BMS design, based on AC techniques

At November's Seattle EVA meeting, Stephen Johnsen made a presentation and showed a video in which he short-circuited a 100Ah ThunderSky cell to test the claimed 2000A limit of the cell. During the test, the cell voltage sagged to 0.5V, but twenty seconds later when he threw the switch OFF, it popped right back up to 3.2V.

Since seeing that, I've been pondering the question of how to tell what a cell is really doing, while it's running in a system. The question that's been rattling around in my head is whether there's a way to tell the difference between a cell that's at 2.5V because it's under heavy load, and a cell that's at 2.5V ocv because it's fallen off the cliff and is in mid-plummet? I mean, obviously, there's the difference in the current going through the system under those two circumstances, but I don't think that's indicative. Let's ask the question a different way:

Is there a way to tell when one cell becomes 'reverse biased' in a running system? Let's say you've got 1/3rd of your pack charge left, and you start driving across Nevada on a flat level road at 50 mph. You're pulling a nice steady 100A off the pack, and you're just cruising along. 25 miles later, your lowest cell hits the proverbial cliff. How can you tell? You still see 100A going through the system, but now the rest of the cells are starting to shove current through one cell that's misbehaving. Is there any sign of this that can be detected before it's too late? Does the voltage across that cell and the current through it in any way indicate that the cell has become a resistor instead of a battery?

In the AC world, you could imagine looking at the phase of an AC device, and measuring whether its phase is leading or lagging the line voltage, to determine whether it's sourcing or sinking power. Is there anything like that on a DC battery? Can you tell by looking at "edges" in the current/voltage domain, to see where voltage edges lead current edges, and where they lag?

Would you learn anything by intentionally pushing a voltage ripple down the wire, and watching what it does as it flows through the system? Or would the capacitance and inductance of the cells just filter such a signal out? What if every BMS board had a PLL oscillator on it, that pushed a small ripple signal down the big pipe, and the next BMS board in the chain was watching that ripple to see how its own cell changed the shape of it? Could that tell you somehow what each cell is doing, and whether it's sourcing or sinking power?

I don't know the answer, or even enough about the field of batteries to predict what the answer might be, but it's an interesting question.

Charger / balancer co-design

In the process of upgrading one EV (lead acid-based) and planning a second (LiFePO4-based), I've spent a lot of time pondering the questions of charging and balancing the cells. I've come to the conclusion that these are not separate tasks. So I find it unusual in the EV community that the two topics are treated so separately. You buy your charger from company X, and then your balancers from company Y, and you pray that they play nice together.

Your charger comes pre-configured with a charge profile from the factory, and the balancer boards have the unsavory job of trying to fight against that charge profile to get all the batteries charged before the charger shuts off. What's going on here? I seem to be missing something.

Another possibility... which is the system being used on the lead acid EV I'm upgrading... is to install separate chargers on every battery. This is very practical if you only have 10 batteries to charge, and don't have to worry too much about discharge characteristics. But when you get into LiFePO4 territory, now you're talking about 45, 50, 90, 100 cells to charge individually? And still need a protection mechanism to protect them as they discharge? It seems like there must be some economies of scale that would be missed by installing separate chargers on every cell.

Yet, in many ways, having each cell charge individually is the ideal.

How can that ideal be accomplished with a series charger and balancer boards? What if the individual cell balancers know the charge profile of the cells, and direct the charging at a local level? The "charger" would become a slave unit, a cooperative partner that provides the total voltage requested right now by the balancers in the pack, with the current limited by the smallest current any single cell needs + the maximum current a balancer board can shunt (after some of them start ramping down).

So, it would be like you split the charger in half. Put "the dumb half" that has the AC windings and the DC regulator in a common spot, and spread "the smart half" that knows the charge profile out to each battery as part of the balancers. The dumb half shouldn't really be dumb... it should be Power Factor Corrected for efficiency, and it should actually be pretty sophisticated with tight control over the voltage and current being supplied to the pack. And the smart half could be pretty simple, using the simplest circuitry available to conduct CC/CV/float charging on a 3.2V cell.

The feedback from all those balancer boards to the charger base is the really tricky part in this equation, to make charging reliable and robust. I don't have a good suggestion there yet. I've been playing in my mind with differential op-amps, and wondering if they'd be sensitive enough to get it right without wandering off into wild oscillations or positive feedback loops.

Are there any chargers out there right now that match this design philosophy? I have heard suggestions that perhaps the Manzanita chargers do this, but their website is very "least common denominator" in this respect. About their chargers, the Manzanita website says (and I quote), "The actual power delivered is a function of input and output voltage." Whereas I'd like to see something that says, "The actual power delivered is a function of input and output voltage, the needs of the individual cells as they charge, and the capabilities of the balancer boards."