^ hmm, I'm not sure I see how your scenario relates to what I was describing. What I was describing I think depends on at least two critical things:
1) at least one cell that's charged to a lower absolute capacity for whatever reason, such as at 1000mAh vs. others at say 3000mAh. Could have gotten there because it has faster self-discharge, or maybe by some other kind of mismatch/deterioration vs. other cells. The important thing is that it's charged to a lower capacity level.
And 2) the pack needs to be taken to neg recal. In the real-world this is a pack that's exhibiting 'pre-mature neg recal' - the BAT gauge de-populating by half the bars for instance and then plummeting to the bottom. This means the lowest cell or cells are getting near-fully discharged in the car. That seems to be the critical factor that leads to different charge voltage curve/behavior -- the voltage will be higher much sooner, it tends to stay high, and I think in some cases, maybe most?, it will reach the BCM's 'pos recal' voltage sooner (~17.0V at rest).
Basically, the other cells will have depressed voltages - their discharge curves will be saggy, their charge curves will be slow to climb. Meanwhile, the cells that discharge completely will have high voltages on both charge and discharge, the latter at least until it runs out of capacity.
I'm not really seeing these 'critical' features in the scenario you describe...
The first couple of cells that drop out were holding between 4-4.3 Ah, or 61-66% of their 6.5 Ah capacity. The second set that dropped out were holding between 5.1-5.5 Ah, or 79-84% of their 6.5 Ah capacity. I know from previous discharges that the 'strongest' of the cells in the pack will hold [about 6.2Ah]... I had grid charged the battery the previous weekend and the car had been used for a week
OK, so you discharge your pack with apparently a very small load, and the first cells drop out at about -4.1Ah and others at about -5.3Ah. And I guess you're using the ~6.2Ah value more or less as a baseline for what the best cells do when fully charged... Or maybe you're saying some cells in this pack would have discharged that amount if you actually kept discharging, or something. So perhaps there's 3 sets of cells at different absolute capacities.
I think I'm following so far...
The gauge was showing around 90%. We know that the car uses a charge window of 20-80%. So the car thought my pack was charged to around 70%, when in fact it contained a spread of cells between 60% and 95%.
OK, I think I see where you're going with this, it reminds me of something I used to do a lot. [Here's a link to a post where I talk a bit about that :
Ideas, facts, wild-*** guesses about background charge?]
Let me paraphrase and you can approve or disapprove of my understanding:
You grid charged to full, you use the car for a week during which time the BCM supposedly establishes what it thinks the charge level is (either your "90%" gauge level or your "around 70%" level). You then discharge the pack 'on the bench' and establish the range of actual charge levels - about 4.1Ah for some cells, 5.3Ah for others, and I guess 6.2Ah for others. The car 'thinks' the charge state is "70%," or rather
you think that's what the car thinks, while your 'on the bench' work says you have a mix of roughly 63%, 82%, and 95%. Good so far?
In general I think that, conceptually speaking, you're treating the "20%-80%" range and the "70%" value too literally, these aren't or simply can't be iron-clad thresholds. On the other hand, your approximately 63% charge state for your lowest cells is just about right, just about what you'd expect... i.e. if the BCM thinks the charge state is about 70%, then it's not very far off the mark, is it?.
No idea where those cells are in the pack, but let's say there's a single stick with a mix of 95% and 60%. On regen, the average voltage across this stick is going to look strong, because the voltage will already be high in the highly charged cells and the regen will top them out as they reach overcharge (it's going to quickly push them to 100%), pushing the average voltage in the stick artificially high. All of a sudden, the voltage will catch up in the 60% cell and push the voltage across the stick way higher than expected. So let's say that stick is the one with the cell holding the least charge, that effect is going to be the greatest and that is the tap which ends up with the highest voltage soonest.
I think this is where I lose you the most. For example, I'm not sure what you mean when you say, "On regen, the average voltage across the stick is going to look strong, because the voltage will already be high in the highly charged cells and the regen will top them out as they reach overcharge..."
Not sure what "look strong" means in this context... So... You have a charge-imbalanced stick, some cells at 60%, some at 95%, you hit it with regen, voltage for near-full cells will... well, first increase the fastest, then flatten out if the charge is allowed to continue.... I think you might mean 'total voltage' or 'tap voltage' when you write "average voltage"... The BCM's looking at tap voltage...
Not sure what you mean when you say "pushing the average voltage
artificially high"... And I guess the same with the rest, like "All of a sudden the voltage will catch up" and "higher than expected."
In general, I can see that what you're describing isn't the same as the idea I was getting at earlier.
I haven't done enough 'work' at the top-end to know with great certainty some of the finer management strategies... In general, lately I've been leaning toward tap resting voltage being the most critical element - what usually determines '75%' and 'too full', 17.0V for the former, 17.4V for the latter. Regen doesn't last forever, it's usually only transient - so most of the time, if the BCM sees either of these resting voltages, that's what rules (well, it's that and nominal charge state; for example, hit 81% nominal and you're done)...
On the other hand, I do think the BCM can/does use slope detection at the top as it does at the bottom, just not sure how that works, exactly. In your scenario, with the 60% and 95% cells in let's call it a tap, the 95% cells would have a very steep slope and thus the tap voltage would increase fast; I think the BCM detects that, measures that, and will disable regen, pos recal, whatever, when that happens...
There's all sorts of stages to 'determining empty' and 'determining full', like check points. On the empty side - which I'm much more familiar with - assist is throttled simply based on nominal charge state, for instance. So, assuming you can bring nominal charge state down to something like 36%, you're gonna see drastic throttling regardless of what the real-time data say (like tap voltage). On the flip side, if the BCM measures steep slope before that nominal - it's going to neg recal - call the pack empty - anyway.
There's analogues on the top end. For example, once background charge is invoked, the pack will keep charging until 70.2% nominal and then quit. That's solely based on the nominal. That's one 'check point'. And then, if you regen more, eventually tap voltage will hit 17V resting, some sliding-scale value under regen load, or perhaps that steeper slope is measured - and we call it a 'pos recal'. Another check point. And then, if you have one of the BCMs that pos recal to 75%, you can charge more, until 17.4V resting is measured or a nominal 81% is reached, whichever comes first, at which point you're really done. And that's another check point...
What does this have to do with your uneven 60% to 95% charged cells in your tap? I guess I don't see that 60% cell triggering any of the top-end management behavior --
unless it were the only cell that's been consistently drained to near empty. Draining to near empty is what elevates voltages, and if all the cells are drained to near empty then they all have the same elevated voltage profile - so no single cell will cause a steep slope or higher voltage sooner rather than later. I think your 95% charged cells would be triggering top-end management...