Battery Box Building Guide v10.5
v10.5, 2026-11-09. Nobotic USV, Cape Dory 25D The main battery's home: a glassed box of 3/4 in okoume plywood holding the 32-cell main bank (EVE MB31 cells, 8S4P: 8 groups of 4 cells in a chain, about 32 kWh). It is glued for good onto a raised 2 in foam shelf in the cabin floor trough (the sunken channel in the cabin floor), forward of the aft bilge sump opening, in the roughly 55 in run between the sump and the companionway bulkhead line; the sump and both bilge pumps stay reachable aft of it. The box is watertight and gas-tight, and a crash-bulkhead cell (Build Guide sec 3.9). The main fuse bolts to the outside of its aft wall, and the watchman (the small box of boards that watches the battery) to the outside of a long side wall. Trade words are explained where they first appear; the full word list is the Glossary. The idea: about 500 lb of cells (about 510 lb installed, with busbars, battery boards, clamping and box) stays put even upside down, nothing touches a terminal thread, and nothing at the battery can cut the power or stop the charging. Nobody is aboard to fix it: build it once, build it right, and glue it down like it grew there.
Rules
- Treat the old brown bedding putty as asbestos (it is from before 1982). For any sanding, grinding or vacuuming in the trough: the P100 rubber half-mask, not a paper dust mask; wet the dust down; the HEPA vacuum. The old putty gets no second chance.
- Gloves for epoxy. Some people become allergic after skin contact, and once that happens it is permanent.
- Pour foam expands hard and sticks to everything, skin and hair included: gloves, old clothes, drop cloths.
- Taped tools and no rings or watches once busbars go live (busbars: the flat metal links between cell terminals). A connected pack can push enormous current into a dropped wrench. Towel over finished busbars while working near them; connect the main cables last.
- Small foam batches, always. One big batch overflows everywhere and cannot be undone.
- Every glue joint lands on a dull, ground, vacuumed, wiped surface. Shiny surfaces peel.
- Fillet first, then cloth. A fillet is a rounded bead of thick epoxy in a corner; cloth over a sharp corner leaves a hidden air gap that becomes a crack.
- Wet the cloth until it is dark and clear everywhere. White patches are dry, and dry patches are weak.
- When in doubt, the bigger fillet, and one more strip of cloth, not one less.
- No bare wood anywhere, ever. Every surface and cut edge gets epoxy.
- The rig's load never lands on this box, its lid or the foam shelf. The old compression post in this run comes out; its load goes around the box, down the three pylon bulkheads (supported from the keel area with okoume) to the trough structure and hull. No pylon box (Rig Base Two Plate Design sec 5).
- Nothing ever touches the terminal threads. The foam-faced battens land on cell cases; the clay check proves it.
- Never mount or tie anything on or across the cells; they swell and shuffle slightly.
- NO-OX-ID on metal joints, silicone dielectric grease on rubber. Never the two swapped.
- Every opening in the box is one-way or sealed: glands and gasket squeezed, the duckbill's lips pointing out into the cabin, the relief stub check passed before the lid goes on for good.
- Tap wires in order, both sets checked with the voltmeter, before either board is plugged in.
Parts List
| Part | Amount and status |
| The box and shelf | |
| 3/4 in okoume marine plywood, BS1088 | 2 sheets, 4 x 8 ft |
| West System epoxy, 105 resin, 206 hardener, pumps | 1.2 gallon kit |
| Cabosil thickening powder, for the peanut butter glue (thickened epoxy) | 1 quart tub |
| 1708 fiberglass cloth (a heavy stitched glass cloth), 50 in wide | About 10 yards |
| Pour foam, two-part closed-cell, TotalBoat or US Composites (grades: Part 2) | Boat 1 done; Boat 2 about 2 to 3 cu ft |
| Limber tube (the drain tube under the foam): 1 in PVC pipe, plus tape to cap its ends | 6 ft |
| Wood screws, #8 x 1-5/8 stainless | 1 box |
| The pack squeeze | |
| G10 end plates, 3/8 in (glass-epoxy sheet), each spanning both stacks: forward 13.5 x 8.5 in, aft 13.5 x 8.0 in | 2 |
| Threaded rods, 3/8-16 18-8 stainless, full pack length, top only | 3; six on hand, three for Boat 2 |
| Rod sleeves, 1/2 in PEX tubing | 6 pieces |
| Belleville washers for the pack springs (cone-shaped washers that act as stiff springs), 3/8 in stainless, 20-count packs | 2 packs on hand; about 30 used |
| Spacer blocks, forward and aft, 13.5 in wide x 2.5 in tall (Part 7 steps 3 and 9) | 1 each |
| Mica heat barrier sheet, 5/64 in (2 mm), or 1/16 in (1.5 mm) if the length is tight; not aerogel; cut to cell face size | 32 pieces |
| Side padding, 3/8 in closed-cell neoprene, for both long walls and the middle strip (not Minicel LF200) | On hand, the 3/8 in sheet |
| Floor foam, 1/8 in dense closed-cell sheet; no fairing (smoothing filler), no rigid plate | About 46.5 x 13.625 in |
| Lid batten foam: medium closed-cell neoprene strip, 3/4 in thick x 1-1/2 in wide, adhesive-backed, 15 ft roll (). Medium is ASTM D1056 2A2, 5 to 9 psi to squash it a quarter. Fallback: Pres-On P8175UL, 3/4 x 1 in x 25 ft. Not soft or polyethylene foam. Three strips take about 11.6 ft | Bought 2026-09-25: a CloudBuyer weather strip, 1 x 3/4 in (the fallback's width), 13 ft, $17.96. Check on one batten that it is medium closed-cell before gluing all three; buy the 1-1/2 in roll only if it fails. The 54 x 16 x 1/2 in Ramede sheet is shop stock |
| Lid and sealing | |
| E-Z Lok threaded inserts (brass sleeves, threaded inside; the knife-thread kind cuts its own way into wood), 1/4-20, 0.500 in long | 24, bought |
| Machine screws, 1/4-20 stainless pan head, 1-1/2 in, with fender washers | 25 |
| Lid gasket: WOD closed-cell sponge rubber tape, 3/8 in thick x 3/4 in wide, 30 ft roll | Bought, $19.21; the rim takes about 12 ft. The Qiveno and MXCELL neoprene sheets on hand are padding, not gasket |
| Cable glands (fittings that seal a cable through a wall), IP68: 2 PG16, 1 PG13.5, 2 PG9 (which cable goes where: Part 5) | 5, from the glands on hand |
| Vent plug: IP68 breather screw (the M12 Gore vent), for daily pressure equalizing | 1, bought |
| The relief stub | |
| 1-1/2 in Sch 40 PVC pipe, a short length | 1, the 1.5 in PVC received |
| Duckbill valve (a rubber one-way valve shaped like a bird's beak), the relief valve: Dometic/Sealand 1-1/2 in toilet duckbill, OEM 385310076 | 1; four on hand |
| PVC coupling, 1-1/2 in; clamps the duckbill flange | 1 |
| The watchman and the wiring | |
| Watchman board: Electrodacus SBMS0, with its cell harness and thermistor | 1; direct from electrodacus.com, about $155 |
| Its two shunts (current-measuring bars), FL-15, 75 mV: 300 A for the pack, 100 A for the solar | 2 |
| WiFi/USB add-on board, Electrodacus "WiFi/USB for SBMS0", with its USB serial lead. Required: without it the base board has no data output | 1; ordered by PayPal, about $38 CAD direct |
| Balancer board: JKBMS JK-B2A8S 2 A active balancer, with its balance harness | 1; Amazon, about $64 |
| Ribbon breakout board: Winford BRK2X6-L-DIN, 12-pin ribbon to screw terminals, for a DIN rail | 1; ordered, $16.05 |
| 12-pin female-to-female ribbon, about 8 in, SBMS0 to breakout board | 1; to buy, |
| Watchman box: small IP65 plastic box with mounting plate (QILIPSU class) and a short DIN rail (a standard clip-on metal rail); no vent, no humidity sensor | 1 |
| Its split gland for the USB lead (it passes the plug): Altelix 4-port PG25 pre-terminated, three ports plugged; hole size from the gland's own thread at fit-out | One of a 5-pack bought (Amazon B0GYT3ZMDL), $35.90 the pack |
| Tap cables: the 22 AWG 12-core shielded cable (SBMS0); the ENERJOUR 18 AWG 10-conductor, UL 2464, tinned, 10 ft (balancer). Wire and Cable Buy List C1 and C7 | On hand |
| Shunt sense wires: a 4-core 20 AWG cable from the shunts to a six-block DIN strip in the watchman box, then cores of a stranded Cat6 patch cord (plugs cut off) to the SBMS0 | In place (2026-11-08) |
| Balancer butt splices, adhesive heat-shrink crimp, sized to the wires (the W0 and W8 ones take two wires in one end, so maybe the next size up) | 9 |
| DS18B20 pack temperature probes (Bill of Materials) | 6 |
| Probe trunk: Mike's 3-conductor 20 AWG cable | On hand (not on the order tracker) |
| Probe splices: red adhesive heat-shrink crimp butt splices, 22 to 16 AWG, plus glue-lined heat-shrink sleeves | 18 + 6 sleeves; the 130-pack bought |
| Probe setup board: DS18B20 screw-terminal adapter board, 4.7 k resistor built in | 1 kit: HUABAN 2-pack, $9.99, bought (Amazon 113-8992596-4977806) |
| Cell terminal Belleville washers: M6 x 12 mm 304 stainless conical spring washers (Part 7 step 13) | 64 needed. Two PATIKIL 50-packs (B0D1VLNR66), both here |
| Terminal nuts: M6 serrated flange nuts | 64 needed; 100 in all: 50 on hand plus a QXSKSLH 304 50-pack bought, $5.69 (Mike, 2026-10-19) |
| Flexible busbar link: braided copper jumper with M6 holes (Part 7 step 13) | 1: one of the copper braid links 2-pack bought (Amazon 114-2434294-4539414, $14.99; Mike, 2026-10-09) |
| Main fuse: 300 A Class T with block and cover, the JIAKUPA kit. Not a Blue Sea 5119 | 1; bought, $22 |
| Small parts and supplies | |
| Thermal (heat-conducting) tape, for taping probes flat to cells | |
| Wire flags: numbered label tape for the 18 tap wires (W0 to W8 on each board's cable), both ends | |
| Modeling clay, for the lid clearance check | 1 lb, bought |
| Harness dressing: glue-lined heat shrink; zip ties with adhesive mounts; split loom; nylon cable clamps on short wood screws. No support strip | Zip ties bought |
| Silicone dielectric grease (non-conducting): rubber and plugs only. Never on a bolted metal joint | Bought |
| NO-OX-ID A-Special: Sanchem's conductive anti-corrosion paste, the one the lightning joints use. Metal joints only: every cell terminal, busbar, tap ring and main cable lug. Never on rubber | |
| Blue Loctite 242 | On hand, two bottles |
| Removable adhesive (removable silicone), for plug bodies | |
| White paint pens, for witness stripes (a paint line across a nut and its stud that shows if it turns) | Bought |
| Click torque wrench, 1/4 in drive, 20 to 200 in-lb | On hand |
| Supplies: gloves, mixing cups, stir sticks, chip brushes, rollers, 80 grit paper, acetone or denatured alcohol, voltmeter, low-pressure manometer (a clear vinyl tube taped to a yardstick) | |
| Not fitted, do not re-propose (the lists: Parts 4, 5, 8 and 9): rim doublers, a lid opening, heaters, gas sensors, an overboard vent, the charge-stop cables and their parts | 0 |
Why It Is Built This Way
The box is the pack plus plates, padding and growth room.
| Length | Width | Height |
| The pack: two stacks side by side, mica barriers included, the 3/8 in middle strip not included (Part 7 step 2) | 46.5 in | 13.625 in | 8.75 in to the top of the terminal threads |
| The box inside | 50.0 in | 14.5 in | 9.25 in deep |
- Length: 46.5 in of cells and barriers, plus 0.75 in for the two 3/8 in G10 end plates, plus 1.0 in forward (the spacer block, 7/8 in wood and 1/8 in EPDM), plus 1.75 in aft (the aft spacer block, cut to the measured gap; the harness ends and main cables sit above it) = 50.0 in. Boat 2's box is 3 in longer.
- Width: the extra is foam padding down each side. Height: the extra keeps the lid off the terminal threads in rough seas.
- The shelf is 2 in, as low as the trough allows. Every inch lower gives back righting ability in a knockdown (the boat thrown flat on its side).
The cells: EVE MB31, 314 Ah rated, 341 Ah typical actual, 2.8 x 6.8 x 8.1 in (71 x 173 x 207 mm), about 12.3 lb (5.6 kg) each: about 395 lb of cells. Each 16-cell stack is about 45 x 6.9 x 8.75 in and about 200 lb. Buy 36 Grade-A cells; the acceptance test (Part 7) keeps the best 32. The other 4 stay ashore at 50 percent charge (3.28 to 3.30 V), terminals taped, as swap-ins during first setup. Never aboard. The lid holds the pack in, upside down. Its 24 screws in epoxied brass inserts hold thousands of pounds, far more than the pack can push upside down under a wave slam. The batten foam, squashed 1/4 in, presses the pack down with roughly 1,000 to 1,900 lb over its three strips (about 210 sq in), more than the pack weighs, so it cannot lift. The unbroken 3/4 in plywood carries many times the pack weight even at 3 g, spread evenly. What plywood cannot take is a point load, like a terminal stud punching into it: the whole reason for the battens and the clay check. The watchman only watches. Nothing gates the pack: power runs to the boat through plain heavy cable, and a dead watchman changes nothing (What Can Go Wrong). Why no battery disconnect and no advanced battery management board: Design Decisions R-4.1A and R-4.1B. The box as built: R-4.1D; the relief into the cabin: R-4.1G.
Before Anything: The One Measurement
Where and when: Parts 1 to 3 are in the trough, 4 and 5 on the bench, 6 glues the box in, 7 to 9 fill and wire it, 10 tests and closes it.
- Hold a level board across the trough, 2 in above the plywood floor.
- Measure the open width between the sloped walls at three places: each end of the box and the middle.
- You need at least 17 in at all three: the box is 16.0 in wide outside, plus room to work the fillets and tapes. Better still, set the mockup or the finished box on 2 in blocks in place and check the side room.
- If one place is under 17 in, raise the height in 1/2 in steps until it passes everywhere. That is your shelf height. Do not mix any foam until this passes.
Part 1: Clean Out the Trough
- P100 mask on. Vacuum the whole trough, reaching the hose into the hollow under the plywood floor through the old holes.
- Pull out the dropped wire brush wheel and all loose junk.
- Wipe dusty areas with a damp rag, not a dry broom.
- Pull the old copper lightning wire out completely; the foam will bury this area, and it is never a strike path again. Keep only 6 AWG or thicker that tests clean, for bond spurs (wires to the lightning copper); scrap the rest (Lightning Ground System Guide sec 4).
Part 2: Fill With Foam
Foam grade: 2 lb/ft3 on Boat 1 (poured), 6 lb/ft3 structural grade on Boat 2. The 2 lb foam carries about 30 psi against under 1 psi from the box; the Part 3 skin carries the box. Check first.
- Hose test: run water in at the end away from the bilge and watch it run out at the bilge sump, the drain path the limber tube must keep. If water pools instead, find out why before burying anything in foam.
- Any lightning conductor that must pass this run goes in conduit, clear of the box footprint, before the foam buries it.
The pour.
- Lay the 1 in limber tube along the trough's low path, the full box footprint and past it at both ends. Tape the ends shut; weight or tape it down so a pour cannot float it.
- Fill the hollow under the plywood floor first: a small cup of foam, equal parts A and B, stirred 30 seconds, poured at once. It puffs up in 2 minutes and is hard in 15. Pour small cups at different holes until the plywood no longer flexes anywhere.
- Draw a level line on both walls at your shelf height, 2 in or what the measurement said.
- Fill the trough in shallow layers, small batches, burying the limber tube as you go, 20 to 30 minutes between layers. Sneak up on the line; leave the last layer just proud of it.
- When fully hard, a few hours, saw and sand the top flat and level to your line (a long board with sandpaper glued on works well).
- Vacuum everything. Untape the limber tube ends and check you can see daylight through it.
Part 3: Put the Hard Skin on the Foam (the Shelf Top)
- Grind a 3 in tall band on both walls, from the foam surface up, the full length of the trough, to dull raw glass. Vacuum and wipe.
- Cut 1708 to cover the whole foam top in two layers, wide enough to run 3 in up both walls onto the ground bands.
- Paint epoxy on the foam and walls. Lay the first layer and wet it through until dark everywhere, then the second. Roll out air bubbles.
- Let it harden overnight. The skin edges are now glassed onto the walls, where the load lives.
- Repeat the daylight check through the limber tube. This is the last time it is easy to fix.
Part 4: Build the Box on the Workbench
Before cutting the walls, do the hardware height check (Part 7 step 4). If it fails, cut the walls 9.75 in tall instead of 9.25. Cut list, 3/4 in plywood. Sheet one: floor and both long walls. Sheet two: lid and end walls.
| Piece | Size (inches) | How many |
| Floor | 53.5 x 16.0 (includes a 1 in lip past each end wall) | 1 |
| Long walls | 51.5 x 9.25 | 2 |
| End walls | 14.5 x 9.25 | 2 |
| Lid | 51.5 x 16.0 | 1 |
| Lid battens | 1-1/2 in wide rips from the same scrap (Part 7 step 11) | 3, cut to the measured pack (the flat cell tops run about 46 in) |
The aft end wall carries the 300 A main fuse, bolted through it from outside on the + side (Parts 5 and 8). The watchman box is bolted to a long side wall, as built (Part 8).
- Draw the wall outline on the floor piece: a 51.5 x 16.0 in rectangle with 1 in of floor past each end. The walls sit flush with the floor's long edges, no side lip.
- Glue and screw the long walls on the lines: a bead of thickened epoxy, then pre-drilled #8 screws up from underneath, about every 8 in.
- Fit the end walls between the long walls the same way, plus screws through the long walls into the end wall edges.
- Check square: corner to corner both ways; equal means square. Let it harden.
- Fillet every inside corner, floor to wall and wall to wall, with thickened epoxy, rounded like a thumb-width caulk bead, then two strips of 1708 over every fillet, wet through. This is the strength of the box; the screws were only clamps.
- Round the outside corners with sandpaper, then one strip of 1708 over each and along the joint where the end lips meet the walls.
- Glass the bottom of the floor with one full layer of 1708, wrapping over the exposed floor edges. The edges are end grain and must be sealed in glass, not just painted.
- No rim doubler: the cells would not fit past one; do not re-propose it. The rim is the bare 3/4 in wall top, capped in Part 5. Boat 2 rule: round every glassed edge to a 1/4 in radius or more before the glass goes on (cloth will not follow a sharp edge).
- Paint every surface of box and lid, inside, outside and edges, with two coats of plain epoxy.
Part 5: The Lid, the Rim Inserts, and the Wall Openings
The lid's jobs: first, keep water out (the gasket); second, keep 500 lb of batteries in the box with the boat on its side or upside down. The pack must never load its terminal threads against anything, so three foam-faced lid battens land on the flat cell case tops: one down the center gap and one down the middle of each stack between its terminal rows. The lid comes off, on its 24 screws, so one or two failed cells can be replaced at annual service or at recovery. Its seal is for moisture and gas, not flooding, but the lid must pass a 24 hr static-head flood test (Build Guide sec 11.2), and gas must still get out through the relief stub (Part 9). The lid.
- Cap the rim: two layers of 1708 over the bare 3/4 in wall top, sanded flat so the gasket lands on a flat glass face. On the lid, 3 in wide 1708 strips along the screw line, top face only, trimmed flush at the edge, no edge wrap. Let harden.
- Try it in scrap first: drill a test hole and set one insert before drilling the real rim. On it, check that a 1-1/2 in screw through lid, washer and gasket does not bottom in the 5/8 in hole.
- Drill the 24 holes with the lid clamped in place. Mark the screw line on the rim centerline, about every 5-1/2 in. At each mark drill the 5/16 in clearance hole through the lid and on into the rim as a pilot. Lift the lid; open each rim hole to 25/64 in, 5/8 in deep (tape the bit as a depth flag).
- Set the E-Z Lok inserts wet: plain epoxy in each hole, thickened epoxy on the insert, then drive it flush with a 1/4-20 bolt, two nuts jammed together and a socket wrench, straight and slow. Wipe squeeze-out before it reaches the inner threads. After cure, run a 1/4-20 tap through every insert.
- Lay the gasket flat, no groove: the WOD tape adhesive down, the full 3/4 in rim width, punched at every insert. One continuous run, its one butt joint on a straight side, never at a corner, the cut ends pressed hard together (a thin silicone bead under the joint is optional). Silicone dielectric grease on its top face. Never open-cell foam or soft weatherstrip foam. It seals out moisture and gas under light squeeze, not flooding.
- The battens are fitted to the real pack later (Part 7 step 11).
- No opening in the lid, no relief-duct fitting. The relief stub goes through the forward end wall (Part 9).
- Fit the lid with the machine screws and fender washers, blue Loctite on the mission bolts. Snug, not gorilla tight: do not crush the gasket flat. Re-snug after a week, when the sponge has taken its set. Paint a white witness stripe across each screw head onto the lid. Check the gasket at every yearly opening.
Drill the wall openings now, before there are batteries below: seven openings plus the watchman box's four bolt holes. None through the lid. Aft end wall:
- Two PG16 glands (7/8 in hole) for the 2 AWG main cables, in line with the terminal posts at the post height measured in Part 7 step 4, so the cables run straight over the aft plate.
- Three signal glands, one cable each: two cables of different size do not seal in one gland, sleeved or not (Mike, 2026-10-13). PG13.5 (13/16 in hole) for the balancer's ENERJOUR 18/10 cable (its nine taps; 0.39 in (10 mm) across); PG9 (5/8 in hole) for the SBMS0's 22 AWG 12-core shielded cable (its nine taps and the XT1 pair; 0.24 in (6 mm) across); PG9 (5/8 in hole) for the DS18B20 probe trunk (Part 8).
- The IP68 breather screw (the M12 Gore vent), high, in a 1/2 in hole.
- Room for the 300 A main fuse outside, at mid-height on the + side; it goes on once the pack is in (Part 8).
- The watchman box's four 1/4 in bolt holes are in a long side wall, as built: Part 8, Mounting.
Forward end wall: the relief stub, 2 in hole saw, high (Part 9). Paint every cut edge with epoxy, then screw the fittings in, bedded in butyl (a sticky rubber sealant).
Part 6: Glue the Box Into the Boat
No mounting holes. The box is glued to the shelf over its full footprint, about 900 sq in, and taped round with four staggered plies of 1708, so no load path depends on a glue joint being peeled.
- Set the empty box on the shelf where it will live and trace around it. Check the side gaps to the walls look even.
- Grind the shelf top dull inside the outline and in a 5 in band around it, and the lower walls where the side tapes will land if the walls are close. Vacuum, wipe.
- Sand the bottom of the box dull. Wipe.
- Spread thickened epoxy over the whole traced area like buttering toast, about 1/8 in thick.
- Set the box into it, press, and load it with weight (cinder blocks, water jugs). Shape the squeeze-out into a rounded bead all around with a gloved finger.
- Let it harden overnight. The box is now part of the boat.
- Tape the box down all the way around: a thumb-width fillet wherever box meets shelf, or box meets wall where the trough is narrow; then four strips of 1708 over it, each wider than the last (3, 4, 5, 6 in), half on the box and half on the shelf or wall. Stagger the corners so the strips overlap. Wet through, roll out bubbles.
- Once hard, paint the tapes and any raw spots with two coats of epoxy.
Part 7: Build the Stacks
The pack is 32 cells in 8 groups of 4, in a chain: two stacks of 16 side by side, squeezed between one wide 3/8 in G10 plate at each end. A cell wants a slight squeeze on its flat faces; the threaded rods and Belleville washers give it, set to the force on the MB31 spec sheet.
Cell acceptance, before assembly
Inside the distributor's warranty window, before any stack is built:
- Clean the terminal pads.
- Test every cell with a 4-wire Kelvin probe (FNIRSI HRM-10 class: separate contacts for current and reading, so the leads do not count) on the pad metal. Record the resting voltage (OCV) and the 1 kHz internal resistance (IR) for all 36.
- Reject any cell over 0.23 milliohm, or drifting below about 3.25 V over a 7-day self-discharge watch.
- Sort the 32 keepers into 8 groups of 4 so every group's average IR matches.
The steps
- Floor: knock down any high spots in the epoxy, then lay the 1/8 in dense closed-cell foam under the pack. No fairing, no rigid plate.
- Side padding: glue the 3/8 in neoprene strips to both long walls now, where the stack sides bear; once the cells are in you cannot reach the wall. 13.625 in of pack plus two strips in the 14.5 in opening leaves about 1/8 in of play. A third strip goes down the middle between the rows as the stacks are built, notched where the stack-to-stack jumper crosses at the forward end. Measure before cutting it: with the wall strips in, measure the opening, then the cells set side by side. On paper two rows (about 6.8 in each) plus three strips are about 14.75 in in a 14.5 in opening, so the middle strip may have to be thinner than 3/8 in; the measurement decides. If it is cut from the Ramede polyethylene sheet instead, check first that it is firm enough.
- Forward spacer block first, glued against the forward end wall at the floor: 13.5 in wide, 2.5 in tall, 7/8 in wood faced with 1/8 in EPDM rubber (60A). The forward G10 plate bears on it.
- Height check, before the walls are cut if at all possible: build up one cell with its busbar and nut on the bench and measure from the cell bottom to the top of the tallest hardware. Over 8.9 in, the walls must be 9.75 in (Part 4). Record the terminal post height too; the two PG16 main cable glands go at that height (Part 5).
- Carry the cells aboard one at a time, 12 lb each. Never move a finished stack (about 200 lb).
- Build both 16-cell stacks in the box, side by side: the forward G10 plate (13.5 x 8.5 in) on its spacer block, then cells face to face, each group of four flipped the opposite way to the last so the series links run along the terminal rows (pack wiring diagram). A 5/64 in (2 mm) mica barrier between every pair of cells (1/16 in (1.5 mm) if the length is tight; not aerogel) stops the aluminum cases chafing and slows heat spreading from a hot cell. The middle strip goes in as the stacks grow. If XT1 (the watchman's thermistor, a small temperature sensor) will sit under the side padding, calibrate it first (Part 8, The temperature probes) and set it against its cell face now. Then the aft G10 plate (13.5 x 8.0 in, lower so the main cables pass over it).
- Three rods, top only, through both plates: 3/8-16 18-8 stainless threaded rod in 1/2 in PEX sleeves cut 1/4 in short of the plates, so bare thread can never touch a busbar or terminal and become part of the circuit. No low rods.
- Bellevilles, all at the aft end: on each rod outside the aft plate, about 10 of the 3/8 in conical washers alternating ()()() make 0.75 in of stack, in a PEX sleeve to keep them lined up; they hold the squeeze as the cells breathe. Set the squeeze by measured stack height with a light starting load, about 1 ft-lb on the two outer rods and 2 ft-lb on the middle one, and check the cell face pressure against the MB31 sheet.
- Cut the aft spacer block now, to the measured gap between the aft plate and the aft end wall at the floor (about 1.75 in), 13.5 in wide and 2.5 in tall, and glue it in. Cut to what you measure, never to the drawing.
- Leave slack for growth: the cells grow over the mission, about equally at both ends, and the aft terminal posts move about 0.15 in. Leave that much slack in the tap rings and the mains at the aft end.
- Fit the three lid battens: each a 3/4 in plywood rip 1-1/2 in wide, glued under the lid, faced with a 3/4 in thick x 1-1/2 in wide strip of the medium neoprene foam: 1-1/2 in in all. Screwing the lid down squashes the foam 1/8 to 1/4 in, so the pack cannot lift even upside down.
- Measure the real gap first, lid underside to flat cell case tops, gasket on: about 1-1/4 in on paper (lid underside about 9.5 in above the floor, cell tops about 8.2 in). Batten plus foam stands 1/8 to 1/4 in taller than the gap.
- Place them by the cells, not the lid edges: mark the flat case tops with the pack in the box, and glue the battens over the marks.
- Keep the foam off each cell's gas vent (on EVE cells usually midway between the terminals, right where the strips run).
- Squashed, a batten's wood sits about 1/2 in above the cell tops, level with the stud tops, so no batten edge may reach over a stud or busbar.
- Leave room for the tap wires: stop the center batten short at both ends (W0, W4 and W8 pass there), and cut a 3/8 in notch in a stack's batten wherever a tap wire crosses that stack (W2 and W6, at cells 8 and 9), so no wire is squashed under the foam (Reference/Battery Tap Wiring 2026-10-03.pdf).
- The foam loses some push over the first year but stays in contact.
- Prove the threads are clear: a pea of modeling clay on the studs nearest each batten, along the pack; fit the lid, snug the screws, open it. Squashed clay means the lid or a batten touches a stud: fix it before going on. Repeat a week later; a stud that cleared on day one can touch once the foam takes its set.
- Bolt the busbars across the posts to make the groups and the chain, by the pack wiring diagram (delivered separately). The tap rings land under these same nuts (Part 8, Landing the taps).
- Series links run along the terminal rows. The one link across the center gap is an insulated flexible jumper (laminated or braided) at the forward end, so the stacks can move a hair without working the terminals. With both pack ends aft and four groups in a row per stack, the chain runs up one stack and back down the other, so it crosses at the forward end. The braid link (Parts List) bends to the span, about 2-3/8 in center to center (worked out from the MB31's terminal spacing and the 3/8 in center strip). If it comes bare, sleeve the braid in heat shrink so it is insulated, lugs left clear.
- Both pack terminals end up at the aft end.
- NO-OX-ID A-Special on every terminal and busbar contact face before the nut goes on. Never silicone dielectric grease on these joints.
- On each M6 welded stud, bottom to top: the busbar; the tap ring where one lands (one lug per tap point, not two: a joint's SBMS0 wire and balancer wire land on two different posts); the Belleville washer; the flange nut, with blue Loctite 242 on the upper threads. The flange nut with the Loctite is the locking nut.
- The Belleville washer (M6 x 12 mm 304) holds the clamp force as the joint heats, cools and settles. Hollow side down, against the ring lug or busbar, the raised middle pointing up into the nut's flange like a small hill: tightening squashes the hill flat, and it pushes back as a spring. Upside down it rocks on its outer rim and does nothing useful. It sits on top of the stack, so no current passes through it.
- Serrated flange nuts: the teeth bite the stainless washer, which is fine. Never let a serrated flange sit straight on a busbar or lug.
- Blue Loctite 242 only, on the top threads only, wiped clean of grease first, never on a face that carries current. Never red (262, 263, 271 or 272): it takes about 500 F, or a force that can snap the welded stud, to undo.
- Torque once, to the cell spec sheet: 37 in-lb (4.2 N-m) for the whole stack, on the click torque wrench on hand. 37 in-lb is just under its rated range (40 in-lb), close enough: nothing to buy. Tape the wrench and any extension so only the socket tip is bare, and cover the next terminals. No second torque later: once the blue sets, the wrench clicks without turning the nut and proves nothing; the Belleville takes the settling.
- Towel over each finished section as you move on. Witness stripes wait until Part 8, after the tap rings land under these nuts.
Part 8: The Watchman and the Wiring
The watchman is two boards in a small IP65 plastic box bolted to the outside of a long side wall of the battery box, never inside it (Electronics and Wiring Plan sec 6.7 and A4):
- The Electrodacus SBMS0 reads the 8 group voltages through its own nine tap wires, the pack current through its 300 A shunt on the main positive, the total solar current through its 100 A shunt on the solar positive, and the pack temperature through XT1 on a cell face.
- The JKBMS JK-B2A8S balancer does 2 A of active balancing across the 8 groups through its own nine tap wires.
Nothing electronic lives with the cells: only the harness ends and the probes are inside the box. The watchman only watches: no wire to the chargers, no load switched ("No charge stop").
Mounting
The watchman box is bolted through a long side wall of the battery box, the same way as the main fuse: four 1/4 in stainless bolts, each hole bored oversize, filled with thickened epoxy and redrilled so no raw plywood is open to water, each bolt bedded in sealant, fender washers inside, nylon lock nuts or blue Loctite.
- The spot, as built: a long side wall of the battery box (Mike, 2026-11-02: "it is mounted on batt box side"). Open: which side and how high; Mike's word or a photo settles it.
- Drill before the pack goes in (Part 5). If after: towel over the busbars, taped tools, rings off (rule 4).
- Its four glands all face down, each cable with a drip loop (the cable dips below the gland before it rises away): the balancer's 18/10 cable in (PG13.5), the SBMS0's 12-core in (PG9), the shunts' 4-core sense cable (PG9; both pairs share its one jacket), and the USB lead out through the split gland (Parts List). The DS18B20 probe trunk does not enter this box.
- The 24 hr flood test checks these bolts with the rest of the box. The box is IP65: it takes spray, not standing water.
Landing the taps
Each board has its own nine tap wires, each on its own ring on its own post: 18 rings in all. Never join an SBMS0 wire to a balancer wire anywhere: the balancer's current on a shared wire would show in the SBMS0's group readings as an error about the size of the slow drift shore watches for (Design Decisions R-4.1B). Any post on a joint is the same joint (the four positives of one group and the four negatives of the next are tied by busbars), so a joint's two wires land on two different posts of it, one ring per stud, under the busbar nuts of Part 7 step 13.
- W0 on the most negative joint of the chain (pack negative), W8 on the most positive (pack positive), W1 to W7 on the joints between groups, in order.
- Which post each ring takes: Reference/Battery Tap Wiring 2026-10-03.pdf. Cells are counted from the aft end, A1 to A16 and B1 to B16; the main cables take A1 and B1, so the W0 and W8 taps go on the next posts. Tap each joint at its post nearest the aft wall (the farthest is about 35 in from it).
- The cables: the SBMS0's nine wires and the XT1 pair are 11 cores of the 22 AWG 12-core (the 12th spare, taped both ends); the balancer's nine are cores of the ENERJOUR 18 AWG 10-conductor (the 10th, white, spare). Wire and Cable Buy List C1 and C7.
- Colors, the same in both cables (the resistor color code): W0 black, W1 brown, W2 red, W3 orange, W4 yellow, W5 green, W6 blue, W7 violet, W8 gray; XT1 on white and the 11th color. If a cable has no core of a color, use a leftover core with colored tape at both ends.
- No fuse of any kind on the tap wires (Mike, 2026-11-08: "I think adding 18 fuses and connections on top of the batteries increases risk"). Each tap wire goes straight to an insulated ring on its bus bar nut, NO-OX-ID on the ring's face. Split loom goes on any tap wire where it crosses a bus bar or stud of another voltage, and every wire is tied so it cannot rub.
- Label both ends of every wire as you land it.
Every pin, post and color, and the order of work: the workbench sheet Reference/BMS and Balancer Wiring Sheet 2026-10-03.pdf (US letter, 4 pages).
Dress the cables
One gland per cable through the aft wall (Part 5), each with its own drip loop outside and silicone dielectric grease on its gland's insert. The two tap cables go on to the watchman box, each into its own gland there; the probe trunk runs straight to the central node (the small control box near the battery that switches loads and triggers releases). Ground the 12-core's shield, by its bare drain wire, at the watchman box end only; at the battery end cut the drain short and tape it. That keeps the balancer's switching out of the SBMS0's wires. Inside the battery box (check the route against the real pack before the clamps go in):
- The jackets stay on to just inside the glands, where each tap cable splits into one branch per stack. As built (Mike's photos, 2026-11-08), the wires run tied to the blue PEX rod sleeves, not in split loom (a slit plastic tube snapped over the wires); split loom goes only where a wire crosses a bus bar or stud of another voltage ("Landing the taps").
- Hold each branch with nylon cable clamps on short wood screws into the long wall, just above the cell tops: screws no longer than 1/2 in in the 3/4 in wall, dipped in epoxy; a clamp wherever a wire leaves the branch, never more than 18 in apart (the ABYC spacing). Never tie anything to the cells.
- Each wire leaves its branch close to its post with a small S-bend of slack, which also takes the 0.15 in of post movement at the aft end.
- A wire crossing a stack runs through the 3/8 in notch in that stack's lid batten (Part 7 step 11). W2 and W6 cross the outer row in the one gap between busbars, at cells 8 and 9. W4 goes around the forward end, over the forward plate, and down the center.
- Nothing over a busbar, a stud top or a cell vent (the spot midway between a cell's two posts).
Inside the watchman box
Build this end first, with both board plugs unplugged. The SBMS0's tap wires land on the ribbon breakout board on the DIN rail, fed from the SBMS0's cell socket by the short 12-pin ribbon. The SBMS0's own ribbon is not cut (Electrodacus: never longer than its 6 ft). Terminal number = ribbon wire number. The 22 AWG cores go under the screws bare (the screws take 26 to 16 AWG), stripped about 1/4 in, no ferrules (crimped metal sleeves). Never more than two wires under one screw. A jumper is a short wire from one terminal to the next.
| Battery wire | Breakout terminal | Jumpers and notes |
| W0 | 1 | Jumper 1 to 2; the 12-core's shield drain on 2 |
| W1 to W7 | 3 to 9, in order | |
| W8 | 10 | Jumpers 10 to 11 and 11 to 12 |
Never skip a jumper or land a wire one terminal low. Pins 1 and 2 are both pack minus and pins 11 and 12 carry pack plus to power the board, so W1 on terminal 2 puts a cell group straight across the SBMS0 (the first SBMS0 overheated this way, 2026-10-21). The meter check before plug-in ("Check, then plug in") catches it. The short ribbon (Mike, 2026-11-02: "pin 1 was stripe, made it backwards and both old and new bms work") runs pin 1 to pin 1, its stripe on terminal 1 of the Winford board. With both plugs keyed the same way, its stripe sits on the opposite edge from the factory ribbon's red wire (Electrodacus's ribbon marks pin 12 in red, Reference/SBMS0 v03d Manual.pdf page 5). A ribbon fits one way at each end and can still be crossed inside, so it is metered before it is plugged in ("Check, then plug in"). Both SBMS0s on hand work; one is on the boat (Mike, 2026-11-08). Check the SBMS0's socket is two rows of 6 pins 0.1 in apart. Tie both ribbons flat to the plate before the boards go on, a dab of removable silicone on each ribbon plug. The SBMS0's green 16-pin plug (push-in): pin 1 PVn, 2 PVp, 3 ADC1n, 4 ADC1p, 5 ADC2, 6 ADC3, 7 to 14 the EXTIO pairs (left empty), 15 XT1-, 16 XT1+. Strip about 1/4 in (5 to 6 mm) and push the wire in to the stop; to release it, press the button beside its hole with a small screwdriver.
- XT1: the 12-core's white and tan cores, on terminals 5 and 6 of the six-block strip, then on to green (pin 15) and blue (pin 16), either way. A thermistor has no polarity.
- The shunt sense wires: through terminals 1 to 4 of the same strip ("The two shunts", next).
The balancer: its 11-pin plug is B0 to B8 plus its own power pair, V- and V+ (10 to 40 V); it does not run on the nine taps alone. Each harness wire joins a core of the 18/10 cable with a butt splice, B0 to B8 on W0 to W8 by pin number (the harness wires' own colors do not matter). At W0 the B0 and V- wires share one end of the splice; at W8 the B8 and V+ wires. Stagger the splices so the bundle stays slim, and tie them to the plate. The board still unplugs at its own 11-pin plug. Its temperature plug and data plug stay empty.
The two shunts: both on the positive side
The SBMS0 reads both shunts through current chips that need 4 to 76 V on the shunt; on the negative side they see 0 V and read nothing (Electrodacus's schematic and drawing, Reference/SBMS0 v03d Manual.pdf). So both sit on the positive side, on the main bus board (its layout: Electronics and Wiring Plan, the main bus board), and the negative side is plain cable:
- The 300 A battery shunt: battery switch, then the shunt, then the + 400 A bar (stud D).
- The 100 A solar shunt: + solar bar, then the shunt, then the + 400 A bar (stud A).
- The negative side: the pack - cable plain to the - 400 A bar (stud 4), and one plain 6 AWG from the - solar bar to the - 400 A bar (stud 1).
Before the first power-up, check both shunts sit on the + side: nothing but plain cable on the - side. Battery switch off and no PV connected while working on the board. Each shunt has two big bolts for the current and two small screws for the sense wires. The sense run, as built (Mike's photo and notes, 2026-11-08): one 4-core 20 AWG cable (black, red, yellow, white) runs from the shunts to a six-block DIN strip in the watchman box, a ring terminal under each small screw. Its cores land on the top of terminals 1 to 4, and the Cat6 cores to the SBMS0's green plug on the bottom, so terminals 1 to 4 are green plug pins 1 to 4. The Cat6 is the stranded patch cord: stranded, never solid (Electrodacus suggests solid Cat5 or Cat6), since solid wire cracks under vibration. Tag the black core +26 V.
| Sense wire | 4-core | Strip terminal | Cat6 | SBMS0 green plug pin |
| Solar shunt, bus side | Black | 1 | White | 1 (PVn) |
| Solar shunt, solar-bar side | Red | 2 | Orange | 2 (PVp) |
| Battery shunt, bus side | Yellow | 3 | Tan | 3 (ADC1n) |
| Battery shunt, battery-switch side | White | 4 | Brown | 4 (ADC1p) |
Terminals 5 and 6 are the XT1 thermistor ("Inside the watchman box"). No fuse of any kind on the four sense wires (Mike, 2026-11-08: "yes no fuses on shunt wires"). At the shunts the sense wires sit only beside each other and the + leads, all within millivolts, so a rub there shorts nothing. The 4-core cable stays clear of the main bus board's minus bars and is tied away from the SBMS0's USB plug in the watchman box (the USB shell is Pi A's unfused minus). Each shunt keeps a cover over it: it is live. Nothing else goes on the shunts: sense leads only. If battery current reads backward (charging shows minus), swap white and yellow. The solar input reads one direction only: if it reads zero in sun, its pair is swapped.
The temperature probes
XT1 gives the pack temperature in the watchman's stream. Trap it against a cell face, not in box air, which can read 10 degrees off the metal. Hold it with something solid: under the side padding, or in a small saddle epoxied to the cell face. Thermal tape over it is the backup, never the mount. Open: the XT1 probe end was found lying loose on a cell top (Mike's photos, 2026-11-08); it is trapped against a cell face as above. Calibrate XT1 in ice water before it is trapped. It is a DROK 10K 3950 thermistor, which the SBMS0 reads with a curve written for a Murata part, so near freezing it reads about 4 F low.
- Probe tip in stirred crushed ice and water, two minutes: that is 32 F.
- Read the pack temperature in the SBMS0 Monitoring menu; note the raw value in the Diagnostic menu.
- Set parameter 29, a displayed value only, to the reading the SBMS0 shows, not to 0 C.
- Write both numbers on the System Settings Sheet.
- Unplug XT1 and record what the SBMS0 and the telemetry show with no probe: the "probe fell off" signature the shore alerts need. Plug it back and check the reading returns. Only if that reading is unclear, ask Electrodacus what the SBMS0 reports with XT1 unplugged (the optional Electrodacus question in Pending Amendments).
The six DS18B20 pack probes feed the boat's own telemetry (Build Guide sec 8.2).
- Where: three per 16-cell stack, one on each of the stack's two middle groups (the cells heat has the farthest to travel from) and one on an end group, each held against its cell face like XT1, reading the metal, not the air.
- Names: each by its factory serial number, so its place on the chain does not matter. Before mounting, read each probe's number one at a time with the spare Nano Every on a PC (Arduino IDE, the OneWire library's DS18x20_Temperature example, data on D10) through the probe setup board. Tag each lead; write number-to-cell in the System Settings Sheet sec 9 and on the node lid. Probe wire colors differ by seller: go by each seller's listing. The trip does not care which probe is which.
- The trunk: Mike's 3-conductor 20 AWG cable, all three wires used: red power, black ground, the third data. No shield is needed for a 1-wire chain this short. It leaves the battery box by its own PG9 gland and runs straight to central node entry 21, never through the watchman box. Inside the box it runs along the stacks past each probe, the probes about 2 ft apart: a daisy chain, not a star, which is also the better layout for 1-wire signals.
Tying each probe in (no solder, and no WAGO 221 lever connectors: the probe leads are thinner than the 24 AWG they take at the low end):
- At each probe, strip the trunk's jacket back about 3 in and cut its three wires.
- Each wire gets one red butt splice (22 to 16 AWG): the trunk wire coming in on one end; the trunk wire going on, plus that probe's matching wire folded double, on the other. A 20 AWG wire plus a folded thin probe wire fits the red splice's range.
- Stagger the three splices about 1 in apart and shrink them, then slide a glue-lined heat-shrink sleeve over the whole stripped section and the jacket ends.
- At the last probe the trunk ends: the trunk wire on one end, the probe wire on the other. Six probes, three wires each: 18 splices.
- The probes' own leads, about 3 ft each, stay full length, so the joints sit together on the trunk, away from the cell terminals; coil and tie the spare lead neatly. Six 3 ft branches plus the run to the node is well inside what one chain handles on a 4.7 k resistor (a guess from common 1-wire practice, not a datasheet figure).
- Tie the trunk and each joint down clear of the terminal rows and the series links.
The cell-temperature trip, the battery's one automatic trip, runs on these probes. Any probe above 150 F, or a rise of more than 10 F in five minutes on any probe, puts the boat in HIBERNATE (loads off) and sends an alert, and the computer in command runs the motor at about the solar output, so the battery's net current stays near zero instead of the solar charging a hot pack (the surplus-spending logic, Build Guide sec 4.1). Not while the drogue is out or in STORM, when the motor stays off. Open: nothing cuts the bank off on the trip; settled in the single-point review. Release comes when every probe has read below 120 F for one hour (System Settings Sheet sec 4). Not 140 F, the cell sheet's discharge limit, which a hot still day in the tropics can reach (Design Decisions R-4.1G). At 131 F (the SBMS0's parameter 30, a displayed value only) shore gets a warning-only alert. Below 33 F on any probe, Pi A alerts shore ("The cold, and no heaters").
Final routing
Run the two main cables straight over the aft G10 plate and out through their PG16 glands at post height. With every cable through its gland and greased ("Dress the cables"), tighten the caps; a drip loop in each cable outside the box. No gas-sensor leads, no heater feed. The + main lands on the main fuse (next) and goes on to the main bus board ("The two shunts"); the - main runs as plain cable to the - 400 A bar (pack wiring diagram). NO-OX-ID on every lug face at the busbars.
Mount the main fuse on the aft wall
The 300 A main fuse is a Class T, a fast fuse made to stop the huge current a battery short can push: the JIAKUPA kit fuse. It bolts to the outside of the aft wall, not to the main bus board. Not fitted: a Blue Sea 5119; do not re-propose.
- Check the room first. Room for the 4.8 x 2.0 in fuse block, its cover and the cable bend has not been measured. Hold the block in place with the + main dressed to it before drilling anything.
- Place it at mid-height on the + side of the aft wall, within 12 in of where the + main leaves its gland.
- If the bolts go through the wall: lid off, a towel over the busbars, taped tools, rings off (rule 4). Keep the bolts at mid-height, where behind the wall there is only the 1.75 in air gap and the aft G10 plate. Bed each bolt in sealant so the box stays sealed, and hold each nut with blue Loctite or a nylon lock nut.
- Land the + main: from the pack post, out its PG16 gland, down through its drip loop, onto the fuse. From the fuse a sleeved 2 AWG run goes to the Blue Sea 6006 battery switch on the main bus board, so the whole run to the board is protected.
- For the first 3 ft, keep the + and - mains apart: separate paths, each in chafe sleeve, clamped, never crossing. A short just past the fuse could pass more than its 10,000 A rating.
- Clamp both mains a few inches from the fuse block, so they do not tug on it.
- Torque the fuse stud nuts and paint a witness stripe across each.
Battery switch off before anyone touches the mains.
Check, then plug in
- Witness stripes: with every ring landed and every nut at final torque, paint a stripe across each nut and onto its stud. The last look in Part 10 checks that no stripe has turned.
- Check the order: with a voltmeter, each neighboring pair of tap wires (W0 to W1, W1 to W2, and so on) should read about 3.3 V, and W0 to W8 about 26 V. Both sets: the SBMS0's at the breakout board's terminals (terminal 1 to 10 for W0 to W8), the balancer's at its harness plug pins. Any strange reading means two wires are swapped.
- The short ribbon: before plugging it in, meter pin 1 to pin 12 at its SBMS0 end, black lead on pin 1: plus about 26 V, never a minus sign.
- Only when all eight pairs of both sets read right, plug in the balancer, then the SBMS0's cell ribbon, then its green plug.
- Take the strain off every plug: at the SBMS0 (both ribbon plugs and the green plug) and at the balancer, tie the wires to the mounting plate right outside the plug, so the plug itself carries no motion. Silicone dielectric grease in each plug and a dab of removable adhesive on each plug body, so vibration cannot walk it out.
- The one DIP switch (a row of tiny on/off switches), six positions, beside the ribbon header on the WiFi/USB add-on board (MTX, STX, SRX, WRX, WTX, IO0): leave it on Normal, as it ships: MTX, STX and SRX ON; WRX, WTX and IO0 OFF. The reverse is only for loading new firmware into the WiFi module. The main board has no switch or jumper; everything else is set on its touch screen.
- Parameter Settings on the touch screen (battery type, number of cells, battery capacity; values: System Settings Sheet sec 1), Store Parameters, then unplug and replug the cell ribbon plug at the SBMS0 so the stored values take (Reference/SBMS0 v03d Manual.pdf page 6).
- Automation menu: EXTIO4 to Type 1 and EXTIO3 to Type 2 (System Settings Sheet). Nothing is wired to either output.
- Parameter 30 = 1182, the 55 C (131 F) high-temperature stop value, and parameter 29 = the ice-water reading from the XT1 calibration (the manual's value is 0 C, 3144). Both are displayed values only: the watchman stops nothing.
- UART (its serial data output) ON in Device settings, at 115,200 (Mike, 2026-11-08), or nothing streams to Pi A.
- "Save device settings," or the settings are lost at the next power cycle.
- The balancer, by phone over Bluetooth (the Jikong BMS app, Android 7 or newer): battery type LiFePO4, 8 cells, before trusting it. The rest stays at factory values: it balances above 3.0 V, when two groups differ by more than 0.01 V, at up to 2 A. It draws up to 1.2 W by its sheet (about 29 Wh a day at most); its real idle draw is not published.
The charge ceilings live in the five solar chargers and are the only automatic ones: user-defined lithium profile, absorption 27.6 V, float 27.0 V, absorption time 30 min, temperature compensation OFF, RX pin at its default (System Settings Sheet). The balancer keeps the eight groups level.
Telemetry to Pi A
The SBMS0's USB port, a micro USB socket on its WiFi/USB add-on board, streams pack data as text; the lead is required. Tie it down right outside the plug. It leaves the watchman box, plug on, through its split gland (never cut and rejoined) and plugs into the powered USB hub (the StarTech 10-port, in its own small box beside Pi A), not into Pi A: 16 ft of lead at most. No USB isolator: the hub runs off Pi A's own 5 V, so one would isolate nothing, and the add-on board has its own isolator between the SBMS0 and its USB socket (Electrodacus manual). Load shedding (turning loads off when power runs short) takes its battery percentage from this stream, because the SBMS0's pack shunt is the only battery current sensor aboard. No CAN and no RS-485 from the battery boards.
The cold, and no heaters
No main-bank heaters: no pads, thermostats, thermal fuses, heater tap or heater wiring in this box. Do not re-propose them. No cold charge stop. Charging these cells below 32 F damages them for good, but the route cannot produce that cold: the coldest monthly-mean sea on the way is about 41 to 43 F (5 to 6 C), near 56 degrees S (Reference/Cold Stop Sea Temperature Check 2026-09-23.md; the battery software notes in Reference, sec 3; why: Design Decisions R-4.1). SBMS0 parameter 29, the 32 F (0 C) low-temperature stop, is a displayed value only. The pack follows cabin air, not the sea: it sits on the 2 in foam shelf with no metal to the hull. A cold spell of several days on Leg 6 in September can take the cabin air, and the pack, below 32 F. So Pi A sends shore an alert when any of the six pack probes reads below 33 F (a message the boat sends home on its own, ahead of the daily report), and the computer in command 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. Not while the drogue is out or in STORM. The 33 F is a software number, changeable from shore. The wing carries no battery. It runs from this bank through the slip ring, so there is nothing on the wing to heat or cold-stop. Not fitted on the wing, do not re-propose: heater pads, thermostats, a snap disc or a cold charge stop (Wing Power System sec 1A).
No charge stop
The chargers always charge (Mike, 2026-11-01). The only charge ceiling is the five controllers' lithium profiles ("Configure the watchman"; System Settings Sheet sec 1); above 90 percent the computer spends surplus on the prop (Build Guide sec 4.1). A dead watchman cannot stop charging: it has no wire to the chargers. Why: Design Decisions R-4.1E. Not fitted, do not re-propose:
- The charge stop. The five Victron VE.Direct non-inverting remote on/off cables, ASS030550320, on hand, stay on the shelf; each charger's RX pin stays at its default. No yellow-wire strip or 1k resistor on the main bus board, no run to the central node (its entry 14 is not drilled), no charging role for relay channel 8 (the channel is spare) and no 0.5 A fuse for it, no 5-day stop or its timer.
- Any charge control by the watchman: no EXTIO4 pause wiring or its resistor; no charge-continue bypass, second resistor or two-core wire to the watchman box; no 10-minute silent-stream rule at the node (the stream lands on Pi A's USB; the node could never hear it go quiet).
- Anything in a charger's VE.Direct port, which stays empty: no remote cable, GX device, dongle or VE.Direct-to-USB data cable. Bluetooth is OFF on the MPPTs (the solar chargers) for power draw: no per-string data, reader board or USB hub for it.
ASS030550320 is Victron's only VE.Direct remote on/off cable, and it is NON-inverting; ASS030550310 is not a real order code.
Part 9: The Relief Stub (into the cabin)
The breather screw handles everyday breathing. The relief stub handles the one bad day: if a cell ever vents, the gas leaves through the stub and its one-way valve into the cabin, forward and away from the electronics, instead of building pressure in the box. The cabin is a sealed hull and gets no vent. The stub is the only opening into the box bigger than a cable gland, so it is one-way: gas out, nothing in. Not fitted, do not re-propose (why: Design Decisions R-4.1G):
- An overboard duct to the transom: no vent hose, swan-neck loop, flapper valve, transom thru-vent or hole, transom duckbill, hose clamps, burst membrane or flame-arrestor cap. It would add a hole that could let water in.
- A gas sensor, in the box or at the cabin deckhead (ceiling). In a damp sealed cabin it would false-alarm far more often than this pack will ever vent. The protection is the stub's aim and the cell-temperature trip (Part 8), the only automatic battery trip.
- Cut the hole with a 2 in hole saw, high in the forward end wall, clear of the forward plate and spacer block. Paint every cut edge with epoxy.
- Glass in the stub: a short length of 1-1/2 in Sch 40 PVC, sanded dull, through the hole, bedded in thickened epoxy and taped to the wall with 1708 on both faces, standing out far enough to take the coupling.
- Fit the duckbill: the Dometic/Sealand 1-1/2 in toilet duckbill, OEM 385310076, inside the stub, lips pointing out into the cabin, flange against the stub end, lips wiped with silicone dielectric grease first. The 1-1/2 in PVC coupling over the stub clamps the flange; nothing else needs a clamp. Not the Whale LV1219 duckbills on hand: those are for the bilge lines (Build Guide sec 3.6).
- No screen on the stub: the duckbill is shut except when the box vents. Not fitted, do not re-propose: a bug cap or a fine flame mesh.
- Aim the discharge up at 45 degrees and forward, away from the electronics, into open cabin space. Nothing is mounted in front of the stub.
Part 10: Test, Then Close
Relief stub check, before the lid goes on
- Blow gently into the stub from the box side: light breath pressure through a short hose, or a shop vac on blow, throttled way down. The duckbill lips must open and air must flow out into the cabin.
- Pull suction the same way. Nothing should come back: the duckbill holds.
Fix any stub that fails either half before the lid goes on for good.
Cracking pressure, measured, not assumed
- Grease the duckbill lips with silicone dielectric grease before this measurement, never after (Part 9 step 3).
- Tee a low-pressure manometer (a gauge for very small pressures; a clear vinyl tube taped to a yardstick will do) into the stub on the box side.
- Raise the pressure slowly with light breath pressure.
- Record the pressure where the duckbill first opens: the pressure the box sees before gas moves. It should be as low as the duckbill allows.
XT1 and cold alert checks
- XT1 ice-water calibration and unplugged reading, as in Part 8 (The temperature probes), before XT1 is trapped; both numbers on the System Settings Sheet. There are no heaters and no thermostat click-test.
- Thermistor reading: before trusting the pack temperature, read it in the SBMS0 Monitoring menu against a thermometer held at the same cell, and note the raw value in the Diagnostic menu.
- Cold alert: put one DS18B20 pack probe in stirred ice water and check Pi A sends the below-33 F alert to shore.
Broken sense wire rehearsal
On the bench, open one sense wire by lifting its ring, and record exactly what the SBMS0 displays and what the telemetry shows. It controls nothing, so nothing else can change. Write it down: the parser learns this nonsense-reading signature, and the shore alerts are built to match what the board really does, not what the manual implies. Reconnect and check the readings return.
Dead watchman test and setup checks
- Dead watchman: unplug the SBMS0 entirely and check nothing changes: all five chargers keep running and every load carries on. Check the quiet-watchman alert goes up. Plug it back.
- SBMS0 settings: check them against System Settings Sheet sec 1, the DIP switch and Parameter Settings rows among them.
- Charger profiles: check all five MPPTs against "Configure the watchman" and the System Settings Sheet.
- Balancer: check the JK board shows all eight groups and its balance current runs toward the lowest group.
- Telemetry: check the WiFi/USB board streams through the hub to Pi A.
- Cell-temperature trip: warm one DS18B20 pack probe past the 150 F trip level (a cup of hot water round the probe tip does it). Check the boat goes to HIBERNATE and an alert goes out. Let it cool and check the release waits one hour after every probe reads below 120 F.
First charge: no top-balance
No manual top-balance (charging all the cells to full, joined side by side, before assembly). Bolt straight into 8S4P, connect both harnesses, and let the JK-B2A8S's 2 A active balancing align the groups on a 5 to 10 A first charge over 24 to 48 hr.
Final look
Look at every busbar nut, ring terminal and gland once more, and check that every witness stripe still lines up straight. Gasket on, lid on, screws snug with blue Loctite. A week later, re-snug the screws and repeat the clay check. Then done.
What Can Go Wrong
At 60 to 70 percent depth of discharge on cells rated for 8,000 cycles (Build Guide sec 4.1), a cell failure at sea is unlikely. For each failure, what keeps the boat going:
- The watchman dies or is unplugged: charging and every load carry on, since it has no wire to the chargers and switches nothing. Shore loses the battery percentage and the group readings; the Pi raises the quiet-watchman alert and sheds loads on a fixed cautious schedule while the stream is quiet.
- The watchman box floods: it only loses readings, since it controls nothing.
- A tap or sense wire chafes: these wires carry no fuse (Mike, 2026-11-08), so split loom where a tap wire crosses another voltage, and ties that keep every wire from rubbing, are what stop a short; at the shunts a sense wire touches only wires within millivolts of it. A broken wire's reading turns to the nonsense signature recorded in the Part 10 rehearsal, which the shore alerts are built to know.
- A cell fades or disconnects inside (the common way these cells fail): its three partners carry on, that group loses about a quarter of its capacity, and with it the 8S4P pack quietly shrinks to about three quarters. The boat keeps running. The watchman sees one group swinging wider than the other seven, and shore watches that build in the SBMS0 telemetry over weeks; reduced range alone will not show it in time. This is the main reason the watchman is aboard. The fix is a cell swap at the next haul-out.
- A cell shorts inside (the rare way): this chemistry usually ends it as a hot swollen dead group rather than a fire. The pack would be crippled and the boat would carry on through the backup bank. Quality cells never frozen, overcharged or dented, and properly squeezed, fail this way about once per million cells over their whole life, far below the voyage's everyday risks for 32 cells over 16 months. The squeeze, the sealed box, the charge ceilings, the 33 F cold alert and the mica barriers push those odds down further.
- A cell runs hot or vents: the cell-temperature trip sheds loads and runs the motor to offset the solar, so the hot pack is not charged (Part 8); gas leaves through the stub into the cabin (Part 9).
- A cold snap: shore is alerted and the motor offsets the solar, so the pack is not charged cold; the boat keeps going on the battery ("The cold, and no heaters").
Open Items
Open questions for this guide are in Pending Amendments v10.6.