Li-ion cells cycle ageing

Not great, not terrible... So far very good results in my opinion considering current price and availability on the market. Yesterday I started my standard 1C discharge cycle life tests. And with those cells I ordered new sample of Moli P42A, so I was able finally to re-test it under 5C (20A) and update the graphs and table.
 

Attachments

  • nominal_test@table_new.png
    nominal_test@table_new.png
    26.8 KB · Views: 1,504
Pajda, your results with the 50S so far has me really curious now how the other high capacity 2170s will stack up in comparison. One result doesn't make a trend, but I wonder if many of these high capacity cells will suffer the same relatively steep capacity decay over cycle life as the 50S is already at 80% capacity in a little over 200 cycles.

From this high-capacity category, I'm also wondering how something like a 50E (highest DCIR) vs a 50S (lowest DCIR) would stack up in a 2C cycle test. I guess I could always order those and run the test myself sometime :) .

Something else I noticed from your charts - if you only look at the capacity decay chart on page 20 of the Samsung 40T appears to crush the Molicel P42A so a pack assembler might go ahead based on that info alone. The other chart plotting performance (V) over time tells a different story where the voltage sag is notably less in the P42A vs the 40T, at least in this initial cycle.

So now having some hands-on experience testing cells and watching this thread, it really comes down to just 2 charts/comparisons I care about when battery shopping: 1) Capacity Decay vs Cycles and 2) Voltage vs Time.

The Voltage vs Time graph is particularly useful when including the initial cycle and a mid-late cycle as the interim performance data can be assumed/extrapolated from the difference. I like 500 cycles for the second comparison, but that's just my own arbitrary preference.

Just to give a concrete example to the readers here of how useful this comparison is - based on my test results with the Lishen LR2170SF you might think that while the Voltage vs Time loses to the Molicel P42A performance when tested at similar current levels, it beats the P42A in terms of total capacity, and you'd be right. However, the LR2170SF only wins if you run your cells/pack below 3.1V/cell (at the testing currents I've compared). So basically, if you plan to run a voltage cut-off at or above this level, there is no category in which the LR2170SF beats the P42A. Not in capacity over time, not in initial performance, and not in late cycle performance.

Of course you can add mud into the mix and turn the clear winner on paper into a much more murky solution by understanding that ebike batteries aren't running at constant max-rated current all the time so at the end of the day it's really up to the individual to assess how much they want to care about these results as price will most likely be the most influential factor anyway. That said, the initial vs late cycle Voltage vs Time chart/comparison really helps complete the comparison picture for cell selection, for me anyway.
 
I would bet that less delta in voltage sag

between New and 200 cycles Old

would correspond to less capacity loss as well

more often than not.
 
john61ct said:
I would bet that less delta in voltage sag
between New and 200 cycles Old
would correspond to less capacity loss as well
more often than not.

You would think, but what I was pointing out above it's the opposite of what you say - the P42A has notably less voltage sag (looking at initial cycle Voltage vs Time only) for basically the entire range vs the 40T; however, the 40T has significantly less capacity loss, at least for the 250 cycles we've seen so far from Pajda's charts. Not only that, but they have virtually exactly the same DCIR values so this measurement isn't at all an explanation for these differences either.

Now there could be a point at which the P42A Voltage vs Time performance is exceeded by the 40T at some cycle count as the P42A capacity decays much quicker in the early cycles, but without plotting these results too, I wouldn't be certain in making any predictions.

In any case, there's clearly a lot going on with regard to internal battery chemistry that lay people like myself aren't going to be able to accurately make any predictions about performance or capacity loss without manually testing (ignoring manufacturing spec data).

An example of what I mean - testing my 7 year old gen 1 25R cells, I paused cycling for 77 days and experienced a decrease in total capacity during the next cycle which leveled off for roughly the next 25 cycles before resuming roughly the same linear capacity decay rate vs cycle count before the pause. This in contrast to my ongoing LR2170SF tests, where I paused testing for 24 hours and experienced minor capacity recovery for the next 7 cycles. Then, I paused cycling for 15 days and had a much greater capacity recovery where 25 cycles later the capacity is still not even close to the lowered capacity it was before the pause.

Why? Age? Maybe. Chemistry? Definitely. Both? I don't know. It's interesting, but I'll leave that kind of work to the material scientists and chemists. Here, I just care about getting good bang for the buck and making sure I end up with a solid pack that'll last as long as possible with good performance.
 
I try to digest all the information and always looking to keep the most watts over time not just the most output. Can not believe we have people like Pajda testing all the time. Feel luck to have access to all the information. I like long lasting batteries but starting to think the time itself will get me not just the cycles.
 
john61ct said:
Yes I was stating that I think cases like that will prove to be exceptional.

Probably, but from my understanding now though I think it's best not to make behavioral predictions about these batteries and instead look to the test results or create them yourself if they aren't publicly available for the cell(s) of interest. A $700+ battery investment might warrant 10% of that on testing to make sure you get the most out of that investment for the next 5-10+ years. At least, that was my thinking going into this. Then again, I spent quite a bit of time on this. I guess you can't put a price on hobbies... 8)

ZeroEm said:
... I like long lasting batteries but starting to think the time itself will get me not just the cycles.

The best information I've found on calendar life aging so far is this aggregate study comparing general battery chemistry and battery life degradation vs SOC and Temperature. It's a shame the 3D graphs are so small/hard to read, but you can still get a general sense of what temps and SOC degrade these batteries most over time (high temps and high SOC generally for the chemistries most ebikers use).

If I had the raw data they used to plot these charts, I would like to see how my 25R results of losing an estimated 2-5% total capacity over ~7 years at ~50% SOC would jive with their characterization for NCA degradation just to sanity check their data with my single data point.

I also found the discussion here to be pretty informative as well.
 
Hi, there are some news:

Samsung 50S is dead :(
The cell has completely lost voltage, so CID protection has been probably triggered due to the excessive heating. It is a pity that the cell did not last up to 300 cycles, where I would stop his torture anyway and do the final test. I think this 50S sample did a good job anyway. According to the catalog, after 250 cycles at 25A CDR, it should still have 60% of the nominal capacity. I think I read somewhere that when introduced it had only 20A CDRs in the datasheet and later this value was adjusted to 25A. It is possible that the value of 25A CDR applies to the 50S2 version, because Samsung, as far as I know, does not release a new datasheet with the new generation of cell model. But this is only a guess.

Sony US21700VTC6A was added to the charts.
However, it has to wait for its 5C torture until one of the P42A or 40T3 dies or completes 1000 cycles, because I want to test it on a slightly more accurate A20 tester. Btw. code on the wrap says that the sample was made in 2021. Usually I get to the samples ca one year old.

Samsung 50E2 sample was added to the table of HE cells.
I have heard rumors that it has been significantly improved intergenerationally in both cycle life and particularly DCIR. Now I can confirm the lower DCIR. The cell has already started my standard 0.5C-1C 100% DoD cycle life test. I guess 50E2 will have very similar parameters with 50G(1), maybe they are even the same cells where the G model is without wrap. It is said that the 50E3 model should already be available.

P42A vs 40T capacity vs. energy delivered
It is a good point, It would be nice from me to show you also energy vs cycle graph, which is actually more practical than capacity trend :wink: . In this particular case, P42A really delivers a little more energy than 40T3 in the first 15 cycles, but then takes over the 40T3 and it consistently deliver about 5% more energy than P42A.

From this high-capacity category, I'm also wondering how something like a 50E (highest DCIR) vs a 50S (lowest DCIR) would stack up in a 2C cycle test.
Actually I perform only 1C and 3C discharge rate tests. 50E(1) at 3C continuous shows similar trend as 50S at 5C. In applications up to 1C continuous there is no significant advantage of using 50S against 50E.
 
Pajda said:
It is a good point, It would be nice from me to show you also energy vs cycle graph, which is actually more practical than capacity trend :wink: . In this particular case, P42A really delivers a little more energy than 40T3 in the first 15 cycles, but then takes over the 40T3 and it consistently deliver about 5% more energy than P42A.

Unless capacity takes a dive soon which doesn't seem likely, 40T3 looks like a superior choice between the 2. It's too bad about the price though.

The forum/we are lucky to get all this data from you as it is. If more happens to show up, all the better :D

Pajda said:
Actually I perform only 1C and 3C discharge rate tests. 50E(1) at 3C continuous shows similar trend as 50S at 5C. In applications up to 1C continuous there is no significant advantage of using 50S against 50E.

Good to know. I was curious about the 50E(1 probably) as they've come installed in an electric unicycle I recently picked up which I don't think see more than 8A/cell.
 
added updated table with P42A final data
added new graph with 5C voltage drop after 1000 cycles for those who survived long enough 8)

Pajda's Comments and Thoughts (PCT):
- other samples of Samsung 50S also began to die by sudden death around 500 cycles even at 1C dischrage rate. if I assume they did not come from bad series, then they still seems to be a good choice for power tools.
- I was thinking about adding Samsung 25R to the test, but the question is if any of you still use 2500mAh HP category?
- Sony US21700VTC6A seems to have similar results to the P42A (a comparison at 1C discharge also show similar results)
- Lishen and BAK "5000mAh" HE category cells seems to be very good after passing 200 cycles.I would especially rate BAK as excellent so far.
 
ginekolog said:
Thanx for lab tests, its good work.

Now real life test.
I have 900Wh Samsung 35E 52V 14S 5P Jumbo Shark battery ( https://fasterbikes.eu/de/akkus/225/900wh-flaschenhalter-akku-52v-14s-5p-jumbo-shark?c=47 ) on Giant Trance X2 + BBSHD. I ride single mountain trails.
I take good care of battery:
- charge to 90% with 5A, discharge to 10% (~45V) with average of 20A (1000W). Peaks 1600W for few minutes.
- storage and charge at 22C
- dont ride with cold battery
- average batt temp is ~40C, max 57C (very rare).

After about 1 year, 4000 km and 100 cycles I measuerd my battery with 200W load. Capacity went down from 17Ah to 15Ah (850Wh to 750Wh). So now I am at 88%. :? This is confirmed with my 200W load calculation and BMS display (BT app).
Internal resistance is ok for now, at 30A it sags for 5.5V. This means 65 mOhm per cell. :roll:

I am still satisfied with performance if it stays this way for some years. From your graphs it seems it hopfuly will :wink:


I did new measurement on 12.10.2021. Battery age is 3 years, ~300 cycles, 9300km on BBSHD. I measuerd my battery with 200W load (resistors) and tool FDY10-H. Capacity is still 15.2 Ah (766Wh, new was 856Wh). So I am at 89%, which is :thumb: This is confirmed with BMS (BT app).
Internal resistance is still ok, at 30A it sags for 5.5V. This means 65 mOhm. So my bike still runs fast as :bolt: :bigthumb:

It is clear that on first year battery lost 11% and now performance is stable. I think it will last atleast up to 500 cycles (5 years in my case).
 
added updated table and voltage/temperatire charts for Samsung 25R8 (Nkon)

Pajda's Comments and Thoughts (PCT):
- Samsung 25R8 shows excellent performance at the start (first competitor which can handle 5C 100% DoD discharge test below 50°C)
- It will be nice to do comparison with BAK N18650CNP-25, which should have the same parameters according to the documentation
 
I realize you have to set standard parameters for comparability

but just want to comment

that I consider a 100% discharge and huge C-rate to be irrelevant, unrealistically abusive for longevity testing.

Maybe it's just me, but I look for cells that deliver lots of cycles charging to 4.15V only and LVC well above 3.0V

and maybe 1C average with 2C peaks.

I acknowledge everyone else may have different parameters, and need to buy the gear to do my own testing.

 
john61ct said:
that I consider a 100% discharge and huge C-rate to be irrelevant, unrealistically abusive for longevity testing.

Yup, It was told at the beginning of this tests. The aim of this test was to show that modern lithium batteries are not as bad as they are still described in old wives' tales spread by a would-be specialists and sites.

john61ct said:
Maybe it's just me, but I look for cells that deliver lots of cycles charging to 4.15V only and LVC well above 3.0V

and maybe 1C average with 2C peaks.

Let me put it this way: Christmas is coming! :wink:
 
Pajda said:
Let me put it this way: Christmas is coming! :wink:
Such a tease you are!!

Very much look forward, and once again thank you for your incredibly valuable work for the community
 
Any experiences with EVE 26v or 33v? According to their datasheets they seem to be pretty good for e bike usage.

As EVE is not so familiar with 18650 cell format and li-ion technology, my biggest concern is if there is much variety in the properties such as IR, capacity, cycle life or the size / dimensions.
 
Not yet. In general, modern chinese production does not lag significantly behind korean-japanese production. In fact, we can already find cells from chinese companies with better overall parameters. When I look at the EVE 33V datasheet, the presented cycle life at 1C @ 95% DoD corresponds to the best small cylindrical cells on the market.
 
I've noted the dimension variations in Lishen cells. Bought a few 2170SF (orange 4500mah 13.5A) and 2170SD (purple 5000mah 9.6A) Lishen cells. Also some spacers which turned out to be too tight for most of them. I found out these came in 21.2mm and 21.7mm, and the latter size fit them all, although some cells were a bit loose.
 
added updated table and charts for BAK N18650CNP-25 (Nkon)
added updated DCIR-10s rise chart

Pajda's Comments and Thoughts (PCT):
- BAK N18650CNP-25 entered the competition decently.
- I managed to pull out of the trash and put another A-20 tester into operation, so three tests can now run simultaneously.
- As a third now started 25R8, after finishing 40T3 the BACK CNP-25 will start.
 
Back
Top