So I finally got around to doing this. I found some used A123 26650 batteries on e-Bay and bought two 11S strings and cut off the circuit board that was on the original pack. The cells were drained to nearly zero but they did take a charge normally and I cycled them a few times and manually balanced them. I took a 8S section of one and hooked it up to my old SmartUPS 700. The float voltage from the built-in charger is 28.0 VDC and it charges at up to 4A at very low SOC, dropping the current as the voltage rises. That is well within the max voltage of these batteries. Normal charging should be up to 3.6V/cell or 28.8 VDC for the 8S pack. I left them for several days with the UPS supply holding the float voltage and it doesn't go anywhere, so that's stable. However, one of the cells droops more than the others, leading to about 220mV imbalance after 48 hours. Eventually, the cells that drop, would push others higher, over their max voltage. A balancing BMS would prevent this by bleeding off the high cells and forcing the pack voltage into the lower cells. The lower pack voltage actually allows a larger voltage range because most of the balancers have a fixed balance activation voltage depending on the target cell chemistry. The LiFePO4 BMS balance boards I've seen start balancing at 3.6V or 3.65V.
View attachment 349276
The load test was very interesting. I set up a load that lit up all the load meter LEDs on the UPS. That was about 385 W according to my Kill-A-Watt. When I pulled the UPS power cord, the battery pack voltage immediately dropped to 25V and slowly dropped to about 24V before I stopped the test after 4 minutes. The amperage draw, according to my clamp meter, was 23-24A. That is only about 70% conversion efficiency. I stopped the test because the batteries were getting hot. In another test at lower power I ran it all the way down to the UPS low voltage shutdown which happened at about 19.8V. The specs say these cells can take 50A max continuous discharge, but that would clearly require some active cooling or significant heat sinking, which is not practical to implement in this kind of UPS application. If I were going to put this into service, it would require a 8S2P pack as a minimum in order to ensure the batteries remained at a safe temperature. Either that or ensure that the connected wattage was lower. Doubling the cells would extend the runtime even further, so it is not really the high power, short runtime solution that I thought it would be.