Caltech/NASA 18650 performance test

Batteries and Chargers Rechargeable Batteries
1

Caltech/NASA tested the performance of 18650 batteries from “multiple high-quality manufacturers, including the MJ1, M36, HG2 models from LG Chem; the 35E, 36 G and 30Q models from Samsung; the BJ and GA models from Sanyo/Panasonic; and the VC7 from Sony/Murata.” They concluded that “the LG MJ1 cell has outperformed the rest, delivering the highest specific energy, best cycle life, and the most capacity even at 9.6 A discharge rate.”

Here are the results of some of the tests,


at “100% DOD” (3.00-4.10V),


at 40% DOD,


and at 20% DOD.

These results show the opposite compared to the popular opinion on the best high capacity 18650 battery. It would be interesting to see the tests that drove popular opinion and compare them to the ones in this study.

2

Those tests were focused on capacity retaining capabilities after exceeding manufacturer cycles claim, while in the flashlight hobby we need a good balance between sustained current output, cell voltage and capacity.

I find very little capacity difference in real world use when using 3,000mAh medium drain cells such as VTC6 or 30Q vs high capacity 3,200-3,500mAh cells. Output capability in the lower SoC is more important than when its full, at least for me.

3

If I read the article carefully, it should be noted, that in all graphs the indicated 100% DoD is related to the range 4.10-3.00V (~90% of actual cell DoD) and the C-rate used in cycle life test is 0.2C for both charge and discharge.

4

Tests from both HKJ and Mooch showed that the GA had the highest voltage under load and MJ1 the lowest, the 35E had also more sag at 10A, since then the 35E had two revisions and the 35E3 now has the lowest DCIR and no more sag at 10A, considering its price is usually lower than GA and that cycle life seems to be better it is IMO the cell of choice for flashlights.

5

Great find! Lots of data to look at in that paper.

This test by NASA focuses on long term performance. Generally in this community we have only had access to test results of brand new cells, thanks to the efforts of HKJ, Mooch, and sometimes a few others. Focusing on different criteria yielded different results.

In HKJ’s tests, the Sanyo NCR18650GA handled high discharge rates slightly better than either the Samsung INR18650-35E or the LG MJ11865.

Note that HKJ tested the Sanyo and LG both once, in 2016. He actually tested 3 sets of the Samsungs - one in 2016, and two in 2021.

At 10A continuous discharge:

  • The MJ1 samples reached the 50 degree C temperature limit at 2500 and 2750 mAh respectively
  • The 35E samples terminated for voltage (planned) at 3000 mAh or more
  • The 18650GA samples terminated for voltage (planned) at slightly over 3000 mAh

At 15A continuous discharge,

  • The MJ1 samples reached the 50 degree C at 1500 mAh
  • The 35E samples reached the 50 degree C limit at 1700-2100 mAh
  • The 18650GA samples reached the 50 degree C limit at 2000 and 2100 mAh respectively

The 18650GA appeared to maintain the best voltage under load, but the 2021 test of the 35E indicated Samsung improved this cells performance over time, so that the latest sample tested was almost a perfect match for the 18650GA in this regard.

So from HKJ’s tests, the 18650GA appeared to have a slight advantage at high drain rates. Not enough to make any of them a bad choice, but it helped set the impression. At lower drain rates (5A and below), the performance of all three cells was almost identical.

I think Sanyo also had a perception as in general more of a battery specialist, where as for Samsung and LG, batteries were just one of the countless businesses they are in. I suspect that also helped give Sanyo a good image.

The NASA tests at low drain rates (0.2C) seem to suggest the MJ1 wins for cycle life, the 18650GA wins for low temperature performance, and the 35E strikes a middle ground in between.

The NASA high drain (9.6A) appears to have been an add-on test performed on the same cells used in the “100%” DoD cycle tests, only for the LG MJ1 and M36, Samsung 35E, Panasonic BJ, and Murata VC7 (“the first batch of cells”). This would explain the discrepancy between HKJ’s test showing over 3000 mAh for the 35E at 10A, where as NASA only saw only 1400mAh for the 35E at room temperature compared to 2900 mAh for NASA’s low drain test of that cell.

6

I noticed that Pajda got very different results on the cycle lives of 35E, MJ1 and GA compared to the paper. Can Pajda please comment on the discrepancy?

7

My takeaway is don’t charge above 4.1v. Do not discharge below 3.0 volts while under load. That means a resting voltage of somewhere between 3.4 or 3.7v depending upon load unless only using very low settings. That’s pretty much what I’ve been doing for a few years with mostly MJ1’s. I don’t want to leave the house with my EDC lights, or any lights, at 3.7v or less. Turbo doesn’t work in any of my lights at 3.7v. I’m not necessarily looking to get 6500 cycles but keeping batteries healthy is a good thing. If your use case requires you to draw it down lower then go ahead and do it. If you don’t have a good reason to, then don’t.

8

I meant Padja found that GA has the best cycle life out of the three, but the paper found that GA has the worst cycle life out of the three.

9

Well, the GA are the best high capacity cells I’ve used for all-around performance. I’ve been using them for years in FET drivers. The 35E3 matches it once the voltage gets down to around 3.8 volts, and from that point on they’re all about equal…3200-3300 mAh at lower current. It’s the higher currents that the GA excels like FET driven or buck/boost drivers that have a higher input current demand. If you want high capacity, high current the VTV6 is still the best, but at a life cycle deficit. The HG2 is old technology and wss good in 2015, but there’s better options. I’m sure Molicel has a rival for that though.

10

Presumably one of the reasons for the different result is because Pajda tested them for 4.2V to 2.5V at 1C discharge rate, while JPL tested cells from 4.1V to 3V at 0.2C discharge rate.

Additionally, they had different test end points. The NASA test showed the GA maintaining around 90% capacity after 600 gentler cycles. Pajda says above 70% capacity after 1000 somewhat harder cycles.

If you look carefully at the charts A and B you posted for JPL’s “100%” DoD test (4.1 to 3V), at 600 cycles, the 35E has maybe a little over 2.6 Ah, compared to its initial 2.90 Ah available over this voltage range. The GA has a little under 2.6 Ah at 600 cycles, compared to what 2.83 Ah initial.

So in the 4V to 3.1V discharge test, they tracked really closely - both of them right at 90% of initial capacity after 600 cycles.

Pajda also indicates he saw the GA initially drop relatively quickly in capacity, but then leveled off. Most of that effect must have been at a greater capacity loss than JPL reached in their tests, but there is a hint of this in the JPL report on page 12, Table IV, where it shows capacity loss by test segment.

From cycles 0-100, the GA lost 48 mAh, while the 35E lost 24 mAh. From cycles 200-500, however, the GA only lost another 9 mAh, while the 35E lost 15 mAh.

Where the 35E ended up looking really good was in the 40% and 20% depth of discharge tests. In those tests, the 35E definitely retained more capacity than the GA. Pajda similarly stated he felt all the cells do well when cycled at less than 50% depth of discharge.

11

Yes, battery cell testing is a thankless job, particularly cycle life testing. The main problem is that manufacturers are upgrading their cell model line and this innovation is not always announced. For example, Samsung is pretty good at labeling new variants, so here we have the genesis 35E, 35E2 and 35E3, or 30Q(136), 30Q(141), 30Q(6) version. And Samsung makes significant progress between variants. So the result which I comment back in 2018 was for 35E(1) version from ca 2016. Unfortunately or rather fortunately? the newer 35E2 version is significantly better in cycle life.

Here you can check the results from 2020 and newer Cycle life tests of High Energy density cylindrical cells | Endless Sphere DIY EV Forum. I am almost certain that in 2-3 years these results will be outdated again due to newer versions will enter production.

12

Thank you for your contributions to battery testing. It’s good to see that battery technology has kept improving.

13

I found a test that got similar cycle life results to the paper.

14

Would be interesting to see how the new(ish) Molicell M35A stacks up.

15

Molicel HE cells are not as good as their HP ones. Both M35A and M50A are significantly behind the LG and Samsung HE cell production line in cycle life. I heard somewhere that Molicel is aware of this and therefore introduced the P45B model, which does not aim on the highest energy density available, but on the best available power density on the market with 3C continuous fast charging capability.

16

Thank you for the information. Good to know for future reference.