Glubux's Powerwall (2016)

There's a lot that goes into manufacturing battery packs beyond the cells. How's your thermal path to ambient in your home wall battery? How is the inter-cell thermal isolation? Is there a path for gas discharge in the event of a cell failure? Is the pack appropriately fused at the cell or module level? When a cell fails, does it take the whole pack with it, catch someone's apartment building on fire and kill a family of 5, or merely become stinky with a hotspot visible on IR?

How good is your cell acceptance testing? Do you do X-ray inspection for defects, do ESR vs cycle and potentially destructive testing on a sample of each lot? When a module fails health checks in the field, will you know which customers to proactively contact, and which vendor to reassess?

Yeah lots of batteries are 18650/26650 in a trenchcoat. The trenchcoats run the gamut from "good, fine" to "you will die of smoke inhalation and have a closed casket" in quality and I think that bears mentioning.

I get that the trenchcoat needs to be well designed and tested, but I am still flat out amazed that you both agree with “meh, most battery packs are made up of rechargeable domestic batteries you find in a kids toy”

I just assumed there was … special stuff in there

There's a lot of risk in creativity when you're selling crap to the public at scale. Way better to just use what everyone else is using.

For a highly engineered battery like a premium EV, there are coolant channels, temp monitoring, voltage monitoring, etc.

Soldering some connectors onto some random cells and knowing they shouldn't go over 4.2v is one thing, but measuring cell health via internal resistance, programming a controller to do temp shutoff and wiring up temp sensors, keeping cells balanced, is a lot of extra work, but critical if you at all care about not potentially burning down wherever they're stored.

Keeping the cells small and just using a hundred of them in parallel (and a hundred of these parallel packs in series to get up to the hundreds of volts needed), thus using ~10,000 cells, in EV batteries limits the maximum damage from one cell going worst-case, assuming your enclosure can contain it.

That being said, it seems there is a slow movement towards larger cells, from 18650 to 26650 or similar. But each cell on its own is still a dumb can of chemicals ready to go boom if you mistreat it.

Where would you put this battery in that trenchcoat gamut? Inside a server rack, fwiw. https://signaturesolar.com/eg4-lifepower4-v2-lithium-battery...

Was definitely one of the harder parts of our solar install to get comfortable with.

There are some pretty huge cells now like the 4680s

I used to joke with my buddy back when he first got his Tesla that we were driving around on "over 7000 vape batteries!", as that was the fad at the time and where most normal consumers recognized them.

Bigger, fewer, more chill cells, fairly robust trenchcoat.

(IIRC, these packs are 16 100ah LiFePO4 cells in a steel case w/ built-in fuse, breaker, and BMS that monitors individual cell health and pack temperature, w/ automatic cut-off if any of that goes out of spec. The weakness is primarily the MOSFETs on the BMS potentially failing shorted. Fortunately, they've added some sort of additional fire suppression beyond just "steel case" in recent-ish versions of these packs)

I can't see what the construction looks like but the mention of 'fire arrestors' gives me a lot of hope. If you haven't designed a battery that can take a cell runaway safely, you haven't done the work, and clearly they've done at least that much.

There's some optimization that happens in the chemistry and construction details for specific uses.

Also with bigger packs inter-cell consistency is really important (good cell integrators will test and bin them by ESR even if they're from the same lot, and using a really reliable cell mfg/vendor is critical because you're selling expensive systems with a number of failure points that scales with the number of cells and you want their process development to be super mature.

For most things cylindrical cells are the right answer. They don't puff up, they're available with protection circuits, they're cheap and highly available, you can get them in a variety of sizes and capacities, even in different chemistries.

Using a custom cell might make sense if you are making a) one megakajillion of a thing or b) you have extreme volume limits which mean you're probably using a pouch cell.

In HW engineering, Not Invented Here syndrome costs you big money. You have to have an actual business case for re-engineering something that already exists plus the capital.

95% with my stuff of the time COTS cylindrical is the answer, which means my shit comes in on budget.

Ah yeah, unfortunately I think we have the version before they added fire suppression, but at least it’s a more relaxed chemistry. Thanks for the analysis!

I think the previous version of this lacks the arrestors, unfortunately, wonder if they can be retrofitted. Thanks for sharing your take!

And really tiny ones, 10400, which is AAA sized.

Don't, uh, buy those unless you're sure.