How does Tesla balance cells, disconnect dead cells, etc?

[4LivesPerGallon]

Anybody provide reliable data (with links) as how the Tesla electric car disconnects dead, faulty batteries from the pack?

Does it balance cells? Strings? How?

If we knew this, we could all have Tesla-reliable BMS's, right?

 

[wineboyrider]

There is a thread somewhere that discussed exactly this idea....not sure of the link.

 

[kd8cgo]

I'm not sure on the thread... This is a very basic document with a high level overview of the Roadster packs. I'm sure there is more out there, I just don't have anything else off hand.

 

[Monstarr]

I wondered how those 2008 first gen Roadsters are performing nowadays.
Turned out they exceed expectations, 80-85% range after 100.000 miles.
http://www.greencarreports.com/news...esla-roadster-is-likely-better-than-predicted

Model S performs even better
http://www.greencarreports.com/news...ow-much-range-loss-for-electric-car-over-time
96% after 50.000 miles and expected to lose another 1% after each 30.000 miles.
That means more than 90% left after 100.000 miles

When is the tipping point?

 

[dmwahl]

All this is from memory, having seen an opened up Model S pack and BMS, and having designed commercial lithium management systems.

I didn't see anything to allow them to disconnect individual dead cells, I suspect they just let them ride along with the good ones. They're in large parallel groups, so a single bad cell wouldn't have a huge effect on the rest. Each cell has a mechanical fuse for over-current protection, which is basically just a wire welded between the cell tab and parallel connection bar.

Regarding balancing, they use TI BQ76PL536A-Q1 ICs (one for every 6S cell block), and approximately 40ohm of shunt resistors per parallel block. There are a couple of things that let them get away with such a small resistance for balancing. First, they use high quality cells, which are probably very closely matched to begin with. Second, they connect 70 or 80 of these in parallel (I forget the exact number). Assuming that the cells are well matched to begin with, the more parallel cells you have the closer to the population average you'll get. The end result is a large "cell" that, thanks to statistics, is a very close capacity match to it's companions. If you had 100% identical cells in identical conditions (obviously not realistic) you'd need no balancing after the initial one. The Tesla cells obviously aren't perfect, but are going to be closer than the consumer grade stuff most people are used to.

It's not just about the BMS. Cells, temperature management, load management, etc also play a role in overall pack reliability. Having worked with the same parts Tesla is using, I can say that we all could have a similar battery system to Tesla's, but few would want to pay for it.

 

[wb9k]

It's very unlikely that the pack can automatically disconnect cells that are welded in place, even if the BMS was capable of identifying ONE bad cell in a parallel group of many. I doubt there is any system out there capable of doing this. As other have pointed out, many cells in parallel dilutes the effect of a single failed cell. Still, if one cell fails bad enough, the most likely outcome is that some sort of error code will be set, telling the user to service the battery. Tesla service folks can likely identify the module at fault, and I would expect they would replace the whole module containing the bad cell. They may or may not repair the failed module after analysis for use as a service part. This is all conjecture on my part, but based on my own experience with OEM EV's and hybrids, and having seen Tesla's modules up close.