Li-Ion Battery 'Guts' Video Lecture (geek alert!)

Carnegie Mellon Robotics Institute Video - "Lithium-Ion Batteries"

http://www.ri.cmu.edu/video_view.html?video_id=60&menu_id=387

June 2010 - Length: 1:15:00

Jay Whitacre, a Carnegie Mellon professor with a joint appointment in Materials Science and Engineering and Public Policy, gives a quick overview of Lithium-ion batteries. Starting at the chemical level, he explains the properties and mechanics of the battery which give rise to macroscopic behavior, especially focusing on issues relevant to electric vehicles and other high-power systems. He then fields questions about the batteries themselves, ongoing research at Carnegie Mellon, and tricks for improving long-term performance, such as pulse-charging and pairing batteries with similar defects.

Click to expand...

Wow, you weren't kidding when you said geek alert! Actually pretty interesting to hear how lithium-ion batteries actually function, even if the chemistry talked about might not exactly be the same kind as in the Leaf.

I watched the whole thing and really enjoyed it. Thanks for the link! I would have liked to hear more about Lithium CobaltOxide designs, but the LiFe points were very interesting.

For anyone who made it to the end -- I was surprised to hear him dismiss battery management systems that monitor each cell. I couldn't tell if he was saying it's not practical to bypass each cell, or even to monitor it at all. He mentioned the Tesla specifically as having only series pack-level monitoring. Anyone know for sure?

In my experience, no matter how well-matched a pack is when new, they diverge over time. If I didn't balance my individual cells (by gating specific cells during finishing charge), they would quickly fall out of line and push one into over/under-charge. Maybe I misunderstood him? Or maybe I'm buying cheap Chinese batteries.

GroundLoop said:

IFor anyone who made it to the end -- I was surprised to hear him dismiss battery management systems that monitor each cell. I couldn't tell if he was saying it's not practical to bypass each cell, or even to monitor it at all.

Click to expand...

I believe he was saying that cell level battery management (the ability to gate each cell) was possible and ideal from a performance point of view, but that it was not economical. More economical is to carefully match the cells into strings, and then do string level battery management.

GroundLoop said:

In my experience, no matter how well-matched a pack is when new, they diverge over time. If I didn't balance my individual cells (by gating specific cells during finishing charge), they would quickly fall out of line and push one into over/under-charge. Maybe I misunderstood him? Or maybe I'm buying cheap Chinese batteries.

Click to expand...

I think the latter.

Cheers, Wayne

GroundLoop said:

For anyone who made it to the end -- I was surprised to hear him dismiss battery management systems that monitor each cell. I couldn't tell if he was saying it's not practical to bypass each cell, or even to monitor it at all. He mentioned the Tesla specifically as having only series pack-level monitoring. Anyone know for sure?

Click to expand...

My take was this: Monitoring/managing each cell is the "right way" to do it from an engineering standpoint but in practice is very cost prohibitive to do in an actual product, so manufacturers don't do it. They instead rely on better matching of cells, monitoring only strings of cells, and apparently a little luck and bet-hedging. This doesn't seem unreasonable to me.
=Smidge=

You're right.

My last batch of batteries were 2.2Ah 11.1v (3S) packs for $8 each. How do they do that?? Not by cell matching, for sure. I balance each cell every charge, abuse the batteries with high discharge rates, and don't sweat it. If I wanted them to last 10 years, I'd be in trouble.

I'm getting 25Wh with a 10v cutoff, so that's only $320/kWh without management, for relatively small cells.

We still need cell-level granularity for minimal pack management, but if we parallel our 40 small cells first before assembling our series string, that 'parallel cell group' can be managed just as easily and effectively as if it was a single cell.

Look at the battery High Tech Systems built for the Killacycle, for example -- 100 series 8 parallel of A123-Systems 26650 cells. They paralleled eight cells first, and dedicated one BMS channel to that parallel group.

Yes - the geek in me would like a management chip on each cell that can signal problems and remove a bad cell from the string if it dies - but since parallel cells seldom die on their own 'in the real world', and since we can swap a cell later if necessary (unlike the folks sending robots to Mars), then parallel management provides much better bang for the buck with little down-side.

I'm with you, GroundLoop - cells will diverge over time. The absolute minimum protection needed for a pack is the high voltage and low voltage detection for each cell (in a 1P string) or each parallel group. It's more difficult to manage a string of cheap yellow prismatics, but even a string of high quality matched cells will diverge during charge and discharge.

These Thunder Sky LiFePO4 cells are highly honed pieces of art:

NOT! :lol: