a bare wire end (the DO sensor, the SBE 45, SDI-12, the wing cable, the 24 V feed) go through ordinary glands in the bottom wall, the wire end fed in from inside.
AIS. The dAISy 2+ has a BNC antenna socket and a mini USB port (Tindie and Starpath listings). The Eifagur stubby VHF antenna’s 16 ft of RG-58 ends in a PL-259; it screws into a Superbat 1 ft SO-239 to BNC jumper, whose BNC end plugs into the dAISy. A flexible jumper, not a solid adapter, so the coax’s weight never levers on the dAISy’s socket. No SMA bulkhead and no SMA to SO-239 jumper. AIS from the dAISy depends on the hub; the em-trak is the second receiver.
Iridium A. The RockBLOCK’s own USB cable (Sheet A11-A) comes in plug on, through a split gland, and plugs into a DSD TECH SH-G01A USB isolator (ADuM3160; 1,500 V power and 2,500 V signal isolation, 400 mA out; on hand), which plugs into a hub port (Mike, 2026-10-15). The isolator is where the lightning path stops: the modem’s ground is tied to the lightning copper through its antenna shield, and only data crosses. Never plug the cable straight into the hub. If the cable is too short, a USB extension goes between the isolator and the hub, never between the cable and the isolator. Drawing: Iridium_to_Pi_Wiring_2026-10-15.
The wing. The wing cable from the wing box (Sheet A14) is any shielded Cat6, all copper, plugs cut off. It lands on a terminal block: blue the trim actuator’s CAN pair, to the DSD TECH SH-C30G (120 ohm switch on); orange the tip meter’s RS-485 and green the compass’s, to the two channels of the Waveshare USB TO 2CH RS485 reader (terminators off); brown the ground, tied to the wing’s 24 V negative in the wing box and here landed on each adapter’s ground terminal. The shield is grounded at the wing box only: here it is trimmed back and sleeved, touching nothing. Both readers are isolated, so a strike on the wing stops at them. The CAN adapter’s lead goes straight to Pi A’s second USB port, never a hub port, so the wing’s control never depends on the hub.
Rules. FTDI chips only for serial, never CH340. Every USB plug clamped within 2 in and staked. No USB isolator on the hub’s own lead (one 5 V supply: it would isolate nothing); the one isolator in this box is Iridium A’s. Not fitted: the DROK 24 to 5 V converter bought for the hub, and a relay to cycle the hub alone. If the hub drops off, Pi A tries a USB reset, then stops its heartbeat so WD-A cycles Pi A and the hub together.
Open: the box itself and each bottom gland’s size; buying the box settles them. Open: which bus bar stud feeds it, the end stud or stud 8. Open: where the MOTU M2 sits; it also settles which split gland takes Iridium A’s cable (port e of the 5-port if the MOTU sits inside, else a second split gland).
What it is. The backup brain, a few feet from Pi A on the settee panel, electrically nowhere near it. Power comes from the emergency bank only: 14/2 from its own 5 A auto-reset breaker on the backup battery board (Sheet A10), in by entry 1 (PG11, 3/4 in hole), then a terminal strip, WD-B’s relay and its own potted Klnuoxj 24 to 5 V converter with TVS, the negative left floating (tied to nothing). That breaker is Pi B’s only protection: no fuse in this box. The relay is a 5 V 1-channel module with Pi B’s 24 V feed through its normally closed contact, ahead of the converter: coil off means Pi B on, so a dead Nano leaves Pi B running. WD-B runs on its own Pololu D36V6F5, fed ahead of the relay, so the relay never cuts it.
Data links. Every data link crosses on something that does not conduct: the CAN HAT’s isolation, the Ethernet jack’s transformer, a DSD TECH SH-G01A USB isolator on Iridium B’s USB cable, and a HiLetgo PC817 optocoupler board on the alive signal from Pi A. Iridium B’s own RockBLOCK cable (Sheet A11-B) comes in plug on by a split gland at entry 7 and plugs into the isolator, which plugs into a Pi B USB port; never the cable straight into Pi B. Open: which split gland for entry 7; none is on the buy list yet. Entry 4 is not drilled.
Lightning detector. WD-B owns the whole lightning role and reads the Playing With Fusion AS3935 board (Sheet A12) at 5 V, straight from its own D36V6F5 rail, so no level shifter: a Belden 6504FE lead under 3 ft, five cores used, in by entry 6 (PG7); the drain lands here only, on the island’s floating negative. Red 5 V, black ground, blue IRQ to D2, green SCL to A5, white SDA to A4. Open: the 5 V surge diode across the lead’s red and black, inside this box; picking a 5 V part settles it. WD-B never cuts Pi B’s power for lightning.
The five isolation rules, drawn. 1) Emergency-bank power only. 2) Every data link optical, transformer coupled or through an isolator. 3) No metal shell or shield ties this box to Pi A: plastic glands, plain UTP (unshielded network cable), so the Alrhso bulkhead jack’s zinc and plastic body touches no wire or shield; its outside face greased and taped (Sheet A6a). 4) No bond stud. 5) Everything else on the emergency bank (Iridium B, detector, WD-B) lives on this side.
What it is. The one team-built node: a Nucleo, screw-terminal boards and a DIN rail where every field wire lands first. It fires the releases, switches the loads, holds the charging stop, drives the Ewellix and reads the sensors. Print the relay map (Compute Bay Plan sec 11, table below) on the lid.
Reading the drawing. Red: relay channel 5, the release enable relay. Amber: channel 8 (charging) and channels 9 and 10. Dashed blue: logic lines. Dashed circle: the release feeder, entry 2a. Entry 29 (top) brings the wet lab’s own 16/2 line in from its 15 A breaker to channel 2; it leaves on one of the six runs aft. Entry 30 (top right) takes the horn’s 16/2 pair from relay channel 13, through its 5 A inline fuse, to the horn low on the instrument post. The charging stop’s 1k resistor sits at the main bus board (dashed box). Not drilled: 20, 21 and 23 to 26 (20 and 25 are spares, drilled only when a cable needs one); entry 15 only if the count needs it. One cable per gland for the two 1-wire chains (Mike, 2026-10-13): entry 18 takes chain 1, the six battery probes’ 20/3 trunk straight from the battery box (Sheet A9 entry 3b, never through the BMS box); entry 17 is drilled for chain 2, the air probes, PG7 or PG9 to suit its cable. Entry 28, the Orion-Tr remote pair, is 14/2 in a PG11 (3/4 in hole).
| Relay ch | What it switches | Contact | Coil driven from |
|---|---|---|---|
| 1 | Starlink power | Normally open | Expander |
| 2 | Science payload group: the wet lab’s own 15 A line, in by entry 29, on to the wet lab box (Sheet A18); on only around each wet lab burst (Claude’s pick) | Normally open | Expander |
| 3 | Nav lights (bow lantern, stern light), sunset to sunrise, and by day too in fog mode | Normally closed: a dead node leaves the boat lit | Expander |
| 4 | em-trak AIS transponder | Normally closed: fails ON | Expander |
| 5 | Release enable relay: feeds the release bus | Normally open: dead, the bus is dead | Native Nucleo pin |
| 6 | Bilge pump 1 manual-on | Normally open | Expander |
| 7 | Motoring whites (coil of the aft junction’s 30 A relay): on only when dark and the motor runs; in fog mode, whenever the motor runs | Normally open | Expander |
| 8 | Charging relay: coil on stops all five chargers | Normally closed: dead, the chargers run | Expander |
| 9 | Camera port UV-C, 5 min on and 10 off, in software | Normally closed: a dead node leaves the UV on | Expander (module 2) |
| 10 | Orion-Tr on/off: 90 minutes a day, and the monthly recharge | Normally open: a dead node leaves it off | Expander (module 2) |
| 11 | Cord puller reverse (test): coils of the four reversing relays | Normally open: dead, the pullers run forward only | Expander (module 2) |
| 12 | Bilge pump 2 manual-on, and its monthly 5-second dry run | Normally open | Expander (module 2) |
| 13 | Horn: its 24 V through a 5 A inline fuse, out by entry 30 to the post; shore command, and one long blast every 2 minutes in fog mode | Normally open: a dead node leaves it silent | Nucleo pin A2 (native) |
| 14 to 16 | Not used as relays. The expander pins of 13 to 16 read the Water Witch states, the release bus readback and the G2 fault | none | Expander inputs |
Steering and sensors. The Ewellix CAHB-22 runs from a Pololu G2 24v13 motor driver (on hand since 2026-09-28), behind a 30 A power-cut relay that is closed when unpowered. Feed in by entry 12, 14/2 on its own 20 A breaker (Electronics plan 7.6); motor and feedback out by 13. Encoder A, wind and the Water Witch states arrive on the aft trunk. Chain 1 carries the battery trip. The opto board’s outputs (Water Witch states, release bus readback) and the G2 fault line go to the expander, on the pins that were relay channels 13 to 16. Not fitted: short relays, branch current meters, a VE.Direct reader, gas sensors.
| Nucleo pin | Job | Nucleo pin | Job |
|---|---|---|---|
| D0, D1 | Wind meter in, out (Calypso CMI1018) | D3 | Relay ch 5 coil (release enable) |
| D2, D10 | CAN board TX, RX | D7, D8 | Release gates ch 1, 2 |
| D4, D5 | I2C data, clock (expander, BME280, humidity) | A3, A6 | Release gates ch 3, 4 |
| D6 | Encoder A | A4, A5 | Release gates ch 5, 6 (after the links come off) |
| D9, D13 | Steering PWM, DIR to the G2 | D11, D12 | 1-wire chain 1 (battery trip), chain 2 (air) |
| A0, A1 | G2 current sense, Ewellix wiper | A2, A7 | A2: relay ch 13 coil, the horn. A7: spare |
Before burn-in, take off the four solder links that join A4 to D4, A5 to D5, and A2 and A7 to the USB serial (check the link numbers in the board manual). The G2 needs its terminal blocks soldered on at the bench too.
The charging stop, beside relay 8. Battery positive comes in by entry 22 on its own thin wire from the main bus board’s positive bar, ahead of the branch breakers, through the 1k 1 W resistor at the bar. Never take it off the node’s own feed: a tripped node breaker would then stop all charging. It runs through channel 8’s normally closed contact and back out by entry 22 on one run to the main bus board, where the yellow wires of the five Victron remote on/off cables (ASS030550320) meet, tied together on a joining strip; all five blacks go to battery negative (Electronics plan 6.7). No fuse: the 1k limits a short anywhere on the run to about 29 mA, and a blown fuse there could stop all charging. Coil on (shore command, 5 days at most unless renewed): all five chargers stop. A dead node, expander or feed drops the relay, and so does the node once neither Pi has been heard for the set time. Measure each run to its MPPT against the cable’s own length; if short, extend with 18/2 tinned wire on a screw terminal strip, no solder.
Channels 9 and 10. Channel 9 (camera port UV-C) leaves by entry 27 to the LM317 regulator and the UV-C lights at the camera port: small 275 nm LED boards above and below the lens, starting at 20 mA with a 62 ohm set resistor (the build: Underwater Camera Port Guide). Channel 10 leaves by entry 28 to the Orion-Tr’s remote on/off.
Release. Channels (Electronics sec 4.3): 1 and 2 the drogue lids (horn A here, horn B on Release Node B); 3 to 6 the four cord pullers; 7 to 9 Boat 2; 10 to 15 retired on Boat 1 (Boat 2 keeps one pull-pin); 16 retired. Relay channel 5 and the channel’s MOSFET must both be on, both from native Nucleo pins. The release bus has its own 12/2 feeder (on hand) on a 30 A breaker at the bar (stud 4), not the node’s 15 A feed, entering by its own entry, 2a in Appendix A3. Open: measure the 12/2 jacket, then size this gland and Release Node B’s (PG11 or PG13.5).
Each puller runs through its own DPDT reversing relay (it swaps the motor’s two wires): 24 V coil, 10 A, each in its own DIN socket with screw terminals, coils on channel 11, no coil diodes needed. The 1N5408 sits on the relay input side (MOSFET drain to the release bus), so it works both ways; the horns need none. Ahead of each relay, one slow-blow 5 x 20 mm fuse in a DIN fuse holder, sized at the bench test; each puller runs 64 s and stops. Release is forward only; reverse needs a shore test command, with ch 5 and the puller’s channel still on. Three cables go straight to the stern, one gland each, with no terminal box (Mike, 2026-10-27): entry 3 the 14/4 fire cable (PG11), both drogue fire pairs, its jacket opened near the transom so each pair runs on to its own potted gland; entries 4 and 4a the two 18/4 cord puller cables (PG7 each), paired by corner, each opened at its quarter to that quarter’s two pullers. Each fire pair’s + is the release bus and its − its MOSFET module; each puller pair leaves its reversing relay. The 18/4 shield drains go to the 24 V negative here only. Colors: Electronics plan A3; drawing Stern_Release_Wiring_2026-10-27.
What it is. The battery watchman, bolted to the outside of the battery box’s aft wall and never inside it. It watches and controls nothing: no wire to the chargers, no relay, no load. The SBMS0 reads all eight group voltages through its own nine tap wires, pack current through the 300 A shunt, solar current through the 100 A shunt, and pack temperature (its XT1 thermistor on pins 15 and 16 of its green plug). It streams them to Pi A through the Electrodacus WiFi/USB add-on board (a micro USB socket, its lead tied down right outside the plug) and the USB hub beside Pi A. That stream is required: the battery percentage comes from it, because the SBMS0’s pack shunt is the only battery current sensor aboard. The JK balancer evens out the eight groups through its own nine tap wires; it talks to nothing and controls nothing. No CAN, no RS-485.
Where. Four 1/4 in stainless bolts through the aft wall on the minus side, opposite the main fuse, at mid-height or above, where the inside of the wall faces only the 1.75 in air gap and the aft G10 plate, never the cells; never on a long side wall. Each hole bored oversize, filled with thickened epoxy and redrilled, each bolt bedded in sealant, fender washers inside, nylon lock nuts or blue Loctite; drill before the pack goes in (Battery Box Building Guide Parts 4 and 8). Four glands and no vent, all facing down, each cable with its own drip loop. It is IP65: spray, not standing water; a drowned BMS box only loses readings.
The SBMS0 side. Its tap wires land on a Winford BRK2X6-L-DIN ribbon breakout board on the DIN rail, fed from the SBMS0’s cell socket by a new 12-pin ribbon about 8 in long; the SBMS0’s own ribbon is not cut. Terminal number = ribbon wire number: W0 on 1 with a jumper 1 to 2 and the 12-core’s shield drain on 2 (the shield grounded at this end only), W1 to W7 on 3 to 9, W8 on 10 with jumpers 10 to 11 and 11 to 12; never more than two wires under one screw. The green plug takes the two shunt pairs (battery shunt on pins 3 and 4, solar shunt on 1 and 2) and XT1 on 15 and 16.
The JK side. Its 11-pin plug carries B0 to B8 and its own power pair, V- and V+ (10 to 40 V). Each harness wire joins a core of its 18 AWG 10-core cable with an adhesive heat-shrink crimp butt splice: at W0 the B0 and V- wires share one end of the splice, at W8 the B8 and V+ wires. Splices staggered and tied to the plate; the board still unplugs at its own plug. Set it up by phone over Bluetooth: LiFePO4, 8 cells.
Entries: one gland per cable (Mike, 2026-10-13: two cables of different size do not seal in one gland, sleeved or not). Entry 1, PG13.5 (13/16 in hole): the JK’s nine taps on its ENERJOUR 18 AWG 10-core (0.39 in across), from battery box entry 3. Entry 1a, PG9 (5/8 in hole): the SBMS0’s nine taps and the XT1 pair on 11 cores of the 22 AWG 12-core shielded cable (0.24 in across), from battery box entry 3a. Dielectric grease on each gland insert. The DS18B20 probe chain runs straight from the battery box to central node entry 18 and never enters this box: entry 6 is deleted and not drilled. Both shunts sit on the positive side of the main bus board, and only their sense leads come in, by entry 3 (PG9): one twisted pair per shunt, each wire on a 30 AWG fuse link at its shunt end, since it sits at battery voltage. Entry 4: the SBMS0’s USB lead, plug on, through an Altelix PG25 4-port split gland, three ports plugged (Mike, 2026-10-14). Powered from the pack per the SBMS0 manual. No humidity breakout (no reader in this box).
Monitoring only. No EXTIO4 wiring, no charge bypass, and no VE.Direct remote cables (they leave from the central node, Sheet A3 entry 22), so entries 2 and 5 are not drilled. The charge limits live in the five MPPTs’ own lithium profiles (27.6 V absorb, 27.0 V float, temperature compensation off). Nothing stops charging in the cold automatically: Pi A alerts shore below 33 F on any pack probe, and shore can send the 5-day charging stop (relay channel 8).
FSESC box: deleted, neither boat. Number kept so later numbers stay put.
What it is. The plastic aft junction box is a patch panel, not a computer: the post’s wind and light cables and the deck sensors land on DIN terminals with one fuse per post feed, about 14 entries (15 if the second trunk is drilled); measure the box before the wires are cut to length. It holds no network parts. Beside it sits the aluminum network switch box (Sheet A6a), which takes every network cable; the Ethernet surge protectors are on their own board near the deck (also Sheet A6a). The entry numbers are kept, so entries 1, 2, 6 to 8 and 20 to 24, which moved to the switch box, are not drilled here.
The 24 V feed (Mike, 2026-10-16). Line 6 from the main bus board runs straight to the network switch box (Sheet A6a, entry 10), so the switch never depends on this box. A short 16/2 jumper, about a foot, brings always-on 24 V from there to this box’s entry 9 for the post feeds, the motoring-white relay and the sensors. A short in a post cable blows only that feed’s own fuse here, and the switch stays up.
Post and cables. The post carries five cables: two cameras, wind, lights and the horn pair (Instrument Post Guide sec 10.1). The wind and light cables come into this box (entries 3 and 4); the two camera Cat6 runs go through the surge board to the switch box (Sheet A6a); the horn pair runs straight from the central node (Sheet A3, entry 30). AIS is below deck. Entry 5 is plugged: no weather pod. The post wind meter’s 5 V regulator (Pololu D36V6F5) sits in this box, just after the wind sensor’s fuse; it gets its terminal block at a workbench, then it is potted.
Switched power. Everything switched arrives from the central node on six 16/2 duplex runs through two PG13.5 glands: Starlink power (relay channel 1), the wet lab’s own line (2; it reaches the central node from its own 15 A breaker, by entry 29 there, and goes on to the wet lab box, Sheet A18), the nav lights (3), the em-trak (4), the bilge pump 1 manual-on (6; pump 2’s, channel 12, rides a 16/2 run or the sensor trunk) and the motoring-white coil (7). The TSW202 is on no channel. The nav lights arrive already switched, on channel 3’s normally closed contact, lit from sunset to sunrise and by day too in fog mode: a dead node leaves the boat lit, and the node lights them by itself if neither Pi is heard for the set time. There is no photocell relay. The motoring whites add on through their own 30 A relay here, its coil on channel 7: on only when it is dark and the motor runs, and in fog mode whenever the motor runs.
Sensors. The two bilge Water Witch states ride the trunk to the central node’s opto board. Pump running is known from the Water Witch 24 V output, so no pump current meter is fitted. There is no high-water sensor. One sensor trunk is likely enough (two Water Witch states, two manual-ons, encoder A, wind); count at fit-out.
Fit-out flag. Lighting leaves this box three ways in the schedule (post lights cable 4, lights group 11, the bow lantern feed 19). Settle the split, and where the stern light’s feed actually rides, when the wires are cut to length; relay channel 3 at the central node switches all of the nav lights either way.
What it is. The network switch box is a CXCESNS aluminum box, the same as the two compute boxes, beside the plastic aft junction box (Sheet A6) at the stern. Every network cable ends here: nine bulkhead jacks (RJ45 couplers; a tenth only if SFP 10 is used), and the 24 V feed in and out, about 11 entries in all. It holds only the TSW202 switch, its Mean Well supply and the SFP module: nothing lightning-related. Keep the box well above bilge water.
Mounting inside. The switch and its supply clip to a DIN rail on a long side wall, not glued to the floor: the TSW202 is 4.8 in (122 mm) deep and the box only 3.9 in. On the side wall the switch reaches about 4.8 in across the 8.3 in floor, plus about 1.2 in of plug, which leaves about 2.3 in to turn the patch cords. Heat goes from the rail to the box wall, the glued face and the hull: about 5 to 15 W in all. Right-angle patch cords at the switch ports (a short right-angle adapter at a port is fine where a straight plug cannot turn), each cord tied down within about 2 in of its plug. Open: which way the right-angle plugs must point; reading it off the switch, with the distance from the rail to the port face, settles it before the cords are ordered.
Wall jacks. Every network cable crossing a box wall comes in on a bulkhead jack, one per cable: the Alrhso panel couplers, zinc and plastic (fine on Pi B’s UTP run: its metal touches no wire or shield). The outside face sits in the hull’s salt air, and the jack’s cap seals only an empty jack. On each outside face: a thin smear of dielectric grease on the plug, plug in, then self-fusing silicone tape stretched to about twice its length, half-overlapped, from the jack’s threads over the plug and 1 to 2 in down the cable, with no gaps, and a drip loop below. Jacks go on side or bottom walls, facing sideways or down, never on the top (the drawing puts them on top only for clarity). Inside faces, within sealed boxes, need no tape. The same goes for the one jack on each of Pi A, Pi B, the Starlink enclosure and the light driver’s dry box: 13 in all.
The switch. The five camera runs, the PoE light driver, Pi A, Pi B and Starlink all meet the Teltonika TSW202 (8 PoE ports plus 2 SFP slots; PoE is power sent down the network cable). Port map: on the drawing. The switch is always on from the house branch, never on a relay: shedding it would blind Starlink and cut the Ethernet between the Pis. Cameras and lights save power by per-port PoE control through the switch’s Web API, read back over SNMP, firmware pinned. PoE budget on the 112 W Mean Well supply: cameras about 60 W set aside, plus the driver at 25 W (802.3at) or 12 W (802.3af); both fit, no supply upgrade.
Surge protectors: on their own board near the deck, in neither box. The five deck and near-deck cables (two post cameras, foredeck fisheye, interior camera, Starlink) each pass a Ubiquiti ETH-SP-G2 (outdoor-rated) on a small board on a bulkhead where they first come inside. Their grounds go to one small tinned bar; one 6 AWG lead runs short and straight from it to the lightning copper, a foot or two at most (Lightning Guide 6.2, 7.4). Open: where the copper runs near that bulkhead, and so which bulkhead takes the board; check at the boat. Short premade Cat6 runs go down to this box’s wall jacks, low in the hull; keep cables before and after the protectors apart.
Network cables are premade, none crimped aboard. Extra coiled near the surge board (small tight strapped coils about 8 in across, the camera side and the switch side apart); none in this box: 6 in to 1 ft patch cords, jack to port. No sharp bends at plugs; check each plug passes its holes and glands, or use a split gland (a gland that opens to close around a cable that already has its plug).
The 24 V feed (Mike, 2026-10-16): line 6 from the main bus board lands here first, on entry 10, on one DIN terminal block from the kits on hand; the switch’s 54 V supply and a short 16/2 jumper on to the aft junction take off from it. The jumper leaves by its own gland, entry 13 (PG9, 5/8 in hole), and enters the aft junction at its entry 9.
Cooling. This box is glued by its back face straight to the hull in the heat-carrying epoxy of Compute Bay 2.3, like the Pi A box; entries go on the other faces, with drip loops. It is wired to nothing and stays off the lightning copper list (Lightning Guide 6.2).
What it is. Main propulsion drive for the ME1616 motor: the Flipsky 75200 Pro V2.0, aluminum PCB, power-button variant, bought 2026-10-03 on Amazon (order 114-2434294-4539414), $160.99, here 2026-10-07; the 2026-09-05 PayPal order shows no payment, and if it turns up too, one goes back. A crimped shorted plug replaces its on/off button, so the board runs whenever battery is present. No precharge (Mike, 2026-09-29): the battery switch makes the first connection. Boat 1 sails with this one VESC; VESC B (Sheet A8) is Boat 2 only. Drawing with every wire: VESC_A_Box_Wiring_2026-10-14.
Plugs. COMM: JST-PH 8-pin, pairs on hand. CAN: JST-GH 5-pin leads, on hand (Kidisoii pre-crimped kit); CANH, CANL, GND only, PWR.ON open. Kill line (Mike, 2026-10-14): on the servo plug’s PPM pin, with the VESC Tool kill switch set to that pin, active low (System Settings Sheet sec 2); low stops the motor whatever CAN or UART are doing. ADC1 and ADC2 carry the motor’s sin and cos, so they are not free for it. The kill line comes from WD-A’s PC817 board (and later Pi A’s kill pin) and shares the UART ground core in the Micro-C run from Pi A. Pin 1 found by meter at the bench check.
Eight holes. CAN; UART + kill; battery + and − (PG16 each, onto the posts, the TVS on the feed); three phase legs (PG16 each, drawn as one run), no ferrite: the phase leads are too short for the 1.0 in clamp (Mike, 2026-10-15); sensor lead. Pi A holds a direct UART line with a hardware kill beside the CAN drop; the kill line is the hardware backstop under the jam logic in the Pi.
Motor sensor. The ME1616 sensor harness is built to the Sep 2 spec (crimp, no solder; SinCos encoder input). A 10k and 18k divider on sin and on cos, unless the bench check shows the tops stay under 3.2 V. The motor’s sensor plug is an Aptiv Metri-Pack 150, 8-way, sealed; its mating plug (an ELITE.CITI pair with terminals and seals for 20 to 22 AWG, $14.81) is ordered. Motor temperature sensor: KTY84/130; check the board has the 10k resistor the firmware assumes.
Capacitors (Mike, 2026-10-14). Four Nichicon 1,000 uF 63 V (on hand) lie in a row on the box floor, stripes all one way. Each leg, left full length, gets a heat-shrink butt splice (22 to 18 AWG, crimped on the bare leg, then shrunk) and an 18 AWG tail, 3 to 4 in, all eight the same length: red on the plus legs, black on the minus (striped) legs. The plus WAGO 221-415 takes the four red tails and a 12 AWG red lead; the minus WAGO the four black tails and a 12 AWG black lead. A WAGO joins everything in it, so plus and minus never share one. Both WAGOs lie flat, levers up; capacitors and WAGOs glued to the box floor with neutral-cure silicone. Each lead, under 6 in with a crimped 3/8 in ring lug, goes to its post. The posts: two insulated MGI SpeedWare 3/8 in posts; each holds three lugs, bottom up: the 2 AWG feed, the controller’s own 8 AWG lead (cut under 6 in), the 12 AWG capacitor lead. Nothing bare but the post lugs; no solder anywhere. Check each capacitor’s stripe faces the minus post before first power-up. Not fitted, do not re-propose: copper strips or any bare bus copper at the capacitors, a G10 capacitor board, a bought RC capacitor pack, stacked WAGOs. Drawing: Cap_Board_Drawing_2026-10-14.
Cooling. The Flipsky sits on an aluminum plate through the box floor, set in the heat-carrying epoxy and wired to nothing. Size the plate before the box is drilled, and check it against Lightning Guide 6.2’s one-connection rule.
Boat 2 only, not fitted on Boat 1. Boat 1 sails with one VESC: losing propulsion costs time, not the trip, because the wing sails the boat. The backup VESC, the phase-transfer contactor, its set and reset coils and position sense are Boat 2 parts; no holes are drilled for this box on Boat 1. On Boat 1 there is one throttle path and the “propulsion_backup” software state is unreachable.
On Boat 2. The backup throttle: a 40 A class VESC on the house bus, feeding the same motor through the phase-transfer contactor, which picks whose phases reach the ME1616. The contactor coil is fed from the main bank (island rule); the set, reset and sense lines pass through this box’s terminals. House bus by design: with VESC A and the Electrak on control and VESC B on house, no single bus failure strands both throttle paths.
Holes. Seven, Boat 2 only.
What it is. Dumb by design: cells and sensors only. Every board that thinks lives next door in the BMS box. No CAN drop, no stop-wire relay, no data isolator at the box. Seven holes (the two mains, three signal glands, the vent and the relief stub), plus the main fuse’s bolts and the BMS box’s four bolts (M). Dashed circles on the drawing are holes that are not drilled.
Aft wall, high. Two PG16 (7/8 in) for the 2 AWG mains, in line with the terminal posts at measured post height so the cables run straight over the aft end plate. Three signal glands, one cable each (Mike, 2026-10-13: two cables of different size do not seal in one gland, sleeved or not): entry 3, PG13.5 (13/16 in hole), the JK’s ENERJOUR 18 AWG 10-core, to BMS box entry 1; entry 3a, PG9 (5/8 in hole), the SBMS0’s 22 AWG 12-core shielded with the XT1 pair, to BMS box entry 1a; entry 3b, PG9 (5/8 in hole), the DS18B20 probe trunk, straight to central node entry 18, never through the BMS box. Each cable with its own drip loop; dielectric grease on each insert. The 12-core’s shield is grounded at the BMS box end only. The Gore vent M12 (1/2 in) as the breathing inlet. Aft wall, outside, mid-height on the + side: the 300 A main fuse (F); bed any bolt through the wall in sealant. Opposite it, on the − side at mid-height or above: the BMS box (M), four 1/4 in stainless bolts, each hole bored oversize, filled with thickened epoxy and redrilled, each bolt bedded in sealant; drill before the pack goes in.
Forward wall, high: the relief stub. A 2 in hole-saw opening with a 1-1/2 in Sch 40 PVC stub glassed through, pointing up at 45 degrees into open cabin space. Inside it, a Dometic/Sealand 1-1/2 in toilet duckbill (OEM 385310076), lips out, flange clamped by a PVC coupling, discharge aimed forward away from the electronics. No screen on the stub: the duckbill is shut except when the box vents, and the cell-temperature trip is the protection. Not fitted: a bug cap or a fine flame mesh. None of these either: burst membrane, overboard duct, transom hole, flame screen.
No heaters. No pads, thermostats, thermal fuses or their wiring. Nothing stops charging in the cold automatically: no stretch of the route has a monthly-mean sea temperature below 35 F, the pack follows cabin air on its 2 in foam, and Pi A alerts shore below 33 F on any pack probe, so shore can send the 5-day charging stop (relay channel 8, Sheet A3).
Heat trip; no gas sensor (do not re-propose). The six DS18B20 pack probes carry the trip: 150 F on any probe, or a 10 F rise in five minutes, isolates the bank, HIBERNATE, alert. Release after one hour with every probe below 120 F. Warning-only alert at 131 F.
The probes and their trunk (Mike, 2026-10-13 and 10-15). Three probes per 16-cell stack: one on each of the stack’s two middle groups and one on an end group, each held against its cell face the way the XT1 thermistor is held, reading the metal, not the air. Each is named by its factory serial number, so its place on the chain does not matter. The trunk is Mike’s 3-conductor 20 AWG cable on hand, all three wires used: red power, black ground, the third data. It runs along the stacks past each probe, the probes about 2 ft apart: at each one its three wires are cut and each joined with one red heat-shrink butt splice (22 to 18 AWG), the trunk wire coming in on one end, the trunk wire going on plus the probe’s wire (folded double) on the other, the three staggered about 1 in and covered by a glue-lined sleeve; at the last probe the trunk ends. No solder and no WAGO. The probes’ own 3 ft leads stay full length, coiled and tied; trunk and joints tied down clear of the terminal rows and the series links.
Taps. Each board has its own nine tap wires, each on its own ring and its own 30 AWG fuse link at the terminal end: 18 rings and 18 links, the two wires of each joint on two different posts of it. Just inside its gland each tap cable splits into one branch per stack, and each branch is held by nylon cable clamps on short wood screws into the long wall, never tied to the cells. Build details: Battery Box Building Guide Part 8.
What it is. The island’s own 24 V: a bought Dyness 24V 100Ah LiFePO4 (25.6 V, 2,560 Wh) in its own plastic case, with a built-in 100 A BMS, bought instead of a team-built 8 × EVE LF50 bank. No box is built and no gland is drilled: the compartment is the enclosure, and every circuit lands on the backup battery board beside it (Mike, 2026-10-15; the layout is Claude’s pick, standing unless Mike says otherwise). No hole schedule and no gland list (Electronics and Wiring Plan A10). Drawing, parts and wire list: Backup_Battery_Board_Layout_2026-10-15.
The board. 16 x 12 in, 1/2 in marine plywood or 1/4 in G10, no bond wire, within a foot of the Dyness, standing up with the chargers’ fins running up and down, somewhere dry (the Orion-Tr is only splash rated); built and wired on the bench and fitted as one piece; bars and breakers under their covers, cables leaving the bottom edge with drip loops. The Orion-Tr and the SmartSolar 75/10 side by side with 2 in of air round each. The 30 A ANL (Renogy, on hand) sits at the corner nearest the + post; its 6 AWG from the post runs in split loom its whole length, as short as the board allows and 72 in at most (ABYC E-11 allows 72 in for a cable in a sheath, 7 in bare); the Dyness’s own BMS also cuts off a dead short. A red and black bus bar pair under covers. The backup minus floats: no bond wire on the board, the bars or any case.
Loads on breakers, chargers on fuses. Pi B and Iridium B each on a 5 A auto-reset stud breaker, as Iridium A is (Pi B worst case about 1.2 A, Iridium B about 0.7 A): a blown fuse would leave either dead for the whole trip. That breaker is Pi B’s only protection; there is no fuse in the Pi B box. The chargers keep fuses, as on the main board, in sealed inline blade holders: the Orion-Tr output 10 A, the 75/10 battery side 15 A (1-1/4 times its 10 A, next size up). The two 50 W panels come in through a sixth NanQue PV surge module (Mike, 2026-10-15; to buy), its ground screw to its own PV minus, never the lightning copper. A spare + and − stud and a spare breaker spot are left. Open: whether Primelink unit B lands on this bank: Build Guide sec 8.3 puts it on the emergency bank, and the Electronics plan lists only Pi B and Iridium B; the spare spot covers it if it does.
Two batteries, kept apart. Only the Orion-Tr touches both. Its input (from main board breaker 13) and its remote wire (central node relay channel 10, out by central node entry 28) keep to a marked strip on the board’s left edge; nothing else on the board touches the main battery. All wire 14/2 (Mike, 2026-10-15), except the two 6 AWG battery leads: the Orion-Tr feed, its remote pair, the deck pair, the Pi B and Iridium B feeds and the board jumpers. 14/2 does not fit a PG7, so Pi B box entry 1 (Sheet A2), the Iridium B box’s power entry (Sheet A11-B) and central node entry 28 (Sheet A3) are PG11 in a 3/4 in hole.
Two recharge paths. The SmartSolar 75/10 charges the bank from its two 50 W panels, even with the main bus or the node dead. The Orion-Tr 24/24-5 isolated charger (output 27.0 V) runs from the main bus 90 minutes a day and after the monthly exercise, switched through its remote on/off terminal by central node relay channel 10 on its normally open contact, so a dead node leaves it off. Its transformer is the only link to the main bank. Meter its case to each minus on the bench before it goes on the board. Restraint is the real work: 42.6 lb at 20.9 in long must hold in a knockdown and a full inversion; chocked shelf, two straps over the top, boots on both M8 (5/16 in) posts. Open with the maker: whether the built-in low-temperature cutoff blocks charging.
What it is. Voice number one. RockBLOCK 9603 on its carrier, fed always on from the main bank through an isolated 24 to 5 V converter (its 5 V side has no wire connection to its 24 V side) and a TVS. Isolated because the stub arrestor ties the coax shield to the lightning copper: the tie stops at the Iridium side, and the main negative is not tied to the copper a second time. The converter (Claude’s pick, 2026-10-09): a Mean Well DDR-15G-5 on a short DIN rail on the mounting plate, 9 to 36 V in, 5 V 3 A out against the RockBLOCK’s 0.5 A at most, 4,000 V between input and output; it stands about 2-1/2 in off the plate. Its idle draw is not published. The box: plastic, with a mounting plate, room on the plate for the converter, the RockBLOCK beside it and a TVS, and about 3 in inside from plate to lid. Open: which box (the TICONN 5.9 x 5.9 x 3.5 in box is not used); Mike’s pick settles it.
The link to Pi A: the RockBLOCK’s own cable (Mike, 2026-10-15). The cable SparkFun ships with the RockBLOCK is an FTDI USB to serial cable (TTL-232R-3V3 with the RockBLOCK’s 10-pin plug); its USB to serial chip is molded into the USB plug, so no other converter is needed. Cut it 8 in from the small plug, inside this box; keep both pieces. Strip each wire 1/2 in, twist it and fold it back on itself. Yellow (pin 1, data out of the modem) to yellow and orange (pin 6, data in) to orange, each in a 2-way lever connector; black (pin 10, ground) to black in a 3-way lever connector with a short black wire to the Mean Well’s −V; the small plug’s red (pin 8, 5 V in) to the Mean Well’s +V. The long piece’s red is cut short and taped, and any brown or green wire at the cut the same. Pin 7 (on/off) has no wire: a floating on/off pin is on. The long piece leaves by entry S; its USB plug goes into the DSD TECH SH-G01A USB isolator (on hand) on a hub port in the USB hub box (Sheet A1a), which it enters plug on, through a split gland. Too short? A USB extension goes between the isolator and the hub, never between the cable and the isolator. Software: a USB serial port at 19,200 baud, flow control off.
Why this way. The modem’s ground is tied to the lightning copper through the antenna shield, so a plain USB link would tie Pi A to the copper; the isolator stops that path, and only data crosses it. Never plug the cable into the hub without it. The modem never takes power over USB: the 9603 needs a supply of at least 500 mA, and the isolator gives 400 mA, so the Mean Well runs the modem and the isolator powers only the chip in the USB plug. The isolated Mean Well stays (Mike, 2026-10-16): the isolator breaks the path to the Pi, the Mean Well the path to the battery; a potted Klnuoxj joins its 5 V minus to its 24 V minus and would tie the main negative to the lightning copper through the antenna stub. Do not re-propose. Not fitted: an ADuM1201 isolator board or a 3.3 V regulator for this line, a Waveshare isolated USB to serial converter, the Ground Control tails cable.
Feed: a 5 A breaker on its own stud at the main bus bar, 16/2 (Mike, 2026-09-28: a breaker comes back after a passing fault; a fuse would leave the voice dead). Coax: the stub arrestor is the only protection (no safing relays; a stub cannot wear out). It lands on the underside of the post’s copper deck terminal, a few inches from the deck stub. The coax comes below through the post’s 1 in deck stub with the GPS and Starlink cables. Antenna: on the aft deck beside the post, under its own clear acrylic dome on a white base plate, rubber gasket, with a 5/64 in (2 mm) weep hole at the low point (Instrument Post Guide 11.1; Electronics 6.5). The dome goes where no solar panel shades it and where it shades no panel.
What it is. Voice number two, the same inside but living entirely on the island: emergency-bank power, its own USB cable to Pi B, its own stub and dome. Its isolated converter keeps the island’s floating negative off the lightning copper, which the stub arrestor ties to the coax shield. The converter, the box and the cut and rejoined cable are the same as Iridium A’s (Sheet A11-A); the same Open line on the box. Nothing is shared with Iridium A, on purpose: a fault that silences one voice cannot reach the other.
The link to Pi B. The cable’s long piece leaves by entry S and enters the Pi B box plug on, by its split gland at entry 7 (Sheet A2). Inside, its USB plug goes into a DSD TECH SH-G01A USB isolator (its status: Bill of Materials), which plugs into a USB port on Pi B; never the cable straight into Pi B. Too short? A USB extension goes between the isolator and Pi B, never between the cable and the isolator.
Feed: 14/2 from its own 5 A auto-reset breaker on the backup battery board (Sheet A10), in by a PG11 (3/4 in hole), sealant-filled. Antenna: forward on the foredeck under its own clear acrylic dome, built like Iridium A’s, with the same 5/64 in (2 mm) weep hole. Its coax comes below through its own small deck hole, so the two chains share no dome, deck hole, cable or fuse (Build Guide 8.3). The dome goes where no solar panel shades it and where it shades no panel. Arrestor lead: 6 AWG wire, about 6 to 7 ft aft to the lightning junction; measure where it lands before cutting the wire.
What it is. One board, one hole. The Playing With Fusion SEN-39003 AS3935 lightning board (Mike, 2026-10-14; bought 2026-10-06) hangs off WD-B in the Pi B box on one short lead that carries its power and its data. It runs at 5 V straight from WD-B’s own D36V6F5 rail, with no level shifter: its datasheet gives 2.4 to 5.5 V for power and signals. It comes tuned at the factory: this board’s value is 88 pF (bag label, register value 11; System Settings Sheet sec 10). I2C address 0x03. It logs strikes; WD-B never cuts Pi B’s power for lightning. The box touches no metal and is glued to a pad glued to the boat. The SparkFun board and the level shifters are not used. Drawing with wire colors: Lightning_Detector_Wiring_2026-10-14.
At the board, on the bench. The one solder job allowed on the boat’s wiring, done on the bench: an 8-way 0.1 in screw terminal block soldered onto the pin row (one of the Hxchen 20-pack on hand), and two short jumper wires soldered on the back between the pads, SI to VDD (picks I2C) and CS to GND, not under the screws. A small dab of West epoxy or high-temp hot glue over them is fine: flux cleaned off first, kept on the pin-row half of the board, off the antenna coil (MA5532) and the small parts beside it, so the factory tuning holds. Then the conformal coat every board gets. The lead’s wires go under the screws on ferrules, one per screw; check the screws tight once more a week or two after wiring. MISO and both Qwiic plugs stay empty (the Qwiic plugs carry no IRQ). Mounted on nylon 4-40 screws, no metal.
The lead. Belden 6504FE (Mike’s roll; 22 AWG, 6 cores, foil shield with a drain wire, about 0.175 in across), under 3 ft, PG7 at both ends. Red VDD to WD-B’s 5 V, black GND to the island’s floating negative, blue IRQ to D2 (Claude’s pick; no WD-B pin map exists yet), green SCL to A5, white SDA to A4. The brown core is cut back flush inside the jacket at both ends. The drain lands at the WD-B end only, on the island’s floating negative; here it is cut back and taped. Open: the 5 V surge diode on this lead, across red and black inside the Pi B box; picking a 5 V part settles it.
Open: bench the false-trigger rate with VESC A running before trusting it here. The fallback is a small forward node with an isolated transceiver.
The broadband dish in its deck enclosure. Ethernet runs to the TSW202 at the aft junction (SFP port 9 through a copper SFP module), a long Cat6 run to the stern; its bulkhead jack’s outside face is greased and taped (Sheet A6a). Power rides relay channel 1 from the central node, so the Pi can shed it or power-cycle it. No enclosure heater: the dish has its own snow-melt heater. Declared sacrificial: no arrestor; losing it costs bandwidth, not the boat.
Deck hole: the instrument post’s 1 in PVC deck stub. The post bundle and the Starlink, GPS and Iridium A cables come below through that one stub and its 1 in liquid-tight flex conduit; no separate standpipes. Only Iridium B keeps its own small deck hole, under its forward dome, so the two satellite chains share no hole. Pass the bundle through a sample stub on the bench before the real one is glassed in; make up the coax connectors after the cables are through.
What it is. Bought devices with their own sealed housings and connectors: no box of ours, no holes of ours. Each takes a tee on its backbone and a cut-cable drop into its own connector. The Ewellix CAHB-22 has no bus: its cable runs to the central node (Sheet A3, entries 12 and 13). The Electrak HD tee is Boat 2 only.
The wing (Wing Power System sec 1A). It reaches the hull through a sealed slip ring in the can (sliding contacts that pass wires across a turning joint), and its wires land in a new small wing box below the deck, where the wing's wires enter the hull. Eight ring wires: the Electrak MD’s 24 V and CAN pairs, the tip meter’s RS-485 pair and the wing compass’s RS-485 pair (on the ring’s two spare power rings). At the tip the 24 V pair feeds a branch on its own 1 A inline fuse: the compass runs on that 24 V (9 to 36 V), and a small converter makes 5 V from it for the tip meter alone (never 24 V on the meter). The compass (WitMotion HWT905, on the tip cap) gives the wing’s turn rate and, against the hull’s heading, its angle; there is no angle sensor in the can. The ring’s three fixed-side cables plug into three hull cables inside the can; these leave by three PG9 glands in the can wall (Wing Can Spec Sheet) and go down through their own deck gland straight into the wing box below the deck (Mike, 2026-10-02), where each ring wire gets a fuse or surge part (three Taidacent modules and the 24 V pair’s 5 A fuse and 1.5KE33CA), so a mast strike stops there. Nothing in the wing box thinks: one shielded 4-pair cable runs on to the USB hub box beside Pi A (Sheet A1a), carrying the actuator’s CAN pair, the tip meter’s and the compass’s RS-485 pairs, each after its surge module, and a ground pair tied to the wing’s 24 V negative in the wing box (the isolated readers’ signal side floats, so it must share a reference with the sensors). The cable is any shielded Cat6, all copper, with a shield over all four pairs, plugs cut off (Mike, 2026-10-09; Wire and Cable Buy List C8): blue the CAN pair (CAN H solid blue, CAN L blue-white), orange the tip meter’s RS-485 (A solid, B white), green the compass’s RS-485 (A solid, B white), brown the ground, both wires together; same colors and labels at both ends. Its shield is grounded at the wing box end only (Mike’s OK, 2026-10-09): the braid (or drain wire) twisted into a short pigtail, sleeved in heat shrink, on the box’s 24 V negative terminal, where the brown pair and the three Taidacent modules’ ground terminals also land (each module’s ground goes to that negative, as the battery bus surge modules’ do, so the boat keeps its one bond wire); any pair foils folded back and trimmed. At the hub box the shield is trimmed back flush and sleeved, touching nothing. The strike comes in at the wing box, so the shield drains there; grounded at the hub end it would carry the surge into the box beside Pi A, and grounded at both ends it would close a loop through the boat’s wiring. There the Waveshare 2-channel RS-485 reader takes the tip meter and the compass, and the DSD TECH SH-C30G isolated USB to CAN adapter talks to the actuator on its own lead to Pi A. RS-485 terminators off. The wing’s program runs on Pi A. The ring is the Rion HC2549, paid $495 (Mike, 2026-10-01).
The wing box’s wall: in, its 24 V feed (16/2 on its own 5 A breaker on the branch bar, then the box’s own 5 A fuse); in, the three hull cables from the slip ring; out, the cable to the hub box. The box (Mike, 2026-10-12): a TICONN IP67 ABS box, 5.9 x 5.9 x 3.5 in, with its mounting plate (Amazon B0B87THLGC), to buy; the three Taidacent modules, the surge diode and the 5 A fuse on a short DIN rail on its plate (that they fit is Claude’s guess; if not, the next size up in the same line). No vent. Open: each entry’s gland; laying out the box and picking the hull cables settles it. Open: which cable each hull cable is; the cable choice settles it. Open: which stud feeds it, the end stud or stud 3; Mike’s call settles it.
What it is. The backup for the central node’s releases. Its own fuse, its own enable relay, its own MOSFET pairs and the same two-condition lock, commanded over the house bus, so no single bus or box failure silences every release. No shorting relays and no opto board: four entries (CAN, 24 V, drogue 1, drogue 2) and five Nucleo pins; entries 5 to 8 are not drilled. It releases the horn-B channel of each drogue box; either node alone cuts a drogue strap.
Same rules as the central node. Unpowered, it cannot release. Two separate conditions: the enable relay closed and the channel MOSFET on. No humidity breakout; no ferrites. No short relays, no opto board, no continuity or confirm readback. The monthly enable relay exercise covers this node.
What makes it independent. Five separate layers; a fault has to cross all five to take both paths to one device. Different bus: the central node is on the control backbone, this node on the house backbone, each with 12 V on its pair, fed once from its own 5 A breaker at the main bus board and its own 24 to 12 V converter at the compute multiport, with its own tees and terminators. Different power: a house-side branch with its own fuse. Different parts inside: separate Nucleo, firmware instance, converter, enable relay, release bus and MOSFET modules. Different holes and routes: its cables leave through their own glands and travel a different path to the same device, so one chafe point cannot take both paths. Different half of the device: each drogue lid carries two channels (horns A and B), and either one alone does the job. Still shared: Pi A, which commands both, and the main bank, which feeds both.
What the two conditions are. Not two brains agreeing, but one brain doing two physically separate acts: close the enable relay, the only thing that puts 24 V on the release bus, and turn on that channel’s MOSFET, the only return path to ground. Idle, the release bus is dead metal and the return is open. An enable timer drops the bus again if a release command never arrives. That defeats a welded contact, a shorted MOSFET, a stray 24 V wire landing on a release terminal, or a wet terminal block. It does not defeat a Nucleo gone haywire, which is why the drogue and JSD paths also need a signed ground command and have no automatic path at all.
The honest cost of the split. Two chains double the odds a lid opens when it must, and double the number of boxes that can open it when it must not. What holds that second risk down is the signed ground command and the enable timer. There is no idle short across the burn wires: the dead release bus and the open MOSFET are the two conditions.
Why it went (Mike, 2026-10-27: “yes, drop the box”). It was planned when the stern took twenty-plus wires: drogue confirms, bridle switches, pull-pin runs and the aux rudder, all deleted or Boat 2 only. For the twelve left, on three cables that would leave it almost as they came in, it only added two screw joints per wire, in a box on the transom, on the one circuit nobody can fix at sea.
What replaces it. Three cables straight from the central node (Sheet A3, entries 3, 4 and 4a). The 14/4 fire cable (0.40 in, PG11 at the central node) runs in one piece; near the transom its jacket opens and each drogue’s fire pair runs on in a sleeve to its own potted transom gland and lid coils, with no joint. The two 18/4 cord puller cables (0.25 in, PG7 each at the central node), paired by corner, each run in one piece to their quarter and open there to that quarter’s two pullers. Wires, colors and the shield rule: Electronics and Wiring Plan A3; drawing Stern_Release_Wiring_2026-10-27.
The 9.5 x 5.5 in box on hand is free for another job. Boat 2’s aux-rudder release, if fitted, runs direct the same way.
What it is. A small node on the house bus that switches each wet lab part one at a time and reads the wet lab’s sensors (Mike, 2026-10-14; the layout is Claude’s pick, standing unless Mike says otherwise). Pi A runs the hourly cycle (Flow-Through Wet Lab Design sec 10) and sends orders over CAN; this box does what it is told and refuses anything unsafe. Plate layout, wiring and tables: Wet_Lab_Box_Layout_2026-10-15; text, parts and hole list: Reference/Wet_Lab_Box_2026-10-14; every part with a photo: Wet_Lab_Box_Parts_2026-10-15.
Where. Beside the wet lab, outside its drip tray and above its rim, glands facing down with a drip loop on every cable; within about 5 ft of the USB hub box if it can be (the FTDI cables are 6 ft). Nothing in it goes to the lightning copper. The box: plastic IP67 with a mounting plate, about 12 x 8 x 5 in: the TICONN IP67 box bought 2026-10-03 if it is that size or bigger, else a TICONN box of that size, to buy. No vent: only the central node has one (Mike, 2026-10-16).
Power. In by entry 2: the wet lab’s own 16/2 line from its 15 A breaker, switched by central node relay channel 2 (Sheet A3) and carried on through the aft junction. Relay channel 2 is on only around each burst (Claude’s pick), so the whole wet lab, this node included, draws nothing between bursts; nothing needs power while idle (valve 2 is shut unpowered, valve 1 stays where it is). The line feeds the XWST converter (24 V steady, 10 A, on hand), with a 1.5KE33CA across its output, and five fuse groups off it, 5 x 20 mm slow-blow in DIN holders (QBGTFAK kit on hand): F1 5 A the Gulper (a guess: read its label), F2 2 A the Iwaki pump and the dosing pump, F3 2 A both valves and their position wires, F4 1 A the DO sensor and the SBE 45, F5 1 A the node, the relay coils and the flow meter. A spare potted Klnuoxj on F5 makes 5 V for the Nucleo and the pressure sensor. A 1N5408 across each pump motor’s two terminals, stripe to +.
The node. A NUCLEO-G431KB (to buy) on an AITRIP Nano screw carrier, staked after burn-in, with an SN65HVD230 CAN board and a Taidacent RS-12V/2S surge part (one of the two bought 2026-10-07 with no other job; its ground to 0 V) on the house bus drop: entry 1, a cut Micro-C cord from its own tee, ahead of Release Node B’s end tee. Before wiring, take off the Nucleo’s four solder links (A4 to D4, A5 to D5, A2 and A7 to the USB serial), as on the central node, and set the relay module’s trigger jumper the same way as the central node’s modules. The node ID, the 60 s silence rule and the relay channel 2 timing: System Settings Sheet.
| Relay ch | Switches | Rule | Nucleo pin |
|---|---|---|---|
| 1 | Iwaki pump (entry 3) | Only while valve 2 reads open | D3 |
| 2 | Kamoer dosing pump (entry 4) | Run time set for 5 mL | D6 |
| 3 | Gulper air pump (entry 5) | Only while valve 2 reads shut | D7 |
| 4 | Valve 1 open line (entry 6) | Never with ch 5 | D8 |
| 5 | Valve 1 close line (entry 6) | Never with ch 4 | D9 |
| 6 | Valve 2 power (entry 7) | Powered = open; off, it springs shut | D11 |
| 7, 8 | Spare | Later sensors |
Inputs. A spare Fafeicy 8-channel opto board (24 V inputs, 3.3 V out to the Nucleo): IN1 the flow meter’s pulses (the meter on 24 V from F5) to D0; IN2 and IN3 valve 1 open and shut to A1 and A2; IN4 and IN5 valve 2 open and shut to A3 and A6 (the position commons on 24 V from F3); IN6 to 8 spare. The pressure sensor runs on the node’s 5 V; its 0.5 to 4.5 V output goes through 10k on top and 18k to 0 V, so A0 sees 2.9 V at most. The Blue Robotics Celsius probe is on I2C (D4 SDA, D5 SCL) at the Nucleo’s 3.3 V, with 4.7k pull-ups to 3.3 V; the resistors sit under the carrier’s screws. CAN on D2 (TX) and D10 (RX). Read each valve’s wire colors off its label before wiring.
What the node refuses on its own. No message from Pi A for 60 s: every relay off, so the pump stops and valve 2 springs shut. The pump runs only while valve 2 reads open and the Gulper only while valve 2 reads shut; valve 1’s open and close lines are never on together. Each valve stroke is timed from its position wires and reported: the wear gauge the Wet Lab Design asks for.
The two science cables. The DO sensor’s and the SBE 45’s cables each carry power and data. Both come in here (entries 11 and 13): power lands on F4’s terminals, and the data wires join the bare ends of their FTDI cables on terminals (RS-485 for the DO sensor, RS-232 for the SBE 45), which leave by entries 12 and 14 for the USB hub box (Sheet A1a). A run over about 5 ft is extended on screw terminals here with leftover 22/12 shielded cable.
Holes. Fourteen, all on the bottom face: PG7 (1/2 in hole) for entries 1, 3, 4 and 8 to 14; PG9 (5/8 in hole) for 2, 5, 6 and 7. Read each cable’s diameter against its gland before drilling (PG7 grips about 1/8 to 1/4 in, PG9 about 5/32 to 5/16 in). To buy: one NUCLEO-G431KB; an N2K single tee for the house bus if the shelf has none spare. Left out, with stand-ins: pump current (no current sensor on the boat; the flow rate stands in) and the dosing reservoir level (counted from the doses given).
What it shows. For every releasable device: which chain can release it, and what has to die before nothing can. Solid lines are the node-A chain on the control bus, on its three straight cables to the stern. Dashed lines are the node-B chain on the house bus, running direct.
The main point. The two drogue lids have two chains; the JSD cord pullers ×4 (24 V worm-gear motors, one per T8 trigger) do not. The pullers are driven only by the central node and reach the stern only on their two 18/4 cables, so the node and its cable are single points in series (paired by corner, a dead cable holds one leg of each drogue). A second path for them would need four more channels and new runs to the stern. They can be run backward for testing from shore through relay channel 11 (Sheet A3).
Channel numbers are owned by the Electronics and Wiring Plan sec 4.3: 1 and 2 the drogue lids (horn A central node, horn B Release Node B); 3 to 6 the four JSD cord pullers (central node); 7 to 9 reserved for the Boat 2 aux rudder (unwired); 10 to 15 retired on Boat 1 (Boat 2 keeps one pull-pin); 16 retired (trim decouple, not on this sheet). Pi A commands both chains, so the Pi and the main bank are the two things this split does not protect against; everything after them is doubled for the two drogue lids.
| Device | Path A | Path B | If one chain is lost |
|---|---|---|---|
| Drogue lid 1 | Central node ch 1, the 14/4, entry 3 (PG11) | Release Node B, direct, entry 3 | Either channel alone cuts the strap |
| Drogue lid 2 | Central node ch 2, the 14/4, entry 3 (PG11) | Release Node B, direct, entry 4 | Either channel alone cuts the strap |
| JSD cord pullers ×4 | Central node ch 3 to 6, the two 18/4s, entries 4 and 4a (PG7 each) | none | No release: node and cable are in series |
| Aux-rudder release, Boat 2 only, unwired (ch 7 to 9 reserved) | Central node ch 7 to 9, its own cable if fitted | none | No release: node and cable in series |
| Pull-pins (ch 10 to 15) retired; bridle switches deleted, no gland | |||
Path A entry numbers are Sheet A3 (central node) entries; Path B entries are Sheet A15 entries. Release Node B has no opto board and no short relays: four holes.
The open questions that touch the wiring live in one place, Pending Amendments v10.4 (the open list). On these sheets they are:
Boxes and glands. The USB hub box itself (to buy) and the gland for each of its entries, and which split gland takes Iridium A’s USB cable (A1a). The split gland for Iridium B’s cable at Pi B entry 7 (A2). The Iridium boxes themselves (A11-A, A11-B). The wing box’s glands, and which cable each of its three hull cables is (A14). Chain 2’s gland at central node entry 17, PG7 or PG9 to suit its cable (A3). The release feeder glands, PG11 or PG13.5, once the 12/2 jacket is measured (A3, A15). The aft junction box size (A6). The wet lab box: the TICONN box on hand if it is about 12 x 8 x 5 in or bigger, else one to buy (A18).
Feeds and parts. Which bus bar stud feeds the hub box and the wing box, the end stud or the stud table’s 8 and 3 (A1a, A14). The 5 V surge diode on the lightning detector lead (A2, A12). Whether Primelink unit B lands on the backup battery board’s spare spot (A10). Where the MOTU M2 sits (A1a).
Routing and fit. Which way the switch’s right-angle patch cords must point (A6a). Which bulkhead takes the network surge board, near where the lightning copper runs (A6a). The aluminum cooling plate through the VESC A box floor (A7). The camera port’s UV-C lights (a factory-made board, or the small boards on trial). Which release trunk carries which cores (A3, A16). The lighting exits at the aft junction (entries 4, 11 and 19 all carry lighting).
The decided items are drawn in on their sheets.