Cycle life test of a Chinese battery, updated 29.03, 450 cycles, test complete

I’ve often come across comments along the lines of “Chinese batteries die quickly.” I generally agree, but the question got me interested, so I decided to look into it more closely—in particular, to run a cycle-life test. More on that in the article.



A few introductory notes.
This article will be updated regularly. At first, I’ll add information every 10 cycles; after 50 cycles, updates will come every 50 cycles. Each cycle takes about 3.5 hours, so initially updates should appear roughly every day and a half, and later about once a week.
Also, since I only have one electronic load and this test will take a very long time, I’ll try to time its use in other reviews to coincide with the end of each 50-cycle stage. The battery will be “resting” in the meantime.
The date of the latest update will be added to the title.

Now, on to the details.
Not long ago, I posted a review of some batteries I bought for a friend. I’d already handed them over, but then I got the idea of running a cycle-life test, so we decided to set one aside “for experiments”—that is, to systematically “kill” it.

So I met up with my friend and picked up one battery.


The battery is styled after the Panasonic NCR18650B. It has a flat positive terminal and is brand new; there are no marks on the negative terminal from being installed in a holder.


For the test, I used an electronic load familiar to my readers, an equally familiar holder, and the battery under test.
Partly because I can only use a basic holder for now, I chose a battery with a flat positive terminal. Anything else simply wouldn’t fit.


Before starting the test, I measured the battery’s internal resistance, since it tends to increase over time.
I plan to check it again every 50 cycles.


For obvious reasons, I couldn’t leave all this on my workbench for so long. I was going to set it up in the room where I shoot videos, but changed my mind and eventually decided to put the test “rig” on the balcony, in the cabinet that houses my home “server,” router, and storage. The temperature inside the cabinet is roughly room temperature, so the results should be representative of normal household conditions.


I connected the load to a computer installed in the same place. All control and monitoring are done remotely via Radmin.


To speed things up, I wanted to run the test in charge/discharge mode at 1C, but the datasheet for the original NCR18650B specifies charging at 0.5C and discharging at 1C. It also specifies discharging down to 2.5 V and charging to 4.2 V, with cutoff when the current drops to 65 mA. I can’t meet only that last requirement: the minimum current I can set on my load is 100 mA.


The first stage of the test is to determine the baseline capacity. To do that, I discharged the battery to 2.5 V at 3.4 A, charged it, then discharged it again. That capacity is the baseline.
The result is 2998 mAh, or 10.06 Wh.

This result surprised me a bit... If you remember, I tested four of the batteries I received, and they had a capacity of around 3350 mAh. I even started checking the electronic load, thinking there might be a problem with it, but no—the battery really does have a capacity of around 3000 mAh. Still, there’s no point turning back; I’ll test what I have.


The load can run a program in a loop. You set the number of cycles, and the results window shows both the test number and the program step number. As I mentioned above, I set it to 50 cycles and started the test.


As it turned out, I can already post the first result after 13 complete cycles, but the next updates will come after 20–30 cycles, and so on.


21 cycles.




30 cycles. I made a classic mistake here. You may have noticed that the capacity seems to have increased slightly in the latest tests. In reality, of course, it has dropped. But since the load is inside a cabinet in a room whose temperature is more affected by the weather outside, the likely explanation is that the air temperature changed by about 12–14 degrees over the course of the test. This affected both the battery itself and the accuracy of the load’s current setting. The test started when it was about -3 outside, then the temperature dropped to -11–12, and now it’s +2.
The difference is small, but it’s there. Of course, I won’t stop the test, since that 2% error isn’t all that significant compared with the overall expected drop in capacity.


40 cycles have already gone by, and the graph looks brutal :) The temperature conditions are roughly the same as at the very beginning of the test, so we can now say the capacity has dropped by about 60–65 mAh.


50 cycles. The first stage is complete. The battery will sit idle for a day or two, since I need the load elsewhere. In the meantime, here are the results from this stage.


The resistance has increased slightly. The instrument doesn't care whether the battery is charged or discharged, but just to be safe, I charged it to the same level as at the very beginning.
It was 35.6 mΩ; now it's 37.3


As I mentioned at the very beginning, I'll post the next results every 50 cycles. The 50th (100th) cycle finished overnight, so here they are.


The resistance hasn't changed much; the difference from the previous measurement is just 0.1 mΩ.


150 cycles.


The resistance hasn't changed at all.


200 cycles. I had a minor mishap here: with 11–12 cycles left, the power went out—and it was a pretty serious outage—so I had to restart the test. Luckily, I save the results to a file periodically, so almost nothing was lost; I entered two results as averages.


The battery's internal resistance has increased, but there could be two reasons for that. I'll see what it looks like after the next stage.


250 cycles. The capacity has dropped noticeably, and the temperature at the end of the test was about the same as at the very beginning.


Internal resistance.


300 cycles. Toward the end, the weather warmed up, and the results started creeping upward. Most likely, this will be the penultimate test, since it'll be hard to maintain the same temperature as at the very beginning from here on, or I'll have to recalculate the results, which I'd rather not do (

By the way, I expected the tests to gradually get shorter in proportion to the drop in capacity, but that hasn't happened in practice. Discharge time has decreased, but charging time has increased, especially in the final stage, where the current is around 100–150 mA.


The resistance has barely changed.


350 cycles. You can see the results starting to change more actively as the air temperature rises. As before, I'm noting the results where the temperature was roughly the same as at the very beginning, so I can later plot another line, straight from the minimum to the maximum.
Also, I did start another run after all, so we'll definitely reach 400 cycles. I originally wanted to go all the way to 500, but I'm having serious doubts.


The resistance is increasing slowly, but steadily. By the way, I noticed that the overall test time is getting longer, and charging takes a very long time. Each cycle now takes almost 4 hours.


400 cycles. At this point I was already thinking about stopping the test, but in the end I decided to keep going. As I mentioned above, temperature has a big effect on the results, and cooler days are just around the corner, so I can take a reading when the result is closest to the actual value.


The battery's resistance is gradually increasing.


450 cycles. I didn't think I'd make it this far, but here we are :)
Unfortunately, the test was interrupted twice. After 10 cycles, the power went out, so I restarted it; then, 20 cycles later, the power went out again. As a result, the screenshot only shows the last 20 cycles. For tests like this, an uninterruptible power supply is definitely a must, if only for convenience.


The internal resistance has increased noticeably. 450 cycles is quite a lot, after all.



Results in text form.
Two entries have been added: -1 and 0. The first is how much charge was drawn from the battery initially, without charging; the second is the baseline capacity.


And as a graph. You can see how the graph keeps fluctuating up and down toward the end, with large swings in daytime and nighttime temperatures. That's why I was planning to stop the test early, even though my original plan was to run 500 cycles. In the end, I'll get at least 450, which is still good :)

I added data for the last 50 cycles to the graph and also made a linear graph. The second graph is based on the reference points I recorded during the test, when the air temperature where the battery was placed was roughly the same as at the start of the test. This gave me a capacity loss of about 1.03 mAh per cycle.



But then I got curious: what if the initial capacity wasn't 3006 mAh, but actually 3244 mAh (at a 1C current), as I measured in my review of these batteries? I think that's what it really was, and that the capacity drop at the very beginning was caused by the temperature change.
After all the tests, I set up the rig on my desk and measured the test battery's capacity under the same conditions I used to test these batteries for the review.

The result was 2762 mAh. Compared with the initial 3244 mAh, that's a capacity loss of 15%, which I think is pretty good. By the way, if you calculate the result using the table I kept throughout the tests, you get the same 15%, so it all checks out.
By the way, Panasonic specifies a capacity loss of up to 30% for its batteries after 500 cycles. I think my result was better because I tested the battery at a lower temperature than the manufacturer does.


I learned a few things from this whole test and figured out what mistakes I made. If I do more tests like this, I won't make the same mistakes again.

That's it, the test is over. I would have liked to reach 500 cycles, but I don't see much point, and to be honest, I'm getting a little tired of it. I hope you'll forgive me for giving up so easily :)