Status: built · in use

15S3P LiFePO4, 2.88 kWh, assembled from bare cells with bolted bus bars and a commercial BMS.

Why not just buy it?

Barrier 3: sold only as a closed system.

E-bike batteries are easy to buy and impossible to build on. They are sold as part of a matched drive system, in a proprietary case, with a proprietary mount, communicating over a proprietary protocol with a controller you also cannot modify. Capacity, voltage, and chemistry are whatever the vendor decided.

Buying one was entirely possible. It would have meant accepting either buying a closed system or buying a much more expensive system that add the feature we needed. We are student, our hours are free :) Furthermore, it also teach us how to make a battery which is nice in itself.

Building it ourselves also let us choose the chemistry.

What we built

Why LiFePO4, and why bolted

Chemistry. LFP removes cobalt from the bill of materials entirely, which addresses the conflict-mineral concern raised in Living Within Limits. It also tolerates abuse far better than NMC, has a longer cycle life, and fails far less energetically, which matters on a vehicle that carries a person and will be charged in a garage. The cost is energy density, and on a vehicle designed around sufficiency rather than range, that is the right trade to lose.

Bolted, not welded. Spot-welded nickel strip is standard practice and cheaper to assemble. It also makes the pack disposable: one bad cell and the pack is scrap, or requires destructive disassembly. Bolted terminals with bus bars mean any single cell can be replaced with hand tools, which is what makes a twenty-year pack conceivable rather than theoritical.

This is the ecodesign argument from Chapter 7 applied concretely: the pack should fail like a mechanical part, one replaceable element at a time, rather than dying whole.

The pack is oversized, and here is exactly how that happened

Chapter 5 argues for small batteries: most daily trips are 20–50 km, and oversized packs carry a real mass and material penalty. At the 10–20 Wh/km this class of vehicle should achieve, 2.88 kWh is roughly 150–290 km of range. That is several times our own stated design point.

The pack was sized by charge current, not by range.

The chain of reasoning went like this:

1. We had never built a pack before, so we wanted comfortable margins everywhere to minimize the risk of having our first pack have rapid uscheduled dissably acompagnated by a thermal event.
2. The cells are EVE 20 Ah LiFePO4, rated 4C for 30-second discharge bursts. Cheap, and available in single quantities from a normal retailer which, as Chapter 7 argues, is itself the reason we could build this at all.
3. Regeneration matters on this vehicle, so the pack has to absorb what the motor can push back when braking: up to 60 A, or 2 × 30 A at 48 V ≈ 3 kW. Which may not be much in term of energy but does greatly reduce wear of the brake pads.
4. Playing it safe, we capped charge current at 1C rather than infering that the cell’s burst rating what for both charge and discharge on our first build.
5. 1C at 60 A means a 60 Ah pack. At 20 Ah per cell, that is 3P.

3P × 15S × 20 Ah is 2.88 kWh. The energy capacity was a by-product of a current requirement. Nobody chose 2.88 kWh.

The fix: three segments instead of one pack

We are thinking about splitting the pack into three independent ~950 Wh segments, each 15S1P at 48 V, rather than one 3P block.

Come argue with us about it.