New 2P4S pack: Bad Cell?

[zandr]

I've been reading and lurking for a while, but I recently built a pack for my RV. Given available space in the house battery box, I went with a 2P4S configuration with MB31s. This will fit neatly in the space previously occupied by a pair of Group 31 AGMs. I'm using a 4S BMS from Overkill (JBD).

The Eve MB31s arrived from a US vendor at 3.295V, reading within 1mV, not that I trust my DMM to that accuracy. I assembled the pack as 2P4S, charged to 14.2V as a pack through the BMS, and then reconfigured to 8P and top balanced to 3.65V connected directly to the power supply. I only have a 10A bench supply, so this is a slow process, but it worked as expected.

I then reconfigured the pack to the intended 2P4S with the BMS, and discharged the pack. My electronic load is rated for 180W, but I just set the load current to 10A to make the math easy. I hooked up a PC to the BMS to log the discharge curve in hopes of calibrating the SoC in the BMS, but the results were a bit unexpected.

One "cell", really a 2P pair started dropping out a little early, and and then collapsed. The glitches line up with me messing with things/taking measurements with a DMM, but I wouldn't have expected that to disturb things quite so much. Even weirder, once that cell hit 2.5V and the BMS turned off discharge, the "cell" recovered. The pack provided 650Ah, running a 10A load for 65 hours. That's also a bit unexpected for a pack with at least one bad cell, though I guess these batteries provide more than rated capacity at low discharge rates?

"Bad cell" is the obvious answer, but I'm puzzled by the recovery, and the magnitude of the glitches when all I was doing while probing around with the DMM. If this were just a bad balance wire connection, wouldn't that affect adjacent cell voltages, and not just one?

I reconfigured the cells that were in "Cell 1" (OK) and "Cell 2" (bad) into a simple 1P4S pack to work out if either or both of the MB31s that comprise that cell are bad, and started charging through the BMS, again at 10A. (So C/30, not C/60 this time!) and so far all of the cells are within 3mv. I'm only up to 3.36V, and this BMS doesn't start balancing until 3.5V by default.

Any guesses or advice here? I'll report back when I see what these 4 cells do under discharge.

 

[sunshine_eggo]

Sensing error. Bad terminal crimp, etc.

Link #2 in my sig.

 

[zandr]

Sensing error. Bad terminal crimp, etc.

Link #2 in my sig.

That doesn't seem to fit the data, so please elaborate. (Note: I'm an FAE for a semiconductor company, but power electronics isn't my thing)

If it were a sensing error due to a bad balance wire:

1. Why wouldn't that affect the adjacent cell voltage?
   The BMS has five leads to measure four voltages, and the bad cell is mid-pack, so shouldn't that affect two cell readings? A flaky connection is consistent with the glitches, though.
2. Why would it "recover" slowly when the load is removed? There should be negligible current in the balance leads when the pack is at rest, and this BMS doesn't balance on discharge by default. That recovery seems like the 2P pair balancing itself, but I can't explain why it would have recovered to the same voltage range as the rest of the pack.

EDIT: Not saying you're wrong, just trying to complete my mental model of what's going on.

 

[sunshine_eggo]

When problem solving, you start with the most probable solution. Your graph shows the cell in question dropping voltage at about 50% SoC, yet it's able to support the discharge until the remaining cells drop to 2.8/2.9V.

This absolutely doesn't make sense. If you have a bad cell, the other must have 2X the capacity to support the bad cell. When things are this "WTF?" it's either something incredibly rare it's never been seen or is almost never seen, OR it's a common problem.

Once you've confirmed there are no connection issues, 1P4S testing is what's needed, and you're doing that. I would put the cells in question at the #1 and #4 position in a 1P4S pack.

"recovery" is normal. Discharged cells rebound when unloaded, just like charged cells sag when charge is removed.

Voltage and balance only correlate in the working range, which is about 3.1-3.4V. Inside that range, even imbalanced cells will have negligible dV. If you have significant dV, you have a sensing issue or a cell with abnormally high resistance.

Would set balancing to 3.40V, 20mV deviation and disable balance only during charge.

 

[zandr]

Yup, fair enough. I don't have any explanations for what it's doing beyond 'bad data', it just wasn't bad in a way that I would expect for that failure mode.

I haven't re-terminated the balance harness, but otherwise reconfiguring the pack required touching every connection. And if it's a bad wire in the harness the issue will stay on Cell 2. The cells in question are now 3 and 4.

I'll apply your recommendations now as the pack is still out of that range.

Thanks for the further detail!