v9.8, 2026-09-23, updated 2026-09-24. Carries every decision made up to 2026-09-24. Open items are in Pending Amendments v9.9. Nobotic USV, Cape Dory 25D. For the software team. What the battery hardware means for code. Hardware detail is in the Build Guide, the Battery Box Building Guide and the Electronics and Wiring Plan. Terms used here.
| Term | Plain meaning |
|---|---|
| BMS | Battery management system. The electronics watching the cells |
| MPPT | The solar charge controller. It takes panel power and charges the bank |
| SoC | State of charge. How full the bank is |
| Group | One of the 8 series blocks of cells in the pack. What used to be called a cell |
| Shunt | A precision resistor in the main negative that lets current be measured |
| Charging relay | Relay channel 8 in the central node. When its coil is powered, all five solar chargers stop |
| Deadman | A timer that cancels a command by itself unless the command is renewed |
The REC Active BMS is gone. Two boards replace it, both in a small IP65 box next to the battery box, neither inside it:
Charging is on by default in hardware. The five solar chargers run whenever nothing stops them, and their own lithium profiles are the only automatic charge ceiling. The central node (the small board by the battery that switches loads and fires releases) can stop all five with its charging relay, on a shore command only, for 5 days at most (section 2). No CAN and no RS-485 from the pack. Telemetry needs the Electrodacus WiFi/USB add-on board (~$38 CAD, ordered from Electrodacus by PayPal 2026-09-02), whose USB-C port streams pack data as text. Reading that stream is required: the battery percentage for load shedding comes from it (section 6), and it is the only per-group data. A silent stream has no effect on charging.
Charging is on by default. Battery positive runs through a 1k 1 W resistor, with no fuse, and on to the yellow signal wire of all five Victron VE.Direct remote on/off cables (ASS030550320, all five on hand, one per MPPT). All five black wires go to battery negative. Above about 6 V on the yellow wire the chargers run; below 1 V or floating they stop. The resistor limits a short to about 29 mA, and a fuse there could blow and stop all charging. The resistor sits at the main bus board's positive bar, with its own thin wire to the central node; the cables and their joining strip sit beside relay 8 in the node box. ASS030550320 is the only VE.Direct remote on/off cable Victron makes, and it is NON-inverting (Amazon B073PDW9S9 is the same cable). A normally closed contact on the charging relay sits in that path. The relay is channel 8 of the central node's 8-channel module, driven through the node's MCP23017 expander (an add-on chip that gives the node extra on/off outputs). Power the coil: all five chargers stop. Drop it: charging resumes. Every failure leaves the boat charging. If the I2C bus dies, the expander's outputs float, the pull-down resistors read that as off, and the relay drops out. A dead node, a dead expander, a reset or a blown fuse all end the same way. So does a dead Pi: if the node hears nothing from either Pi for a set time while it holds a charging stop, it drops channel 8 by itself. The time is a software number (System Settings Sheet v9.8 sec 1). Only something alive and deliberate can stop charging. Shore commands it, through the Pi, since nobody is aboard. The same relay is the service disconnect at the dock. Deadman: 5 days. A stop command expires by itself after 5 days unless renewed, and charging resumes. The Pi carries the timer. After a Pi reset the timer must read expired, never held. The 32.2 kWh bank against roughly 800 Wh/day of hotel load is about forty days of reserve, so 5 days with no charging is well inside it. "Charging commanded off" and the time left go in the daily report. The five MPPT profiles are the only automatic charge ceiling: absorption 27.6 V, float 27.0 V, absorption time 30 min, temperature compensation OFF, RX pin = Remote on/off (System Settings Sheet v9.8).
| State | What the chargers do |
|---|---|
| Relay coil off (normal) | Charging on the MPPT profiles |
| Shore command "stop charging" | Relay coil on: all five stop, for up to 5 days unless renewed |
| Deadman runs out with no renewal | Relay drops out: charging resumes |
| Node, expander or I2C bus dead; a reset; a blown fuse | Relay drops out: charging on |
| Neither Pi heard for the set time while a stop is held | The node drops the relay: charging resumes |
| SBMS0 unplugged or dead | No effect on charging. The serial stream goes silent (section 4) |
Not fitted: the watchman's charge pause on EXTIO4, and the node's charge-continue bypass. Do not re-propose. A dead watchman stopped all charging, and the bypass's 10-minute silent-stream rule could never have been built: the SBMS0 stream lands on Pi A's USB, not on the node.
No CAN bus stack for the BMS. Remove REC CAN parsing, DVCC assumptions, and any BMS heartbeat-over-CAN logic. No load control from the BMS. The Pi decides shedding by SoC, and the central node's relay board does it. The Oceanic 3478 module is discontinued and gone; its loads moved to the relay board. The Oceanic held state on its own; the relay board goes dead if the node dies, so nothing that keeps the boat alive runs through it. No charge decisions. The Pi only passes on a shore stop (section 2) and never stops charging on its own. It can never enable charging, and the BMS still switches no loads. The MPPT VE.Direct ports carry the remote on/off cables and nothing else, and MPPT Bluetooth is OFF (too much watt-hour draw). So there is no per-string reader board, no per-string yield data, and no USB hub for a reader. No heaters (deleted, do not re-propose): no pads, thermostat modules, thermal fuses or heater wiring. Nothing for software to see or control. No automatic cold charge stop. Charging a LiFePO4 cell below 32 F plates lithium and damages it for good, but the route cannot produce that condition. No stretch of Legs 6 to 9 or the Drake approach has a monthly-mean sea temperature below 35 F; the coldest is about 41 to 43 F (5 to 6 C) near 56 degrees S (Reference/Cold_Stop_Sea_Temperature_Check_2026-09-23.md). SBMS0 parameter 29, the 32 F (0 C) low-temperature charge stop, is a displayed value only. The pack follows cabin air, not the sea, and cabin air can dip toward freezing in a cold snap (Leg 6 in September is the likeliest place). So the pack temperature is logged, and Pi A alerts shore when any of the six pack probes reads below 33 F (section 5). Shore can then send the 5-day stop.
Cell-temperature trip, the one automatic battery trip (Mike, 2026-09-15). Any of the six DS18B20 pack probes above 150 F (65 C), or a rise of more than 10 F in five minutes on any probe: isolate the bank, go to HIBERNATE and raise an Iridium alert (Routing Guide v9.8 11.7). Release when every probe has read below 120 F for one hour. Not lower: the EVE MB31 cell sheet allows discharge to 140 F, which a hot still tropical day could reach (cabin near 104 F), and cells do not heat themselves until far above 150 F. The rate rule is the real detector of a cell fault. Numbers: System Settings Sheet v9.8 sec 4. Log the trip and the release, and put the hottest probe in the daily report. No gas alarm and no switching freeze. Both combustible-gas sensors were deleted 2026-09-07 (do not re-propose): a heated gas sensor in a damp sealed cabin false-alarms far more often than this pack will ever vent, and a freeze on a drifted sensor would park the boat for the rest of the voyage. No gas input on the node, no gas alert, no freeze state; do not add one. Read the SBMS0 serial stream, forward it, and watch that it keeps coming. The USB serial lead is the one data path, and the data is required (section 1), so the liveness check comes first. Physical: USB-C on the WiFi/USB add-on board (the Wi-Fi half is never configured), through the DSD TECH USB isolator at Pi A, showing up as an ordinary serial device. Protocol: plain text lines at a fixed interval, carrying 8 group voltages, pack current, SoC, temperature(s), and status flags. The format is in the SBMS0 manual and in the community "missing manual" at github.com/af3556/sbms0. An afternoon of parser work, not a week. Forward as telemetry:
Shore rate of once per few minutes is plenty. Log locally at every reading. Per-group data is the early warning. The common cell failure is one group drifting away from its seven siblings over weeks, and a venting cell shows first as a group-voltage spike. Nothing aboard acts on either: shore watches the trend and can send the 5-day stop, and the 27.6 V absorption ceiling is the backstop. Liveness: "SBMS0 serial stream stale" is a shore alert, not a log line. It means the boat has lost trend data, venting detection, the group-drift warning and the battery percentage. Alert at about an hour of silence, escalating after a day. Loss of BMS telemetry must never cause the Pi to change any load or charge state, except the fixed careful shedding schedule of section 6. Alert and carry on.
If a sense (tap) wire breaks, the SBMS0 may read a nonsense group voltage that poisons the telemetry. At the Gate 1 bench test, open one sense lead on purpose to learn what it looks like, and teach the parser to flag it.
| Part | Source | Price |
|---|---|---|
| Electrodacus SBMS0 (base board; received 2026-08-27) | electrodacus.com, direct | ~$155 |
| Electrodacus WiFi/USB add-on board for SBMS0 (ordered 2026-09-02) | electrodacus.com, direct | ~$38 CAD |
| JKBMS JK-B2A8S, 2 A active balancer | Amazon | ~$64 |
| DSD TECH SH-G01A USB isolator (on hand) | Amazon | ~$20 |
VE.Direct remote on/off cables ASS030550320: five on hand; buy none. All five are fitted, one per MPPT, in parallel on the charging relay's normally closed contact, beside relay 8 in the central node box, fed through one 1k 1 W resistor at the main bus board. Sources: Pending Amendments v9.8 (charge control, the cold-pack alert, the dead-Pi release), Reference/Cold_Stop_Sea_Temperature_Check_2026-09-23.md, the v9.6 Change List 2026-09-15 in Archive (temperature trip, XT1 calibration), Electronics and Wiring Plan v9.8 secs 6.7 and 9, Battery Box Building Guide v9.8 Parts 8 and 10, Reliability Review Five Lenses v1.0 (battery findings 2, 3, 11).