mass life cycle testing

Due to a cronic lack of decent testing results i concluluded that someone has to bite the bullet for this so i am planning on buying a 8 channel battery tester in order to finally get to the fine details on what is the best way to make your cells last as long as possible.

so far my plans have testing of the panasonic PF and GA cells as i have a bunch of those on hand.

my plans are to have each channel do different runs.
i base my testing currents on factory limits and what loads they might see during actual usage in scooters with different sized battery packs.
for these tests i dont use all 8 channels so i have a few left for weird/unusual/torture tests.

so far i have these tests figured out for my use cases:
4.2>3 at 2A discharge and .7A charge with 5A peaks each 5 min for 10 seconds
4.1>3 at 2A discharge and .7A charge with 5A peaks each 5 min for 10 seconds
4.05>3 at 2A discharge and .7A charge with 5A peaks each 5 min for 10 seconds
4.05>3 at 3A discharge and .7A charge with 5A peaks each 5 min for 10 seconds
4.05>3 at 4A discharge and .7A charge with 7A peaks each 5 min for 10 seconds

i plan to run 1000 runs for each channel or until they basically die, whatever comes first.

i would love to hear some cycle testing parameters from you guys and reasons why it should be included.
if you have cells that you want to see tested and want to send them to me i will test them after these have finished their runs you can send me a message for shipping details.

results will be public and posted here.

Temperature is one of the biggest factor of lifecycle. Change of room temp will change your results

cwah said:

Temperature is one of the biggest factor of lifecycle. Change of room temp will change your results

Click to expand...

testing will be done in my home. it is always between 19~21c. that is pretty negible difference in temperature considering the load they are under 24/7.

flippy: I appreciate this activity and look forward to the results

I think that your fundamental setings are fine, you will only wait very long for the results. That is the reason why I personally using 0.5C charge - 1C discharge tests and then comparing the cells between each other. Your test settings will be much more close to the real life use of cells, but be prepared for about 7-9 months to reach 1000 cycles.

There are some "non written" standards for cycle life testing.

1) charging cut-off current: I am using 100 mA for all cells and so some of them are not 100% charged (a lot of 18650 cells have nominal cut-off current aroud 50-65 mA)
2) rest-time: It is common practice to add some rest time between each charging and discharging, I am using 10 min (15 or 20min is also common)
3) Inserting "nomial capacity test": If you would like to compare capacity fade with the nominal value in the cell datasheet you should insert the nominal capacity test at the begining of test and then each 50 or 100 cycles. Nominal capacity test is described in cell documentation and it is usually 0.3-0.5C charge and 0.2C discharge at 100% DoD. This nominal capacity test also gives you some information about internal resistance change.

+1 on temperature control - it's probably the number one factor in cycle life. If you can hold the cell temperature as close to 20'C it will live a long life. Ambient can be constant, but the cell will heat up and cool down a lot as it cycles. It's the same benchmark everyone else uses, so there's not a lot of point in going overboard with regulation.

+0.9 on state of charge - charging a cell up to 4.15 V instead of 4.20 V will make a big impact on longevity, but putting numbers to it would be good too. Plus it charges a bit quicker More results sooner. But it does also impact on...

+0.8 on cut-off current. If you use a power supply like I do, it will jam the set current into it no matter what the voltage is. This is not normal; usually it will taper off from 4.1 V as the voltage approaches 4.20 V. Again, this can add another 15 or 20 mins to the charge cycle.

For reference, I've been cycle testing a pouch cell from HGB which the University of Nottingham guys are using on their bike. I'm charging at 1 C (10.0 amps) and discharging at about 1.3 C. The charge takes about 50 mins and the discharge about 36 mins.

I stop the cycle tester every 150 cycles and do a full 4.2 V to 3.1 V discharge test for capacity at the same 1.3 C rate. It's in my shed, so the overnight minimums are around 15 degrees Celsius and maximums are about 32 degrees Celsius.



So 1000 cycles to 80% capacity is entirely viable.

thanks guys,

i dont live in a distopian wasteland after a atomic war ( ) so temperature will be pretty much unchanged all year. my home has a very stable temperature, never under 19 or above 21 (i have aircon in my house case it does get hotter) so temperature is basically sorted. i am not going to actively cooling them, that would negate the purpose of the test as the cells would not be cooled during usage anyway.
i might stick a probe on the most abused cells and stick the Dt temps on the datapoints as a reference.

the 4.2v run is just a baseline. all others will be kept at 4.05V@100mA cutoff charging and the discharge current setting is the cutoff, no tapering off the discharge current. as soon as it hits 3 volts (prehaps it might be slighty changed to 2.8V) at the set current the battery is considerd empty. at that point you wont drive far with the power left in the cells anyway and i am trying to make it realistic, nobody will completly drain their battery or they are pushing a 70+ kg scooter home. so keeping it between the goal posts is more real world.

i plan on 10 min rest time between each cycle.

i plan to make all data points availiable so people can make their own plots for wear and capacity loss calculations if they want to.

personally i dont care about the rated capacity, i care what i can actually take out of it during regular use.
so i pay close attention to the actual capacity measurements from each run. that is the true capacity. what you can take out at 0.2C ro whatever is useless in a ebike or escooter setup. what are you going to do? drive 5kph because the battery died? no, this is a torture test to see how long it takes for the cells to crash below 80 and 70% capacity from the first run. that is my benchmarks for wear. beyond that point is academic for me.
this gives acual usage calculations were i can base my warranties and claims on for the packs i build. having actual data to support my warranty and expected lifespan helps me and my customers make a informed decision. note that some of these customers are not ebike users but companies that only care about hard facts and statistics, they dont give a shit about marketing wank, just reliabillty and life expectancy.
not looking at the business side of this i want your input to help this community at the same time. it might take a while but i hoppe it will be useful information and hopefully help people get more from their batteries.

the tester i want is rated for 10A charge and discharge so i have plenty of room to work in.

and if the tests take 9 months to complete then i am very happy, because that means the cells are REALLY good and with those loads they will live very long lives in just daily runs. 70% remaining capacity is the original limit for me but i probably will keep going and see how hard it takes a dive after that.

i have begun my testing with 4 NCR18650PF's and 4 NCR18650GA's.

each group has the following loads:

4.2>3 at 2A discharge and .7A charge
4.1>3 at 2A discharge and .7A charge
4.05>3 at 2A discharge and .7A charge
4.05>3 at 3A discharge and .7A charge

anyone willing to place bets when and wich one hit 70% of starting capacity first?

In order of most capacity remaining:

4.05>3 at 2A discharge and .7A charge
4.05>3 at 3A discharge and .7A charge
4.1>3 at 2A discharge and .7A charge
4.2>3 at 2A discharge and .7A charge

However you need to calculate 'equivalent full cycle life'

The number of equivalent full cycles is obtained by dividing the cumulative discharge (Ah) with true battery discharge capacity after every 50 or 100 full/partial cycles and adding the quotients over the entire cycle testing period.

So lets say the 4.05 V top of charge to 2.9 V spent delivers 2600 mAh, and the 4.2 V to 2.5 V delivers 3000 mAh.
After 100 cycles the equivalent full cycles for the lower SOC experiment is:

2600 mAh / 3000 mAh * 100 cycles = 0.866 * 100 = 87 EFC's

Yes and no.

Rated capacity is a useless figure compaired to the actual capacity you can draw from it in the test use case.
The first cycle sets the benchmark capacity. I mark the cell as worn when i lost 30% of that initial capacity.
Doing a 0.2C discharge test is fun and all but it does not mean anything in the real world.

In a full cycle you might get 2500mAh from it at 0.2C but if you load it down you only get 1600mAh in your application its kinda a useless number.

jonescg said:

+1 on temperature control - it's probably the number one factor in cycle life. If you can hold the cell temperature as close to 20'C it will live a long life. ....

Click to expand...

jonescg has been after this for a long time.
I trust his judgement like I trust Steves.

-methods

Accelerated aging can be used to move this test along.
Contact someone on LinkedIn and get a profile

By running at 20C, 40C, 60C, 80C...
By increasing the rate of thermal expansion/contraction
By pushing this limit or that limit...

You can get a reasonably accurate 10 year run into 10 weeks

...

As for simple battery cycling... there is enough data out there already to make on nauseous. I would spend some time digging thru white sheets or contacting battery nerds from EV companies like Tesla *

What is more interesting (to me) is prediction around "epic failure"
Basically... how much larger can we expand the "safe box" without worrying a (SNIP - GOT A CALL)

flippy said:

Yes and no.

Rated capacity is a useless figure compaired to the actual capacity you can draw from it in the test use case.
The first cycle sets the benchmark capacity. I mark the cell as worn when i lost 30% of that initial capacity.
Doing a 0.2C discharge test is fun and all but it does not mean anything in the real world.

In a full cycle you might get 2500mAh from it at 0.2C but if you load it down you only get 1600mAh in your application its kinda a useless number.

Click to expand...

True, you need a benchmark C-rate for capacity, and for most cells that's 1 C, but I see 18650s often quote 0.5 C or less. As long as all capacities are tested at that benchmark you're in the right realm.

of course you can use whatever reference test you want for capacity, as long as it is repeatable.
but since most manufacturers have adopted the 0.2C discharge rate as the defacto standard ,anything different means you will not be able to compare to any other test data, only your own

Hillhater said:

of course you can use whatever reference test you want for capacity, as long as it is repeatable.
but since most manufacturers have adopted the 0.2C discharge rate as the defacto standard ,anything different means you will not be able to compare to any other test data, only your own

Click to expand...

there lies the problem. all their datasheets has just lab grade wanking going on. 0.2C discharge? give me a break. totally unrealistic.
i want to know how they preform in REAL LIFE, not in a lab. and on that there is no real data out there.
why do a 0.2C 4.2V>2.5V run when that cell is never ever going to see that kind of cycle in his entire life? all batteries that i build have a mandatory 4.05 or 4.1V limit and a 2.8V (under load) bottom limit. how would the data from a full cycle run be useful in the real world? nice to compare it to whatever samsung or panasonic is telling you on the datasheet but its still completly useless data in the real world.

methods said:

Accelerated aging can be used to move this test along.
Contact someone on LinkedIn and get a profile
By running at 20C, 40C, 60C, 80C...
By increasing the rate of thermal expansion/contraction
By pushing this limit or that limit...
You can get a reasonably accurate 10 year run into 10 weeks
...
As for simple battery cycling... there is enough data out there already to make on nauseous. I would spend some time digging thru white sheets or contacting battery nerds from EV companies like Tesla *
What is more interesting (to me) is prediction around "epic failure"
Basically... how much larger can we expand the "safe box" without worrying a (SNIP - GOT A CALL)

Click to expand...

yeah, no. that hard loading of cells is not real world and it does not tell you how long the cell actually lasts in your use case, just on how long it lasts when you are torturing it. not useful data. but 1 cell is getting a 4A discharge wich is still well in spec for all cells but does wear them out faster. i just want to know HOW fast.

and yes, i plan to load the 4A cells right up until (SNIP - WHAT IS THAT LASSIE? THAT frocker IS STUCK IN THE WELL AGAIN?)

Have at whatever you like.

The principles of Accelerated Aging apply

The second order principles apply too (like COOL DOWN TIME) so get back to us when you are done calculating how long it will take to life-cycle the cells.

At higher discharge rates you will require more cool-down period
If 0.2C tests can be run back to back
If 1.0C test can be run near back to back
I would guess aggressive tests may take 2X as long? (not looking at data)

So to do 3000 cycles you are looking at 6000 hours
Per test rig

6000 hours / 24 hours = 250 days
YES?

So 1000 tests with cool down would be a third of that - maybe a hundred or so days?
Not as bad... but highly likely the test will get disturbed during that period

All I was trying to convey to you is that the trivial case has been long solved and that the methods and ways of accelerated aging would increase your value as an engineer (or aspiring engineer). The methods and ways would be of value to us too... hence my suggestion to fish around for expert donations of profiles

So...
Have fun

-methods

(not myspace profiles, facebook profiles, or LinkedIn profiles... "Aging Profiles")

Just for clarification

-methods


EDIT: The above is conservative and calculates in margin. Yes I get that a 3C discharge happens in 20 minutes and that it will probably return to nominal temp in 20 minutes - so still a 1 hr ish test? But thats PER STATION and most people dont invest in 4 or 10 stations... so round number rules. You may very well have 10 stations and the ability to get good science done fast. No idea.

sorry but no, physics dont work that way. (not exactly anyway)

doing a 20C does not mean the cell lasts 20 times shorter then a 1C rate. just like running a engine at redline does not make it last half as long as when it runs at half the rpm it can. same goes with cell chemistry, doesnt matter if it is lead or lithium in that regard.

accelerated aging works in specific use cases and needs to be preformed highly specifically. due to the nature of batteries you cant simply throw more C rates at it to "fake" cycles.
tesla proves that point every day by having cells in their cars that handle 5 times the cycle count panasonic's datasheet says.
how come? they also get hammerd but they also get active temperature management and only work in a specific voltage range and have the best BMS system money can buy. panasonic (or any other company) does not give you any data on 4.05>3V voltage cycles because that real world data simply does not exist. otherwise i would have found it already. there are a few papers behind paywalls that discuss the idea but as long as tesla's internal testing stay internal de dont know shit and need to do our own testing.

this is why i am doing these tests. now i dont -need- to keep it running until the cells die (but i probably will) but seeing the capacity loss under a certain load over time is what really is important. and i dont care that a worn cell can do 400Mah more at 0.2C when fully cycled because that is a meaningles figure when you are going to load it down at 2C and it instantly tanks below 3v under load. that is what is missing here. how good does a cell preform under real loads when it has done 500 cycles or more. that's helpful data and shows what the best cell is for you when you have used your 1200 euro battery for 2 years...

ps: i got 8 tests running. 4 with PF's and 4 with GA's.

here is the first 6 runs:

https://docs.google.com/spreadsheets/d/e/2PACX-1vRwD0YBdXAsu4puwHLjiFK0dQkpvleo2tzkmFNaZpmGErK-QCV8rs482RHU0IgzGHwqACdr9PRSbzMZ/pubhtml

you can see that the 4th run (full voltage range@4A) the PF's are only 200~250mah short of the GA's that cost considerably more.

yes, i know this will take effing ages but it will be real world and useful to this community (any beyond).

for the record: in 2 weeks i will take the system offline to install a UPS to keep the server and tester running when shit happens. data is automatically stored but its too much hassle.
UPS will be driven by a nice WAY oversized (lithium) battery.

flippy said:

sorry but no, physics dont work that way. (not exactly anyway)

Click to expand...

What physics does not work what way?
You are sounding like Luke.

Probably a miscommunication?

I made no statement that accelerated aging was simple and directly related to C-Rate - I just posted margined estimates on how long I thought non-accelerated aging would take (at a given C rate which directly relates to hours), and the means of accelerated aging I was thinking of would be extreme thermal cycling under vibration load - on top of the electrical cycling at whatever rate you chose - so as to stress the mechanical bonds (like seals) and chemistry.

I am familiar with Teslas R&D department and battery testing - have poked around in there and looked at the setup. No different than any setup (just bigger and better). I spent a decade at a lab where everything gets accelerated aging tests.

On to other things - just trying to get you to do free Phd research for us on accelerated aging profiles.

-methods

I just posted margined estimates on how long I thought non-accelerated aging would take (at a given C rate which directly relates to hours), and the means of accelerated aging I was thinking of would be extreme thermal cycling under vibration load - on top of the electrical cycling at whatever rate you chose - so as to stress the mechanical bonds (like seals) and chemistry.

Click to expand...

there is the problem. you can make ballpark estimations but depending on how fast you do the age acceleration you go pretty fast beyond the point where the results are no longer relevant. that is key here. and especially with batteries you go beyond resonable estimations pretty fast. low C rate wear levels are simply not publicly (or paid) to read. the data is at tesla's R&D department but no way they will give analysis data they pull fromt he BMS in all their cars and just give them to me. so this is why i am biting the bullet (and bill) to buy my own testing equipment and to the testing myself.
i doubt it will be Phd level but the data will be more then useful for at least the people here and more importantly: help me make better battery designs moer suited to the customers specifications.

one thing i wont do is crank up the settings just to see a result faster. that would invalidate the entire thing.

Awesome - have fun.

If you are looking for aging profiles you can find them at social gatherings like "lunch" and the like. You just have to make enough rounds (picture bit swarming - there are only 100pcs to the puzzle) to get enough random pieces, to frame out the border (bounds) and internals, to guess at the big picture and get close.

Of course - it would be costly and lame to try and bruit force this information from scratch.

Stoked to hear that you are spending your time and money finding out

As for lunch - I suggest:

PB&J
Shrimp coctail
and vodka redbull

That's the "bait shop special" which is one of my favorites.... a lunch of champions.

-methods

yeah, i think you lost me there. i am talking about cycling cells, not what to have for lunch.

Flippy, i assume you are aware that there have been several detailed threads on cycle life tests for various 18650s.
They are tricky to find again (dumb search system) but i believe Padja, Okashira, youyoung, electricbike, and a few other posters were involved..
Examples
https://endless-sphere.com/forums/viewtopic.php?f=14&t=73883
https://endless-sphere.com/forums/viewtopic.php?f=14&t=85113&p=1278596#p1278596

They may help you avoid problems, or even short cut some tests.
Good luck anyway.
HH

flippy said:

yeah, i think you lost me there. i am talking about cycling cells, not what to have for lunch.

Click to expand...

Thats what happens when you skim
Let us know when you have some pics to show progress

-methods

What are the top three choices when shopping for an 8-channel 18650 cell tester?

Affordability would be nice, but a reasonable amount of accuracy is required to make the results useful. I'm thinking 1A drain, then 2A, 3A, etc...