diy solar

diy solar

New LF105 cells seem to have a sudden drop in capacity, bad cells? Something else?


New Member
Oct 23, 2023
Downeast Maine
First off, I'm off grid, so I don't have the ability to easily charge batteries besides through solar and doing a full discharge is risky because it leaves me without power if I get the weather predictions wrong.

I bought 8 LF105s from ezealco (here), hooked them up to a JK BMS, set conservative numbers on a Victron MPPT and have been using them for about 2 weeks. They arrived all at exactly 3.291 volts, had perfect QR codes and untouched packaging and appear to definitely be brand new cells. I didn't top balance because the volts were identical and the JK BMS has active balancing. Even at a peak of 3.45 it hasn't been more than 0.033v delta between them. So while maybe they need more balancing, they can't be far enough off to effect total capacity in a significant way.

Up until a few days ago, the lowest I had drained them according to my shunt was down by 20 or 25Ah from fully charged. I've had at least 2-3 times where they got all the way up to 3.45v (my absorb setting, which I hold for 30m for balancing) and I reset my shunts based on that. Because I'm off grid, with a single battery, doing a capacity test is hard to schedule -- especially in the winter (I'd have to do it in the morning right before the sun came up, get it finished before peak solar, and then be able to reclaim enough storage to get through the evening and possibly the next day depending on the weather).

In any event, in the last couple days we had limited enough sun that I was finally effectively capacity testing them (albeit it at low C rate). This evening, the battery was suddenly and unexpectedly down to 24v (my LVD). My shunt indicates about 65ah used; JK BMS says about 55ah (BMS seems to be off calibration because it consistently reads about .03v lower than the SCC, the shunt, or my voltmeter).

In either event I should be nowhere near 24v. I'm clearly in the knee and falling off rapidly and by-the-voltage it's obvious that the battery is pretty well fully depleted. At the moment the BMS hasn't got to its cutoff yet, but cells read from 2.939 to 2.973V.

So, it APPEARS that the capacity would be around 55-65ah (depending on whether the shunt or the JK BMS is more calibrated). Bad cells? All of them equally bad? This doesn't make much sense.

Further, in my Victron BlueSolar history it shows that yesterday (after 2 days without any substantial solar activity) the battery reached absorb (8*3.45V which I have set to run for 30 minutes to support the balancing function on the JK BMS). This makes utterly no sense, as neither the shunt nor the jk bms SOC climbed in any significant way yesterday; they should have needed 55-65Ah to get to absorb and should be full then. So, it seems like the batteries voltage raised unusually.

Based on my history with lead acid, that sort of behavior (reaching absorb prematurely with no actual Ah going in first and depleting prematurely) would indicate a failing battery. But I can't imagine that Ezealco sent me 8 perfectly identical bad cells.

Is there a chance this could be something other than bad cells? And if it is bad cells, how do I make that case to the manufacturer (particularly with being off-grid and having limited ability to bench-charge or discharge in a super controlled manner)?
Curious if there's some chance that the cells were never fully charged -- I've been doing an absorb at 27.6v (3.45vpc) for 30m (one day though the victron charge controller got stuck on absorb for hours though). Is it possible that I need to take the volts closer towards 3.65V initially to be sure they're truly fully chargedz? I had thought that by definition, if an LFP cell is at 3.45 it's 100% (or close to it) SOC because the rapid knee and that any charging about 3.45 does nothing but add wear and tear.

Also these cells are tested by EZealco and they provided a spreadsheet with actual tested capacities, so that should mean they were full cycled at least once already...
Try setting them to 3.5V/cell and hold them there until charge current drops to 5A.
@Partimewages On one day by accident it was held at 3.45 for a few hours. But, in general, I was following off-grid garage and other conservative values that indicated 3.45 for 15m up to 1hr was plenty to ensure balancing, and that going over 3.45vpc served no purpose other than to trigger balancing. Are they incorrect or am I missing an important detail?
@AntronX I will try that tomorrow, weather permitting. I'm still a bit confused though because the sources I had read when setting up the SCC indicated 3.45 was a good conservative number for the cells once in use.
Crossing my fingers that this is just a problem of being insufficiently charged in the first place. But I'd still like to a find a source that explains why they'd not be at 100% SOC at 3.45V. Is it surface charge or something (the chart from EVE looks like it should be darned close to 100% at 3.4 and elsewhere on the forum people indicate that 3.0-3.4 represents the nearly 100% of the capacity).

So much research and yet seems like I may have made a rookie mistake. Just would like to find a good source to clarify this so I'm not stabbing in the dark here and actually learn a bit.
Post a picture of QR codes of your LF105 cells. Were they perfectly flat on the sides?
That QR looks smaller than one on my cells.


  • 20231211_204629s.jpg
    153.6 KB · Views: 5
I assume that your Cells were not fully charged. It would be nice if you had some data of actual Cell voltages while they are in absorb. You can watch the absorb and see when the current into the Cells tapers off to less than 5 amps. That will indicate a full charge as long as your Cell delta is good.

Possibly some of the Cells were above 3.4 Volts and some were below. If a Cell is below 3.4 volts then it is not fully charged. If all of the Cells are at 3.45 Volts and you hold that voltage for a few hours then they should be charged. The Cell delta should be around 5-10 mv. If you see the Cell delta up around 100mv then some of the Cells are not balanced. Again, data is very important for proper diagnostics.
Last edited:
I'm highly doubtful there's anything wrong with the QR codes. The phone somehow blurred or couldn't capture one in the corner.

Definitely working way up to being sure of a full charge. But it will take days with the weather. I drained them to disconnect and am instead counting amp hours on the way up. And yes, previously I monitored voltages in JK BMS app and had reached above 3.4 on every cell (and closer to 3.45 on almost all). Cell delta also was about 10mv, but at the floor it varied more.

Anyhow it's going to be a few days before I find out whether they take in 100Ah or not. I've increased my absorption and will keep an eye on the absorb time and try to keep it going until my shunt claims 105Ah or the current drops.

The hard part is doing the capacity test by draining afterwards.

Also the voltage calibration was pretty far off on the JK BMS. I had avoided changing it because I wasn't sure what was more accurate the BMS or the other devices, but since inverter, charge controller, voltmeter and shunt all agreed it was +.3 higher, I went with the consensus. Amps are still off and it's terrible and registering low amp draws. But SOC on BMS doesn't have to be perfect, I trust the shunt to be more accurate.

Will report back any discoveries when solar charging is finally complete in a couple days (winter in Maine with a small array).
Ok, so here's the update.

On the charging, I had to push my SCC almost all the way up to 3.65VPC to get the shunt to report up to 105Ah. The JK BMS itself has terrible current estimation. It's low resolution, unsteady, and jumps all over the place -- plus it seem consistently off by a good amount. The voltage is also noisy under load (and it's the BMS sensing, not the cell itself -- my voltmeter is rock steady in its readings). Connection are all good as far as I can tell.

Anyhow, so I did push it up from a reset of 0ah @ 2.5vpc up to 105ah@3.65vpc according to my shunt.

After this I had to wait for the right conditions to do a discharge test. The discharge test not only didn't get a full 105ah (95 at best), but it seemed like the voltage drop at the cell under load was totally out of spec. I was doing a discharge of 40A (so about 0.4C), beginning at about 56Ah remaining according to shunt (or 50ah according to BMS). The voltage dropped to an average of 3.072VPC (totally jumpy in BMS though totally steady with voltmeter).
Now, according to my spec sheet, at ~50% under 0.5C load I should have been easily above 3.2VPC.

Screenshot from 2023-12-20 15-26-26.png

Of course, any time I stopped the load the battery would immediately recover its voltage to a more expected value.


But they shouldn't be dropping like this under a 0.4C load to my knowledge or according to spec sheet.

In any event, this discrepancy only increased. The cells hit my 2.5V under voltage protection when the shunt read 11Ah. I could eek out a few extra Ah by lowering my discharge to 10A.

Had I been able to supply a 0.5C load (I don't have anything in my house that can consume that much power), I'm pretty sure it would have cut out much earlier. With the 0.4C I got a measured 94Ah. The first 50% discharge was done via lower C rates across several days of cloudiness at my usual usage level (which is only maybe 10ah a day and no more than 100watts ever).

The same thing with seeming to follow an exaggerated curve under load is happening on charge: it reaches 3.45 or 3.5V under maybe a 0.1-0.2C charge when it's at like 75Ah according to shunt (and lower according to BMS). I can only get it up to 105Ah by increasing the absorb up 3.65V or holding absorb for many hours. It's like the voltage response whether during charge or discharge is exaggerated relative to what LiFePO4 should be. All of this to me seems like bad cells, but I can't see how they would send me a bunch of perfectly equally, clearly brand new bad cells, specifically when they supposedly provide a three year warranty on the cells and a spreadsheet with tested values for each cell's serial code all within spec.

So, now I guess I have to try to convince them the cells are bad or follow whatever test procedures they want to have this verified. I'm also pretty leery of continuing with the JK BMS. Not sure if I got a dud or what -- it was way out of calibration on both voltage and current and seems to have tons of noise on the cell voltages under any significant load.
Note that the unusually low voltage relative to SOC and the final cut off voltage under load was verified at the cells themselves, so while there could be voltage drop issues in the system as a whole, that variable should be eliminated from the test results. Voltage at cells was steady and predictable, just way lower than it should have been for 0.4C at 50% (or even 40% adjusting for final measured capacity).
Capacity tests are done at multiple temperatures (thus the included chart for temperature). This cells particular data sheet only happens to show 55c/25c/-20c lines at 1C.

13C is definitely not -20C. Sure, it's a bit under 25C (less different than 55 to 25 though). But also, I'm doing 0.35C or 0.38C to be precise, not 1C nor even 0.5C.

Further, the cells were warmer by the time I reached cut off. About 20C. Here's a screenshot a few minutes before cutoff with 18C:


So, yes, it's not 0.5C@25degC test conditions, it was probably something like ~0.38C@~20degC. I'm pretty sure it's a bit of a wash or a couple Ah at worst.

But I'm not splitting hairs between 103Ah and 105Ah here, I'm seeing results that are not within the ballpark of the SOC chart at a discharge rate below that provided in the product literature (for example, see the above screenshot with 2.874VPC average -- clearly well into the knee, despite reading 24Ah remaining on the BMS and maybe 30Ah on the more accurate shunt at the time).

My goal is merely to determine if there is something substantially under-performing about these cells. I'm not looking for a lab test here, I'm trying to determine whether I got bad cells.

(Also note, the temperature sensors are just resting on top of the cells in free air currently and haven't been taped to the cells; the actual cell temp was likely a good bit higher than the 20degC reported at the moment of LVD).
Note you can see the erratic readings from the BMS there too -- the voltmeter didn't report that much variation. Nor does the shunt have the same amperage. The BMS is way out of wack on calibration of voltage/current and a heavy load exacerbates that -- this variation isn't seen on the shunt (which is after the BMS or the voltmeter which is before).