Results of Capacity Testing 18650s for 1,600+ Cycles (2023)

**Go to the bottom of this post for my most recent results** 1,605 cycles as of August13, 2020

I'm going to be dedicating one of my Opus' to longevity testing some cells. It has 4 slots, so I am going to be testing 4 cells. Originally, I was going to only test generic cells, but I came up with the idea of testing a genuine cell as the 4th cell to act as a kind of "control." The "Generic" cells will be the ones that I have the most of: (1) CJ blue 2000mAh cells; (2) ASO purple 2000mAh cells; and (3) THLD blue 2000mAh cells. These are all new-old-stock cells from around 2011.The fourth cell will be a genuineone that was originally rated for 2200mAh and tested with almost new capacity. I was thinking of using either an LGAAS31865 from 2010;a Samsung ICR18650-22F from 2012; or a Sanyo UR18650A from 2015. I'd rather avoid using the Green Sony's or red heater Sanyos from before Sanyo was bought out by Panasonic. I'm leaning towards going with the Samsung, but let me know if you think some other cell would be a better representative of the "genuine" cells most people use for their power walls.

I'm going to use the Charge Test function at 1000mAh and record the capacity for each of the cells after each cycle. These will all be full cycle 4.2V to 2.8V discharges, so it's a lot more abuse than a typical powerwall will give these cells, but it will also help speed up the results. I'm going to aim for at least 100 full cycles.

I know that many people open up generic packs and are surprised that cells are testing at their rated capacity. I also know people question their longevity like in the articleMike/LithiumSolar cited posted by BatteryHookup. Also, I think hbpowerwall was longevity testing some Chinese cells, but I don't know what happened to that.So I'm hoping to, as scientifically as possible, give an answer to the age-old question: "Generic" Cells, are they ok to put in my powerwall?

Let me know if there is anything else I should consider before starting this test. Once I start cycling, I'm not going to make any changes.

EDIT: (1/30/19)I ended up going with the LG battery for the genuine cell. The Samsungs retested below 2000mAh and I want to keep the capacities as close as possible. Also, adding a picture of the cells being tested and their first capacity test.

EDIT: (2/11/19) Started cycle 34 this morning and after breakfast came back to find the Opus fan not working. Before unplugging the Opus, I tested the cell temps. They were between 65 and 72*C! One spot on the back of the Opus was 85*C! I guess the overheat protection on this unit is not working because the cells were all still discharging at 1A! So I waited for everything to cool and when I plugged in the Opus, the fan kicked on again like normal. Anyway, I went ahead and discharged the cells and entered an asterisk in the log for cycle 34. Hope the high heat doesn't affect the test results.

EDIT: (2/12/19) For cycle 35, I ran a 120mm computer case fan blowing air over the cells from right to left. I was worried that the Opus fan might not come on, so this was mostly for backup. Well for this cycle, I thought either my Opus was toast, or the cells were permanently damaged, because Cycle 35's results, as compared to Cycle 33's results went like this: LG - 3.8% lower; CJ - 3.4% lower; ASO - 3.0% lower; and THLD 3.8% lower. Then for Cycle 36, I took the case fan away, and ran the test like normal. Results were back to normal ranges, and for Cycle 37, were almost identical to Cycle 33 (within 1mAh!). So looks like no permanent damage to either the cells or Opus. I'm carrying on with the test just like before. I'm just really shocked how much cell temperature can make a difference to the test results!

EDIT: (2/17/2019) Hit a milestone today, 50 cycles complete! Took a little longer than I'd like, but I'm making progress. Cells are changing slots again in the morning for Cycle 51. Nothing else new to report.

EDIT: (2/27/2019) Currently running my 79th cycle! That's like over a kilowatt of electricity put into and taken out of eachcell :-DMy new ETA for results of 100 cycles is Saturday, March 9th, 2019. Not really an update, but I did look at the LG datasheet, and turns out I'm testing it like they recommend, go figure! 1 Amp charge and 1 Amp discharge, but they do say to test at 500mA to get true capacity? The data sheet says that the LG Battery should have greater than or equal to80% of its capacity after 300 cycles, not 500 like other name brand cells are good for. This was not a new cell, it had about 90% of its capacity left (2200mAh * 0.9 Opus correction factor = 1980mAh?). Based on that, it should lose about 6-7% of its capacity during this test by Cycle 100, and be at about 2070mAh. We will see!

EDIT: (3/9/2019) As promised, here are the results!! I have run 105 cycles, 5 more than the 100 set in my goal. The reason for that was to get an average of the first 5 cycles and the last 5 cycles, and compare those. Because I was switching slots every 25 cycles, the heat on the right 2 cells was affecting the results a little. Cells have more capacity at higher temperatures, and a few degrees Celsius seemed to make a little difference. That's why I used the averages ofCycles 101-105 in comparisonto Cycles 1-5, as those cycles were all in the same slots. Ok, here are the results:

And the degredation:

So with that, obviously, I have to run this test to at least 205 cycles or until I get bored of it. And it seems like I have more questions than answers at this point. To me, I'm not surprised the LG lost the most capacity. It started this test at about 90% capacity when you factor in the tester correction factor. Also, according to the data sheet, it is one of those 300 cycle rated cells, not a 500 cycle one. I was expecting 6% degradation to be honest, so it did better than I thought it would. Now onto the generics. Why did they do so good? Is my Opuswearing out or something? It was surprisingly consistent throughout this whole test. The jagged lines on the charts are only because my resolution is between 2100 and 2220. If it was 0-2200, it would appear very smooth. I have some theories about why generics and regular laptop cells die fairly quickly. Mainly (1) keeping them at full charge most of the time, which I have seen in other tests that itabsolutely kills the cells quickly; (2) temperature in a laptop is usually pretty warm as there is usually no active cooling for batteries and computers are known to generate heat; and (3) especially in generics, the BMS's are terrible. Almost every generic pack I have opened uses thin gauge aluminum wiring, nickel-plated steel for nickel strips, some lack a thermister for temperature monitoring, and the spot welds are usually really easy to remove and countless packs have a disconnected cell in them. Bad quality control might be why people avoid generic packs. But the cells seem to be well manufactured. Time will tell.I will have to keep this test going to see if capacities will fall off a cliff. Or maybe, just maybe, generic cells are actually good. P.S. I did take pictures of every cycle's results and I have the results of every test written down, if anyone is interested, I can upload that as well.

EDIT: (4/2/2019) Just wanted to post an update that this test is still going and I'm running Cycle 175 as I type this. I was going to do an update around Cycle 130 but didn't feel I needed to. Around Cycle 130, the spring furthest to the right on my Opus started sticking a little and I missed one result. Cell fully discharged and a few minutes into the recharge, it went "null." Happened a couple times, but the other times I wrote down the results before taking a picture. Anyway, I took apart the Opus, put a little dielectric grease on the track, and everything was fixed. It's been uneventful since that time. I'm starting to see some degradation in the THLD cell, and the LG has declined the most, but I was expecting that since the LG was a previously used cell not at 100% capacity when the test started. Anyway, I'm hoping to post the next set of results by April 15th.

EDIT: (4/15/2019) Ok, it is time to release the next set of results. Finally seeing some degradation! Here's the chart:

And a little closer look:

You can see a definite trend. Just going to repeat a few things if you don't want to read from the beginning. Since I am rotating cells every 25 cycles, and the right two slots get warmer than the left two slots, when a cell is on the right side, it benefits from heat adding capacity, and when it goes from slot 4 to slot 1, the capacity takes a hit. Also, the missing data for the THLD cell in Cycle 129 is from the slot being sticky and going null on the recharge. I didn't get a reading, but when I took the cell out and put it back in, it was at 3.29V, so I know it had gone through a full cycle. Also, because I am rotating every 25 cycles, I wait until 5 cycles into the next rotation to compare to the first 5 cycles, so that there is no slot selection effect.

With that being said, here is the degradation observed over 205 cycles. The first row is the average of Cycles 1-5 and the second row is the average of Cycles 201-205 for each cell. The third row shows the percentage of degradation of eachcell.

Please keep in mind that the LG cell did not start out this test as a new cell, and already had a number of cycles or cycle equivalents on it before the test got started. The other cells are 8 year old cells that were never used. And as most of usknow, degradation is not linear, it accelerates.

Anyway, I'm continuing this test. I just finished Cycle 213. I would have posted this earlier, but I didn't really have a chance.

EDIT: (6/3/2019)Sorry for my leave of absence, it says my last login was the 22nd of April. Ouch. Just took a break from this forum and used forum time for other stuff. I did keep this test going. I've dedicated my Opus to this test for 1/3 of a year now. Wow. Yes, I just finished my 350th cycle! Didn't think I would make it this far, but I'm really glad I did. Here are some results:

And here is the degradation (second number is Cycles 301-305)as calculated by the method mentioned by Wolf back in April. Makes more sense to do it this way. Assuming I reach 405 cycles, I will use this same formula.

It seems the more data I gather, the more questions I have. Like, why am I seeing such small degradation? Is there something wrong with my Opus? With my testing method? For example, the data sheet for the LG cell rates the cell as 80%+ remaining capacity after 300 cycles. The LG cell that I'm using in this test came from a used Laptop pack that had an unknown number of cycles and cycle equivalents. And yet, 350 cycles after an unknown number of previous cycles and like 8 years after the cell was manufactured, the LG cell is still doing great! Oh and the generic cells are not doing bad either. I mean the THLD cell has seen 6% degradation, and Cycles 348-350 were the first times this cell saw less than 2000mAh, but this is like a generic cell that nobody has heard of, unlike a number of people who have CJ or ASO cells. I was expecting 100 cycles before this cell was toast and it's nothing like that.

EDIT: (6/23/2019)Time for an update! Just finished Cycle 405, and here are the usual charts I post. I can see that the CJ cell is starting to accelerate its degradation, but the LG and the ASO are still doing great!

And here is the degradation chart:

I feel like I really have to go for 505 cycles just because 500 seems to be the magic number that most cells are rated for. Will also probably put together a video with pictures of all the cycles once I hit that mark.

On a side note, I also started safety testing my generic cells. Not the ones used in this longevity test, but other cells from the same manufacturers. I finally learned what PTC does when I was testing for CID. Turns out that when you try to short out a cell, or use it for a higher discharge rate than what it is rated for, you will not trigger CID, but rather PTC. I was getting strange results shorting my cells, but that was the PTC protection working. Spoiler: CJ and ASO have PTC protection. I'll save the rest for another thread in the future. Stay tuned!

EDIT: (8/10/2019)505 Cycles! Here are the results:

And the degradation:

One thing that these capacity tests do not show is IR. Even though the LG cell is doing pretty well, it definitely has the highest IR. It takes the longest to charge, and it has the highest "bounce" voltage on the first test in the morning. The THLD cell, though it lost a lot of capacity, still has the tightest voltage and is the fastest to charge. I could literally do 4 tests a day if it was only the THLD cell, but the LG cell is slowing me down. It's getting to the point that I have like a half-hour window to take a picture of the third result of the day between when the THLD cell is fully charged after its third cycle and the LG cell finishes discharging its third cycle.

EDIT: (9/9/2019) 605 cycle update: Degradation on the CJ and THLD cells looks like it is accelerating. The LG and ASO cells are still degrading fairly linearly. The test is getting a bit slower, I'm doing about 2 cycles a day instead of the usual 3 cycles. I'm going to continue this test, but I'm starting to get a bit fatigued by it. I have been doing this since January!

EDIT: (10/14/2019)Update time! 705 cycles. Chart is self-explanatory. We are now past the 70 or 80% that every datasheet I have ever seen usually stops at. I am also re-uploading the degradation numbers at each 100 cycle interval. I was doing the math with the Wolf method, but was getting over 100% degradation, so I'm reverting to how we all do state of health: tested capacity over initial capacity. I'm starting to wonder if the THLD cell will make it to 805 cycles? I'm also considering changing the Title of this thread to "Long Term Capacity Testing Cells." I think a lot of people who would benefit from this information are skipping over this thread because they have an uninformed negative bias toward generic cells (I used to be one of those people).

EDIT: (11/24/2019)

First cell to be completely degraded: CJ! It has been testing under 100 mAh since Cycle 776, but it is still putting out about 64 mAh for 30 straight cycles. I think I'm done testing this cell at 1000mA, so now that I have an official number for Cycles 801-805 for my data, I'm going to run the cell at 500mA discharge for Cycle 806 to see what capacity it has at a lower discharge rate. And maybe 300mA for Cycle 807 and 200mA for Cycle 808 if it doesn't take forever. Then after Cycle 808, I'm going to fully charge it at a 200mA charge rate, and test its voltage at weekly intervals to see if it self discharges. The THLD cell looks like it's going to be fully degraded pretty soon, too, so I think I will do the same thing to it after 5 or 10 sub-100mAh cycles.

You may notice that the ASO and LG cells seem to be dipping quicker than before, but I don't think they are degrading any faster. I think they are showing lower capacities because the CJ and THLD cells are not generating heat next to them like they did before. I wish I had temperature correction like rev0's equipment, but I don't, and this is one of the downsides of the Opus.

Also, like any true research, data creates more questions than it answers. For instance, the THLD cell started dropping capacity much earlier than the CJ cell, but the rate of decline was flatter. The CJ dropped like a rock, from about 1000mAh to <100mAh in just 75 cycles. Could it be the chemistry? Electrolyte? Cobalt content? Age?

I also wanted to post this chart. It shows how much energy each cell has stored in its lifetime. The CJ cell was able to store 5 kilowatts of energy in its lifetime!

Lastly, I just want to mention that I have received the cell cycler from rev0 and am excited to start another long-term test after I do some other testing like high-drain generics and possibly some end-of-life (50% of original capacity) genuine cells to see their degradation much more quickly than with new cells. We'll see. Stay tuned?

EDIT: (2/5/2020)I've finally hit over 1,000 cycles. Took 53 weeks, but here we are! And sorry for not updating at 900 cycles, I had everything ready, but it was right as the New Year hit, and I never got around to posting it. Right around 900 cycles is when the THLD cell ran out of steam, so I ran that one up to Cycle 922. Cycle 923 was at 500mA and it tested at 190mAh; Cycle 924 was at 300mA and it tested at 300mAh; and Cycle 925 was at 200mAh and it tested at 402mA. Anyway, here are my usual charts:

Also, I know someone asked if the cells were self-discharging. I have been taking measurements of the voltage of the CJ and THLD cells every 25 cycles and there's not much of a story to report. As of February 4, the CJ cell is still sitting at 4.15 Volts, and the THLD cell is also at 4.15 Volts. They are not self discharging at all, despite being end-of-life cells.

I'm coming to the conclusion that the ASO cells are most likely OK to mix into a powerwall with LG, Samsung, Sony, Sanyo, and BAK cells. They seem to have decent Cobalt content, at least on par with LG's value cells. I'm also pretty sure they are not manufactured by Samsung, as is commonly believed. They are just too different with the positive cap and with the top cell crimp. Lastly, I have verified that they have both PTC and CID protections, so they are on par with genuine cells with regard to safety, capacity, and lifetime.

EDIT: (4/12/2020)Update time! Sorry I never got around to posting at 1100 cycles, I had everything ready, but didn't get a chance to log in here and post it. Here is the degradation chart, at least:

So, I have made it past 1200 cycles! Right at Cycle 1175, the Opus fan finally gave up (it was moving, just not enough to remove much heat), and you can clearly see a spike in the results. I have several replacement fans on hand, so it didn't cost me any time. Otherwise, it's been pretty uneventful. The THLD cell is sitting at 4.13V and the CJ cell is sitting at 4.14V after almost 4 months, so they are not self-dischargers, despite being end of life. I'm not going to be checking those voltages anymore. The ASO and LG cell are continuing their very slow, almost linear, decline. On the one hand, I'm wishing they would just die already so I can end this test and not feel guilty about ending it early, but on the other hand, I'm rooting for these cells to keep going because the results of this test are so much bigger than the fate of 2 random cells. Anyway, here's where we are at:

And the degradation:

EDIT: (6/19/2020) I skipped the update at 1300 cycles, but here I am at 1400! Here's the graph:

And where we are at SoH:

I also set out to see if degradation is linear or exponential. The CJ and THLD were clearly exponential. The ASO and LG cells, I'm not sure. Here is a chart showing the rate of change. For example, the data point at 200 is the % increase in degradation from Cycle 201-205 over the degradation in Cycle 101-105, and so on. The LG looks fairly linear, while the ASO seems to be going exponential.

EDIT: (8/13/2020)First off, I made a mistake. I write down every result on a notepad, and then do data entry into excel when I hit a milestone and am ready to report an update. Well, for some reason, I wrote Cycle 1528 twice (...1527, 1528, 1528, 1529...). What I decided to do is average the results of the 2 cycles and use that number as the entry for Cycle 1528. To make up the test cycle, I will skip over Cycle 1666, assuming I get to that point, because I don't like that number :angel: .

Second, I kind of discovered something that I wanted to run by the community. I never really paid attention to the THLD cell's "INR18650" designation, as I understand that INR/ICR/IMR/NCR don't actually signify chemistry and are just naming conventions, but it got me thinking. If the THLD cell is labeled as an INR cell, and it actually is an INR (Lithium Manganese Nickel) vs. being an ICR (Lithium Cobalt Oxide), then the results of my degradation testing don't signify quality, but rather chemistry. It's not that the CJ and THLD cells are Chinese junk, but rather, they are low-cobalt-content INR cells that typically have a shorter life than their cobalt-rich ICR cousins. That also got me thinking again. If the CJ and THLD cells were truly INR cells, why did their PTC protection kick in to limit their short-circuit discharge to just 2 amps and why did the cells still heat up at that low of a discharge rate? My theory is that, since much of the cost of INR cells comes from having thicker copper and aluminum collectors, the Chinese manufacturers used thinner current collectors, while also using the low-cobalt INR chemistry. That would explain how they got their cost down. So high discharge chemistry without the ability to do high discharge. What do you all think about that theory?

Top Articles
Latest Posts
Article information

Author: Jerrold Considine

Last Updated: 12/16/2022

Views: 5915

Rating: 4.8 / 5 (58 voted)

Reviews: 81% of readers found this page helpful

Author information

Name: Jerrold Considine

Birthday: 1993-11-03

Address: Suite 447 3463 Marybelle Circles, New Marlin, AL 20765

Phone: +5816749283868

Job: Sales Executive

Hobby: Air sports, Sand art, Electronics, LARPing, Baseball, Book restoration, Puzzles

Introduction: My name is Jerrold Considine, I am a combative, cheerful, encouraging, happy, enthusiastic, funny, kind person who loves writing and wants to share my knowledge and understanding with you.