Cycle life testing : LG INR18650 MJ1 vs Sanyo NCR18650GA

[youyoung21147]

***************Updated April 2017 for 800 cycles*****************



Both cell models are the current state-of-the-art in terms of high capacity 18650 Li-ion.
Although 18650GA gained popularity in the ebike community due to better pricing/availability (true in 2016, not in 2017), the LG MJ1 has been found superior by NASA in their applications (safety, cycle life and impedance)
Samsung 35E is a very promising cell also, but not included in this comparison because of its higher price back in Sept 2016.

In spite of extensive bibliography, I couldn't find any relevant or comparable information about how they would behave in a proper electric vehicle application (commuting ebike, motorcycle, cargo lift...).

I decided on the following parameters, which I think represent a pessimistic scenario of someone educated to battery care but also placing high demand on the batteries because of the application.
- Charge voltage 4.01V (yields 3.98V rest voltage, should enhance cycle life greatly. Satellites use 3.85V)
- Charge current 1.3A to 75mA (full charge in a little less than 3h)
- Discharge current 6A continuous (imagine climbing a hill with a DH ebike, or cells in the middle of the pack which cannot cool down properly)
- End of discharge : 5A at 2.8V (yields 3.2~3.3V rest voltage)
- 30-minute pause after charge and discharge operations (to let the cells cool down and stabilize the chemistry, which Sony has found to slightly improve cycle life)

CULPRITS
Unfortunately SD card crashed during the first cycles, so initial capacity is not known. Paper and pen used for the data shown.
I tested a new 18650GA in the same conditions and found 2451, 2400mAh for the 2 first cycles. I don't have any new LG MJ1 cell yet.

If I had to do it again I would probably chose Kokam's recommended parameters for cycle life (4.1/3.3V), because these results are disappointing

CONCLUSION
- This cycle represents a ~70% DoD scenario for the Sanyo cell, and ~66% for the LG cell
- After 300 cycles, 18650GA performs better in this scenario despite being a C rank cell. It's not being aged faster in spite of being discharged more.
- After 800 cycles, the GA suffers from increased internal resistance, reducing its capacity in this usecase. The LG is more steady but loses more capacity under slow discharge.
- Both cells are disappointing in terms of cycle life, having lost ~15% of their initial capacity after only 50 partial cycles. Things calm down and the trend becomes linear after that but past that the cells premium price becomes less interesting. A 18650PF still holds 2500mAh after 300 cycles @ 100%DoD, 4A (according to datasheet)

Will keep updating the results and hopefully reach 1500+ cycles !

 

[Hillhater]

...
- Both cells are extremely disappointing in terms of cycle life, having lost ~15% of their capacity after only 300 partial cycles. A 18650PF still holds 2500mAh after 300 cycles @ 100%DoD, 4A (according to datasheet)....

you should not compare your results with some other data sheet info , compiled under different test methodology or protocols.

....Samsung 35E is a very promising cell also, but not included in this comparison because of its higher price back in Sept 2016.

the Samsung 35E seems to be similar , if not lower cost than these two cells ..https://eu.nkon.nl/rechargeable/18650-size.html

 

[Sunder]

The only disappointing thing I can see in all of this, was the test methodology.

Basically, the "loss" is less than the margin of error, which means we can't conclude that the battery lost any capacity within the testing.

Eye balling it does suggest that perhaps the averages are falling - but again, all within the margin of error, so it could be coincidental.

Appreciate the effort - I wish more people would do this - but a better methodology is needed.

 

[Overclocker]

thanks for the effort but you really need to improve the methodology

the sampling interval is very erratic. why?

i have attempted some limited cycling using an icharger 106b+. sampling interval of 8, which occurs at approximately the same time during the night, so minimal temperature fluctuations. the trend was always downwards. i've never seen it go up.

 

[youyoung21147]

thanks for the effort but you really need to improve the methodology

the sampling interval is very erratic. why?

i have attempted some limited cycling using an icharger 106b+. sampling interval of 8, which occurs at approximately the same time during the night, so minimal temperature fluctuations. the trend was always downwards. i've never seen it go up.

Sample rate is erratic because I have a problem with SD card on iCharger. So whenever I stop the test to charge my ebike batteries, I report the number of cycles and last recorded discharge mAh value in Excel.

Up-downs are correlated to temperature. This is very obvious : when the heating is on during the day and when I removed the cooling fan after first 120 cycles, the capacities go up. Also I noticed that capacity goes slightly up when the cells rest for a week or so without cycling (like 20-30mAh)

I have updated the graph with a more realistic scale : keep in mind the fluctuations are in the 50mAh range so margin of error, even if high, is not unacceptable to estimate ageing trend. At 500 cycles I will perform full charge/discharge cycle per manufacturer specification : this is how the cells are actually rated.

 

[TRRRR]

Thanks for the data YouYoung. I think it actually shows quite a bit of information and from the way I read it the cells have so far only lost about 6-7% of capacity after 350 cycles which seems ok.

Have you done the full cycle tests 4.2v-2.5v after 500 cycles?

Do you have any updates to the data?

 

[TheBeastie]

Great to see more cycle testing, we don't get to see many of these.
Last one I remember was from Electricbike https://endless-sphere.com/forums/viewtopic.php?f=14&t=76048&hilit=+cycle#p1148499

Anyway, your results seem about right to me.
To me when the specs say "Max Discharge 10amps" etc this really means max discharge and it won't instantly die or burst into flames but you can forget about it having a lot of cycles left in it.
Just look at this chart made by the manufacturers themselves trying to push this point across for the very latest 20700 cells, the capacity left after 500cycles in 8a vs 12a discharge is massive.

 

[neptronix]

I like this idea, but not the proprietary testing methodology. The results cannot be used to compare other cells. Using a shallow cycle depth is the biggest problem with tests such as these. And the data is erratic.

A similar testing regimen for the GA exists, but it is using 100% SOC. Sanyo's test sheet shows that it hits 80% capacity in around 125 cycles. LG does not provide the same kind of detailed graph for their cell.

 

[Zulu80]

So the GA has better cycle life at 300 cycles than the MJ1 ? I just got a battery with these Sanyo GA cells

 

[peleap]

good topic

 

[youyoung21147]

Will update the graphs tomorrow with about 700 cycles.

Surprisingly the LG now performs comparatively better in the "proprietary" test cycle, but the GA recovers more capacity in the standard test (0.2C). Probably, the LG predominantly suffers from capacity loss and the GA suffers from internal resistance degradation.

Depending on the application, you could prefer one or the other cell but honestly they are tight in performance. LG cells are said to perform better in the cold but I have no data to back this. In the meantime, NASA has found them safer because the side walls are thicker. Price is IMO the most determining factor in choosing one or the other. In August 2016 the LG was considerably more expensive but the trend seems to have reversed now.

In fact the saddest thing about these cells is that they lose capacity heavily in the first 100 cycles, rather than gradually.

 

[RuiP]

I signed up in this forum to thank you for your effort!!!
Did you go past 850 cycles?
Would be great to see those 1500 cycles!

 

[DVDRW]

Nice tests.
Building Arduino based 2x8 channel cycle tester myself.
All truth will come out

 

[youyoung21147]

I stopped testing at ~1100 cycles.

Interestingly the capacity retention under manufacturer's testing conditions is excellent. Less than 15% degradation in 1000 cycles.

For LG MJ1 the Coulombic capacity degradation is predominant
For Sanyo GA the internal resistance degradation is predominant

The LG cell seems better suited for cycle life at high discharge rates ! Totally opposite conclusion compared to 0-cycle testing.
The GA would perform great in stationary applications, if it was not so expensive ! It also seems to be a great choice for commuting ebikes : partial charging/discharging and moderate discharge rate.

But for relatively high discharge rate applications both cells are a no-go. Too little usable energy left because of the partial charging and dramatic increase in internal resistance. I am now testing lower capacity cells at higher charge voltage to see how they behave.

Will post the results when I have gathered sufficient data !

 

[Pajda]

Thanks for this test results! Very useful for me.

 

[madin88]

Thanks for sharing the test results and your work.

Compared with the MJ1 the GA has lost a lot more of the usable capacity between 4.00 and 2,8V, but between 4,2 and 2,5 it is above the MJ1.
kinda strange.

To improve lifetime of my batteries i personally charge to 4,1V - 4,15V intead of the 4.00V of your tests, and i stop discharging at about 10% SOC (idle voltage of about 3-3,3V dpending on the type of cells).
Most high current cells like Samsung 25R are almost empty at 3,3V while high energy cells still have plenty of mAh at 3,3V.

Are you able to meausre internal resistance?

the Vapcell YR1030 would be a nice device:

https://de.aliexpress.com/store/pro...d-batterie-ladeger-t/1800579_32835523320.html

https://www.youtube.com/watch?v=sULMqNfGcUA

 

[TRRRR]

Thanks for the updated results

"The LG cell seems better suited for cycle life at high discharge rates"

Could you confirm this by doing a complete discharge at 6A? ie 4.2 to 2.5V @6A

 

[daveching]

Hello people,
Would you by any chance have any updated (current) life cycle testing data?

 

[Abhineil46]

Thanks for the data YouYoung. I think it actually shows quite a bit of information and from the way I read it the cells have so far only lost about 6-7% of capacity after 350 cycles which seems ok.

Have you done the full cycle tests 4.2v-2.5v after 500 cycles?

Do you have any updates to the data?

What was your Current Discharge/Charge Rate.?