My experience pretty much parallels what Steve has seen, and I haven't seen the case where a 0V cell can eventually "kill" other cells that are in parallel with it. I do still have packs that were constructed using cells that have been fairly well trashed in RC heleicopters, and some of the cells definitely were "stressed", with reduced capacities, but the packs still work pretty well. The trick is that you just can't let any cell, or group of paralleld cells, get down to 0V, or they die.
What I've found is that it is pretty easy to pick out stressed a123 cells. just by checking the voltage of the cells, about 10-20 minutes after a full charge. With healthy cells, a "surface" charge builds up across the electrodes so that if you measure the voltage, it will read somewhere between 3.50V-3.60V. As soon as even a light load is placed on the cells, even for just a few seconds, this surface charge will disappear, and the measured voltage will drop down to somewhere around 3.4V. Stressed cells drop down to somewhere between 3.30V-3.40V within a few minutes. the surface charge phenomena is pretty much gone.
I've not had a case using healthy cells in parallel first, and then in series, where a block of cells died, or even got stressed. I even had a case where I dropped a pair of pliers that ended up dead-shorting a couple blocks of 4 cells in parallel, in a pack I was building, and even though the short lasted long enough to burn some pretty big pits on the pliers, the cells only lost about a tenth of a volt, if that, and they did not get stressed.
the only time I have actually killed cells in an ebike pack is in a setup where I was acually using four separate 16s1p strings, that where then connected in parallel. One day, I inadvertanly forgot to hook up two of the four strings, and then went out on one of my "normal" rides. All of a sudden, about 2/3rds into the ride, I noticed the power suddenly died, like someone turned off the switch. I actually checked the controller wires, because I thought maybe I lost a phase, or something. but then I looked at the Wattsup, and noticed the pack voltage was down to about 37V. That's when I discovered two of the four 16-cell strings were not connected. I swapped these "full" ones for the drained strings, and bingo, full power again.
I ended up with several 0V/dead cells in each of the two failed 16-cell strings. Most of the cells were around 2.5V, but a few were down around 1.6V or 1.7V. The ones that were still up at 2.5V recovered just fine. The 3-4 that were under 2V ended up being "stressed". The few that actually now read 0V are truly dead.
What was interesting about what happened is that I had literally no nitice that anything was wrong. Literally seconds before the pwer died, I was going up a slight hill, and it was working fine. I coasted down the other side of the hill and then applied power, but nothing was there. The reason for this is because of the extremely high "C" rating of a123 cells, which are rated at 30C continuous and over 50C for 10 second "bursts". with even two of these 2.3 Ah cells in parallel, that translates to 138A continuous and a whopping 230A for 10 seconds. That means a typical ebike 40A peak load is nothing for even a 1p configuration of these. This means there is very little voltage drop, even under load and even with me only having two of the four strings hooked up. The voltage stay about the same, all the way up until about 10 seconds before the end. The more balanced the cells are, the less "notice" you get. Anyway, this experience is why I pursued doing the LVC board. Also, the reason I picked the 2.7V version of the TC54 detector chip si to try and catch the end-of-duration voltage "dumping" as soon as possible. The lower voltag version of the chip would still keep the 0V condition from happening, but I was worried that the lower cutoff voltage might let cells get stressed, even if they weren't killed. Bottomline is it works quite well. I can run my a123-based packs down to where the LVC starts tripping, and the Ah used readout on the Wattsup show the rated capacity has been used, sometimes a bit more.
With other LiFePO4-based cells, with lower C ratings, I think the 2.1V TC54 chip is the one to use. This is because the voltage drop under load is greater, so towards the end of the duration, the cell voltage could dip down to under 2.7V, under high loads. With my LiFeBatt-based packs, it takes 80-85A loads for this to happen, but using the 2.1V versions, I can run all the way down to where the LVC starts tripping, and the capacity used is usually right at 10Ah.
-- Gary
