Behaviour of cells in parallel take 2

Wolf

Moderator
Joined
Sep 25, 2018
Messages
2,004
Another experiment has taken place at the Wolf's PowIRwall location.
Previous posts on this subject are at https://secondlifestorage.com/showthread.php?tid=7846
I dragged out my 4p testing rig and decided to test 4 ICR18650-28A cells that had a reasonable IR at ≈ 50mΩ but their mAh testing was in the low 2000mAh with the SKYRC tester.

cell1 4.176V 52.74m? 2089mAh
cell2 4.178V 51.20m? 2141mAh
cell3 4.178V 53.73m? 2072mAh
cell4 4.180V 53.30m? 2043mAh

As predicted the discharge of all cells was relatively even as the IR and mAh were closely matched so no cell was stressed more than the other. Temperature did rise about 4C toward the end. Temperature on cell 1 and cell 4 was less as they did not have a cell on both sides such as cell 2 and cell 3.

image_yplnej.jpg

Since these cells have a spec sheet of max 4.30V full charge and cutoff at 2.75V. I will run this test again with those parameters and see what happens under those circumstances.
Pictures of my testing lab.

image_molgzr.jpg

image_negchu.jpgimage_yhawkv.jpg


Wolf
 
Last edited:
Hey Wolf

Im currently in the process of assembling my 80-100p packs to get an even overall capacity.
something that came into my mind was the following:

I try to bring my cells together by internal resistance and not by there capacity to minimise the loadsharing and balancing curent between the 80 cells in one pack.
But the thing is the following:
If we have 2 cells with the same internal resistance but different capacity how does the loadsharing look like?
Cell 1 has 1500mAh
Cell 2 has 3000mAh

Do you have a nice graph to show that? But maybee I just need another coffee to find the solution ;)
 
PAF,

Good question.

Let me see if I can locate 4 cells with that criteria and run a test.

Wolf
 
the 3000 will in average have 2x the current. BUT this depends on the internal of the cells and will if different vary over time.
Ie in beginning one have 1.5x more than the other but in the end perhaps 2.5x

Several tech sheets out there talk about this and its also pure electrical theory.
 
A Comprehensive Approach for the Clustering of Similar-Performance Cells for the Design of a Lithium-Ion Battery Module for Electric Vehicles

https://www.sciencedirect.com/science/article/pii/S2095809919307829

The SVC algorithmIn general, a support vector machine (SVM) is adopted for clas-sification (supervised learning). SVC is a slightly different algo-rithm from an SVM. In fact, SVC is an unsupervised learningclustering algorithm.The main idea of SVC is to map data space to a high-dimensionalfeature space using a Gaussian kernel function. Next, a sphere witha minimum radius is obtained and the sphere contains most of themapped data[41,42].

and

Analysis of the Effect of the Variable Charging Current Control Method on Cycle Life of Li-ion Batteries

https://res.mdpi.com/d_attachment/energies/energies-12-03023/article_deploy/energies-12-03023.pdf

Both of these are worth a read ..
 
PAF,

I have managed to find 4 cells of similar IR but widely differnt mAh.


Here they are:

image_xnkxxe.jpg

They both have a maxV of 4.2 and a 2.5V cutoff so we will run them trough the gamut.
Let see what the Discharge tester will show.

Wolf
 
Those are made to be charged to 4.35V, so you can actually get a little more mAh out of them.
 
Hm maybe I got the wrong datasheets.
This is what I came up with.
Maybe you have some different ones that you could share with us.

UR18650H

image_yxbshi.jpg



LGABC41865 (INR18650C4)


image_bombro.jpg


Let me know if I got it wrong.

Thanks

Wolf
 
OK here are the results of the 4 cells with similar IR and different mAh results.
The discharge current was set at 2.6A cause that is as high as my ZH-YU goes which is 650mA per cell in a perfect world.

After cell insertion there was the normal jockeying for position between cells as they started to equalize.

image_pybvqf.jpg


When discharge started the V drop on all cells was pretty even and to my surprise the mA draw was also pretty even. Cell 1 was a little more aggressive but all in all not bad.

image_ffkvlp.jpg

The first crossover occurred at ≈ 3.73V Cells 1&2 (LGABC41865) relaxed a bit and gave over to Cells 3&4 (UR18650H)

image_yqtnqs.jpg


The next crossover occurred at ≈3.57V and Cells 3&4 started to give up

image_fwtqwu.jpg


And finally at the end 3.0V Cells 1&2 where taking the brunt of the discharge at 655mA and 1.5A respectively while 2 & 3 where exhausted.

image_prevva.jpg


What's an interesting finding is that cell 2 with the lowest mΩ reading of 49.4 was slow to start (not by much) but the best performer in the end.
That difference is a miniscule 4mΩ between the lowest and highest cell.

Full chart with temperature. (Note: Cell 2 also was the highest in temp not surprising as it was carrying 75% of the load in the end.)

image_lreure.jpg



Charging charts coming up

Wolf
 
Last edited:
Hey Wolf, thanks for this test, its realy interesting altought i dont understand the results completely.

From the start down to 3,5V it is as I would have expected it, maybe a slightly off with the cell 2 cause of the lower IR togehter with the lowest initial current draw.
But the end is crazy, why is there such a big difference?
Is it due to the different chemistry?
Does the IR change with the SOC?
Why is there also this big gap between Cell1 & Cell2? They are the same cell with a close mAh test and "only" 10% difference in IR.
Additionaly Cell 2 has the highest "bounce back" when the dischargetest is done.

This test brings up more questions than it answered. :huh:

First of all I will check if i can get more Informations from the datasheets.

PS: by checking your Screenshots I saw, that Cell 1 is labeled as "18950" Cell in the fourth row.
 
PAF said:
PS: by checking your Screenshots I saw, that Cell 1 is labeled as "18950" Cell in the fourth row.
Details details,
Always the small details that get me into trouble. Fixed

But the end is crazy, why is there such a big difference?
Is it due to the different chemistry?

No I dont think its the chemistry I think that the UR18650H have run out of juice and if it weren't for the LGABC41865 holding the Voltage up they would have died a long time ago. ≈1850mAh 650mA =2.8 hrs.The test started at ≈19:40 and ≈ 22:40 they were done, just supplying a mere pittance of ≈260mA to the overall draw. They did hold up their weight though for those 3hrs.

Does the IR change with the SOC?

Good question. I am recharging the cells as we speak and will run the test again with the cells in reverse slots as in 4,3,2,1first to see if there is a variance in my tester and also to confirm the results. I will test the IR of the cells after the discharge has occur and see if there is a difference. I do know that the IR of a charged cell vs a discharged cell can vary quite a bit. We shall see.

Why is there also this big gap between Cell1 & Cell2? They are the same cell with a close mAh test and "only" 10% difference in IR.
Additionally Cell 2 has the highest "bounce back" when the discharge test is done.

Ah yes the ending that is a good question and unfortunately I do not know the answer to that. One thing is for sure though at that time the UR18650H cells are really out of the picture and the LGABC41865 are supporting the current draw of 2.6A at that low voltage and SOC anything can happen as the cells are giving up their last gasp of energy. IR certainly can have a huge effect at that time.The bounce back is not unusual and especially at that discharge rate and that low of a voltage.
Also they are used cells so....................who knows.



In the meantime here is the charging curve of the cells.


image_ljycjc.jpg


Wolf
 
Last edited:
Looks like it should be and as expexted as well :)

As i tend to say IR is not the sollution to everything and it doesnt 100% state that it will work like the cell next to it. Its simple since they arent used the same. But IR do give a rough estimate on how they can potentially perform. But 2nd hand cells are always tricky.

For it to behave close to equal they must have been abused to the same degree and not over abused either. Because over abuse they start to drift.
Manufacturer challenges is what they call it :)
 
Also this is an extreme case scenario just to prove a point that you really shouldn't mix cells with that much of a mAh variation.
Although I must admit the IR was definitely an influence in the first 3 hrs as the cells discharged pretty much even.
I have charts that show that with all 4 cells being the same.The discharge curve is very even.See the first post

After the 3 hrs though.. there we get to see the true colors of the cells in this scenario.
Essentially you are talking a difference of ≈1Ah between cells
I dont know of anyone who would put that combination into a pack.



Wolf
 
Last edited:
The combination can be ok. BUT you need to define that the worst cell never exceeds your max basically. For instance 3ah with 1ah cells the 3ah will take most load. Can it take it is the question?

With that said best match is ALWAYS capacity and Ir and type. The closer to its edge the more important it becomes. Nothing new there.
 
Exactly
If you do decide to use this type of setup you are limited to the lowest mAh cell as that cell after empty will stress the rest of them to take up the slack. So for instance you have a 1000 mAh cell in parallel with a 3000mAh cell and you draw 1A on the "Pack"and their IR is the same.
you have 2 hrs. before the 1000mAh cell is depleted and will now be drawing from the 3000mAh cell putting more stress on that cell.



So here is the chart with the same cells in reversed slots also the IR of the cells after the discharge.
Other than the IR of the cells changed a bit after discharge esp. the LGABC41865.


image_cfzcpt.jpg

The results are pretty much the same no surprises. How boring.............
Maybe I will test low capacity with low IR and high capacity with high IR next?

image_tftcui.jpg


Wolf
 
Last edited:
Test of 2 low IR low capacity Sanyo and 2 high capacity high IR LG.

image_iqrclc.jpg


So we can see that the IR is not very critical initially as both type of cells respond to the load in a pretty even manner till about 2hrs into the test.
at 3.56V the separation starts to show up and grow from there.


image_feeyfe.jpg


Will post recharge graph when done.

Wolf
 
Its interresting that the difference comes only in effect after the 3,7 "flat part" of the discharge curve.
Lets try to get my tought into a different language :)

conclusio for myself and maybee also a bit of a theory: the Internal resistane doesnt really matter at such low C rates if you use the capacity from the smallest cell as your "weakest link in the chain". That explains why most people in here have no or only little trouble with there Powerwall that uses only capacity checked cells.
the IR is nice to preevaluate if the SOH of a cell is still good but for the actual battery build the capacity is more important.

just checked quickly your Moli discharge graphs and this does not completely support the upper theory because the Moli delivers under 3,6V more current.
next try (theory): the IR doesnt matter from 4,2 down to somewhere around 3,6-3,5V at such low C rates. But under this "magical" line the IR does play a roll as we can see in the Moli tests and also in this 2 tests. This brings me back to one of my earliest questions: does the Chemistry play a role. I think now "yes", because if the discharge curve is flatter you have a "higher" voltage in a low SOC and therefore this cell delivers more current.

Can you make a test with low Cap & high IR and high Cap cell with low IR? (If you have done this test already please give me a gentle push into the correct direction.)
Do we see a real difference in the initial discharge current or is it just as we have seen here because the current that we draw from the cell is to low (load is ~6Ohm vs 0,05Ohm)?

thank you wolf for all this testing, these are independent tests which are bringing up some interesting questions.

@daromer: if you have the answers to my questions/thought please throw them in, this way wolf has more time to process cells and safe cells from the recycler :)
 
All things play in. And thats i promotes to test the cells at a current high enough. They Will then later differ during a Charge or discharge process but as Long as you stay well below thats not a factor that Will cause issues.

The difference Will also over time level out IF ir increases

So lets say you stay att 300mA max the Effecta is even less. Add on top that your average is around 50mA of lower... Add upp that a powerwall should not be cycled fullt but just small dod cycles .. Id you add upp all those factors the problem that vmcan be seen during above tests are for most People irrelevant.

In ebikes, or high performance packs its something else
Example when i did Stone racing we Always tested all our packs for ir. We even pre-heated the packs before a race ;) quite a difference. And why we preheat? Beacause its related to discharge capabilities... :)
 
Back
Top