Soliciting advice for 18650 cell choice in 2021 (+ 25R pack performance recap)

[flippy]

havent found a genuine hg2 yet

because you need to look for HG4. its the "rebranded" HG2 after a couple morons blew their hands off.

 

[eMark]

havent found a genuine hg2 yet

Are the only HG2s available salvaged or fakes ... can still retrieve the original LG HG2 datasheet, but LG HG4 datasheet isn't available when googling :?

because you need to look for HG4. its the "rebranded" HG2 after a couple morons blew their hands off.

Why it's "rebranded" HG4 instead of HG3 was because it was two morons, not just one :wink:

LG HG4 datasheet (2013-05-30) ... https://www.batteryspace.com/prod-specs/10679.pdf ... so it's not like it's a recent HG2 replacement. Fake HG2s are cheaper than HG4s ... probably why more than one moron still buys them thinking they're saving money.

Isn't my intention to be anymore flippant than flippy :wink:

CORRECTION: The above datasheet is HE4, not HG4 (moron-ish is catching). LG HG2 2015-01-28 datasheet ... https://www.batteryspace.com/prod-specs/9989.specs.pdf ... the original release of the LG HG2 was in 2014, production became available in Summer 2015. Apparently LG HG4 uses the same datasheet as HG2. HG4s are hard to come by (out of stock), but HG2s are plentiful. Can't even Google a LG HG4 pdf datasheet ?? Is it identical to HG2 datasheet :? :?

 

[eMark]

What's up with the latest + 25R8 as if the previous 25R grade A cells (if any still exist) weren't good enuf at half the cost of today's 25R8 or 25R6? Is it marketing ploy to justify an increase in price of $6.99 ... thus separating 25R grade A cells from the 25R grade B (fake A) cells? Will this new coding system prevent 25R8 and 25R6 grade B cells from reaching suppliers and discounting them enuf to attract buyers?.

Here's the two latest 25Rs for Pajda to test to determine if this BatteryStore 25R8 at $6.99 is better than IMRs 25R6 at $6.99 ... https://www.imrbatteries.com/samsung-25r-18650-2500mah-20a-battery/ ... maybe Pajda can tell us which one is the better buy (25R6 or 25R8) at $6.99 or 100 at $650. :wink:

Will Samsung's new coding system for 25R (and 30Q) with the price increase discourage DIYers from buying cheap grade B (fake A) cells? Who's going to pay $6.50 in quantities of 100 for grade A cells when you may be able in a few months to buy 100 25R "8" or "6" grade B cells for half the price.
"

https://www.18650batterystore.com/collections/samsung-18650-batteries/products/samsung-25r-18650

One would think Pajda just shakes his head having to get his hands on the latest 25R8 knowing there's no significant change from the 25R6. Someone should start a thread in the coming months on who supplies the best 25R "8" or "6" grade B cells for those that can't afford $650 for 100 25R8 or 25R6 grade A cells ... :wink: : :wink:

 

[flippy]

please dont make the letters bigger and bold. use Ctrl+mouse wheel if you have trouble reading webpages. :thumb:

cell changes like this are fairly common. especially with samsung. generally the number changes when the production process has been changed to some degree. it can just be different type of steel casings.

and it helps (for a while) to weed out the re-wrappers that managed to inject themselfs into the offical channels.

 

[eMark]

Do you still use a mouse wheel? Haven't used a mouse since getting my chromebook 10 years ago :thumb:

Apparently you're not aware of the lengths the Chinese will go to pass along inferior grade B cells as grade A ...
https://batterybro.com/blogs/18650-wholesale-battery-reviews/30828867-is-it-a-fake-18650-battery-featuring-the-samsung-25r ... and this article was written in 2015. Six years later it's next to impossible to know for sure what is the real McCoy (grade A) and what is a counterfeit grade A cell (inferior grade B).

it can just be different type of steel casings.

That and other little changes just makes it all the more difficult to separate the real from the counterfeit. The Chinese are some of the best when it comes to a counterfeit that passes as the real McCoy. How many DIYers are going to take the time required (as evidenced in the above article) to identify a grade B (fake A) 25R cell? Even if they do what lengths will they have to go too to prove the cells they received are fakes and want their money back. Then again who do they have to blame but themselves when they bought cells at what seemed to good to be true.

Even IMR customer service was kept in the dark (on purpose) about the switch from 30Q "141" to 30Q "6 KH1T" when "141" is still shown on their 30Q webpage ... https://www.imrbatteries.com/samsung-30q-18650-3000mah-15a-battery/

It's anyone's guess about a supplier's rationale when for example BatteryBro sells the latest 25R8 while still showing 30Q 136 cells for sale. Guess they didn't have time to update 30Q "136" photo or they acquired some 30Q "136" cells at a price that they couldn't pass up :wink:

Bottomline: Buyer Beware! There will always be more than a few builders that will buy cheaper 25Rs thinking they're getting a good deal based on advertising claiming the 25Rs are authentic (authentic grade B). Isn't that why the Chinese go to such lengths to disguise grade B cells as grade A cells because there will always be a buyer (for authentic grade B cells) :wink:

 

[eMark]

just keep your 25R, strip the packs apart, use a dremel with a blue grind wheel to clean up the spot welds, but a ton of new 25R's and equalize the old cells with the new cells so each P group has the same amount of old cells, then top off with new ones.

if you actually babied your 25R's with slow charging i doubt they have been hurt in their capacity much. no need to toss good cells if you are looking to save a buck and are willing to spend some hours stripping and cleaning cells.

pro tip: make the pack 20S and halve your current. :wink:
(the controller does not care about volts, just amps. more volts means less amps means more efficient.)

flippy's suggestion certainly has merit ... thus my 25R posts. The new 25R date code (and generation 5, 6 or 8 ) on the shrink label will hopefully stymie Chinese 25R fakes from flooding the market. The real 25Rs like 25R8 was $4.50, but no longer available at that price. Price increases on all Li-ion cells is blamed on car manufacturers demand and Suez Canal Blockage. It will take time (if ever) to see genuine 25R8s at $4.50 again like at Bulk Battery ... https://www.bulkbattery.com/batteries/18650/samsung-25r-18650-2500mah-20a-battery/ ... or 25R6 at LIION Wholesale was $5.99 ... https://liionwholesale.com/collections/batteries/products/samsung-inr18650-25r-battery-genuine-tested-20a-2500mah-flat-top-wholesale-lot?variant=31113493053509 ... now out of stock. Customer service rep expects delivery of 5000 25R mid May, but not sure if 25R6 (as previously in stock) at $5.99 or newer 25R8 at $6.99.

25R8s are again in stock at their (Bulk Battery) Battery Store ... https://www.18650batterystore.com/ ... with the current price (as with other US suppliers like IMR) at $6.99. Their customer service rep estimated purchase of at least 100 25R8 cells would be $650.

 

[Pajda]

Several problems were mixed together here. In my opinion, there may or may not be a connection between them.

New marking from Samsung
I think that the change in marking was aimed at simplifying the orientation in the released versions of cells for customers, nothing more. It was confusing to decide if the "136" or "141" is newer version. So now it should be simple and clear. But I noticed that some new cells have both marking, for example 40T3 have new style "3" with old style "181" marking together.

Availability of different Samsung cell versions at the same time
I think that Samsung is not producing more than two versions of each model at the same time, so after the new version is entered mass production it should be widely available in the "near time"("near time" is at least one year in battery world) and the older version will be discontinued soon. So if 30Q6 entered the mass production in Q1 2020 it should start appearing at distributors stocks in Q1 2021 and in Q3-Q4 it should dominate the market, where the "141" version will slowly disappear. I think that very little distributors have much older "136" version still in stock, It does not matter how much outdated picture they have in their store, I believe that they send you "141" or even the newest "6" version.

Fake vs B-grade cells
A distinction must be made between Fake and a B-grade cell. I must say that I do not remember if I ever got fake cell from reliable distributor. Yes, I ordered several "110% Authentic" cells from Aliexpress and many of them were Fakes. But one should know what to expect from B-grade. In most cases, these are cells that have not been electrically sorted into the almost identical parameters. This has a practical impact in case you make a battery of them without further sorting. B-grade cells can then disbalance significantly in the battery quickly. But if you torture only one B-grade cell alone it usually perform very close to A-grade cell.

 

[vanturion]

Fake vs B-grade cells

Pajda, can you spot the self-discharge quality of B-grade cells from the manufacturer by waiting a couple weeks after the batteries arrive, presumably at some 3.4-3.6V level storage voltage, and taking individual measurements before and after this waiting period to compare drift? Or do you need to charge them to a higher level first to then check self-discharge rates after some time? Basically just wondering if it is more pronounced at higher voltage levels.

I don't know if it's just good timing or what, but I happened across some 21700 Tesla/Panasonic cells on ebay for an incredible $3/cell at bulk! Based on the seller's flawless rep, it looks like a good, low-risk acquisition. Check it out if you're interested. Even comes with free cell holders, crazy. Looks like I'll be assembling a second pack at 20S7P for long range jaunts.

Frankly, I should've been checking if the 21700 cells would've fit earlier as the specs/performance/claimed-cycle-life of the Molicel P42As is more appealing than the P28As that I went with.

25R8s are again in stock at their (Bulk Battery) Battery Store ... https://www.18650batterystore.com/ ... with the current price (as with other US suppliers like IMR) at $6.99. Their customer service rep estimated purchase of at least 100 25R8 cells would be $650.

I don't really find the 25Rs as attractive today now having spent some more time looking around. Back in February 2015, I paid $3.60/cell, but shipping was a bear costing an addition $120 from China. To pay more for essentially the same performance (assuming a blue label would be closely comparable too) is just no good! Neglecting cycle life of course.

In any case, $3/cell for essentially one of the highest energy dense batteries is unbeatable in this market. Hoping they test well.

 

[vanturion]

I don't know if it's just good timing or what, but I happened across some 21700 Tesla/Panasonic cells on ebay for an incredible $3/cell at bulk! Based on the seller's flawless rep, it looks like a good, low-risk acquisition. Check it out if you're interested. Even comes with free cell holders, crazy. Looks like I'll be assembling a second pack at 20S7P for long range jaunts.

Quoting myself, had to update with this quick end to the prospective 20S7P salvaged Model 3 21700 battery pack. The performance was not good, significant voltage sag at 10A... (charged CV to 4.2 w/ .10A cutoff)



compare this with Mooch's tests, significantly less capacity...



and significantly higher ending temperature at same 10A discharge.



I don't have any prior experience testing cells, but it was still surprising that these Model 3 cells would already be so degraded considering how little time they've been in production/use and Tesla's hearsay reputation for long battery lifespan. I'm not sure the best way to take a DCiR measurement(s), but using the EBC-A20 Resistance Test at 1000mA on several of the cells put each one in the high 30s mOhms so I expected similar (bad) performance from the whole lot. That said, the ebay owner is probably the most stand-up seller I've ever run across and is accepting a return for a full refund. It was worth a shot.

 

[Pajda]

vanturion please show us your cell fixture, from your FLIR picture it is not clear. A common issue causing lower measured values than expected (higher voltage drop) is an improper 4-wire connection. A must have accessories to ZKETECH testers is a BF-2A cell fixture.

https://www.aliexpress.com/item/32824553260.html?spm=a2g0o.detail.1000060.2.30a612c8NjAKjh&gps-id=pcDetailBottomMoreThisSeller&scm=1007.13339.169870.0&scm_id=1007.13339.169870.0&scm-url=1007.13339.169870.0&pvid=df4244b5-e0cc-44a7-81d1-69483f86c05d&_t=gps-idcDetailBottomMoreThisSeller,scm-url:1007.13339.169870.0,pvid:df4244b5-e0cc-44a7-81d1-69483f86c05d,tpp_buckets:668%230%23223416%232_668%230%23223416%232_668%23888%233325%2315_668%23888%233325%2315_668%232846%238108%231977_668%235811%2327189%2386_668%232717%237566%23875_668%231000022185%231000066059%230_668%233468%2315617%23886_668%232846%238108%231977_668%235811%2327189%2386_668%232717%237566%23875_668%233164%239976%23870_668%233468%2315617%23886

 

[vanturion]

I see. Since the two wires were combined on the alligator clip, I assumed one connection was fine. I just used dimpled .15mm x 8mm nickel plated copper strips to sandwich/clamp the cell.



I actually picked up a cheap 4-wire fixture with the same order,



but didn't want to separate the alligator clips to use it if it wasn't necessary.

So basically if the recorded voltage sag was reading in excess, the cell I tested would have tested with more capacity (it recovered to 3.2V+ after load was removed), but it also would have an even worse ending temperature from the additional heating still indicating significant cell degradation. Couldn't I just compare the ZKETECH reported voltage under load simultaneously with a multi-meter probes on the copper strip to sanity check voltage drop error in the clamp fixture I used?

One more thing I wanted to ask - I'd like to run a reasonably equivalent cycle test with my old unused 25R to your 3000mAh 18650 cycle test cell chart and would like some clarification on your DCiR tests. For your DCIR 10s and IEC values, at what SOC/voltage are you calculating these values, 50%? Are you just programming in the pulses and manually calculating resistance after from the .csv values? Also, do you take an average of several pulses? So for my resistance tests, it would be 500mAh/.2C in 10 seconds and 2500 mAh/1C in 1 second.

 

[Pajda]

I am just saying that using a clamps is not reliable method, I know about many examples (myself included) when the measurement was significantly affected by this. The only reliable plane contact is the screw secured connection. The best approach is always isolate voltage measurement signal and current path. For cylindrical cells there is no reason why not use BF-2A cell fixture design or at least ZKETECH design with fully isolated needle for voltage measurement. I am not saying that your results are wrong, but if you compare your measurements with others and these are quite different, there is a significant chance that one of the methods is wrong.

DCIR 10s method by Samsung SDI
DC impedance is measured at 50% SOC(state of charge) state. (I am using 25 °C)

After Standard charge, discharge with constant current of 0.5C for 1hr followed by 30min rest time (Check voltage at this time – V1). Discharge with constant current of 0.5C for 30sec (and check voltage at 10sec – V2)

Initial DC impedance = (V1-V2)[mV] / 0.5C[mA]

DCIR 10s method is a "single stage" method and so can be conducted with cheap equipment. You can program recipe for DCIR 10s measurement with EBC-A20 tester but the DCIR value must be calculated from the .csv values. Results from DCIR 10s method are close to the ACIR 1kHz method. You can find also DCIR 30s method which vary only in the length of discharge time.

DCIR IEC is "two stage" method by IEC 61960. This method is implemented in "better" testers and unfortunately it is imposible to conduct it with ZKE. I am using 25°C ambient temperature and 50% SoC for measurement.

 

[vanturion]

I am just saying that using a clamps is not reliable method, I know about many examples (myself included) when the measurement was significantly affected by this. The only reliable plane contact is the screw secured connection.

OK I can see that, especially for repeatability when swapping cells in and out of the fixture over time. I guess it's hard to claim a result if you don't follow the testing standard for separating current vs voltage probes in the first place. I cheaped out not going with a BF-2A.

Results from DCIR 10s method are close to the ACIR 1kHz method.

Thanks for your explanation, I found some of your old posts with docware on the budgetlightforum, and I think I understand DCIR measurements better now. Ideally I'd have better equipment for direct comparison with docware and your results, but I think without spending more the best I can do is to have consistency with the DCIR-10s method you described to characterize the increasing DCIR trend over the test cycles. Actually I have a couple more questions regarding this if you don't mind:

1. Since the DCIR-10s and the ACIR 1kHz are close, do you think it's worthwhile to even bother calculating the 10s method and instead use the AideTek SM8124A (something I already have) at 50% SOC every 25 or 50 cycles after rest period? If all I care about is the trend, do you think the ACIR test would be sufficient in revealing it?

2. What do you think about the EB Tester software "Battery Resistance Test"? This would be even more convenient to program 1C or even 10A/15A discharge and record resistance at the same 50% SOC to develop a similar degradation curve. While it doesn't appear to output csv data with this test, it looks like it just applies current over 1 second before calculating the resistance.

In any case, going with either of these would be less useful to compare with your charted 18650 test results, but the same %-change over time could at least be observed and compared especially as it relates to capacity loss.

 

[Pajda]

1. Since the DCIR-10s and the ACIR 1kHz are close, do you think it's worthwhile to even bother calculating the 10s method and instead use the AideTek SM8124A (something I already have) at 50% SOC every 25 or 50 cycles after rest period? If all I care about is the trend, do you think the ACIR test would be sufficient in revealing it?

This is not about the method itself (if the method principle is clearly described) but about people "laziness" ACIR method main drawback is, that it is not supported by majority of cell testers/cyclers (even professional ones), so you need to use another device to conduct it when you are running cycle life tests. Particularly several cells at once, you really do not want to go to the cell fixture after each 50 cycles and make changes to the connection (I only know one person who can handle this - docware :wink: ) So ordinary people are able to run only one different recipe after 50 cycles, particulaly if it can be done remotely. Of course professional testers can program all in one recipe.

2. What do you think about the EB Tester software "Battery Resistance Test"? This would be even more convenient to program 1C or even 10A/15A discharge and record resistance at the same 50% SOC to develop a similar degradation curve. While it doesn't appear to output csv data with this test, it looks like it just applies current over 1 second before calculating the resistance.

This method is not clearly described and cannot be automated via ZKETECH testers, on the other side DCIR 10s method (including charge, 50% discharge, 30min rest time, 10s (1min) 0.5C load) can be automated in one recipe with nominal capacity test check, you only need to calculate the DCIR value from saved data. So if you are looking at my 5C graph, I am only periodicaly switching between two recepies of ZKEtester. One for 49x load cycles and one for 1x nominal capacity + DCIR 10s test.

I added sample recipe for EBC-A20 to attachements, you only need to rename it from .txt to .stp

 

[vanturion]

This is not about the method itself (if the method principle is clearly described) but about people "laziness" ...

Shoot, called out haha.

I added sample recipe for EBC-A20 to attachments, you only need to rename it from .txt to .stp

Thanks, looking at this I need to ask one more thing about 50% SOC.

Why is the DCIR-10s calculated after .5C for 60 mins rather than after charging to a set voltage value (like something between 3.66 to 3.77 depending on the battery)? Since we already know the capacity will be decreasing over time wouldn't 50% SOC be a moving/decreasing target as capacity shrinks? If we always use the same .5C discharge and time interval, it's probably going to record a higher resistance than would be measured from targeting a constant voltage value for each resistance cycle test (my assumption).

Thinking about it some more - it's like doing it this way ensures the DCIR measurements are more linked or dependent upon the capacity loss. However, since we're already accurately characterizing the capacity loss, it really isn't telling us anything new. By establishing a DCIR-10s measurement targeting the same voltage for each measurement interval, we could perhaps get better information about for predicting how the lifecycle count might affect voltage sag as the cell is cycled. This all assumes that an increasing DCIR measurement trend is still established over time while targeting a set voltage value (I don't have any testing experience yet). What do you think?

Just asking to help figure out how I want to set things up before I start cycling a 25R.

 

[Pajda]

Thinking about it some more

I think the biggest problem is that you think about it too much :wink: . We can argue for a long time about the advantages and disadvantages of each method but at the end we can conclude that using a larger hammer is also a plausible method for dealing with problems.

So maybe unwanted advice is to choose any method you like (which is reasonably repeatable) and begin to compare measured results on a relative basis. I mean "under this particular test method shows best results this sample..." Forget absolute values.

 

[vanturion]

I think the biggest problem is that you think about it too much :wink: . We can argue for a long time about the advantages and disadvantages of each method but at the end we can conclude that using a larger hammer is also a plausible method for dealing with problems.

So maybe unwanted advice is to choose any method you like (which is reasonably repeatable) and begin to compare measured results on a relative basis. I mean "under this particular test method shows best results this sample..." Forget absolute values.

Haha, it's been said. But hey, is that not why we are here? What do they call it.. learning through adversity.

No that's good advice, since I'm curious anyway about what the results will be holding voltage constant for interval DCIR tests and about directly comparing to standard discharge cycle calculated DCIR at this same 50% SOC voltage, I think I'll give it a shot. Worst case, I resort to the hammer.

 

[Pajda]

I try to say that if you are interested in testing the cyclic life of cells, then do not be afraid to use any method that you find interesting. There is nothing such as wrong test method, just a wrong conducted one.

Back to the DCIR vs SoC measurement issue. If you look into docware's Well of Wisdom starting on page 4, you will find his DCIR tracking results for the most interesting cells on the marked and you should notice that almost all of them have a wide plateau of almost constant DCIR around the 50% SoC. So actually I do not expect a significant error in getting "absolute" value of DCIR due to the SoC drift caused by the lost of the capacity during cycle life test. And we're back to laziness, if you do not have at least partly automated measurement method you have to sacrifice so much time and so everybody sometime thinks about bailing out of this life episode..

 

[vanturion]

...If you look into docware's Well of Wisdom starting on page 4, you will find his DCIR tracking results for the most interesting cells on the marked and you should notice that almost all of them have a wide plateau of almost constant DCIR around the 50% SoC. So actually I do not expect a significant error in getting "absolute" value of DCIR due to the SoC drift caused by the lost of the capacity during cycle life test...

Makes sense, good info. Ended up using .5C discharge from full for 1hr to find 50% SOC voltage, and I decided to repeat the measurements 2 for .5C and 1 for 4C with appropriate wait breaks for future measurements. I'm only running blocks of 50 cycles because I want to take a thermal image at end of 50 cycle discharge and have to perform manually. So not fully automated, but that's OK for this first time around.

Speaking of, started the test last night after attaching the leads to the 4-wire test fixture and the first cycle yielded 2,440 mAh out of 2,500 mAh rating. Seems like a good result after sitting around at 0 cycles and 3.59V for 6+ years. I'll post the results to docware's cycle aging thread when I have something to show.