Inconsistent testing results

I have purchased a few HW586 testers and over the last few days I have done some testing of a tester.

I am a bit disappointed with the results.

First a bit about the testing:
I used a single Sanyo cell (red wrapper, green ring)
I used the same HW586 tester, with 1 resistor, set to test down to 3 volts. I thought that 1 resistor might stress the cell less, and give more consistent results.
I used the same TP4056 charger
I tested voltage of the cell before the test and after the test. The time when the test finished until when the voltage was checked varied. I am not sure if the voltage may have bounced back more when longer times between testing finishing and voltage checking was completed.
The test was done in an undercover area outside in Sydney. Temperatures varied a fair bit between daytime and nght time. Charging and testing was done whenever I saw that the last step was complete, so some charging/tests were at night and some in the daytime.

Here are the results

Test Start Voltage Capacity End voltage
1 4.21 2752 3.55
2 4.23 2965 3.48
3 4.20 3029 3.48
4 4.20 2985 3.46
5 4.21 3058 3.43
6 4.22 2860 3.58
7 4.22 2890 3.59
8 4.17 2809 3.51
9 4.20 2937 3.52
10 4.23 2912 3.59

Highest result: 3058
Lowest result: 2752
Difference : 306

From looking the the cell database I think the cell is a UR18650FK which was originally rated at 2100mAh, so the results are already way too high. The high numbers would not worry me too much as long as the results were repeatable, then the sorting of cells would still be ok.


From memory, I think Pete (hbpowerwall) did some repeated testing of cells and the results also looked a bit inconsistent, but I dont think it was as much as above.

I am not happy that my testing could be this inconsistent. My plan to follow others process of binning in 100mah increments looks pointless if the testing results are so variable.

Is day vs night temperature going to make a difference (or that much difference) to the test results?
Is this amount of inconsistency normal in testers?
Is the hw586 I used no good?
Are hw586 testers in general no good?

Can anyone shed some light on this?

Ambient temperature does make a difference, but I don't think it would make a 10% difference. I found a between 3 and 4% difference on an Opus when I cooled the cells with a computer case fan vs. just natural cooling. And maybe a 1% difference between slots on the same charger due to heat. If your results are 30-40% above what they should be, then something is wrong with the tester.

300mAh difference between tests in one thing, generally way too high results is another. The inconsistency is a problem, the unrealistically high results not so much if you can find a baseline that you can work with. If this is connected however, i.e. inconsistent results because the offset differs, then you're pretty much screwed and will never get decent results that you can work with.

Example, just in case it was hard to follow my thoughts:

Let's assume the cell has 1800mAh of capacity left. The tester reports 1700 to 2000, so the results are inconsistent by 300mAh. And on top of that there is an offset by 1000mAh because of some reasons. If the offset is always 1000mAh then you only need to deal with results varying by 300mAh. If this is the case can be found out by comparing the results from another tester.
If it is, then this is still pretty hard to work with, because 300mAh can make a huge difference. I wouldn't worry too much about 50mAh or even 100mAh if necessary, but 300mAh is massive.

However, if the offset isn't fixed, then the results are completely worthless as you have, basically, at least one, but possibly two random factors.

IF ur test is 1000mAh above you dont havet it correctly setup. It should be within 100mAh from actual value. Since the value is that off then the error also multiply.

Fix that and you Will be within 100mAh spec.

Temperature do make difference to. And the bounce back need to be meassured same time after other wise you Will see what you see.


So... Fix the resistor

daromer came in between i was able to answer, and i must admit his hint that the hardware may be troublesome as well is valid.

What i would have pointed out is two general things (even when i repat myself)

1) Chargers are no measurement equipment. If you take Litokala, Opus or whatever, they are neither calibrated nor accurate at all.

2) The charging process itself is not a stable complete predictable thing. How the software in the chargers really defines end of (dis)charging, taking accuracy in accounts, is unknown. Just imagine the Voltage difference full/empty is roughly 1 Volt, makes 1 % as 10 mV as necessary "accuracy", when the current alone makes Voltage drops of 0.2 something volts. All with running switching converters.

Both can be prone to temp differences, circuit inaccuracies and several other parameters.

In summary i do not expect much from a charger, perhaps a little bit more than you have experienced. But a 5 % would never trouble me.

In your case, i can note one more thing - third is the cell itself - i learned on this board ( and have found that on my checks as well) that first loading of a longer stored cell maybe different. if you take the first measurement out of consideration, it gets better, but still not to the 5 % I mentioned.

All in all, thats one of the reasons because i have critizized the sorting into 100mAh groups, because the accuracy of the emasurements is hardly enough to do it accurate. On the other hand, the average you find in your group should more or less match the group values, so no problem on combining them into banks with regard to the groups.

Sorting in 100 ranges is the best way if we look at statistical outcome.... But you need to random otherwise it wont work. Im One of those prooved that in reality with many thousands of cells


But that require that the discharge test is done correctly and within that Range.

Sorting in 1mah Range and expecting that accurazy on otherhand Will never work out and its proven over and over again on packs by many users here

Sorting in 100mAh ranges is actually the right way to do it BECAUSE the accuracy of the whole process isn't good enough. At least for how most people do it. If you could then you should work with even smaller ranges.

I follow both your opinions, as it levels out statistical over a bigger number of cells.
Still, you may have underated cells dismissed and overrated cells taken.
Thats life.
Still, i am trying, as the OP, to vealuate better.