I've been looking at the probable Model 3 and Powerwall/Powerpack 2 architecture based on the pictures wk057 provided of the P100D pack architecture, and I think I've previously overestimated the amount of cells needed per pack. Which also means that I've underestimated the capacity per cell, at least for Tesla Energy.
First some assumptions, I've assumed that Tesla will be using the same cooling architecture as the 100 kWh pack, which means cooling in between the cells, but significantly more compact coolant loops. I've also assumed the powerwall battery contains two modules, identical to the ones found in the Model 3 (eight in number), totalling around 700x700 mm.
I estimate that a gen 3 module would have 30 cells in each row and 16 rows, 480 cells in total. Probably arranged in a 12s40p, but potentially it could be arranged in a 48s10p configuration. This means the pack module could have two cooling loops just like the 100 kWh pack.
Both the more traditional 12s40p configuration and a 48s10p configuration offers some interesting properties, especially the 48s10p configuration, in my view.
12s40p
In this case, the Powerwall has a 100V battery voltage with two modules in series. The powerpack can have a 800V voltage with 16 modules in series. (~800V jives with
AudubonB's observation of 900V markings.) The Model 3 would have a 400V pack with 8 modules in series.
48s10p
In this case, the Powerwall 2 could have a 400V battery voltage with two modules in series, which means you don't need a lot of DC conversion to reach the specified DC output voltage. The powerpack can have a 800V operating voltage with a 4s4p module configuration. The Model 3 could have a 400V pack with a 2s4p module configuration. This allows for different pack sizes without doing anything to the number of cells in each module. You could have one pack with six modules (2s3p) and one with the full eight modules (2s4p). (On the largest pack, you could also have support for 800V charging, if you could switch between a 4s2p configuration and a 2s4p configuration.)
Each module would end up slightly larger than a gen 1/2 module, roughly 690 mm x 330 mm vs 685 mm x 280 mm. And the total Model 3 pack would be roughly 1.6 meters by 1.2 meters, containing 3840 21-70 cells.
A powerwall would contain 960 cells, which means that with 14 kWh, it would have 14.5 Wh/cell. This would mean a ~34% improvement in battery chemistry over the Powerwall 1, which is a lot, but not impossible, as the NMC chemistry hasn't been close to the NCA chemistry in terms of energy density. There could be significant untapped potential there.
Assuming a ~10% improvement in chemistry for the NCA cells, or 20 Wh/cell, the base 2s3p pack could have 57.6 kWh, and the top spec 2s4p pack could have 76.8 kWh (this would be the same ragerdless of pack configuration.) This would work out to ranges in the area of ~250 miles and ~330 miles. I know this is less than some people are hoping for, but it's still acceptable for Tesla, and the available data supports it.
The benefit with the 48s10p configuration is the following:
One module architecture across the Powerwall/Powerpack and Model 3 with two pack sizes, changing only the cell chemistry and number of modules per product. This should simplify production substantially.
