Battery innovation · Open Pack Designer
Design a battery pack from the cell up. Layout, mass, fast charging and a simulation of every cell, with the evidence behind each number.
4,416 cylindrical cells, 96 in series × 46 in parallel, about 76 kWh. The engine builds it from one parts list: 446 kg against 480 kg published, with no tuning.
See it in 3D →Panasonic 2170 NCA, 4.8 Ah, 70 g, from a cell teardown. Every one is run on CellBench's electro-thermal model in the pack twin.
Zoom into the cells →250 kW fast charge, 10→80 % in about 25 min in the model. Checked against three published Supercharger sessions: all 11 timings and powers within tolerance.
See the HV bay →Glycol-filled aluminium ribbons run between the rows of cans, inside full-height gel potting, as the teardowns show.
See the cooling →Battery explorer
This is the engine's own geometry for the Model 3 Long Range pack, not an illustration. Drag to turn it, scroll or pinch to zoom.
What the designer does
Enclosure, mass and bill of materials come from the same parts the 3D view draws. Pick prismatic or cylindrical cells, set voltage and space, and every arrangement that fits is listed, smallest first.
BMW i3 pack mass −2 %, Tesla Model 3 −7 %, both untuned.
The pack twin runs each cell on CellBench's equivalent-circuit model, wired through the busbars and cooled through the hardware the engine drew: discharge, charge, fast charge and eVTOL missions.
A real CALB 58 Ah cell, fitted from one pulse test and one slow discharge, predicts 30 drive cycles on 10 other cells to 10.6 mV on average.
Each result says whether it was measured, from a datasheet, assumed or borrowed from a similar cell. Anything never compared with a test is marked not validated, never a pass.
Pack-level thermal and safety results are not yet validated against pack tests.
Through life
Each cell carries its own lithium inventory and both electrodes. The model tracks SEI growth, lithium plating, and active-material loss on each electrode, so it shows why capacity is lost, not only how much.
LG M50T, 17 development-test cells: lithium and active-material loss within 2.5–3.3 points RMS.
Fast charging plates lithium first along the electrode edge, where the anode overhang is thin. A two-zone model, sized from the overhang measured in CT, gives the charge rate at which each cell starts to plate.
Behaviour check only: not yet validated against measured plating.
Ageing coefficients do not carry over between cell types. One cycling test at a single condition calibrates a new cell; every other temperature, window and rate is then predicted.
On 74 held-out Sandia cells: 3.3–7.8 points of state-of-health error, against 7.6–12.7 uncalibrated.