AW IMR Cell Testing

[Battery Guy]

Greetings Everyone

I purchased a number of AW IMR cells a month or so ago for the purpose of doing similar tests but had not gotten around to it. Well, last Friday I found myself standing in front of my Maccor battery tester and inspired by DFiorentino's recent post showing his test results on a variety of IMR cells. So I grabbed my box of AW IMR cells and went to work. I know that this is somewhat of a duplication of DFiorentino's efforts, but I already had the cells, so I figured I might as well run the tests.

Here are the cells that were tested:

14500 (rated 600 mAh)
16340 (rated 550 mAh)
18350 (rated 700 mAh)
18490 (rated 1100 mAh)
18650 (rated 1600 mAh)
26500 (rated 2300 mAh)

Here are some of the test details.

Test Setup

All tests were performed on a Maccor Seriers 4300 battery tester. The Maccor has separate leads for voltage sensing. If connected to a cell properly, this eliminates voltage errors due to contact and lead resistance. All cells had two nickel tabs spot welded to each lead. One tab was connected to the voltage sense lead and the other to the current leads. In this way even the resistance of the spot welded nickel tab should not affect the results.

Test Procedure

All cells were subjected to one initial "break-in" cycle where they were charged at a C/2 rate to a C/20 cut-off, allowed to rest for 1 hour then discharged at C/2 to 2.5 V. This first break-in cycle is not reported in any of the curves below.

Following the break-in cycle, all tests were performed with the following procedure:

1.) charge at C/2 to 4.2 V with a current cut-off of C/20
2.) rest for 1 hour
3.) discharge at specified rate to 2.5 V
4.) rest for one hour
5.) repeat steps 1-4 with a different discharge current

Limitations

One cell is hardly what I would call a statistically significant sample. There is certainly going to be variability in cell performance, so please recognize this limitation. Unfortunately, I do not have the time to repeat this work on multiple samples. These results are what you might expect for new cells. Exactly how the cells degrade will depend on the cell design, manufacturing quality and your use/abuse pattern.

In other words, your mileage may vary.

Presentation of the data

It is always difficult to figure out how to present a lot of data so that it is useful for a broad audience. In an attempt to cover my bases and provide something for everyone, I am presenting Ragone plots and discharge curves. I decided to present the discharge curves in two ways. The first series of plots represents one cell size per graph and shows discharge curves for the following conditions:

0.5 A, 1 A, 2 A and 4 A (14500, 16340 and 18350)
0.5 A, 1 A, 2 A, 4 A and 6 A (18490)
0.5 A, 1 A, 2 A, 4 A, 6 A and 8 A (18650)
0.5 A, 1 A, 2 A, 4 A, 6 A, 8 A, 10 A and 15 A (26500)

The second series of plots is an attempt to compare the performance of the cells in a normalized way. These plots show the C/2, 1C, 2C, 3C, 4C and 5C discharge performance normalized to the rated capacity of the cell.

Ragone Plots

If you are unfamiliar with Ragone plots, see my Intro to Ragone Plots thread.

Ragone plots are great ways to represent a lot of information in a compact form. If you have a regulated light, then Ragone plots are going to probably provide you with all of the information you need. I discuss how to use these Ragone plots in post #24 further down in this thread.

The first plot below shows the actual power vs energy response for all of the cells. The second plot is normalized to cell volume.

Please note that these plots are currently incomplete. I still need to test the 18650 and 16340 cells, and I need to run the 18490 and 26500 cells at higher discharge power to establish their power limits. I will update these plots as additional data is collected.

Constant Current Discharge Results per Cell Size

Constant C-Rate Discharge Results Normalized to Rated Capacity

I find this last plot showing the 5C rate comparison very interesting. It is clear that the 16340 cell does not hold up as well as the others. You also see that performance of the 14500 and 18490 are very similar, as are the 18350 and the 26500. It might mean that these similarly performing cells are made by the same supplier, or it might mean nothing at all.

I hope that you guys find this useful. If there are any additional discharge conditions you would like to see, let me know and I will do my best to add them to these plots.

Cheers,
BG

 

[Kestrel]

Very nice, thank you.

I find this last plot showing the 5C rate comparison very interesting. It is clear that the 16340 cell does not hold up as well as the others. You also see that performance of the 14500 and 18490 are very similar, as are the 18350 and the 26500. It might mean that these similarly performing cells are made by the same supplier, or it might mean nothing at all.

Any similarity/differences in appearance regarding the positive nipples? :thinking:

 

[Battery Guy]

Very nice, thank you.


Any similarity/differences in appearance regarding the positive nipples? :thinking:

I did not look close. When I get a chance I will take some photos and post them.

 

[kiely23+]

very good!
thank you... :thumbsup:

 

[badtziscool]

thanks for this very informative post. I think the most interesting thing is the comparison between 16340 and 18350. There's been quite a few legos put together that is using a 16340 to power triple and quad xp-g dropins and it's clear from the graphs that pushing the cell at 4A is just too much. Of course some of these dropins are capable of going well over 4A.

 

[45/70]

Nice job, BG.:thumbsup:

I've been using LiCo cells for quite a few years, but only recently have acquired four of AW's 18650 IMR cells. I have just never needed what these cells are best at, offering good performance at high current rates (1C+). It will be interesting to see how the IMR 18650 cells perform.

In the meantime, I am now also in need of IMR 16340 and 14500 cells. Unfortunately for this particular application (Jetbeam RRT-0, which can utilize either size cell), they have to have at least a "button top", if not an actual "nipple". This kind of narrows the choices down quite a bit.

I'm torn between using AW's "protected" LiCo cells, or using IMR cells. The LiCo cells work great, but I really don't like running them at 1.7A. On the other hand, in addition to this light incorporating the "cheap" solution to polarity protection (a recessed positive contact, instead of a diode), it also offers no over discharge protection, as most of my other lights do. These two "features" are unbecoming of a $100 pocket light, IMO.

Anyway, thanks for performing all the work involved and providing the resultant information. I will keep an eye out for the 18650 update. From experience, I'm expecting they will also perform well. I'll also add, that considering the size of a 16340 cell, I'm not too surprised they falter a bit at high current.

Dave

 

[Battery Guy]

In the meantime, I am now also in need of IMR 16340 and 14500 cells. Unfortunately for this particular application (Jetbeam RRT-0, which can utilize either size cell), they have to have at least a "button top", if not an actual "nipple". This kind of narrows the choices down quite a bit.

With the exception of the 26500 cell, all of the AW IMR cells that I tested had a relatively small button top for the positive terminal. I remember this clearly because it was very challenging to spot weld two separate leads to positive terminal for voltage sensing.

Lighthound now has AW IMR 18650 cells in stock. I ordered two yesterday for testing, and hope to have the results posted in a week or so.

Also, when I purchased the IMR cells I also purchased samples of the rest of AW's line of LCO cells. I hope to perform similar tests on these and post the results in a new thread in a few weeks.

Cheers,
BG

 

[45/70]

With the exception of the 26500 cell, all of the AW IMR cells that I tested had a relatively small button top for the positive terminal. I remember this clearly because it was very challenging to spot weld two separate leads to positive terminal for voltage sensing.

Yeah, as I said, the selection of 16340 or 14500 IMR cells with nipples, is pretty much limited to either AW's, or AW's.🙂 I believe there may be some "CrapiFire" type cells available, but more and more have pretty much given up on cells of this type. I've sure bought and used enough of them to determine their quality, ie. low.

I did check into shao's cells in the MarketPlace, but unfortunately, the only cells that have nipples are a few of the 16340 cells, and this seems rather random, as most do not seem to have them. On top of that, the 16340 cells don't seem to perform very well. All of his other cells are flat tops, or minimal button tops.

I always appreciate tests of new cells. There are many to research that help with making a decision when purchasing cells for specific applications. On the other hand, there are very few tests of used cells, some, but not many. It is my opinion that these tests better reveal the true ability and performance of any chemistry cell.

I don't now whether you've seen this thread that I started a while back. It involves testing some older used AW cells and, in most cases, some used, but not as much, or as old "CrapiFire" cell types. The results are typical of what I've experienced in my nearly 7 years, using LiCo cells in flashlights/torches. As I state in the thread, considering both the number of cycles, and the age of the cells tested, one could easily come to the conclusion that the AW cells would last at least four times as long as the competitors presented.

Dave

 

[xul]

In the first graph,
note that discharge current, I, is not a linear function of measured capacity
so to predict capacity C from I
take the log of I to make the relationship linear


Measured ....discharge ........log of I
capacity, C ....current, I .........LI
0.35 ............2 .................. 0.301029996
0.25 ...........4 ...................0.602059991


Have Excel calculate the slope and intercept
-0.332192809 =slope
0.45 =intercept


So C should equal
-0.332192809 times Log(I) +
0.45


confirm
I LI ................C
2 0.301029996 0.35
4 0.602059991 0.25


so for a discharge current, I, of
3
take the log
0.477121255 =log I

then the capacity should be C =
-0.332192809 times
0.477121255 plus
0.45
equals
0.29150375

and not
0.3
as you would expect from a linear relationship
between C and I

In this case the difference is only
2.8 percent but this may be
larger with other batteries

 

[Battery Guy]

xul

Sorry buddy, but you totally lost me. What exactly are you trying to point out in the first graph?

Cheers,
BG

 

[xul]

xul

Sorry buddy, but you totally lost me. What exactly are you trying to point out in the first graph?

Cheers,
BG

Sorry.
I need to take a writing course. 🙁

 

[Battery Guy]

Sorry.
I need to take a writing course. 🙁

I doubt that very much. I just did not understand what the calculations were for. Looks interesting. Please explain.

Cheers,
BG

 

[xul]

Well, battery testing graphs may show current levels of 0.1, 1 and 10A. This much variation, two orders of magnitude, is better understood using a log scale rather than a linear scale.

The trouble comes if you want to know performance for an intermediate value of current. If you linearly interpolate between values you will be significantly off, i.e., halfway between the 1A line and the 10A line is 3A, logarithmically speaking, not ~5A.

My spreadsheet is for doing this tedious calculation, back and forth between linear and log scales.

Nailing down battery performance can be really difficult, at least for the reason of the exponential relationships between almost everything.
And we all think linear.

Too bad the sliderule people are out of business. These things have a one-for-one correspondence between linear and log scales.

 

[Battery Guy]

Well, battery testing graphs may show current levels of 0.1, 1 and 10A. This much variation, two orders of magnitude, is better understood using a log scale rather than a linear scale.

Ah, I understand now. I was thinking about testing these under constant power conditions and plotting Ragone plots, which are log Power versus log Energy. Maybe I should do that.

 

[xul]

Ah, I understand now. I was thinking about testing these under constant power conditions and plotting Ragone plots, which are log Power versus log Energy. Maybe I should do that.

With switch-mode regulators constant power is an option and it's generally different than constant resistance or constant current.

Decide what the goal of your testing is and then design your test around that goal. If constant power regulators are commonly used then test results that relate to this mode will be useful for many people.

 

[Justin Case]

Well, battery testing graphs may show current levels of 0.1, 1 and 10A. This much variation, two orders of magnitude, is better understood using a log scale rather than a linear scale.

The trouble comes if you want to know performance for an intermediate value of current. If you linearly interpolate between values you will be significantly off, i.e., halfway between the 1A line and the 10A line is 3A, logarithmically speaking, not ~5A.

My spreadsheet is for doing this tedious calculation, back and forth between linear and log scales.

Nailing down battery performance can be really difficult, at least for the reason of the exponential relationships between almost everything.

Assuming an exponential relationship, such as Arrhenius, then it seems you ought to take natural logs, not base 10 logs. Using base 10 vice base e results in a constant shift error. Also, if different mechanisms operate at low vs high discharge current, then you will also have different activation energies, which can cause further problems with the interpolation, since it really becomes an extrapolation. I don't know if this is actually the case, but unless you have knowledge of operative mechanisms behind the exponential relationships, it is a risk. If I were to assume the same mechanism across the board, then I might use something like Lagrange interpolation or divided differences, and take advantage of the other data points available to me, not just the two adjacent ones.

Your check that the slope and intercept values are correct also seems unusual. Of course the back check will give you exact results. You conducted a linear regression using two data points. Well, any two points define a straight line so of course you'll get back the same x,y points.

 

[xul]

If I had the software I'd do more sophisticated analyses.
I tried to download a free program, I think it was 'R', but something went wrong. Computers are great when they actually work. 🙂

 

[Battery Guy]

I just started the tests on the AW IMR 18650 cells today, and hope to post results by Friday. I also started constant power discharge tests on the other cells, so I hope to post a partial Ragone plot on Friday as well.

Cheers,
BG

 

[xul]

Check the variation between several identical cells.
If you see a +/- 5% variation between cells and a +/- 5% variation between brands, the cell variation may be masking the brand variation.

 

[Battery Guy]

Check the variation between several identical cells.
If you see a +/- 5% variation between cells and a +/- 5% variation between brands, the cell variation may be masking the brand variation.

As stated in the limitation section in the original post, I acknowledge that a sample of one is far from being statistically significant. Unfortunately, I do not have the time or resources to do testing on multiple cells.

Cheers,
BG