Andy at Off-Grid Garage woke on the morning of his 24 September video to a dark house. All three packs in his 48V battery shelf (304Ah, 280Ah, 280Ah) had tripped on undervoltage overnight. The load was an EV charger drawing about 2.8kW. It was set to stop at 10% state of charge, and it took that reading from a Victron Smart Shunt.
The shunt got it wrong because of what it can't see. When cell 12 in the middle pack fell to about 2.4V, that BMS disconnected and the bank lost a third of its capacity. The shunt only counts amp-hours at the bank terminals, so it kept calculating against the original capacity. Near the end the pack BMS estimated 0.3% at 44V and the shunt said about 15%. That was the number the EV charger was using.
The more interesting finding is what was wrong with cell 12. On charge it had always read highest, so the active balancer kept pulling energy out of it, for months. Andy took off its terminal connections, cleaned the aluminium terminals, applied a thin layer of MG Chemicals 847 carbon conductive paste and refitted the tinned-copper busbars at 4Nm. Cell 12 then became the lowest cell. Charging it by itself to 3.45V took 48.17Ah, and it was still absorbing at 4.4A when he stopped. That is roughly 17% of a 280Ah cell.
His reading: the cell was fine and the joint was not. The BMS measures voltage through the connection, so under charge current the resistance of an oxidised joint adds to the cell reading. By our arithmetic, 1mΩ at 80A is 80mV — enough to make a mid-pack cell look full. The balancer can't tell the difference.
Two caveats from Andy himself. First, re-making these joints has fixed the problem before, only for it to return after a few months, so the paste still has a long-term test ahead. Second, only two cells have been treated so far, and at least three others in the pack read low.
