I see the logic for using both 3.2 V and 2.8 V as the testing termination voltage. OTH, I have tried to duplicate the data sheet conditions when testing. To basically verify the manufacturers claims. So if the data sheet says 2.5, that is what I use. But even the data sheets from different manufacturers and even different cell types from the same manufacturer can have different termination values listed. It certainly would be nice to have a standard that everyone uses.
Would that be as measured under load with sag or OC?
How could the light measure the open circuit voltage without turning itself off for long enough for the cell to recover?
I have always thought that LVP was triggered when the voltage reached a specific value under load.
1 Thank
load, flashlight has no way of detecting OC. Since they usually start throttling around 3V, however, it might be pretty close to OC.
Realistically a single value cut off voltage is “good enough” for most purposes, but I do think it is a little crude.
I don’t have a good suggestion, but would a 2.5V cutoff and some function related to discharge rate (C) work? I guess it might have to be a % of max discharge else you’re going to have strange values for very high discharge current cells.
I’m not sure if it works this way, but I was thinking that for buck drivers, there need to be a 0.3-0.5 V difference between the battery voltage under load and the voltage that the LED requires under load?
For higher outputs the sag can be a few hundred millivolts and the LEDs can require regulated 3V or higher to maintain reasonable output, so with these two combined the light may kick into unregulated reserve at fairly high OC voltage. In which case the higher cutoff may more realistically represent the battery’s regulated runtime capacity (low current/voltage reserve notwithstanding)?
According to Mooch, comparing discharge graphs rather than single-value indices may be the best way to predict runtimes at a range of currents. I think I like his conservative choice of 3.2 V, and the consistency that it provides for his Wh ratings. Squeezing slowly even more juice at low SOC is a little like a bonus safety/reserve feature?
2 Thanks
I am so annoyed. Yesterday I wanted to check in on the discharge progress of the JP40 at 0.5A, accidentally misclicked and cancelled at ~3500 mAh.
Then I restarted and during the night we had a thunderstorm that knocked power for a second. It stopped again.
JP40, 0.5A measurement, the third. Beginning now.
1 Thank
Per aspera ad astra ! 
For some reason it gives me comfort to know that this kind of thing doesn’t only happen to me… 
1 Thank
Yeah, Mooches e-scores are at 3.2V, which is also a metric I will give for comparability. The nice thing about 2.8V discharge - I can simply cut my data at 3.2V for a e-score compatible metric. 2.5V is already “damaging cells” territory for some cells, nothing I’d like to push.
Haha 
Yes it is, but some data sheets list that level in the specs. I certainly don’t use that limit unless the data sheet specifies it…
1 Thank
Ampace JP40 is done, EVE 58E is currently running.
EDIT: Replaced plot on moochs request 
I added an “E-Score” calculation, which is the same as my total discharge energy, but cut off at 3.2V (same voltage @Mooch uses for E-Score ratings). I am not sure how exactly it compares to his measurements. If they contradict, take mine with a grain of salt. I have way lower quality testing gear 
In this case I am a bit higher (he has 13.7 Wh E-Score for JP40 at 5A), but we will see once I have more data of different cells.
You can clearly see the JP40 is a high discharge cell - barely any difference between 0.5A and 4A curves. Not really the right cell for my setup. It needs a beefier test load (check Moochs review). Still quite interesting to compare once data for other cells I have is ready. The steep drop-off from basically 3.4V down to 2.8V is interesting. I have not seen that with any of my other cells so far.
I will restructure the start post these days, move all old data into an archive and add the new data.
EVE 58E seems pretty interesting so far, IIRC I was around 5400 mAh at 0.5A, 4.2V to 2.8V. Curious to see how it scales to higher currents, and how it compares to the LG M58T.
5 Thanks
Thank you for posting your results!
Always great to see more testing done.
I wouldn’ t worry much about the differences between our results, it’s less than 5% for the JP40. What’s a lot more important IMO is that each of us is being very consistent with our testing so all my results can be compared against each other and all of your results can be compared against each other. That’s where our testing is most valuable for others IMO.
I’d like to ask if you could use something other than my “E-Score” name for your energy values, to help reduce confusion. Especially with mine being switched over to a 2.8V cutoff soon. Thank you!
3 Thanks
Thank you! Okay, will do, I will rename it, and add a footnote saying it’s a comparable metric to yours.
1 Thank
Thank you. 
I swapped the plot of the Ampace for one that does not contain the word “E-Score”. Also some minor formatting improvements to the plot itself.
EVE 58E measurements are done. This cell packs some serious capacity. LG M58T is now running.
3 Thanks
Start post updated with some new information. Moved the old tests to a “Legacy test” spoiler. Click here for the list of all results (currently JP40 and EVE 58E): ebastler's low current battery tests
Is there a CV phase at the end? It’s odd that different current graphs converge to one point.
That is quite odd indeed, but no, no CV phase. The charger keeps going at nominal discharge current until the LV threshold is met, then stops. Seems like this cell simply hits the same capacity at different currents, and only the voltage (and thus energy) changes. This would definitely change at higher currents.
1 Thank