Sanyo NCR18650GA 3500mAh: a new cycle-life test, or how to kill a battery quickly. Updated 20.06.2020

Yes, friends, the cycle-life test is back, and this time we have a genuine Sanyo NCR18650GA battery. But this isn't just a test: we're testing it under conditions the manufacturer doesn't recommend—in other words, we're deliberately killing the battery and seeing how it holds up.


The batteries were bought by one of my regular readers at the Batterex store in Kharkiv, batterex.com.ua, specifically for this cycle-life test. We agreed on the test parameters and battery model in advance.

There were five batteries in the package: a pair of Sanyo NCR18650GA cells for the test and an LG HG2 as a "bonus" for me. I'll review that one too, but a little later.

They were shipped by Nova Poshta and packed in nothing but a piece of cardboard, though it was reasonably secure.


Along with the batteries, the package contained a store business card with a discount code for the next purchase, valid until the end of the year, as well as a business card for... a butcher shop :)


I've already reviewed these batteries, and they were sent by the same person—many thanks to him. That time, though, the batteries were unused but not new. This time, I have cells I can confidently add to my "database".

The capacity is 3450mAh, with a minimum of 3350mAh (at 25°C). The Rated capacity is listed as 3300mAh (at 20°C), while the seller's page says 3500mAh. So take your pick; I'll be measuring them myself.


They look like the familiar Sanyo batteries: a distinctive red color and a blue ring around the positive terminal.


A distinctive feature of batteries like these is their hard-to-read marking. It's printed on both the negative terminal and the heat-shrink wrap, and it's especially difficult to make out on the latter. But if you look closely, you can read:
Sanyo
NCR18650GA 9405

The last four digits indicate the production date. In this case, that's April 5, 2019, so the batteries were just over a year old at the time of the test.


Dimensions and weight.


As usual, I ran a cycle to "normalize" the batteries before testing. By the way, I noticed two unusual services on the store's website:
Charge before sale
Charge and boost capacity before sale

I can just about agree with the first one, although if someone is buying a battery, you'd assume they already have a charger. I'm very skeptical about the second, though: there's no point in "boosting" cells like these. Any difference would be within the measurement error of the instruments.


The standard charging current is 0.5C (1.68A). The current from my charger fluctuated a little toward the end of the CC phase, but averaged out to the set value. Charging took about 2 hours 50 minutes, compared with 2 hours 40 minutes for the previous cells.


The manufacturer included some nice graphs in the datasheet, but for currents that aren't multiples of the battery capacity. I don't think that makes much difference, but I had to measure at different values.


I usually use currents that are multiples of the battery capacity. In this case, that's 0.2C, 0.5C, 1C and 2C, plus the maximum of 10A, which is essentially close to 3C.
At 0.2C, I got a capacity of 3320mAh, or 12.07Wh. The test was run at about 19°C, so the result is consistent with the manufacturer's stated Rated capacity of 3300mAh (at 20°C).


Battery temperature at 2C (6.7A) and 10A. Interestingly, this battery got a little hotter than the previous ones. For example, back then I measured temperature rises of 27.6 and 44.5 degrees, while this time they were 36.6 and 55.7.


The second unit; the results are similar



The test results for the two units are so similar that I even made a comparison graph at currents of 0.2, 1C, and 10 A.


And here is a comparison of the new batteries with the previous ones. In both cases, I used the second unit; the comparison is also at currents of 0.2, 1C, and 10 A.


The battery's impedance before testing, after the “normalization” cycle, and at the end.



And now, the reason I went to all this trouble.
The manufacturer provides cycle-life test results at a charge rate of 0.5C and a discharge current of 6 A. The charge current is listed rather oddly as 1.68, 0.99 A.
In any case, the person who sent me the batteries was interested in testing them at 3/3 A. I suggested 1/1C, i.e. 3.35/3.35 A, and that's what we settled on.


For the test, I used the same EBC-A10 electronic load as last time, but I took the previous mistakes into account and put it somewhere with minimal temperature fluctuations. It's basically in the center of the apartment, in an unventilated room (a storage closet). I also used a proper four-wire battery holder, and recalibrated the load before the test.
Control and monitoring are handled over the local network, using the same computer as a year ago.


Last time, the tests ran with a charge current of 0.5C, a 2-minute pause, a discharge current of 1C, and a 1-minute pause.
This time, I started the test with a charge current of 1C, a 1-minute pause, a discharge current of 1C, and a 2-minute pause.

As you can see, the pauses were the same length, but arranged differently. In my opinion, the pause after the discharge cycle is more critical, since the battery is still warm afterward, and then charging starts at full current. Once charging is complete, the battery is cool because charging ends at a very low current, so the pause isn't as important.

I initially set it to 50 cycles, but noticed that the capacity was literally melting away before my eyes. I talked it over with a friend, and we decided it might make sense to increase the pause. Cycle 24 was just finishing at that point when the power went out.

Yes, before starting the test, just in case, I set the protection trip threshold very high for the whole apartment, but the power was cut off throughout the entire building....


In the end, I kept the ratio between the pauses the same, but increased both pauses tenfold. The settings now look like this: charge at 1C, pause for 10 minutes, discharge at 1C, pause for 20 minutes.
Everything would have been fine, except the electricians cut the power again right at cycle 50...


Since I had to stop the test anyway, I decided to “avoid doing things twice” and checked the battery's internal resistance. As it turned out, it had increased noticeably, from the initial 22.8 mΩ to 29.0.

But the most interesting thing is that increasing the pauses had no effect on the battery's capacity loss from cycle to cycle.


All right, I set it to 25 cycles and resumed the test, but apparently I hadn't run out of problems yet. During discharge on cycle 22, the electricians showed up to replace the meter. There was about half an hour left in the cycle, followed by a 20-minute pause, but who was going to wait? Honestly, I was speechless. They replace the meter once every 16 years (or so I was told), and they had to show up right now.
I hope my EBC-X arrives soon so I can give it an uninterruptible power supply.


By the way, I want to point out what the charge cycle looked like on a relatively worn battery: charging at 1C now takes almost as long as charging a new one at 0.5C.


After that, things went more smoothly; cycles 72 through 100 completed without any problems.


I did make a small mistake—or maybe it wasn't really a mistake, but more of an oversight. Each run of mine ends with a discharge cycle, which is very inconvenient when measuring resistance.
So, during the pause on cycle 6 of the 101–150 range, I pulled out the battery and measured its resistance. It had increased quite a bit more, now reaching one and a half times the initial value: it was 22.8 mΩ, and now it's 33.9.


Throughout the test, I record all the data: cycle number, charge capacity in Ah/Wh, and discharge capacity in Ah/Wh, so I can plot various graphs later. But I can already show you something at this point.

The arrows on the graph mark interruptions to the test:
1. Red — power outage and change to the pause durations
2, 3. Blue indicates a power outage.
4. Green indicates a load calibration check and, after a few cycles, a measurement of the battery’s resistance during the pause, i.e. without interrupting the test.

You can see that changing the pause duration made no difference. What interested me more was that at a certain point the graph became a straight line. I even checked the calibration during the pause after the hundredth cycle: with the current set to 3.35 A, the multimeter read 3.355 A, so everything was working correctly.
Then I remembered that the datasheet also shows a flat section like this; it just occurred earlier because of the high charging current.

By the way, note how smooth the graph is. In this case, that’s thanks to the stable test conditions. I attribute the brief spikes to the load warming up and interruptions to the test.


So, after 150 cycles in total, I stopped the test. I could have continued, of course, but I don’t think there’s any point.


Capacity dropped to about 1200 mAh from the original 3250 mAh, while the charging cycle time barely changed, remaining around 2.5 hours.
Resistance increased to 44 mΩ from the original 22.8 mΩ. Even after just 20 minutes of discharge, the battery heated up to about 45 degrees.


Now for the final graphs.
Capacity change in mAh: red shows how much went into charging, blue how much was recovered.


The same data, but in Wh.


So, what can we say? By the end of the 150th cycle, the battery had really lost almost 2/3 of its capacity. I suspect the high charging current. As I mentioned earlier, the second battery is being tested with the same charge and discharge current settings, i.e. 1C (3.35 A), but both pauses have been increased to 1 hour.
The first cycle is almost finished. I’m curious to see how it performs, since the first sample may indeed have been damaged by the very short pauses.


Sorry for the delay; I ran into a few problems, so I’m publishing the report for both sections at once: cycles 1–50 and 51–100 for the second test subject.
The first section completed 49 cycles, then the power went out, but overall things went relatively smoothly.


By the end of the fiftieth cycle, the battery’s resistance had increased to 30 mΩ.


Cycles 51 through 100 went without a hitch; the electricians didn’t get in the way.


But resistance increased by another 10%, reaching more than 33 mΩ.


And here’s the most interesting part: the graph. The first sample is blue, the second red. Since testing of the second sample is still underway, its graph isn’t complete yet, but it’s already making for grim reading. The test took almost three weeks in total. For roughly the last 5–6 days, the temperature rose slightly, by just a few degrees, but that could still have affected the battery’s service life.
So this is where things stand now.


Another 50 cycles have passed, and the results are far from encouraging. Our electricians got in my way again, but all the discharge tests ran correctly.


Impedance rose to 54.5 mΩ, noticeably higher than for the first sample.


As a result, the charging cycle has turned into a nightmare. Almost as soon as charging starts, the battery voltage rises to 4.2 V, after which the current starts to drop. The whole cycle takes more than two hours!


But the discharge graph is, of course, more informative.
It shows the batteries’ internal resistance at four points: right at the start and then every 50 cycles.
The arrow marks the change in air temperature. It was stable at first, but rose sharply in early June, which naturally affected the temperature in the apartment as well. The difference was about 5–7 degrees compared with the temperature during testing of the first battery and the first 75 cycles of the second one.


Now for the conclusions.
1. A high charging current definitely affects battery life, especially if it exceeds the manufacturer’s specified value.
2. The pause duration between cycles makes almost no difference. The key is not to let the battery overheat. So a 20-minute pause after charging is enough, and after discharge even 10 minutes is plenty; I think it could actually be reduced further. The thing is, the battery is cool at the end of charging, unlike at the end of a discharge cycle, and heat is the enemy of batteries.
3. Temperature. It’s very important. The upside is that at higher temperatures, a battery can deliver more energy; the downside is that this comes at the cost of a shorter service life. The graph shows that the second battery held up for a while, but then started to give out and dropped off much more sharply than the first toward the end. Of course, the sample size isn’t very large, but still.