Yes dendrites could develop later on. But the process of dendritic growth is quite slow and gradual.
Because of that, before the cells start getting too hot and threaten thermal runaway, what will happen is that the pack will run out of balance faster than the 40mA my BMS can balance. And shortly after my BMS will trip protection and I will be safe. This is actually why BMSs have a lower balance voltage - 1mA is of course not enough but they tend not to exceed 50mA so they will catch damaged cells very early removing all the risk.
As for deep discharge, you might enjoy this paper : https://www.nature.com/articles/srep30248 which also includes a short literature review.
The TL;DR is that damage is relatively limited during deep discharge even to 0v. There is plausibly possible capacity degradation but it not very different from accelerated wear and tear, though the authors did not detect any.
What is really dangerous though is discharging a battery pack to 0v - when that happens it's easy for some cells to gain a negative state of charge, which is when the nasty stuff that actually causes fires and damage happen.
Copper dissolution according to this paper in more modern batteries does not even really happen until -12% state of charge. Indeed in modern higher capacity batteries they found no detectable damage at all until -12% soc.
Because of that, I am very confident from my setup - the risk of internal short circuit is incredibly low given that I know that the cells were always in balance, and any slow-forming internal short circuit will trip the protection much before there is any fire risk.
It's very interestinf though from reading the paper that the deep discharge risks essentially disappeared at 0v from pure coincidence as capacities increased though, I'd have expected somewhat the reverse :)
Agreed that your setup is pretty safe though, especially if you have already cycled the pack many times and see no deviation between the cells.
I've figured out the layout, it - unfortunately - won't be possible to put the pack in the frame triangle, there are too many details that will become messy, space is constrained, mounting options are limited and wiring the pack would be a nightmare with a lot of odd groupings of cells which would have a lot of potential for mistakes, so I'm going with the 'boring' option of making a rear carrier based box that will house the battery. A port at the bottom will connect a cable with a custom made plug from an old pack to the port on the bike, which can then still do double duty to charge the pack.
From a center-of-gravity perspective it's not ideal, but it does allow boosting the pack to a full 200 cells in 10S20P configuration, which should be more than enough to do my runs and have room to spare. Can't wait to get it tested :)
>I'm not quite sure what 'modern' means in this context but all of the batteries that I've seen so far (Samsung, LG, Sony) caution in their MSDS against deep discharge both from a functional risk as well as from a safety perspective and there are a number of pretty knowledgeable people online who are on the record with their reputations at stake cautioning against this
Basically by modern they mean higher capacity than the ones in the 2006 study that found dendrites at 0V - those were early 2000s era tech, shoddy 760mAh cells. They found that the higher the specific capacity, the higher the resistance to Cu becoming dissolved.
As far as constructor experience I do agree with you, it's preferable to avoid it. Though one factor is that people with extensive commercial experience will probably have had to work for very cheap, low capacity 1600mAh cells that are definitely more prone to this, and quite old arrays. I definitely recommend against it too, it just seems to me that it won't result in a fire anymore. I wouldn't take the risk though with a cell I know was discharged to 0 or beyond.
I should do a write-up on battery pack failures, I've collected a small mountain of data on this, it might be useful to others.
Many thanks for the rear carrier information, this is something I will have to really deal with somehow, the frame gap on this bike is just too bloody small and I don't want to permanently alter the bike in case I want to sell it. I've done quite a bit of cycling with kids on the back so I'm used to having a lot of weight there (and one on the front too, total bike weight in excess of 130 Kg), but it's something that will have more effect the faster you go.
What's definitely out is an off-side pack or to have it split across the carrier bags, that's asking for trouble, both from a mechanical point of view and an electrical one (you'd be running quite a few wires between the two sides). At least a carrier one is going to be centered on the bike.
Cells are arriving tomorrow, for now I've decided on a 10x20 grid (so a few more cells ;) ), 21 x 41 cm plus some room in front to house the BMS. the box will be fairly low, about 90 mm altogether so that should help keep the COG low, the carrier is rather low on this bike so on the whole it might not be that bad. I'll be sure to be extra careful with wet roads and I categorically refuse to cycle when it's icy (I have a bunch of steel in one leg from a previous accident so falling again is really not an option).
At least my adapter works, so I won't need to butcher the bike wiring harness.
If you want to take this out-of-band: [email protected]
So that's why you really need that Bosch BMS. Then the bad news: the Bosch BMS is on the edge of what's safe, the FETs are underpowered (just a hair under rated maximum current and voltage), not cooled, there is a 40V Schottky diode there which regularly operates at or even slightly above it's ratings, there is a single chip DC-DC convertor which also isn't cooled properly and which tends to blow. So I don't particularly trust the Bosch BMS from a safety perspective, I've seen a whole bunch of these now (about 15) and a good number of them had blown up. And that's with the standard battery. The one that I'll be hooking up to it is four times more powerful. Any short circuit protection in the old BMS is going to be terribly undersized so I need something in series with the Bosch BMS to ensure that if there ever is a short (say in the motor) that it's the battery that gets disconnected well before the wiring goes and I simply do not trust the Bosch BMS not to fail 'closed' in which case the whole thing will turn into a giant fuse. That's the kind of scenario that I can really do without.
On more than one pack that I've opened I've found evidence of small fires, typically BMS balancing wires that shorted out (there are no fuses there, obviously) and that could have ended quite badly. The wires from the pack to the motor are substantially thicker so then the question is what cuts out first: the BMS or the wires. I simply want to play that as safe as possible and aim for the disconnect to happen fast enough that the wiring doesn't light up.
Another option is a very fast fuse in line with the battery terminals, that's definitely a possibility as well (and easier to mount and less chance of it in turn breaking than yet another BMS in series). Anyway, work in progress and to be continued, nice to find someone on HN who has a lot of knowledge about this!
Waiting for a while to see if there is a self discharge is a very good method to spot flawed cells before you assemble a pack but such defects can also develop at a later date, so please do keep a good eye on your pack, especially if - as I suspect - you are using reclaimed batteries, you can't be 100% sure that they have never been deep discharged.
Safety is one of the main reasons why I decided to go with brand new high capacity brand name cells, yes, it's a lot of money, but even so, the whole thing will cost about as much as a 600 Wh commercially available pack for this bike and have more than 3 times the capacity (and a better form factor to boot, and will last a lot longer).