Electronics CAN Sources list

System Settings Sheet v10.5

v10.5, 2026-11-09. Every setting aboard the Nobotic USV (Boat 1, the Cape Dory 25D), with where it is set and why. This sheet is the home for every setting, limit, threshold, timer and address; the other guides name a setting and point here. The idea: one place for every number the boat runs on. Change a number here first, then in the device, then log it. If a guide disagrees, this sheet wins on settings and the guide wins on wiring.

1. Charging (the five main-bank MPPTs)

The solar chargers (MPPTs, the Victron boxes between panels and battery), the emergency bank's charger, and the SBMS0 watchman. The five main chargers always charge. There is no charge stop; do not re-propose one (Mike, 2026-11-01). Their lithium profiles are the only charge ceiling; above the surplus level the Pi spends the surplus on the prop (section 4). The SBMS0 watchman only watches.

SettingValueWhereWhy
Profile typeUser-defined lithiumAll 5x Victron SmartSolar on the main bus board (2x 100/20, 3x 75/15)LiFePO4 8S (eight cells in series)
Absorption voltage27.6 V (3.45 V/cell)sameWith float, the only charge ceiling. No SoC target or heat cap; do not re-propose. Worst case, a 300 mV drift between groups tops one at about 3.7 V: ages cells, no damage
Float voltage27.0 V (3.375 V/cell)same
Temperature compensationOFFsameLiFePO4 wants none
Absorption time30 minsame
RX pinLeft at its factory default, on all fivesameNo remote on/off cables are fitted (no charge stop), so each VE.Direct port is empty
BluetoothDISABLED after first setup, on all fivesameToo much draw, and one less way in. So no per-string data; total solar current comes from the SBMS0's 100 A solar shunt
Pi-silence timerTO BE SET. If the central node (the small control box by the battery that switches loads and triggers releases) hears neither Pi for this long, it turns the nav lights on (channel 3) until a Pi is backCentral node firmwareA dead Pi never leaves the boat dark. Channel map: Compute Bay Plan sec 11
Low-temperature charge stopNone. Not set in the MPPTs; SBMS0 parameter 29 is displayed only-The route's sea never gets that cold: nowhere on Legs 6 to 9 or the Drake approach is the monthly-mean sea below 35 F (Reference/Cold Stop Sea Temperature Check 2026-09-23.md; Design Decisions R-4.1). The cold-pack alert watches the pack instead (section 4). No heaters; do not re-propose them
Emergency bank chargerSmartSolar 75/10, same profile: absorption 27.6 V, float 27.0 VEmergency bankThe Dyness battery's own BMS is the backstop above that
Emergency bank standby level95%, not lowerEmergency bankWhen the main bus is dead it carries the backup island, charged only by its own two 50 W panels (Electronics and Wiring Plan A10), so it starts as full as it can
Emergency bank monthly exerciseDischarge to 25% into the main-bus loads, then recharge to 95% from the main bus: about 50 Wh of conversion loss, logged by the BMS and sent to shore. Open: how the bank discharges into main-bus loads with no power path joining the islands, and how the Dyness BMS reports it. Open: the 25 percent floor sits below the backup converter's cutoff at about 30 percent (section 4); picking one floor settles itEmergency bank; Orion-Tr (next row)Build Guide sec 4.3
Orion-Tr float chargerVictron Orion-Tr 24/24-5, output 27.0 V, float only, switched by central node relay channel 10 through its remote on/off terminal; normally open, so a dead node leaves it off. Runs 90 minutes a day; as long as the monthly exercise recharge needs (about 14 hours, an estimate); longer while Pi B commands with the main side alive (Pi B awake costs about 60 to 85 Wh a day from the emergency bank); and Pi A can run it while bilge pump 2 runs, to keep the emergency bank topped upMain bank to emergency bankElectronics and Wiring Plan A10; Compute Bay Plan sec 11. It idles at about 1 W: switching it cuts that from about 24 Wh a day to about 1.5, plus about 0.5 Wh for the relay coil. Left off by a dead node, the emergency bank charges from its own SmartSolar 75/10 and panels

SBMS0 watchman settings

The SBMS0 (the watchman board that reads every cell group and both shunts) controls nothing: no charger, no relay, no load.

SettingValueWhereWhy
DIP switchNormal: MTX, STX and SRX ON; WRX, WTX and IO0 OFF, as the board ships. The reverse (Program ESP32) is only for loading new firmware into the WiFi moduleThe one DIP switch: six positions on the WiFi/USB add-on board, beside the ribbon headerThe board's printed table (Reference/SBMS0 v03d Manual.pdf page 15, the photos). The main board has no switch or jumper; every other setting is on its touch screen (manual page 6; menus on pages 30 to 39)
Parameter SettingsBattery type 1 (LiFePO4); number of cells 8; battery capacity 1256 Ah. Then Store Parameters, then unplug and replug the 12-pin cell plug so the values take. If the capacity field stops short of 1256, enter its top and record it hereSBMS0 touch screen (manual page 6)The 8S4P pack's rated figure, 4 x 314 Ah. The typical 341 Ah would give 1364 Ah; the rated figure runs the percentage out a little early, the safe side for load shedding
EXTIO4 / EXTIO3Type 1 / Type 2. Nothing is wired to either outputSBMS0 Automation menuNo EXTIO4 charge pause; do not re-propose it
Parameter 301182: the 131 F (55 C) high-temperature charge stop. Displayed onlySBMS0Nothing acts on it (section 4)
Parameter 2932 F (0 C), the low-temperature charge stop. Displayed only. Entered as the SBMS0's own XT1 reading in ice water, not as a rated zeroSBMS03144 in the manual; firmware unconfirmed. The Part 10 bench check measures the XT1 readings (next row). Near freezing the DROK 10K 3950 XT1 probe reads about 4 F low. Where XT1 sits and how it is wired: Battery Box Guide Parts 8 and 10
XT1 readingsTo measure at the Part 10 bench check: the SBMS0 display with XT1 in ice water (about 28 F if the 4 F offset holds), and with XT1 unpluggedSBMS0 displayRecord both here and in Build Guide sec 11. The first sets parameter 29; the second lets the parser tell a broken lead from a cold pack
Battery shunt value0.2500 milliohm: the 300 A / 75 mV FL-15 on the positive side, between the Class T fuse's output and the + 400 A bar. The SBMS0 reads shunts only on the positive side: its current-sense chips (LMP8481) need 4 to 76 V on the shunt (Reference/SBMS0 v03d Manual.pdf pages 3 and 44)SBMS0The manual's rule: 0.075 V divided by the rated amps. Reads both directions
PV shunt value0.7500 milliohm: the 100 A / 75 mV FL-15 on the positive side, between the + solar bar and the + 400 A barSBMS0Same rule. Reads one direction only. Solar is about 50 A at most, half the rating, inside the manual's two-thirds limit
Save device settingsREQUIRED after the aboveSBMS0Or the settings are lost at power cycle
USB serial to Pi ARequired: the battery percentage comes from it (section 4). The lead plugs into the USB hub beside Pi A, whose lead runs straight to Pi A, with no USB isolator (Compute Bay Plan sec 4.1). Serial speed 115,200 (Device settings, UART on; Mike, 2026-11-08), the same in the softwareWiFi/USB add-on boardReadings only. No CAN, no RS-485

2. Propulsion (primary VESC, Flipsky 75200 Pro V2.0)

The motor controller (the VESC, set in VESC Tool): its limits, heat protection and kill switch. CUT-BURST is the short full-power run that cuts a line wrapped on the prop (Build Guide sec 5.3).

SettingValueWhy
Battery voltage max28.0 VAbove absorption. Open: where it sits against the battery bus surge modules' clamping voltage; their label, read on arrival, settles it (section 4)
Battery voltage min (cutoff end)22.0 VPack floor
Battery current cap, normalAbout 85 A (2 kW down to about 23.5 V; at 25.6 V, 2 kW is about 78 A)Mission ceiling
Battery current cap, CUT-BURSTAbout 180 A, only on command, 10 sBuild Guide 5.3
Motor phase current cap, normalAbout 180 A
Motor phase current cap, CUT-BURSTAbout 300 AThe 75/300 rating; stall torque about 55 ft-lb (74 N-m)
Regen termination27.8 VRegen (the turning prop charging the battery) tapers from 27.6 V. Not 28.4 V: that is above battery max
Motor thermistor type, in VESC ToolKTY84/130, the native option. Never the NTC or beta path, and never KTY83/122, the next one in the listThe ME1616's KTY84-130 rises in resistance with heat. Set as NTC it reads about 254 F at room temperature and faults at first power-up; set as KTY83/122 it reads about 104 F when the motor is really at 248 F (Reference/Flipsky and ME1616 Research 2026-09-24.md sec 7.2)
Motor thermistor bench checkMotor at room temperature, sensor unplugged: about 600 ohm across the two sensor wires at 77 F confirms a KTY84-130 (about 1000 ohm is a KTY83, about 1100 a PT1000, about 10,000 an NTC). Plugged in, VESC Tool within about 2 F of the NXP table for the measured ohms passes. More than 5 F off: the board's temperature-pin resistor is not the 10k the firmware assumes; settle that before trusting any limitSame file, sec 7.2. Whether the Flipsky board carries the 10k is not known
Motor temp derate / shutdown (VESC Tool: Motor Temp Cutoff Start / End)212 F / 248 F (enter 100 C / 120 C). Pi A alerts shore at 194 F (90 C)Mike, 2026-09-24. A straight-line derate from full motor current at 212 F to zero at 248 F; at 248 F the output switches off, the prop freewheels, and it restarts by itself once the reading drops. Never above 248 F (120 C): the sensor board inside the motor is rated 221 F (105 C) for its connector and 257 F (125 C) for its other parts, and it sits at the motor's end, away from the winding the thermistor reads, so it runs cooler than that reading (inferred). Not lowered: a 221 F shutoff would cost motor power on hot, hard runs for no gain (Mike, 2026-11-02). The ME1616's winding and magnet limits are not published (Reference/Flipsky and ME1616 Research 2026-09-24.md sec 7)
Motor coolantNothing to set: no coolant loop and no coolant probe. The water jacket is filled three quarters with 50/50 coolant and both ports plugged (Build Guide sec 5.2); the motor temperature limits above are the heat protectionMike, 2026-10-29
Controller transistor temp derate / shutdown167 F / 194 F (enter 75 C / 90 C). VESC Tool calls these MOSFET Temp Cutoff Start and EndThe factory 85 / 100 C suits skateboards that coast and cool; this drive runs nonstop
Acceleration Temperature Decrease0 percent. Factory is 15 percentThe most important row. At 15 percent it quietly drags the drive limits down: thrust drops in heat with no reported reason
Over temperature faultNothing to set: it does not latchThe drive faults, clears itself after 500 ms, and faults again while still hot
Sensor plausibility checkPi A alerts shore if motor temp reads below 32 F (0 C) or controller temp below -4 F (-20 C), at any timeA broken, shorted or wrongly typed thermistor reads cold, and cold never trips anything
Field weakeningOFFIt pushes the motor past its normal top speed; Motenergy's coolant limit drops with it, and the firmware's heat derate does not rein it in as hard as the torque current
Controller transistor temp, loggedSent to Pi A over CAN at every sampleA slow climb at the same power and sea temperature is the early warning of aging capacitors
Motor position sensorSin/cos: sin on ADC1, cos on ADC2 (the COMM plug), both 3.3 V inputs; sensorless is the fallbackThe sensor (an RLS RMB29 board, analogue sinusoidal, RMB29AC) runs on 5 V; both signals are centered at 3/5 of the supply, 3.0 V, and swing 1.1 V plus or minus 0.2 V, so each runs about 1.9 to 4.1 V, worst case about 4.45 V (RLS data sheet RMB29D01_07, page 4, Reference/RLS RMB29 Datasheet RMB29D01 07.pdf). A 10k and 20k divider (two resistors that scale the signal down) is always fitted on sin and on cos in the harness (Mike, 2026-11-06), so each signal runs about 1.3 to 2.7 V, worst case top about 3.0 V, under 3.3 V. No filter capacitors (Mike, 2026-11-06). Sensorless is fine at cruise, weakest at full torque from a standstill
Kill switchVESC Tool kill switch on the servo plug's PPM pin, active low. The opto-isolated kill line from WD-A (Pi A's watchdog) (and Pi A's own kill pin, when assigned) lands there, its ground core on the servo plug's GND beside it. A 10k resistor in the harness from the COMM plug's 3.3V to PPM holds the line high (motor allowed) whenever the PC817 is off, so a loose kill wire cannot stop the motor. WD-A holds the line only while Pi A is hung but still powered; it lets go once Pi A is switched off or has handed over, so Pi B can drive the motor over the control bus (Compute Bay Plan sec 4.1)PPM is the free pin: ADC1 and ADC2 carry the motor's sin and cos. Active low is inferred from the VESC 6.x firmware's kill switch choices. Plan: Box Plans v10.5/VESC A Box v10.5.pdf
Backup VESC (cold standby), BOAT 2 ONLY (Boat 1 sails with one VESC)Motor and battery caps about 12-15 A, about 300 W, soft-startCruise-only limits

3. Steering (Ewellix CAHB-22, single actuator; the Thomson Electrak HD is BOAT 2 ONLY)

The steering actuator's facts, its drive at the central node, and the numbers the boat steers by. Actuator facts are from the Ewellix CAHB series catalogue (Reference/Ewellix CAHB Series Catalogue.pdf, page 24, this exact unit) unless marked.

SettingValue
ActuatorEwellix CAHB-22-F4E-3000529-BAA0PT-000, bought 2026-09-10. 24 V DC, 11.8 in (300 mm) stroke, self-locking screw, built-in position feedback, no CAN. Eyes, pins, washers and the play in the stack: Steering System Handbook secs 1, 2.1, 3 and 4 (no side load on the rod)
ForceRated push and pull 2,250 lbf (10,000 N). Maximum about 3,150 lbf (14,000 N), capped by the clutch, which trips somewhere between the two
Speed0.51 in/s (13.0 mm/s) with no load, 0.40 in/s (10.2 mm/s) at rated force
Duty cycle20 percent: 85 seconds on, 340 seconds off. Plenty for this boat's short, far-apart moves
TemperatureFull performance only from 32 to 104 F (0 to +40 C). It survives -40 to 185 F (-40 to +85 C)
Current8.5 A at rated load and 10.9 A at clutch trip. The SKF CAHB-22 E datasheet (PUB MT/P2 17210 EN, March 2017; its four load versions read 9.5 / 11.0 / 11.0 / 10.0 A) gives 10.0 A rated at 24 V for this 10,000 N version, which confirms the 20 A steering breaker (Electronics and Wiring Plan sec 7.6)
Heat switchBuilt in (the T in the part number), so a stall has its own protection
WiringMotor leads 14 AWG. Position potentiometer (a variable resistor that reports where the rod is): three 22 AWG wires, about 254 ohm per inch of stroke (10 ohm/mm by the catalogue), about 3,000 ohm over the full stroke. Measure it on the bench and scale the node's input to the measured span
Other catalogue factsBacklash about 0.024 in (0.6 mm). IP 69K / 66M. 250 hours salt spray to ISO 9227:2012. Holding force is a graph against stroke, not one number
Rod lifeEwellix publishes none, at any load. The voyage is about 7 miles (11.2 km) of rod travel on the 190,000-move budget, at 11 to 40 lbf (half a percent to two percent of rated), so rod travel sets no limit. Do not quote a rod-travel life; the 3.1 mile (5 km) limit is the Thomson Electrak HD's figure, not this actuator's
DrivePololu G2 High-Power Motor Driver 24v13 (Pololu item 2992) at the central node, ordered 2026-09-27: 6.5 to 40 V in, 13 A with no heat sink, its own current-sense output (about 40 mV per amp), its own current limit set at 30 A from the factory. PWM (fast on-off pulses that set the motor's power) at 20 kHz. No CAN, no J1939. Wiring: Electronics and Wiring Plan sec 5.6
Drive moveSet direction, ramp PWM from 0 to full over about 1/4 s, hold full, set PWM to 0 at the target reading. PWM at 0 brakes the motor and the screw holds; a dead Nucleo leaves PWM at 0. Keep the 20 percent duty limit and a 30 s cap on any one move
Stall stop (starting numbers, Claude's guess, set at the bench)Ignore the first 0.3 s of a move. Stop on: over 9.5 A for 1/2 s; over 18 A for 50 ms; no wiper change for 2 s while driving. After any stop, PWM to 0 and open the power-cut relay; if current still reads over 1 A, the drive has failed on and the relay is the fix. Replace 9.5 A with about three times the no-load current once it is measured. After a stall stop the node retries the move after a pause (set at the bench), so one false trip cannot end steering
Passage-mode deadbandPlus or minus 10 deg (the heading error allowed before a move), as set (Mike, 2026-09-23). A software number: change it if the boat weaves or the battery draw looks high. Checked by one hour of logged compass heading under self-steering once the boat floats
Passage-mode minimum move interval3 min, on the same terms
Passage-mode move styleWhole-error moves, under 300 moves/day. Not fitted: the halved plan of 150 moves/day. Do not re-propose it: no rod-travel limit calls for it
Precision-mode deadband / intervalPlus or minus 3 deg / 20 s. Budgeted about 200 h per voyage
Storm / streamed JSD (Jordan series drogue, the long line of small cloth cones streamed from the stern in a storm)Rod held at center by the self-locking screw, drive unpowered, zero moves (no brake: the screw holds)
Authority scheduleRudder deflection cut above about 8 kt boat speed
Position sourceWhichever is healthy of the actuator's own feedback and encoder A on the stock; either may steer. While both work they cross-check, and disagreement past a threshold set at first setup raises an alert
Software rudder limit25 deg each way (Mike, 2026-09-28), inside the actuator's own ends of stroke at about 27 deg each way with the rod eye at 13 in. The hardware stop blocks at about 28-1/2 deg sit just past those ends: Steering System Handbook sec 3
Soft limitsScaled to the measured wiper span: read the wiper at each end of stroke on the bench and scale the node's input to that span
Rudder rateAbout 1.85 deg/s with the rod eye at 13 in on the tiller arm (rod 0.40 to 0.51 in/s); zero to 25 deg in about 13 s. Enough: at 2 to 3 kt the rudder's force comes from water speed, not swing rate
Jam-detection stall timerSet at the bench; start from the 2 s no-wiper-change stop above. Longer than the Thomson's (Boat 2): a move takes about two and a half times as long here, so the Thomson timer would false-trip on a healthy move
Jam responseEffort commanded but no motion on either feedback: the current-sense trip cuts the drive within seconds, then the power-cut relay (closed when unpowered) opens
ExerciseWeekly full-stroke sweep

4. Battery protection and environment

Battery temperature alerts and trips, charge-level limits and load shedding, the main fuses, surge modules and bilge pump settings, and what the boat does on the emergency bank alone.

SettingValueWhere
Heater thermostatsNone: no heaters; do not re-propose them. No automatic cold charge stop either (section 1)-
Cold-pack alert33 F. Any of the six pack temperature probes below 33 F: the computer in command alerts shore at once, ahead of the daily report, and runs the motor at about the solar output so the battery's net current stays near zero. Discharging cold is allowed; charging cold is the harm (the surplus-spending logic, Build Guide 4.1). Not while the drogue is out or in STORM (motor off). A software number, changeable from shoreThe computer in command; battery software notes (Reference folder) secs 3 and 5
Cell-temperature trip, level150 F (65 C) on any of the six DS18B20 pack probes (Mike, 2026-09-15): HIBERNATE the loads, Iridium alert, and the computer in command runs the motor at about the solar output, as for the cold-pack alert and on the same terms, so the solar does not charge a hot pack. Not the EVE MB31's 140 F discharge limit, which a hot still tropical day could reach (cabin near 104 F); cells do not heat themselves until far above 150 F, so hours of warning remain. Open: nothing cuts the bank off on the trip; settled in the single-point reviewPi + central node; battery software notes (Reference folder) sec 4; Routing Guide 11.7
Cell-temperature trip, rateA rise of more than 10 F in five minutes on any probe: same action. The real detector of a cell faultsame
Cell-temperature trip, releaseWhen every probe has read below 120 F for one hoursame
Cell-temperature warning, alert onlyAny pack probe above 131 F (the level of SBMS0 parameter 30, displayed only): shore alert, no actionPi A
SoC sourceSoC (state of charge, the battery percentage) comes from the SBMS0 watchman's 300 A pack shunt, read over its USB serial link: the only battery current sensor on the boat (its 100 A shunt reads total solar current). If the stream goes stale the Pi sheds on a fixed safe schedule and raises the quiet-watchman alert; stale numbers drive nothingBMS Notes sec 6
SoC load-shed ladderShed at 70 / 50 / 30 / 15 percent; below 5% the relays and node stages open what they switch (most loads have no switch). What each step turns off: the ladder belowPi + central node relays
Surplus level / floor90% / 10%. Above 90% the Pi spends the surplus on the prop: spending, not a charge cap. With no charge stop it is the one way to keep the pack off the top. Shore can lower it for a high cell group or a cold or hot pack; not in STORM or with the drogue outBuild Guide 4.1, 5.3
Main fuse and motor branch300 A Class T (JIAKUPA), bolted to the outside of the battery box's aft wall; 250 A MEGA on the motor branch under itElectronics and Wiring Plan sec 7.6
Battery bus surge modulesTwo modules of one class: the Boartechs 48-70 V class, Amazon B0DFBPBW1R, bought 2026-09-27. Plain, no status contacts. Not the stepped MOV ladder of about 35 / 40 / 45 / 56 V DC: no US seller offers it. Their working voltage sits far above the 27.6 V absorption and the VESC's 28.0 V maximum (section 2); read the label on arrival to confirm. The PV surge modules are plain too, no remote signalMain bus board (Electronics and Wiring Plan sec 7.6)
Bilge pump breakers5 A Type-1 auto-reset. Every small branch that can take a 24 V stud-type auto-reset breaker has one (bought 2026-09-07); no blade-fuse blocks. Pump 2's (line 15) sits on the backup battery board, since pump 2 runs from the emergency bank at all timesBuild Guide 3.6; Electronics and Wiring Plan sec 7.6 and A10; Compute Bay Plan sec 5.2
Bilge pump 2 float switchJohnson Pump Ultima Switch 36303: on at about 2 in of water, off at about 3/4 in, 3-second delay. Fixed in the switch; nothing to setPump 2's trigger height on the aluminum angle, above Water Witch 1
Bilge pump 2 exerciseMonthly, 5 seconds, run by Pi B through pump 2's manual-onPi B (Build Guide 3.6; Compute Bay Plan sec 4.2)
Bilge pump 2 runningWD-B (Pi B's watchdog) watches pump 2's running line and wakes Pi B at once; Pi B alerts shore (a red alert, section 5)WD-B, Pi B

Load-shed ladder

What turns off at each battery level. SoC is read from the watchman's stream (above); the Pi and the central node's relays do the shedding. Open: the step at which the wing's trim and its rate damper shed (the Wing Trim Actuator Installation Guide puts trim late in the ladder); setting it in the software settles it.

SoCLoads turned off, and policy
Above 70%All active
50 to 70%Cameras on schedule only, Starlink less often. Science runs FULL to the 10% floor in the tropics and doldrums, where solar far exceeds the about 845 Wh/day hotel load (what the boat's own systems use); it is trimmed only when SoC nears the floor (a long sunless crossing of the ITCZ, the doldrums band near the equator, or a low-light Southern Ocean leg)
30 to 50%Propulsion 350 W at most, science off, Starlink minimal. The dAISy AIS receiver stays on: it has no switch (Compute Bay Plan sec 11). Open: whether science keeps running here in the tropics and doldrums, as the row above says (full to the 10% floor, though HIBERNATE starts at 15%); Mike's word settles it
15 to 30%Propulsion 200 W at most, compute hourly, Starlink emergency-only
5 to 15%HIBERNATE: Iridium every 2 hr, watchdog only
Below 5%The relays and node MOSFET stages open what they switch; most loads are wired straight to the bus with no switch, and there is no main-bank series disconnect. When the main side dies, the emergency bank carries Pi B and, through the backup converter, steering, the wing, the central node and the drogue (the next table; Compute Bay Plan sec 5.2). Open: no low-voltage cutoff on the main bank; settled in the single-point review

Emergency bank alone: steer or drift

When the main side is dead, the emergency bank (the Dyness backup battery) feeds the backup-fed bar through the backup converter, and the boat steers or drifts, anywhere on the route (Compute Bay Plan sec 5.2). Shore can change these levels by signed command (Build Guide sec 8.3).

SettingValueWhere
Backup converter onOnly when the main side is truly dead, never just because Pi B took command. Pi B switches it by its remote on/off, and can power-cycle the central node by cycling itPi B; the Victron Orion-Tr Smart 24/24-17 isolated
Backup converter cutoffAbout 30 percent left in the emergency bank, set as the converter's own input low-voltage cutoff: below it, drift. Open: the exact voltage; Victron's manual settles itThe converter
Backup converter restartThe converter's own restart voltage: above it, steering starts again by itself. Open: the exact voltage; Victron's manual settles itThe converter
Steer or driftSteer while the converter runs; drift once it has cut off. The converter's cutoff and restart are the switch; no reading is needed. Shore can order steering on a good dayPi B
Steering on the backupNormal passage mode (section 3: plus or minus 10 deg deadband, at least 3 min between moves). The converter is on only for each rudder move and a wind reading now and then, so the node and rudder have power about 15 s at a timePi B
Wing check on the backupOnce an hour: the wing box powered to check and trim the wingPi B
DriftRudder held where it is by its self-locking screw; wing flap set straight once, just before the converter cuts off, since the wing box is fed from the bar that converter feeds (the wing weathervanes, no push); converter off; Pi B shut downPi B
Drift wakeEvery 3 hours, on Pi B's own clock alarm, for about 3 minutes: start, a GPS B fix, one Iridium B position report and mailbox check, shut down (Claude's pick)Pi B
Lights levelNav lights dusk to dawn (the channel 3 rule, section 5) whenever the emergency bank is above its steer levelPi B; central node
AIS levelTO BE SET: the em-trak on, day or night, whenever the emergency bank has power to spare, off when it does not. It keeps the converter and node on, roughly 6 W in all (a guess), about 150 Wh a day, so in practice in the trades, not in winterPi B, from WD-B's voltage reading
Fail-ON channels at power-upEach time the node powers up on the converter it keeps channels 3, 4 and 9 (nav lights, em-trak, UV-C) off unless wantedCentral node; Pi B
Drogue on the backupOne horn at a time, with steering paused for the burn (Compute Bay Plan secs 5.2 and 10.3)The computer in command
Emergency bank voltage readingWD-B reads it on a spare analog pin through two resistors (a divider, from the kit on hand) and passes it to Pi B over their serial line; no shunt, no USB parts. It informs shore and Pi B; the converter acts on its own voltage settings. Open: the pin and the two resistor values; the WD-B pin map settles itWD-B

5. Cameras and network

The network switch, cameras, lights, horn and fog mode, the relay-switched loads (UV-C, cord pullers, wet lab), and when Starlink is on.

SettingValue
Network switchONE switch on the boat: Teltonika TSW202 in the network switch box at the stern (bought 2026-09-07). ALWAYS ON from line 6, on its own 20 A breaker on the branch bar. Never on a switched-load relay channel: shedding it blinds Starlink and cuts the Pi-to-Pi link. Not fitted: the Netgear GS308EP and GS305EP, or a switch in the Pi A box
TSW202 port map1 to 5 the five cameras; 6 the PoE light driver; 7 Pi A; 8 Pi B; SFP 9 Starlink on a 10GTEK SFP-T copper module; SFP 10 spare. Ports 1 to 8 carry power; a copper SFP module does not
TSW202 firmwarePIN IT: fly exactly the version the bench test ran on (Teltonika renamed a setting between 1.01 and 1.05). Version: to be recorded at the bench test
Port controlTeltonika's documented Web API (endpoint /api/ports_settings/config) turns a port off and on. After every command, read the state back over SNMP (the standard RFC 3621 MIB, read only), so command and answer travel different paths. Record the exact working request and the firmware version at the bench test. Ports 1 to 5 off is the zero-camera state. The switch runs TSWOS, not RutOS
PoE switch supplyMean Well DDR-120B-48 set to EXACTLY 54.0 V, trimmer sealed with paint. Cameras plus lights fit with room over
Switch boot stateALL ports ON
Camera IR LEDsOFF in settings
Camera dutyOne watching, others slept, by per-port PoE control from Pi A through the Web API. The foredeck fisheye stays on, for the boat check that turns Starlink on (below)
Interior PoE lightsPort 6, the PoE Texas driver with two disc lights; 25 W or 12 W mode, both fit. They give the interior dome camera color
Nav lights (channel 3)On from the computed sunset to sunrise (from GPS position and the date; a missing fix or bad clock lights them), and in daytime too while fog mode is on. Channel map and wiring: Compute Bay Plan sec 11
Motoring whites (channel 7)On only when it is dark (the channel 3 sunset-to-sunrise rule) AND the motor is running; in fog mode, whenever the motor runs
Fog mode, on by itselfWhen fog is likely: the air within about 2 F of its dew point with humidity about 97 percent or more (the dew point worked out from the HygroVUE 10's temperature and humidity), or the air's dew point above the sea temperature from the wet lab (sea fog). In daytime the forward camera confirms it by checking whether the horizon has faded; not at night, with the IR LEDs off
Horn (channel 13)Sounded by signed shore command, and by fog mode (below). Normally open: a dead node leaves it silent. Its coil is driven from the expander pin PB3, not a native Nucleo pin; channel 12 has no driver. Channel map: Compute Bay Plan sec 11
Fog mode, from shoreShore can force fog mode on or off at any time by signed command. Near shore, the shore watch can also act on National Weather Service dense fog advisories
Fog mode, what it doesChannel 3's side and stern lights come on in daytime too, the motoring whites come on whenever the motor runs, and the horn (channel 13) sounds one long blast of 4 to 6 seconds every 2 minutes. A false alarm only costs a little power. The logic is in the software repo
Foredeck cameraA 360 degree outdoor PoE fisheye: Hanwha XNF-8010RV, bought 2026-09-30 (open box, eBay, $132); wired PoE (802.3af Class 3) from the TSW202. The Reolink FE-P on hand is indoor and stays as the cabin camera, never on deck
Underwater camera UV-C cycleAbout 5 min ON / 10 min OFF, switched in software, not a hardware timer, by channel 9: the first channel of the central node's second 8-channel relay module (driven by the MCP23017 expander). Normally closed contact: a dead node leaves the UV on nonstop. Rest of the channel map: Compute Bay Plan sec 11
Cord puller reverse (test only)Relay channel 11, normally open, drives the four DPDT reversing relays that swap each puller motor's wires. The node refuses "reverse" unless a shore test command is active; a shore test is reverse on, run puller N, reverse off. The release lock still holds: the release enable relay (on its own 30 A module) and the puller's own release channel must both be on. Normal release is forward only (Compute Bay Plan sec 10.3). No puller exercise at sea: the shore test is the only test (Mike, 2026-09-26)
Cord puller run timeEach puller runs 64 s (the 4 s expected pull plus 60 s), then stops. No stall sensing: after the trip the worm gear holds the line and the motor stalls against it for the rest of the 64 s, which may wear out or burn out the motor; Mike accepts that (2026-09-26). Fire guard: a slow-blow fuse per puller, on the plus side between the release bus and its reversing relay (Electronics and Wiring Plan sec 4.1)
Wet lab line (relay channel 2)On only around each wet lab burst: Pi A turns it on about a minute before and off after, so the whole wet lab, its box included, draws nothing between bursts (Claude's pick). Nothing needs power while idle: valve 2 is shut when unpowered and valve 1 stays where it is. When the wet lab box is silent, Pi A still switches the line on schedule and holds it on for the whole solo burst; a bilge alarm cuts it, and it stays off until shore says. Channel map: Compute Bay Plan sec 11; the cycle: Flow-Through Wet Lab Design sec 10
Wet lab box silence ruleAt power-on, no message from Pi A for 60 s: the wet lab box (Electronics and Wiring Plan A18) runs one burst on its own from its saved settings, the solo burst (Flow-Through Wet Lab Design sec 10). In the middle of a burst, no message for 60 s: it turns all its relays off, so the pump stops and valve 2 springs shut. At all times it runs the pump only while valve 2 reads open and the Gulper only while valve 2 reads shut, and never powers valve 1's open and close lines together
Starlink normal cadenceA 5-minute window every hour, adjusted by state of charge (the load-shed ladder, section 4); every 15 minutes while the lightning detector hears strikes (section 10). Shore can override it by signed command (the Starlink cadence override, Build Guide sec 8)
Starlink on: harborOn the whole time from the dock at Brunswick Landing Marina to Buoy STS, and on the tow or run back in at the end. A command takes about 20 to 40 seconds by Iridium and well under a second over Starlink, but only if Starlink is already on: it takes minutes to connect from off (a guess)
Starlink on: near landWithin 20 nm of land
Starlink on: AIS trafficWhenever the AIS receiver shows a vessel whose closest pass is forecast inside about 3 nm within the next 15 minutes. Any vessel sending AIS counts, big ship (Class A) or small boat (Class B); AIS beacons on buoys and shore stations do not (Claude's pick). An AIS rescue beacon (SART or man-overboard) turns it on and alerts shore
Starlink on: camera sees a boatWhenever the Hailo-8L (Pi A's vision chip) spots a boat on the foredeck fisheye, at whatever range it manages; the forward post camera can feed the same check when it runs
Starlink on: red alertsA steering fault, a bilge pump running, a battery temperature alarm or trip, every GPS lost (GPS A, the em-trak's GPS and GPS B), a computer failover or handover
Starlink on: hands-on momentsA tow hooking up or letting go, a drogue release, and the Canaries arrival from 48 hours before Spain's 12-mile line until the tow has the boat
Starlink off againEach case ends by itself when it is over, not when shore clears it, after a 10 minute hold so Starlink does not flick on and off (Claude's pick), then back to the normal cadence. Over means: out of the 20 nm zone; the AIS vessel past; the camera no longer seeing the boat; a red alert's reading back to normal; 30 minutes after a drogue release; a tow, when shore marks it done, with a 4 hour limit in case shore forgets (the boat cannot sense a towline); the Canaries arrival, when shore marks that the tow has the boat
Starlink: not an on-ruleNear charted shipping lanes with no ship showing (AIS traffic and the camera cover those); low battery (Starlink is the first load shed); storms (the lightning rule already runs it every 15 minutes)
Commands over StarlinkSigned commands, the failover reset and hand-back come by Iridium or by Starlink, each checked on the boat the same way. The command list (among them a new route, a keep-clear zone, remote-manual control, deploy drogue, handing command over and back, run pump 2 (never stop it), and the emergency bank levels of section 4): Build Guide sec 8.3
Remote-manual controlHeading, rudder angle, motor power and wing trim by signed command, for the lake trials and close-in work over Starlink. When the link drops, the boat returns to its route. Link-drop time: TO BE SET

6. Alerts (shore)

What the boat tells shore on its own, and the level that sets each alert off. Most point to the section that holds the number.

AlertThreshold
SBMS0 serial stream staleAbout 1 h of silence; escalate after 24 h. Eyes and battery percentage are lost; charging does not depend on the stream
No-charge health checkNo charge current for a full solar day with sane bus voltage. Catches failed MPPTs, PV wiring loss and a wrong profile
Group spread / excursionSpread threshold set at first setup; any group above 3.45 V/cell average
Motor and controller temperature plausibilitySection 2
Bilge duty trendRate of rise in pump duty cycle (how much of the time the pump runs). Pump 1's Water Witch runs its pump a few seconds each time its feed comes on (maker's sheet, Reference/Water Witch 217 230 230-24 Install Sheet.pdf), so the duty-cycle software ignores that run (Build Guide sec 3.6)
WatchdogWD-A cycles Pi A's power on 5 min of heartbeat silence (section 10)
HandoverEach handover, take-back or hand-back between Pi A and Pi B (section 10)
Bilge pump 2 runningWD-B wakes Pi B at once and Pi B alerts shore (section 4)
Steering feedback disagreementSection 3
IMU disagreementIMU A1 and A2 apart past their limit (section 7): Pi A falls back to the GPS track and gyro and alarms shore
Hailo offPi A tells shore when it starts without the Hailo driver after the set fault count (section 10)
Flap given way (wing)Commanded rod position and measured wing angle disagree: the tail flap has swung toward straight against its give-way spring under wind load. After repeated alerts the wing's program, on the hull's Pi, backs off the flap command (Wing Trim Actuator Installation Guide sec 6)
Cell temperature trip, warning and cold packSection 4: trip at 150 F or 10 F in five minutes; warning at 131 F; cold pack below 33 F

7. Navigation (heading from the IMUs; two receivers for position only)

How the boat knows heading, position and wind: the IMU (motion and compass sensor) offsets and calibration, the GPS receivers, and the two wind meters.

SettingValueWhereWhy
IMU heading offsetDegrees, one per BNO085 (three: IMU A1 and A2 in the Pi A box, IMU B in the Pi B box): the angle of each IMU's forward axis (the arrow on its plate) against the bow, measured at setup against a known bearing; zero if mounted square. Heading = IMU yaw corrected by this offset, fused with the trackBoat software config fileNo ZED-F9P moving-base heading pair (Mike, 2026-09-08); the two NEO-7M receivers give position only
IMU roll and pitch level offsetsDegrees, one roll and one pitch per BNO085, beside its heading offset. Boat afloat, still and trimmed level at the dock: read each IMU's roll and pitch and store them as its offsetsBoat software config fileThe Pi A box is glued to the sloping hull, so its plate and IMUs sit tilted. Uncorrected, every heel, knockdown and STORM-entry trigger and the post wind meter's tilt correction read off by that tilt
IMU compass calibrationEach BNO085's own compass calibration, with the boat afloat and its own cables live: a slow full circle under motorEach IMUThe motor, batteries and wiring change the readings from the bench
IMU disagreement limitTO BE SET. When IMU A1 and A2 disagree past it, Pi A falls back to the GPS track and gyro, flags it and sends shore an alarmPi APi B sleeps most of the time, so A2 is A1's only check in normal running; the GPS track is a weak check at 2 kt
GPS receiverhiBCTR NEO-7M, one per island (bought 2026-09-09 as a 4-pack; two are Boat 2 stock). TTL serial at 3.3 to 5 V, SMA antenna. Single-band GPS with GLONASS, good to a few yards-
Pi A's second position sourceThe em-trak B921's own GPS, on the house bus Pi A reads (inferred that it sends its position on NMEA 2000). Pi A turns the em-trak on by releasing relay channel 4 (its normally closed contact) when GPS A fails. GPS A lost and the em-trak with no fix: Pi A hands command to Pi B (section 10). Open: where the em-trak sits and whether its GPS sees the sky therePi A
Wind sensor primacyPost unit (Calypso CMI1018, UART to the central node) primary. Wing-tip unit (CMI1017, RS-485 down through the slip ring to the hull) the cross-check, turned into boat angles by the wing angle from the wing compass (WitMotion HWT905); the tip meter stays (Mike, 2026-09-26). The central node's serial port (USART1, receive only, on pin D0) reads the post meter, so either computer gets it through the node. The meter sends on its own at the data rate set once with Calypso's PC app, so the node never writes to it. The wing reader (one Waveshare USB TO 2CH RS485 on the USB hub) reads the tip meter, so only Pi A has it (Compute Bay Plan sec 10.2; serial settings: section 9)Boat softwareDisagreement flags icing and fouling
Wind meter supply5 V to each Calypso meter. Never 24 V. Post meter: its own Pololu D36V6F5 regulator (50 V max in). Wing-tip meter: the DROK sealed 24 to 5 V converter at the tip (bought 2026-09-26), fed from the slip ring's 24 V pair on its own 1 A inline fuse, feeding only the tip meter. Wing compass (HWT905, RS-485): 24 V, not 5 V (9 to 36 V, under 40 mA; its datasheet v20-0707, sec 3.1), from the tip branch's 24 V on the same 1 A fuse, ahead of the 5 V converterPost meter: in the central node, right after the wind meter's 1 A fuse F1 on the node's 24 V, so the meter is on only while the node's feed is on. Wing-tip meter and compass: at the tipBoth Calypso meters take 3.3 to 18 V and their inputs carry clamping diodes, so 24 V destroys them. 5 V, not 12, because the post meter's serial data may swing to its supply rail (Calypso's manuals do not state the data level). The compass would not run on 5 V
Wind motion correctionTilt correction from the Pi A IMUs' attitude, after their level offsets, on both sensors (a 2D ultrasonic tilted 20 degrees under-reads about 6 percent, always low). The swinging correction needs rotation rate plus one tape-measure offset from the center of motion to the post head: to be measuredBoat software

8. Wing trim

The tail flap's give-way spring, the trim actuator's limits and link, and the wing compass. The give-way spring at the tail flap is the weak point on purpose: a wind-load fuse that resets when the wind eases (Wing Trim Actuator Installation Guide secs 6 and 7). The wing carries no battery, panel, radio or computer. Its trim actuator, tip wind meter and compass are wired to the hull through the slip ring, and the wing's program runs on the hull's Pi (Wing Power System sec 1A). So there are no wing settings for a radio (the LoRa link and its quiet-radio rule), a pod watchdog, a charge controller or hull-link boost, a pack cutoff or power steps.

SettingValueWhereWhy
Give-way spring setting4.0 in-lb at the flap hinge (about 2.3 lb on the pushrod at the horn's 1-3/4 in hole), 5.6 in-lb at 30 degrees of give. Set by the spring being wound 74 degrees between its pins. Nothing to adjust: only a different spring changes itThe give-way spring (McMaster 9287K47, 302 stainless torsion spring) between the horn and the flap, in the flap's front-edge pocketReference/Flap Spring Horn Build 2026-10-31
Test 1, rod lockedRod locked at full extension; pull the flap's back edge toward straight with a hanging scale: it starts to move and reaches near straight. Open: the pass band with the give-way spring at the flap (about 1.45 lbf at the back edge to start, about 2.0 lbf at full give; worked out, inferred); a worked check written into the Wing Trim Actuator Installation Guide settles itBench, before the wing closes and at each refitThe only test: with no centering spring there is no power-off spring-straight test (Mike, 2026-10-30)
Actuator soft limits0.875 in of rod each side of mid-stroke (flap 30 degrees each way) for normal sailing; mid-stroke is flap straight. Full travel (0.985 in each side) swings the flap only to about 34 degrees, inside the 40 it clears, so there are no stop pinsThe wing's program and the Electrak MDWith the 2 in horn
Bias springAbout 1/2 lb at about 1 in of stretch, horn to the rear tail clamp bolt. It takes up the actuator's end play (up to 0.047 in, about 3/4 degree at the flap) one way, so the flap does not chatterAt the tail
Exercise sweepWEEKLY full-stroke sweep in very light wind, not daily; if no calm window comes in a week, a small health check insteadThe wing's program, on the hull's PiThe MD's brushed motor and potentiometer are why it matters
Loop rateA few degrees at a time, a few times a minuteThe wing's program, on the hull's PiWing Trim Actuator Installation Guide sec 1
Damper deadbandTO BE SET: a small band in which the rate damper does not move the flap, so it does not make endless tiny movesThe wing's program, on the hull's PiWear lands on every pin, hinge and the PEX (Reference/Boom Tail Actuator Review 2026-10-30.md)
Hold still when heeledNo flap moves above about 45 degrees of heel and for a minute after (a guess)The wing's program, on the hull's PiThe MD is sealed IP67 and IP69K standing still, IP66 while moving (distributor listings of Thomson's data)
Actuator protocolJ1939 CAN at 250 kbps (section 9) to the Thomson Electrak MD (MD24A025-0050CNO): control PGN 61184, status PGN 126720 every 100 ms. Its own short CAN link through the slip ring, kept off the boat's two buses, with a 120 ohm terminator at each end. No switched feed: the MD sleeps under 2 mA. On CAN silence it stops and holds. Pi A: a DSD TECH SH-C30G isolated USB to CAN adapter on a hub port in the USB hub box; the adapter's own 120 ohm switch, set on, is the hull end's terminator. On bus-off (the CAN port shutting itself down after errors) Pi A restarts the port by itself (SocketCAN restart-ms about 100 ms), and alarms shore when the actuator's 100 ms status message stops (Compute Bay Plan secs 4.1 and 10.2). Pi B, when in command: its own SH-C30G in the Pi B box, a third drop on the same link; the two Pis never both command it (Compute Bay Plan sec 4.2). Open: where Pi B's adapter joins the link and its terminator setting; settled at fit-outThe Electrak MD; the link runs down through the slip ring to the wing box below the deck, where the wing's wires enter the hull (Electronics and Wiring Plan A17)The link stays off the hull buses (Mike, 2026-09-26)
Wing compass calibrationOnce, on the hard or in calm water: WitMotion's magnetic calibration over the compass's RS-485 pair, turning the wing slowly by hand through two full turns. Then the angle table: turn the wing to the marks every 30 degrees on the mast plate and can lip, and at each save the compass heading minus the hull heading: 12 numbers, the wing angle table in the hull software. Compass address as it comes; its speed set to 115,200 baud once at its set-up (section 9)The wing's program, on the hull's PiReference/Wing HC2549 and Compass Install 2026-10-01.md
Cold protection on the wingNone: no heater pads, thermostats or snap disc anywhere on the wing; do not re-propose them. The wing has no packs to protect-Mike, 2026-09-17 and 2026-09-26

9. Buses, serial ports and addresses

Bus speeds, serial port settings, device addresses, and the fixed names the Pis give each USB device.

SettingValueWhereWhy
CAN bit rate250 kbps on both buses; VESC A set to 250 kbpsBoth hull busesElectronics and Wiring Plan sec 5
Wing CAN link bit rate250 kbps, the J1939 norm, set on both DSD TECH SH-C30G adapters' CAN ports (Pi A's in the USB hub box, Pi B's in the Pi B box)The wing's own CAN link, Pi A or Pi B to the trim actuator (section 8)Reference, Wing RS-485 Readers and Timing 2026-10-07
CAN HAT terminator jumpersOFF on both Pi HATs (the CAN boards on the Pis). The bus ends are the molded terminator plugs on the end tees, two per busPi A and Pi BCompute Bay Plan sec 4.1
VESC controller IDBOAT 2 ONLY: clear of any J1939 address, since Boat 2's Electrak adds J1939 to the control bus. Both boats share one software buildVESC ToolElectronics and Wiring Plan sec 5.5
Node numberingNode addresses, message IDs, release channels, relay channels, probe serials, I2C addresses and opto inputs: record here at first setup, and print on each node's lid. Three nodes: the central node, Release Node B and the wet lab boxCentral node, Release Node B, the wet lab box (A18, on the house bus)Electronics and Wiring Plan sec 4.3 and A18
DO sensorThe dissolved-oxygen sensor: Modbus RTU, address 1. Later Modbus sensors join at 2, 3 and onIts FTDI USB-RS485-WE-1800-BT cable, on a hub port in the USB hub boxElectronics and Wiring Plan A1a
SBE 45RS-232, on its own serial line, never the DO sensor's: it answers any character, so each Modbus poll would set it talkingIts FTDI USB-RS232-WE-1800-BT-0.0 cable (the 0 V version: no power out on its wires; the SBE 45 takes its power from the boat) on a hub port in the USB hub boxElectronics and Wiring Plan A1a
Iridium A and BEach modem: a USB serial port at 19,200 baud, flow control off from the factory, through the FTDI chip in its RockBLOCK cable's USB plug and a USB isolator. Its on/off pin (pin 7) has no wire: floating, it reads as onIridium A on Pi A's second USB port, through its isolator inside the Pi A box; Iridium B on a Pi B USB portCompute Bay Plan secs 4.1 and 4.2; plans Box Plans v10.5/Iridium A Box v10.5.pdf and Box Plans v10.5/Iridium B Box v10.5.pdf
Pack probes (DS18B20, 1-wire chain 1)Six, three per 16-cell stack: one on each of the stack's two middle groups (where heat has the farthest to travel out) and one on an end group, each held against its cell face like the XT1 thermistor, reading the metal, not the air. Each is named by its factory serial number, so its place on the chain does not matter. At setup, before mounting, read each number one at a time: the spare Nano Every on a PC (Arduino IDE, the OneWire library's DS18x20_Temperature example, data on D10) through a DS18B20 screw-terminal adapter board with its 4.7 k resistor built in (HUABAN 2-pack kit). Tag each lead; write number to cell in the table below and on the node lid. Wire colors differ by seller: go by each seller's listingCentral node pin D11 (Compute Bay Plan sec 10.2)The trip itself does not care which probe is which (section 4). Battery Box Building Guide Part 8
BaroVUE 10 and HygroVUE 10 (SDI-12)Both ship at address 0. Set the HygroVUE to address 1 (send 0A1! to it alone) before joining them; the BaroVUE stays at 0The SDI-12 USB adapter in the USB hub boxElectronics and Wiring Plan A1a
USB serial namingA udev rule (a Linux naming rule) keyed to each FTDI adapter's serial number keeps each sensor's name across reboots: the DO sensor, the SBE 45, Iridium A, and the wing reader's two channels (tip meter and compass) each get a fixed name (Iridium B the same on Pi B). Record the names here at first setup. FTDI silicon only, never CH340 or CH341Pi AElectronics and Wiring Plan A1a (its rules)
FTDI USB latency timer1 ms on every FTDI chip (Linux sets 16 ms unless told), set by a udev rulePi AThe 16 ms default would add up to 16 ms to each wing reading
Wing compass (HWT905) serialModbus RTU, address as it comes (0x50). 115,200 baud, set once at its set-up (it ships at 9600). Pi A polls registers 0x39 to 0x3F (turn rate and heading) 20 times a secondThe wing reader's compass channel, in the USB hub boxAt 9600 baud a reading would be about 45 ms old when the command goes out; at 115,200 about 5 ms. Detail: Reference, Wing RS-485 Readers and Timing 2026-10-07
Tip wind meter (CMI1017) serialNMEA 0183, streamed once a second. Its baud is not published: the wing program tries the standard rates at start-up (4,800, 9,600, 19,200, 38,400, 57,600, 115,200) and keeps the one that gives sentences with good checksumsThe wing reader's tip meter channelSame file
Wing line terminatorsOff: the wing reader's 120 ohm jumpers off. At 115,200 baud the run is far too short for echoesThe wing readerCompute Bay Plan sec 10.2
Wing reading age limit100 ms. Every wing reading is time-stamped; the rate damper skips any older reading and leaves the flap where trim puts it for that cycle. The wing program runs as its own high-priority process. With Pi B in command there are no wing readings (no tip meter or wing compass), so the damper is offThe wing's program, on the computer in commandThe damper's end-to-end limit (Wing Trim Actuator Installation Guide, the rate damper)

Pack probe serial numbers (record at setup)

Fill in at setup from the pack probes row above; the same numbers go on the central node's lid.

ProbePlaceSerial number
1Stack 1, first middle groupTo record
2Stack 1, second middle groupTo record
3Stack 1, end groupTo record
4Stack 2, first middle groupTo record
5Stack 2, second middle groupTo record
6Stack 2, end groupTo record

10. Lightning watch and watchdogs

The lightning check, the watchdogs that restart a hung computer or node, Pi A's safe mode, and the handover between Pi A and Pi B. The boat does not shut down for lightning; there is no go-dark mode (Compute Bay Plan sec 13). WD-B never cuts Pi B's power for a strike. The lightning numbers are settable by signed ground command; strike strength is not used.

SettingValueWhere
Satellite lightning checkEvery Starlink window: the last 2 minutes of strikes within about 100 miles. Every 15 minutes while the detector hears strikesPi A; Compute Bay Plan sec 13.4
WD-A heartbeat timeout5 minutes of silence from Pi A, then WD-A hard-cycles Pi A's 24 V feed through its relayWD-A (Compute Bay Plan secs 4.1 and 4.3)
Pi B wake scheduleAbout 5 minutes an hour awake to copy state over Ethernet, on a clock alarm Pi B sets before each shutdown. WD-B's power cycle is the fallback if Pi B does not come back; WD-B also wakes it at once when pump 2 runs or the main-island-alive line dropsPi B; WD-B (Compute Bay Plan secs 4.2, 4.3 and 5.2)
Pi B start-up settingsPOWER_OFF_ON_HALT=1 and WAKE_ON_GPIO=0 in Pi B's EEPROM (its start-up chip), so a shut-down Pi B draws about 0.01 W, not 1.2 to 1.6 W (Raspberry Pi's figure for the Pi 5; inferred for the CM5). WD-B's power cycle still wakes itPi B (Compute Bay Plan secs 4.2 and 5.3)
Central node hardware watchdogOn: the STM32's own watchdog restarts a hung node. Timeout TO BE SET in the node firmwareCentral node (Compute Bay Plan sec 10.2)
Lightning detectorPlaying With Fusion SEN-39003 (AS3935), run at 5 V. Tuning value 88 pF (register value 11), set at the factory and read off this board's bag label. I2C address 0x03. Strike signal (IRQ) on WD-B pin D2 (Claude's pick; no WD-B pin map exists yet)WD-B (Compute Bay Plan sec 13.1)

Pi A: safe mode, the Hailo switch and the clock

What Pi A does when it keeps crashing or its Hailo vision chip faults, and how fast its processor runs.

SettingValueWhere
Safe mode restart count3. On WD-A's third restart of Pi A within about an hour, WD-A first sets the safe-mode wire and Pi A starts from its safe copy. In safe mode Pi A runs steering, CAN, the WD-A heartbeat, its IMUs, GPS A, Iridium A and the wing's trim. Only a signed shore command ends safe mode (Ken, 2026-10-07)WD-A and Pi A (Compute Bay Plan secs 4.1 and 4.3)
Hailo fault count2, a guess. At two Hailo faults (the driver fails to load, or Pi A is reset while the chip is busy) Pi A starts without loading the Hailo driver and tells shore; a signed shore command turns it back on. The count is kept on Pi A's eMMCPi A (Compute Bay Plan sec 4.1)
CM5 clock capAbout 2.0 GHz, set in config.txt, below the CM5's 2.4 GHz top, so it runs cooler and lasts longerPi A (Compute Bay Plan sec 4.1)

Handover between Pi A and Pi B

One computer commands at a time (Compute Bay Plan sec 4.3).

SettingValueWhere
Handover triggersPi A hands command to Pi B when it loses GPS A and the em-trak's GPS has no fix, Iridium A (three missed windows), the USB hub, both IMUs (A1 and A2), or the central node or the backup-fed bar. Pi A asks WD-A to drop the main-island-alive line, and hands over only after Pi B confirms it is upPi A; WD-A
One commanderBefore taking command Pi B checks over the control bus or Ethernet that Pi A is really silent; Pi A stops sending control messages the moment Pi B claims commandBoth Pis
Take-backIf the computer in command goes silent, the other takes command at once. A watchdog check runs on Pi B while it commandsBoth Pis
Automatic hand-backWhen Pi B took over only because Pi A lost a part, and Pi A reports that part working again for 1 hour. At most once a day; after that shore decides. Never after Pi A crashed or hung: then only shore hands backPi A; Pi B
Hand-back by shoreAny time, by signed command over Iridium or StarlinkPi A; Pi B
Failover resetSigned command over Iridium or StarlinkBoth Pis
Backup converter while Pi B commandsOff while the main side is alive (section 4)Pi B
Pi B self-checkOnce a week while the boat runs on Pi A: Pi B checks its own GPS B, IMU B and wing link, in softwarePi B

Sources: Reference/Backup Review 2026-11-01.md, Pending Amendments v9.8, v9.9, v10.1 (the 2026-09-26 and 2026-09-27 decisions) and v10.2, all in Old, Reference/Ewellix CAHB Series Catalogue.pdf, Reference/Cold Stop Sea Temperature Check 2026-09-23.md, Reference/Wing Slip Ring Options 2026-09-24.md, Reference/Flipsky and ME1616 Research 2026-09-24.md, Reference/Wing RS485 Readers and Timing 2026-10-07.md, Reference/SBMS0 v03d Manual.pdf, Reference/AS3935 PWF SEN-39003 Wiring 2026-10-14.md, Reference/Wet Lab Box 2026-10-14.md, the plans Box Plans v10.5/Iridium A Box v10.5.pdf, Box Plans v10.5/Iridium B Box v10.5.pdf and Box Plans v10.5/VESC A Box v10.5.pdf, Pending Amendments v10.6 Part 1A, Build Guide secs 3.4, 3.6, 4.1, 4.2, 4.3, 4.5, 5.1, 5.3, 8.4, Electronics and Wiring Plan secs 4 to 7 and Appendix A, Compute Bay Plan secs 4, 10, 11 and 13, Serial and Network Summary v9.5 (port map, firmware pin).