Cycle life tests of High Energy density cylindrical cells

[Pajda]

"Standard Power Rate" test (1C continuous discharge)
note: the 1C discharge test is performed at ca 90% DoD (4.15V-3.00V)

Test eqiupment:
- 2pcs ZKETECH EBC-X0510
- BF-2A four-wire cell fixture (10A version with Nickel plated contacts)
- tested in room with relatively stable temperature 23 ±2 °C

Remarks and comments:
- sister thread to the High Power cylindrical cell tests with the same settings Cycle life tests of High Power density cylindrical cells
- the reduction of tested DoD to about 90% (4.15V-3.00V) is based upon requests from the ES community
- from a purely technical point of view, even this small reduction of DoD significantly reduces the differences between high-quality and lower-quality cells.
- it means in practice that almost all cells shows results very similar to each other at 70% DoD or lower (see docware results), but you miss the information that some of high-quality cells perform very well even at 100% DoD (as was shown in 5C @ 100% DoD "torture" test of modern HP cells)
- you will certainly notice the "wawes" in the capacity test at 90% DoD, these are caused by the insertion of a nominal capacity test at 100% DoD. This behaviour is tied with the internal design of HE cells as the HP cells does not shows this effect significantly.
- tester accuracy is 0.2 % ±0.01 A. The stability is relatively good.
- do not much look at the absolute values, particularly at low currents in nominal capacity tests, trends are here important results.
- each 50th cycle is inserted datasheet nominal capacity test with 10s DCIR test (see dots in figure)
- minimal charging CV phase cut-off current is set to 100 mA due to the ZKETECH testers limitation

____
"Table" with values measured on precision battery tester (not ZKETECH). Those test were performed in temperature chamber under 25°C, TC is using forced air flow to maintan consistent temp. This 25°C air flow acts as cooling force on the cell surface, so the measured temperatures under high load will be affected by this.

*Temperature is measured by the K-type thermocoupler, attached to the cell body using Kapton tape. Some cells are sold as Naked (without tube) and these shows significantly higher temperatures than cells with tube due to the effect of the tube thermal insulation.

added LG INR18650-M36 to the chart (X0510 tester, cell supplier: NKON)
added Lishen LR2170SD to the chart (X0510 tester, cell supplier: NKON)
added BAK N21700CG-50 to the chart (X0510 tester, cell supplier: NKON)
added LG INR21700-M50LT to the chart (X0510 tester, cell supplier: NKON)
added BAK N18650CP-35 to the chart (X0510 tester, cell supplier: NKON)
added Panasonic NCR18650T1 to the chart (A20 tester, cell supplier: Vapcell, *Naked)
added Samsung INR21700-50E2 to the chart (X0510 tester, cell supplier: NKON)
removed Samsung INR21700-50S from the chart (X0510 tester, cell supplier: NKON)
added Lishen LR1865HB to the chart (X0510 tester, cell supplier: NKON)
added EVE INR18650-33V to the chart (X0510 tester, cell supplier: NKON)
added LG INR21700-M58* to the chart (X0510 tester, cell supplier: NKON, *Naked) *NKON refers the cell as an M58T model, but on the body is marking M58
added Samsung INR21700-50G to the chart (X0510 tester, cell supplier: NKON, *Naked)
added Samsung INR21700-48X(1) to the chart (X0510 tester, cell supplier: NKON)
added BAK N21700CD-53 to the chart (X0510 tester, cell supplier: Vapcell)

All in One:



21700 size


18650 size

 

[Pajda]

"Standard Power Rate" test (1C continuous discharge)
note: the 1C discharge test is performed at 100% DoD (4.20V-2.50V)

Test eqiupment:
- 2pcs ZKETECH EBC-X0510
- BF-2A four-wire cell fixture (10A version with Nickel plated contacts)
- tested in room with relatively stable temperature 23 ±2 °C

Remarks and comments:
- tester accuracy is 0.2 % ±0.01 A. The stability is relatively good.
- do not much look at the absolute values, particularly at low currents in nominal capacity tests, trends are here important results.
- each 50th cycle is inserted datasheet nominal capacity test with 10s DCIR test (not shown in the graphs)
- minimal charging CV phase cut-off current is set to 100 mA due to the ZKETECH testers limitation

____
"Table" with values measured on precision battery tester (not ZKETECH). Those test were performed in temperature chamber under 25°C, TC is using forced air flow to maintan consistent temp. This 25°C air flow acts as cooling force on the cell surface, so the measured temperatures under high load will be affected by this.

*Temperature is measured by the K-type thermocoupler, attached to the cell body using Kapton tape. Some cells are sold as Naked (without tube) and these shows significantly higher temperatures than cells with tube due to the effect of the tube thermal insulation.

added LG INR18650-M36 to the chart (X0510 tester, cell supplier: NKON)
added EVE INR18650-35V to the chart (X0510 tester, cell supplier: NKON)
added BAK N18650CP-35 to the chart (X0510 tester, cell supplier: NKON)
added Lishen LR1865HB to the chart (X0510 tester, cell supplier: NKON)
added EVE INR18650-33V to the chart (X0510 tester, cell supplier: NKON)
added Vapcell INR18650 F36 to the chart (X0510 tester, cell supplier: Vapcell)
added Vapcell INR18650 F38 to the chart (X0510 tester, cell supplier: Vapcell)
added Vapcell INR18650 N40 to the chart (X0510 tester, cell supplier: xxx)

added Tenpower INR21700 50ME to the chart (X0510 tester, cell supplier: NKON)
added LG INR21700 M58 to the chart (X0510 tester, cell supplier: NKON)
added Vapcell INR21700 F60 to the chart (X0510 tester, cell supplier: Vapcell)
added Samsung INR21700 53G(1) to the chart (X0510 tester, cell supplier: NKON)

21700 size
5300+ mAh samples comparison





18650 size

 

[Pajda]

100% DoD vs. 90% DoD comparison (1C continuous discharge)

Test settings: 100% DoD (4.20V-2.50V) vs. 90% DoD (4.15V-3.00V)

Purpose of this test:
- To show the "unexptected" behavior which can be encountered when cycling modern lithium cells at lower DoD than 100 %.
- The generally accepted statement is that lithium batteries do not suffer from so-called "memory effect" (often seen in NiCd/NiMH technology).
=> The practical problem of this effect is an unwanted (but recoverable) loss of capacity/energy of the cell .
- Tricky thing is that some lithium cells show significant "memory effect" and some does not show it at all.
- It seems that this problem is connected with HE cell technology as HP cells does not show it.
=> The way how to recover the lost capacity/energy is to perform single "formatting cycle" which is one full (100 % DoD) charge-discharge cycle (same as with NiCd/NiMH technology)

Test eqiupment:
- ZKETECH EBC-X0510
- BF-2A four-wire cell fixture (10A version with Nickel plated contacts)
- tested in room with relatively stable temperature 23 ±2 °C

Remarks and comments:
- tester accuracy is 0.2 % ±0.01 A. The stability is relatively good.
- do not much look at the absolute values, particularly at low currents in nominal capacity tests, trends are here important results.
- each 50th cycle is inserted datasheet nominal capacity test with 10s DCIR test (see dots in figure)
- minimal charging CV phase cut-off current is set to 100 mA due to the ZKETECH testers limitation

 

[Pajda]

Thanks to Tumich (ES member) I got samples of LGX E78 (78 Ah) pouch cells which are used in several EV platforms. One of them is VW MEB platform, so you can find them in VW ID.x or Škoda Enyaq family vehices (but recently they start to use also prismatic cells from CATL, so it is hard to tell)

The sample of LGX E78 pouch cell have this parameters:

supplier: LG Chem
format: pouch
L: 550 mm (595 mm with unfolded side terminals)
W: 9 mm
H: 98 mm
capacity: 78Ah
nom.volt: 3,65V
cathode: NCM712
weight: 1090g
vol. energy density: 590 Wh/l
grav. energy density: 265 Wh/kg
DCIR 10s (50% SoC): 1.65 mOhm
DCIR IEC (50% SOC): 1.44 mOhm
0.2C nom. capacity (4.20-2.50V): 79.19 Ah
0.2C nom. energy (4.20-2.50V): 290 Wh
1C capacity (4.15-3.00V): 72.39 Ah
1C energy (4.15-3.00V): 254 Wh

Test conditions:

This cell si huge and unfortunately it will not fit in my temperature chamber, so the test below was conducted in room temperature. It must be said that the temperature comparison with other samples is misleading for E78 as it was not "cooled" with 25°C forced air in TC.

 

[john61ct]

1C rate and 3.0V cutoff are great afaic for building a list of "good enough" performing cells from which we can look for **great value** deals.

reduction of DoD significantly reduces the differences between high-quality and lower-quality cells.
- it means in practice that almost all cells shows results very similar to each other at 70% DoD or lower (see docware results), but you miss the information that some of high-quality cells perform very well even at 100% DoD (as was shown in 5C @ 100% DoD "torture" test of modern HP cells)

The longevity factor in cycles to 70-80% SoH is IMO the critical factor worth paying more for

rather than focusing on which cells are "the best" according to power and DoD levels I consider unrealistic, even abusive.

Thanks in advance for taking this category on!

 

[Pajda]

added LG M50LT to the test.

Q: What to expect in this thread until next christmas? :wink:

I would like to add two more cells from 18650 category. The idea is to have one branded cell in each format and two cost@avalability effective alternatives from "less recognized manufacturers(for now)". So from category 18650 it is definitely BAK CP but then I am not sure about the second competitor, EVE 33V looks interesting?

 

[e3s]

can you try eve 26v? They are inside some xiaomi scooter battery packs

 

[pomah]

added LG M50LT to the test.

Q: What to expect in this thread until next christmas? :wink:

I would like to add two more cells from 18650 category. The idea is to have one branded cell in each format and two cost@avalability effective alternatives from "less recognized manufacturers(for now)". So from category 18650 it is definitely BAK CP but then I am not sure about the second competitor, EVE 33V looks interesting?

Looking forward to the results, just built my new ebike battery using the M50LT but was unsure for a while do lack of knowledge in regards to this battery.

 

[Pajda]

can you try eve 26v? They are inside some xiaomi scooter battery packs

Probably not in the near future. The reason is that this cell does not fit into the HE or HP category. But that doesn't mean it's a bad choice, it just deserves its own category, something like "good price-performance ratio".

 

[mighty82]

This is very nice results. I like the internal resistance staying at a low level. In my experience IR seems to become a problem before capacity loss. Had this problem especially with nca cells.

 

[Pajda]

Small update after a while

Panasonic NCR18650T1 (aka Tesla Plaid cell)
- For those who are familiar with Tesla/Panasonic cell performance it is no surprise. Cell shows similar results (capacity fade) as another cells from 18650 line used by Tesla

BAK N18650CP-35
- A cell promising very good price-performance ratio in the 18650 HE category

LG INR21700M50LT
- The prices of this LG cell have recently returned to pre-crisis levels and therefore it returns again to the competition of the best price-performance ratio in the 21700 category. And one remark "L" or "LT" M50 cells shows significantly better cycle life performance than "T" model.

Lishen LR2170SD
- I came across a problem with the used ZKE tester channel and so the values of the nominal capacity test inserted every 50 cycles (dots in the graph) are skewed, but 1C discharge values are not affected. At the end of the test, I will try to correct it based on the initial and final values.

 

[liveforphysics]

The cell you're calling a Plaid cell is not actually sourced from a Plaid Model S right?

Thank you for your hard work and sharing results!

 

[Pajda]

The cell you're calling a Plaid cell is not actually sourced from a Plaid Model S right?

Yup, tested sample was sourced by Vapcell as original Panasonic NCR18650T1. The reasons why this type is referred as "Tesla Plaid" are two: First, it has the same unique mechanical design of positive terminal and bottom vent with Tesla "three scratches" marking as actual Plaid cell. Second, that in the past I have tested original Tesla cells, salvaged from wrecked cars and compared them with original Panasonic "industrial" models. For example, the cells from TMS "85" with its parameters such as DCIR and cycle life corresponded almost exactly to the results of original Panasonic NCR18650BE and later BM models. The same was observed in 21700 format.

 

[vanturion]

Chart is showing some pretty significant capacity recovery from the end to the beginning of each 50-cycle run with the 50E2. What would you say is the average time between each run to get this effect?

I see you have the same effect with most of the other cells tested here too, although to a lesser degree (neglecting the Panasonic NCR18650T1 results). On the other hand, this capacity recovery behavior doesn't really manifest noticeably in your High Power cell tests with any of those cells. Do you think it's because of the lower discharge rate here and/or subsequent less heat generation, the truncated DoD, perhaps a combination of these, or something else I'm missing?

As for the Panasonic, that one looks like it's on it's way out fast and with only 1C 90% DoD! That's actually remarkably unimpressive. You know, I haven't ran across a Panasonic I would even want to spec in a PEV battery build in lieu of other options. And yet they've seen prolific use in Tesla's for years now. It's odd.

 

[Pajda]

Small update after a while

Samsung INR21700-50S
- Nkon few weeks ago listed them in their store. I was able to order samples from there before they went out of stock. This cell is interesting in terms of representing the "best energy-power performance ratio on the market" (at least until the LG H50 will be widely available), but its price is completely out of the question. Previous 50S samples coming from Vapcell show significant cycle life problems even at lower 1C rate at 100% DoD. Lets see if there will be any significant change at 90% DoD.

I also found Lishen LR1865HB 3350mAh newly listed on Nkon store. But right now when the LG cell prices (M36) are stabilising it is not as interesting competitor. But still will try to add it to the chart.

 

[Pajda]

Chart is showing some pretty significant capacity recovery from the end to the beginning of each 50-cycle run with the 50E2. What would you say is the average time between each run to get this effect?

I see you have the same effect with most of the other cells tested here too, although to a lesser degree (neglecting the Panasonic NCR18650T1 results). On the other hand, this capacity recovery behavior doesn't really manifest noticeably in your High Power cell tests with any of those cells. Do you think it's because of the lower discharge rate here and/or subsequent less heat generation, the truncated DoD, perhaps a combination of these, or something else I'm missing?

This is fun when working with electrochemical cells that you are never sure in advance of what you will measure. But the fun stops when someone puts a knife to your neck to tell them the essence of the phenomenon. I observed two major phenomena, which I would call the "memory effect" and the "self-healing effect". In addition, these can be tied together.

So called "memory effect" is what you can see in this particular 90% DoD chart. The waves are caused by the insertion of nominal capacity test, but the root cause here is application of full charge voltage 4.20V instead of 4.15V used in the cycle life part of the test. For example if both nominal capacity test and cycle life test will use the same charging voltage, there will be no waves as you can observe in 5C chart. Both C-rate and the discharge voltage limit (DoD) does not have impact on this effect. In this case, I would not talk about the self-healing effect, because the "recovered" energy is still present in cell at that time, it is only not used. And of course no generalization, you can see that LG M50LT does not show this effect at all.

So called "self-healing effect" can be described as after prolonged period of the time you can measure significantly higher energy, particularly at higher C-rates (recovery of DCIR) than that was available after finishing the use of the cell. This phenomenon is mostly talked about in connection with Tesla/Panasonic NCA cells.

As for the Panasonic, that one looks like it's on it's way out fast and with only 1C 90% DoD! That's actually remarkably unimpressive. You know, I haven't ran across a Panasonic I would even want to spec in a PEV battery build in lieu of other options. And yet they've seen prolific use in Tesla's for years now. It's odd.

All I can say is that in my synthetic cycle life tests all Panasonic/Tesla production behave in similar way and shows significantly worse results than other competitors. But this does not mean that in real world application they cannot perform well, or at least good enough.

 

[vanturion]

But the fun stops when someone puts a knife to your neck to tell them the essence of the phenomenon

Wow, doing research must be pretty tough in Czech! :wink:

So called "memory effect" is what you can see in this particular 90% DoD chart. The waves are caused by the insertion of nominal capacity test, but the root cause here is application of full charge voltage 4.20V instead of 4.15V used in the cycle life part of the test.

Ah, thanks, I missed that. Explains why the memory effect is so much more pronounced with some of the cells here vs not present on your high power tests. A good reminder for me not to directly compare charts without care.

For example if both nominal capacity test and cycle life test will use the same charging voltage, there will be no waves as you can observe in 5C chart. Both C-rate and the discharge voltage limit (DoD) does not have impact on this effect. In this case, I would not talk about the self-healing effect, because the "recovered" energy is still present in cell at that time, it is only not used. And of course no generalization, you can see that LG M50LT does not show this effect at all.

OK, funny, first thing I wanted to do is ask a question about generalized behavior on the self-healing effect, but I get it more now, it's not possible to generalize this nuanced behavior between different cells and possibly even generations of cells. Well let me put it a different way if you can answer--

Is it fair to say that the memory effect only creates a temporary wave or capacity gain and doesn't effect the slope of the capacity degradation curve of a bench-tested cell OR does it actually contribute to slowing the capacity degradation curve to some small degree in these type of tests? In other words, does the memory effect matter in the long run for slowing capacity degradation as a kind of technique or can it be ignored then?

So called "self-healing effect" can be described as after prolonged period of the time you can measure significantly higher energy, particularly at higher C-rates (recovery of DCIR) than that was available after finishing the use of the cell. This phenomenon is mostly talked about in connection with Tesla/Panasonic NCA cells.

Ah, I initially thought that every 50th cycle you had a 24-hour break or so in which these comparatively less abused cells (vs 5C test) were experiencing some measure of "self-healing", but moving forward with the understanding that the HE_1C_cycle_life_test@capacity chart is only demonstrating the memory effect, for cells that can experience self-healing then, what kind of time-frame are we talking then for this behavior to manifest noticeably from what you've seen? Days, weeks, months?

All I can say is that in my synthetic cycle life tests all Panasonic/Tesla production behave in similar way and shows significantly worse results than other competitors. But this does not mean that in real world application they cannot perform well, or at least good enough.

This is such a good reminder, I definitely feel like I've been over-relying and forming conclusions about these cells based on these type of bench tests alone. It's as you say, some cells may perform better in these tests, but with regard to most ebikers or PEVs use in general, most people aren't constantly cycling their battery packs without breaks, and time between cycles can be sometimes days or weeks long. It's possible that some batteries that seem to perform worse in these bench tests, may actually perform better with 24-hour breaks or longer in between cycles. Or best with active heating & cooling which is probably the case the Panasonic cells.

It takes a while to beat this kind of nuance into my head. 8)

 

[john61ct]

Normal usage by users who care about longevity also never goes near 100% DoD, that is murderous practice.

If your LVC stop-discharge is calibrated to say 3.4V isolated at rest, I bet that is **way** above 90% DoD

Just a 5% higher average DoD can double or triple lifespan.

But that factor must vary wildly by model...

 

[vanturion]

Normal usage by users who care about longevity also never goes near 100% DoD, that is murderous practice.

If your LVC stop-discharge is calibrated to say 3.4V isolated at rest, I bet that is **way** above 90% DoD

Just a 5% higher average DoD can double or triple lifespan.

But that factor must vary wildly by model...

For sure. It all depends on the application and how long you want the batteries to last vs capacity use/range, but I'm all for preventing unnecessary abuse of the lithium-ions. Maybe we can start a "Practice Safe Voltage Limits" club

 

[calab]

I remember reading from Justin or someone of similar caliber, in general terms (not high energy or high discharge or high charge) that its about how long the battery stays at high voltage.

If its depth, then whats the info on leaving it empty vs charging up right away.
To many variables for testing I'd guess, to many wide variances in how often it would be used.

Normal usage by users who care about longevity also never goes near 100% DoD, that is murderous practice.

If your LVC stop-discharge is calibrated to say 3.4V isolated at rest, I bet that is **way** above 90% DoD

Just a 5% higher average DoD can double or triple lifespan.

But that factor must vary wildly by model...

 

[john61ct]

There are lots of factors affecting longevity.

Yes LI should only get charged to high SoC just before loads will be pulling it back down, stored in the 40-60% band when not cycled.

Not charged too high in the first place, average DoD is a critical one.

Not discharged too fast.

Not fast charged in cool temps, but cool storage is great.

Of course they all interact, vary by model, near impossible to quantify.

And many use cases do not allow such coddling, many racers get one season then replace and accept that.

Anyway, let's start a new thread to get into the weeds any further rather than diverting this one further.

 

[bmwupgrade]

Have you tested 50s? Have photo? Expect them to have it at gen2 now

Small update after a while

Samsung INR21700-50S
- Nkon few weeks ago listed them in their store. I was able to order samples from there before they went out of stock. This cell is interesting in terms of representing the "best energy-power performance ratio on the market" (at least until the LG H50 will be widely available), but its price is completely out of the question. Previous 50S samples coming from Vapcell show significant cycle life problems even at lower 1C rate at 100% DoD. Lets see if there will be any significant change at 90% DoD.

I also found Lishen LR1865HB 3350mAh newly listed on Nkon store. But right now when the LG cell prices (M36) are stabilising it is not as interesting competitor. But still will try to add it to the chart.

 

[Pajda]

Have you tested 50s? Have photo? Expect them to have it at gen2 now

I recently moved actually running 50S test results from this HE chart to the HP one in the second ES thread. The tested "green tube" sample comes from NKON (you can find photos here) and it seems to be Gen1. I found some pictures of 50S also in a purple tube.

 

[Pajda]

Small update after a while

Lishen 18650HB (3350 mAh) and EVE INR18650-33V (3100 mAh), both samples from Nkon added to the test this week.

Finally, I have back access to my Temperature Chamber, so I would add other test results within a week.

 

[Pajda]

Small update after a while

added final graphs and values after 1000 cycles for BAK CG50 and Lishen 21700SD