Diagnosing Lithium cells?

[dmwahl]

I came across a really good research paper some time ago where they tested this on Panasonic production cells

Could you post a link to this? I'd be interested in reading it.

[Siwastaja]

Batteries from the factory are charged for maximum shelf life. Measure a new one, for LiPos they're within a couple of hundredths of a volt off 3.8V. Look at the Lithium 'storage' setting on any hobby charger, it charges/discharges to 3.8V.

No source for the claim that they are especially charged for "maximum shelf life". They are meant to be integrated into products as soon as possible. There can be other reasons for the typical 50% shipping SoC. "Good enough" shelf life is experienced at 50% to 60% SoC, no doubt about that. It doesn't mean it couldn't be even better at 30% SoC. Also, I have really seen some that are shipped clearly below 50% (3.5 something volts), but cannot remember which brand, now. It was some of the leading Japanese or Korean 18650.

Finally for a real-world experiment take 3 identical LiPos. Discharge one to 3V, fully charge one and leave the other at 3.8V. Leave them for 6 months on a shelf.

I have more than 50 cells sitting at 3 different temperatures, charged to different levels, two different chemistries from four manufacturers. Now that I'm fired from the Uni research position due to some office politics, I'm not tied anymore into current scientific standard of selling copyrights to a corporation or restricting access to the scientific work from non-paying members of public, so I can publicly post the results. We'll see!

3V(open-circuit) sample is overdischarged, don't do it. I'm not surprised if it puffs. Even if it did store well at exactly 0% SoC, even a small leakage will cause it to overdischarge. Also, 0% is not specified as exactly as 100% is (which would be a CC-CV method). For this reason only, I'd try to avoid even short-term storage below 10% SoC. I would stay away from the clear "knee" part of the curve, which would start at about 15 - 30% SoC, depending on chemistry.

[Siwastaja]

I came across a really good research paper some time ago where they tested this on Panasonic production cells

Could you post a link to this? I'd be interested in reading it.

Sorry, I read a version on physical paper and I have forgotten where I put it, can't remember the authors, and it will be behind some paywall anyway. The title was something like "effect of regenerative braking on battery life". Their main test subject was to test short charge pulses during discharge, and they showed that they actually benefited the cell life (possibly by reducing the total SoC window, because the amount of energy discharged stayed constant), busting the myth that short charge pulses are detrimental to the batteries. But they had other interesting findings, too, related to shelf life.

[doobedoobedo]

You're right, but if I was making batteries I'd send them out at the pretty much universally accepted storage charge. You never know how long they're going to be sitting on a distributor's shelf after all.

I'd love to see evidence that it really doesn't matter so long as it's between the two steep drops in the discharge curve.

[Artlav]

Anyone else like collecting and looking at the data?

Well, here are some curves of good Sanyo R1112 cells, nominal 2400mAh at 4.2V, with a few years of laptop duty behind it (found at above 3V).
One cell, 3 charges, 4 discharges
The capacity stays the same within the noise.

These can be though of as a reference for what a good cell should look like (and also confirms that the rig is repeatable to within 1%. ).

3 charge cycles of one cell.
Not as consistent as discharge, apparently dependent on how long the cell had to recover after the discharge.


4 discharge cycles of one cell.
Start automatically exactly 5 minutes after charge termination, and so super consistent.


4 of these cells charging at once.


And 4 of them discharging at once.
ESR is around 80-90mOhm on all of them.


I'm now cycling a couple more of Sanyos, that were found at 2.94V, and they look great so far.
And a couple of ICR18650-24E, that were found at 2.63V.
These are the same as the BAD cells before, and they show capacity degradation after 3 cycles as well. Looks like a pattern.

I guess that the low voltage being bad is cell brand dependent. Good ones can survive a dip into 2V just fine, lower quality or different chemistry ones would degrade.
Hopefully will know for sure once i go through the whole box in a few weeks. 

The title was something like "effect of regenerative braking on battery life". Their main test subject was to test short charge pulses during discharge, and they showed that they actually benefited the cell life

Sounds like this one?
http://www.a3ps.at/site/sites/default/files/downloads/evs28/papers/A3-01.pdf

Never charge li-ion cell, new or used, below 0 degC. Very low currents may be acceptable, but don't risk it.

I had a good cell running in a weather station through a Russian winter (as low as -30*C, while being solar charged every day at about C/50-C/100.
By spring it was still alive and have the same capacity as it had in the summer.
So yeah, low currents work i guess.

I have more than 50 cells sitting at 3 different temperatures, charged to different levels, two different chemistries from four manufacturers.

I'd love to to see your results.
I guess i can bend the dump resistors towards the cells and compare them at different high temperatures, but that feels like a recipe for disaster.

Even if it did store well at exactly 0% SoC, even a small leakage will cause it to overdischarge. Also, 0% is not specified as exactly as 100% is (which would be a CC-CV method). For this reason only, I'd try to avoid even short-term storage below 10% SoC. I would stay away from the clear "knee" part of the curve, which would start at about 15 - 30% SoC, depending on chemistry.

Hm...
Should i program the rig to give the cells some charge after the last cycle?
As of right now the cells go into the done bin essentially discharged to 0%, but they all recover to about 3.8V after a few hours.

[Siwastaja]

Sounds like this one?
http://www.a3ps.at/site/sites/default/files/downloads/evs28/papers/A3-01.pdf

Yes! Thanks! Apparently they also tested 5-month storage at 0% SoC (not specified more accurately), and it was the best storage condition. (I'm still not recommending storing at 0%, but by all means go for 30% instead of the "traditional" 50%).

I had a good cell running in a weather station through a Russian winter (as low as -30*C, while being solar charged every day at about C/50-C/100.
By spring it was still alive and have the same capacity as it had in the summer.
So yeah, low currents work i guess.

This is not surprising. Still, it's generally explicitly forbidden by all manufacturers. In reality, lithium plating is a function of charging current (lowest best), voltage (i.e., termination SoC, lowest best) and temperature (highest best), but manufacturers go the easy way and define the temperature limit as a step function - no charging below 0 degC at all, and full charging current acceptable above 0 deg C. Unfortunately, I have a strong "educated feeling" that charging at full specified charging currents slightly over 0degC will shorten the cell life considerably, especially near 4.2V. On the other hand, there is nothing wrong to do a occasional, very slow charge slighty below 0 degC. Avoid charging fully, however.

Hm...
Should i program the rig to give the cells some charge after the last cycle?

I was thinking the same with my rig (which is a non-dissipative bidirectional converter), but never did it. It would be important only if you go really close to 0%, or slightly below by accident. 0% is kinda fuzzy, anyway. You can also use "safer" limits like 3.0 to 3.1 volts ending voltage, but then the resulting capacity reading is slightly pessimistic, especially for cells with higher DCR.

As of right now the cells go into the done bin essentially discharged to 0%, but they all recover to about 3.8V after a few hours.

This sounds extremely strange. Double-check the numbers. Empty cell should show about 3.3V open-circuit voltage. 3.8V is near 50%. Extemely weak cells with huge internal resistance could do this, having enough resistive voltage drop to trigger your low-voltage cut-off limit while they in reality are at 50%, but you would notice this from very crappy capacity (like <1000 mAh for any 18650) and abnormal heating for such a short cycle.

What is your low-voltage cutoff limit? It also depends on current. 2.8V at 0.5C is generally good for laptop cells, but increased DCR leaves some juice behind unless you lower the cut-off voltage (possibly not recommended) or add a CV phase to the discharge. (CV discharging at 2.8V may not be recommended either, so.....)

[Artlav]

This sounds extremely strange. Double-check the numbers.

Just checked, and yes, only bad cells are like that.
The Sanyos got up to 3.4V overnight, the so-so are at 3.6-3.7, while the bad ones are at as much as 3.9V, even though all were discharged to 3.0V.
DC ESR is around 100mOhm for all of them, so i'm confused.

Hm, with low capacities the discharge current (average of 0.9A, 3.9Ohm load) is getting to 1C and above, while the good ones are getting 0.5C. Maybe that's the problem?

[Siwastaja]

The Sanyos got up to 3.4V overnight, the so-so are at 3.6-3.7, while the bad ones are at as much as 3.9V, even though all were discharged to 3.0V.
DC ESR is around 100mOhm for all of them, so i'm confused.
Hm, with low capacities the discharge current (average of 0.9A, 3.9Ohm load) is getting to 1C and above, while the good ones are getting 0.5C. Maybe that's the problem?

DCR is not a constant - it's a function of temperature and SoC. DCR can really shoot up when the cell is nearly empty. Let's say it is 100 mOhm when measured from full cell, then it might even be as much as 400 mOhm from an empty cell.

If you are discharging at 0.9A, then there is a resistive voltage drop of 400mOhm*0.9A = 0.36V. So when you are measuring terminal voltage of 3.0V, the open-circuit voltage would be 3.36V, which is more than 0%.

There are other effects in play too (some of which I don't fully understand, either), but this should give you an idea why the increased DCR makes it harder to fully drain the cell, and why cells with increased DCR work better in slow-discharge applications. You are measuring the voltage after the internal "resistor", and the bigger the "resistor", the higher the "actual" cell voltage still is. So you are just not fully discharging them.

"Power" type RC toy cells (which have rather low DCR) may specify 3.0V or even more as the stopping voltage, but these 18650 energy cells are specified lower, typically between 2.5V and 2.8V. To fully discharge them, you need to go to lower than 3.0V, or accept that you are not getting full capacity.

Your assumption of "higher C rating" is also correct; as the cell now has less active material that does the work, it also has less surface area to do the work. The remaining bits are working harder, which you can measure as an increased DC resistance.

So there are two mechanisms (approximately):
- Loss of active material - reduces capacity and increases DC resistance linearly. Caused by cycling.
- Electrolyte reaction - reaction products produce "gunk" that slows down ions from getting where they need to go - increases DC resistance [sometimes gunk can even completely block parts of the structure, which can be seen as capacity loss]. Caused by calendar aging.

Both may be equally important in a typical used laptop battery. In a badly aged battery, "real" full capacity can only be measured really slowly. It is often better to try to simulate your intended usage.

[Siwastaja]

Attached is my (incomplete) study of DCR vs. SoC on several (new) commercial cells. Your mileage may vary, because there is no one valid standard of measuring DCR, and also because temperature affects the results --- DCR is obtained here as a difference of three curves obtained on three different cycles with different temperature characteristics because of heating effect of DCR itself! There is something unexpected going on because Samsung ICR18650-26H is getting better results than expected, compared to newer NCA cells which perform better in other tests. Oh well. Hope this gives you something to think about!

[amyk]

Sounds like this one?
http://www.a3ps.at/site/sites/default/files/downloads/evs28/papers/A3-01.pdf

Yes! Thanks! Apparently they also tested 5-month storage at 0% SoC (not specified more accurately), and it was the best storage condition. (I'm still not recommending storing at 0%, but by all means go for 30% instead of the "traditional" 50%).

Very interesting. Figure 6 in that paper gives very different numbers from the BatteryUniversity, namely:

At 25 degrees and 40%, BU gives 96% after 1 year whereas the paper has at 45% equivalent of 96.4% (98.5% after 5 months, assuming the trend is exponential). They agree on this.
At 40 degrees and 40%, BU gives 85% after 1 year whereas the paper will be equivalent to 93.0%

At 25 degrees and 100%, BU gives 80% after 3 months, or 41% after a year, whereas the paper will be equivalent to 94.6% after a year.
At 40 degrees and 100%, BU gives 65% after 3 months, or 17.9% after a year, whereas the paper will be equivalent to 88.4%.

Who has stored lion cells at 100% and 25 degrees for a year? Do they end up having less than half their original capacity afterwards?

A few more interesting things to read, which seem to be closer to the paper above than BU's numbers:
http://ramasamy.uga.edu/pdf/2005%20-%20JPS%20-%20Calendar%20Life.pdf
http://www.che.sc.edu/faculty/popov/drbnp/WebSite/MSA-calendar.pdf
https://www2.unece.org/wiki/download/attachments/8126481/EVE-06-05e.pdf?api=v2 (NCA cells, after 3.45 years at 100% or 50% and 25 degrees, with almost no degradation, but slight degradation with 40C and very rapid degradation at 60C.)
https://books.google.com/books?id=LxwRBwAAQBAJ&pg=PA407&lpg=PA400&ots=iP97D3lkzq
https://hal.archives-ouvertes.fr/hal-00876555/document (also no degradation at 30C, slight at 45C, and significant degradation at 60C.)

It looks like temperature is the biggest factor. Don't store batteries in hot places, but if you must, then keep them around 40% charged. Otherwise 100% charged is fine.

Investigation on internal resistance vs aging: http://www.iaea.org/inis/collection/NCLCollectionStore/_Public/32/026/32026613.pdf

...and then there's this design note from Linear, showing only 80% remaining capacity when stored at 100% for a year.
http://www.linear.com/docs/28778

[Siwastaja]

Very interesting. Figure 6 in that paper gives very different numbers from the BatteryUniversity, namely:

The problem is that "Battery University" has been totally discredited numerous times, and is mostly matter of joke in the industry. I have stopped considering anything posted there even as a grain of salt. Not worth of your time. I have the feeling they have tried to up the quality a bit and have removed some of the most horrid urban myths (some of which may have first originated from them, such as "li-ion has 3% per month self discharge" myth). Still, try to find other sources if at all possible. Even though they may have valid data sometimes, they are not just reliable enough and it's impossible to know whether to trust them "this time" or not.

In any case, thanks to the inspiration from this thread, I continued my research. I originally planned that I'd let the cells sit for a year, but they are a bit overdue now! Just had other things to do. But I started to do "measurements after" right now. I have all kind of samples from full to empty, at different temperatures. I charged them in July-August 2014.

[Artlav]

Your mileage may vary

And it seem to.
The way i measure DCR is to turn off the load, wait 100us, then measure the voltage and compare it with the one before the load.
It works, verified both on a scope that the rig does it right, and with a resistor in series with the cell.

However, measured llike that the DCR does not ever change.

I.e. here it is for a bunch of so-so cells that recovered to 3.7-3.8V after discharge to 3.0V:


The resolution isn't that great, since the measurement limit is about 5mOhm, but it should be good enough to tell 50mOhm apart from 100mOhm, or the whole 400mOhm.
But what i get is a flat line, essentially, and looking at the raw data the voltage drop goes from 90mV full to 65mV empty.
That's why i'm confused about the cell not discharging all the way.

Maybe there is more than just ohmic resistance?
I.e. equivalent parallel capacitance of some sort?
If so, then i have no idea how long to wait - with the load disconnected, the voltage just keep going up without stopping for minutes.

Who has stored lion cells at 100% and 25 degrees for a year? Do they end up having less than half their original capacity afterwards?

I'm pretty sure i did. A quadcopter i made crashed a few years ago with the battery still almost full, and was sitting in a box for several years before i bothered to rebuild it.
It still flies about as long as before, so the battery didn't suffer any noticeable capacity loss.

The problem is that "Battery University" has been totally discredited numerous times, and is mostly matter of joke in the industry.

Which raises the question - how can i know about it? As in, how are people supposed to find out that it is discredited?
The info on it sounds kinda true, and the only mentions of it being discredited is the wiki talk page and this thread.

[Siwastaja]

The way i measure DCR is to turn off the load, wait 100us, then measure the voltage and compare it with the one before the load.

The problem with this is that you are not measuring DC resistance, but AC impedance at fairly high frequency - 10 kHz, which is even higher than the industry standard 1 kHz for AC impedance measurements. The problem with AC impedance again is that the number has little practical usage in battery system design. It can be useful in cell diagnostics when manufacturing cells, however. Even the pure capacitance of the cell due to its "electrolytic capacitor" type physical construction is enough to give short, 100 us current peaks with lower ESR.

It is indeed true that the AC impedance does not shoot up like DC resistance does at low SoC - AC impedance is most of the time smaller than the DC resistance.

DC resistance, however, is the important one as it explains (or models) the voltage drop that causes your battery management to stop discharge at low-voltage cutoff limit. It also defines resistive heating, and you can indeed feel many 18650 heat up considerably faster at the end of discharge -- due to the combination of increased DC (i.e. ionic) resistance, and enthropy heating. At high discharge currents, I have measured many 18650's to first slowly rise in temperature to about 45-50 degC, and then jump over 60 degC during the last 10% of the SoC. But it's hard to quantify the amount of resistive heating at play here, because it's partially chemical enthropy - and the cell actually cools down when you start charging the empty, heated cell right away - it cools faster than with no current!

Maybe there is more than just ohmic resistance?
I.e. equivalent parallel capacitance of some sort?
If so, then i have no idea how long to wait - with the load disconnected, the voltage just keep going up without stopping for minutes.

You are exactly on the right track here. There is only "ohmic" resistance, because ohm by definition is the unit of resistance, but there are several sources for the resistance - in the DC measurement, ionic resistance is by far the most important, and it is caused by the ions slowly "swimming" and struggling to find their homes in the anode and cathode lattices. This is orders of magnitude slower than electrons shooting in the copper and aluminium current collectors.

Methods to measure DCR vary, but I have been getting comparable - i.e., "consistent enough" results ranging from 3-4 second step to 20-30 second step. Most of the AC effects go away by simply waiting for one second. I also do it by using two (sometimes three) different current levels, instead of stopping charge/discharge completely. My current methodology is: every 120 seconds of charging or discharging at current I:
10 second step at I1 = 0.85*I, V1= voltage measured at the last second;
10 second step at I2 = 1.3*I, V2=voltage measured at the last second;
10 second step at I3 = 0.85*I, V3=voltage measured at the last second.
R = ((V1+V3)/2 - V2) / ((I1+I3)/2 - I2)

That graph I posted is different, though; it's R=dU/dI on whole discharge curves at once! It's the best way to avoid AC effects completely, but you get heating effects, because at higher current, the cell runs hotter, and hotter cell has lower DCR. Heating effects don't affect the pulse method in the same way, because of the thermal mass.

The problem is that "Battery University" has been totally discredited numerous times, and is mostly matter of joke in the industry.

Which raises the question - how can i know about it? As in, how are people supposed to find out that it is discredited? The info on it sounds kinda true, and the only mentions of it being discredited is the wiki talk page and this thread.

I can't say, really! I can't "disprove" them, I don't have time, I'm not interested, etc. I have seen some valid discreditations but they have been lost somewhere in the vast land of the Internet. Also, their site seems to change quite a bit. It is well possible that most of their info is valid now. It's just my $0.02 that I don't trust them, which is the case with many people on this field, too. Scientific community also isn't interested in random websites calling themselves "university", they refer to peer-reviewed journals (which are sometimes jokes, too, but that's a different subject altogether)...

But I remember for sure that the "3% per month self discharge" myth originated from there.

Finding really credible sources is difficult! You basically need to go to the very sources which describe the exact methods in a replicable manner, and make your own conclusions.

[amyk]

Finding really credible sources is really difficult! You just really go to the very sources which describe the exact methods in a replicable manner, and make your own conclusions.

...or test them yourself, as is happening in this thread.

Another interesting study I found, involves cycling and storing cells for over 10 years:
http://www.dtic.mil/get-tr-doc/pdf?AD=ADA515369

They do degrade, but not as much as you'd think - there's a cell from 1994 in there that has over 100K cycles and appears to still have more than half its original capacity left.

[Siwastaja]

I am finishing the characterization of the first sample of my self-discharge/calendar aging test:
Samsung ICR18650-26H
Capacity at start = 2554 mAh
DC resistance at start = 70 mOhm
Charged fully to 4.20V, C/40 stop for storage
Stored for 1.36 years at between 21 to 25 degC, on average close to 22.5 deg C
Capacity has INCREASED by 0.5% per year (so no practical effect; I expected at least some loss of capacity!)
DC resistance has increased by 7.8% per year (as expected)
Reversible leakage (i.e., self discharge) was 4.2% per year (slightly more than I expected, however this was a full cell).
Info like this coming from other samples with NCA chemistry instead of LCO, different storage SoC and different temperature, but it takes time because I cycle them three times at 0.5C to average capacity and DCR, so over 12 hours per cell to test them.

[MagicSmoker]

...
The way i measure DCR is to turn off the load, wait 100us, then measure the voltage and compare it with the one before the load.
It works, verified both on a scope that the rig does it right, and with a resistor in series with the cell.

However, measured llike that the DCR does not ever change.
...

The correct way to measure the DCR of a cell/battery is to compare the difference in voltage at two different non-zero currents close in time to each other (ideally by bouncing back and forth between the two test currents, then taking the average of the voltages at each). For example, if a 4000mAh cell reads 3.250V at 1A and 3.200V at 2A then the DCR = (3.250 - 3.200) / (2A - 1A) = 0.05R

[KL27x]

^That sounds like a way better method that could remove other potentially unstable variables!

Regarding Battery University, call me a paranoid conspiracy theorist. But I think a lot of the information on that site was encouraged by some interested parties.

Ten years ago, it seems like anytime a thread like this popped up on the internet, you'd get 20 morons saying the same thing. Repeating something out of a datasheet (that did nothing to support their views in most cases. Views which largely consisted of how you needed to do XYZ and alpha to use li ion batteries, and that if you don't use a dedicated charging IC and balance cells and have an EE degree, you'll burn your house down). And I'm sure that 99% of them believed everything they said, simply because they heard/read someone else say the same thing in a convincing manner. You wouldn't believe the number of safety experts who had no first hand experience or knowledge.

And then... a few days later a picture of a burned down building or a smoking airplane would show on the front page of the news machine with a story about the dangers of li ion batteries. Or someone would post such pictures in the thread, with a vague backstory. I don't recall ever reading about casualties, curiously. As if in the entire existence of li ion batteries exploding and burning, they have never killed anyone? Maybe that would be too scary? For some reason we were all very well educated on the dangers of li ion batteries, while leaving all the important bits quite vague and ambiguous. The general populace was encouraged to use these batteries, but only in the devices that originally contained them. This was safe. When done, we were supposed to toss the environmentally dangerous "heavy metal" containing and "dead" batteries in a recycling bin.

Meanwhile, there is probably 0.3 laptops and 0.7 smart phones with these incendiary devices built-in for every man, woman, and child on every passenger flight in America, not in the cargo hold, but on peoples' persons. At a time when you couldn't carry a pair of nailclippers in your carryon.

Maybe it's because you can't sell as much milk if everyone has their own cow? And think about this: What business is Battery University involved in? How do they make their money? Aren't they in the business of selling various flavors of milk? Last I checked, they sold batteries and/or advertised links to such stores. So you might say the people behind Battery University were an interested party. You might even say they had a vested interest in the dissemination of this kind of knowledge.

In the 1970/80's a sci fi (pop sci?) writer named Robert Heinlein wrote of an invention called the Capstone. It was a portable energy storage device that could store and release massive amounts of energy with near 100% efficiency. It didn't generate any energy. It didn't blow up cities. It was simply a really good battery. The inventor never patented it, because patents expire. Instead he kept it a trade secret. And the Capstone empire quietly grew through shell corporations until it has its fingers in everything. This wasn't a particular book, as I recall, but rather a consistent theme throughout one of his major fictional worlds.

Heinlein didn't predict the internet. With the internet, you don't need patent or trade secrets. If you have enough resources, you could actually strategically misdirect, subtlely encourage, and anonomously support the vast majority of morons on it. Where did all these li ion experts come from? An unsubstantiated website. A mysterious "I worked for [company]. I did research/design and have degrees in this and that. And it's more complicated than that. Because of [this vagery, that  vagery, and the other vagery in super complicated technical speak, which if you dissect and understand, isn't really saying anything]... and you'll burn your house down" - poster who appeared out of nowhere and then disappeared. Then sit back and watch all these armchair experts appear and repeat what they read, because they know how to use Google. And they get an emotional reward from remembering what they learned... or rather what they were taught.

[Siwastaja]

Yes! I remember the times, just some 5 years ago, when internet forums were flooded full of this shit:

1) "lithium-ion cells are extremely complex and difficult to charge and require special algorithms provided by special ICs" -- false; li-ion is the simplest battery type, practically the only type that can be charged by CC only/voltage stop, or CC-CV, and can also be partially charged. There only is voltage limit and current limit, nothing else. Any proper lab PSU does the exact algorithm. Li-ion is the closest real battery to the "ideal" battery. This claim seems to be a mixup with NiMH batteries, or just pure stupidity.

2) "lithium-ion cells blow up instantly when you do anything wrong" -- false, as we can clearly see. OTOH, it's essential to have good safety precautions and know what you are doing. Li-ion cells are hightech, and it really is dangerous when manufactured by some random no-name sweatshop; or even the good cells if they are severely misused.

3) "There is this LIPO thing which is completely different OMG!!1"  -- false, there is no such a thing as "lipo" at all. This is a complete joke in the industry. This misinformation existed in Wikipedia, too. I started the great cleanup process of the Wikipedia "LIPO" page about a year ago, and the time seemed ripe for it; it didn't seem to cause much backlash.

And every time something completely dumb was going on, magically, Battery University website was used as a reference a lot.

That's why I still cringe every time I see anyone use Battery University as a reference; I won't even bother clicking the link. Not interested.

But I can see that things are much better now, in many ways. People are using SMD components now and not recommending that you should place every DIP IC in those bloody DIP sockets anymore. And people are talking less shit about li-ion, because they are actually using and testing the cells, and hence getting real-world data.

BTW, for a LCO cell charged to only 4.00V instead of 4.20V and stored at room temp, the self discharge went down from 4.2%/year to 1.4%/year and DCR rise from 7.8%/year to 2.1%/year. No noticeable loss of capacity in either sample. Yes yes, I'm writing some kind of paper after I have all the samples tested.

[KL27x]

BTW, for a LCO cell charged to only 4.00V instead of 4.20V and stored at room temp, the self discharge went down from 4.2%/year to 1.4%/year and DCR rise from 7.8%/year to 2.1%/year. No noticeable loss of capacity in either sample. Yes yes, I'm writing some kind of paper after I have all the samples tested.

Now that's cool. It's good to see some actual data.

This data is suggestive that SoC may, in fact, affect DCR, which is interesting. It's also interesting to put this into perspective.

At 2.1% decrease in DCR per year, a battery would have a DCR half life of 33 years.

And stored at full voltage, the half-life would be reduced to 8 1/2 years... But with a self-discharge rate of 4.2% in the first year, the voltage of the battery wouldn't stay full the whole time. In some number of years it would hit the same curve as the 4.00V cell. So in storage, we might think that batteries would last a pretty long time in either scenario. Did you charge them and store them floating?

Because, OTOH, if a cell were to be kept topped off at 4.2V all the time, the decrease in the DCR could logically be presumed to have been way, way higher over a year. Considering the 4V scenario resulted in 2.1%, that 7.8% decrease appears to be the average of an expotential curve that started out who knows how high.

[Siwastaja]

Yes, my method was charge-and-forget; probably most of the self-discharge happened rather quickly, which helped the cell to store better. A floating test would be interesting, too. Anyhow, it's widely accepted that when floated, you don't want to float at 100%, but rather at 90% or 80%. This has been a standard feature in laptop battery configuration for years.

[amyk]

Yes! I remember the times, just some 5 years ago, when internet forums were flooded full of this shit:

1) "lithium-ion cells are extremely complex and difficult to charge and require special algorithms provided by special ICs" -- false; li-ion is the simplest battery type, practically the only type that can be charged by CC only/voltage stop, or CC-CV, and can also be partially charged. There only is voltage limit and current limit, nothing else. Any proper lab PSU does the exact algorithm. Li-ion is the closest real battery to the "ideal" battery. This claim seems to be a mixup with NiMH batteries, or just pure stupidity.

At least one of the articles I linked to in my previous posts has a study of cells kept on float charge at 4.2V; it seems capacity does not degrade much as long as they're not being kept at high temperatures.
2) "lithium-ion cells blow up instantly when you do anything wrong" -- false, as we can clearly see. OTOH, it's essential to have good safety precautions and know what you are doing. Li-ion cells are hightech, and it really is dangerous when manufactured by some random no-name sweatshop; or even the good cells if they are severely misused.

I think the Chinese are largely responsible for debunking these points, because of the huge amount of products they've created which use lion cells yet have ridiculously simple charging circuitry; nothing more than a 4.2V regulated supply with current-limiting resistor. Granted, some of their cells do come to an exciting end, and that's what makes the news, but that neglects the huge number of others which operate uneventfully. 18650s are quickly becoming the new AA as can be seen by the numbers of products using them (replaceably) now. I'm optimistic that in a few more years, they'll become even more popular once the mysticism around them disappears.

[Artlav]

Here is something curious.
13.1.1 and 13.1.2 are Samsung ICR18650-22B, 11.3.1 and 11.3.2 are Samsung ICR18650-22F.
Both have nearly identical specs, and both got stable capacity over 4 cycles.
Both were found at mid-3V voltage.

And yet, they are quite different - the F got half of it's rated capacity.
The interesting thing is how the curves change - the slope is capacity dependent like that.
It really is as if they have different DCR.

In any case, thanks to the inspiration from this thread, I continued my research. I originally planned that I'd let the cells sit for a year, but they are a bit overdue now!

Glad to be an inspiration. 

That graph I posted is different, though; it's R=dU/dI on whole discharge curves at once! It's the best way to avoid AC effects completely, but you get heating effects, because at higher current, the cell runs hotter, and hotter cell has lower DCR. Heating effects don't affect the pulse method in the same way, because of the thermal mass.

How would that work if you know voltage and current along the graph, and they are not the same?

I tried to find it by matching the graphs by capacity spent/accepted, then doing (Vc-Vd)/abs(Ic-Id), and i'm getting something that does not look right at all.
It is proportional to the cell's crappyness, however.

The edgy green line is the computed DCR, at 100mOhms per bright line:

There is only "ohmic" resistance, because ohm by definition is the unit of resistance

I mean that in a sense that we are measuring more than just a resistor.

Most of the AC effects go away by simply waiting for one second.

The correct way to measure the DCR of a cell/battery is to compare the difference in voltage at two different non-zero currents close in time to each other

Ok... I got to try that.

Finding really credible sources is difficult!

Yeah, i think i understand how stuff like antivaxers spread now. It's just not obvious what is true and what isn't when you didn't do a lot of searching.   

1) "lithium-ion cells are extremely complex and difficult to charge and require special algorithms provided by special ICs"

Having done an exercise similar to this thread's one to a bunch of NiMH AAs, i cringe on that one as well.
With Nickel, you don't "charge" a battery, you feed the current into it that makes some chemical reactions go at a rate almost independent of the charge level.
The only way to know you are done is to look for tiny variations in voltage or charge fast enough for the battery to heat up rapidly when the active material gets depleted.

While Li-Ion is pretty much a capacitor - current in is dependent on the state of charge. Limit it and done.
Just WTF.

2) "lithium-ion cells blow up instantly when you do anything wrong"

Now that is an easy one to believe in.
I had at one point (2009-ish) crashed a quadcopter enough to puncture the LiPo. One tiny little hole, and yet it was smoking in 10 seconds and ablaze in 15.
Teaches you to respect these...

And now, here is this guy who tries to make this kind of cells go off again, and again, and again with no luck.





Kinda makes you wonder if maybe the technology changed over the years?
Or i was just "lucky".

[Artlav]

Wait, no. I was stupid.
I matched the capacity backwards - discharge from top to bottom with charge from bottom to empty. 

When done right, i get something much more interesting.
Here is the above one, at 100mOhm per bright line:


And a comparison of the two cells i talked about earlier, at 1 Ohm per bright line.
The DCR goes up to several Ohms near the end (but i guess that's method artifact by then).
So yeah, it kinda adds up with the incomplete depletion.

[Siwastaja]

How would that work if you know voltage and current along the graph, and they are not the same?

You plot the voltages on common X axis, which must be charge (mAh). Your current integration must be fairly accurate. You need to start at the same point, i.e., have fairly well standardized 100% charge before you do the discharges. Then, you will get two voltage curves which show different amount of voltage sag compared to (imaginary) Open Circuit Voltage, but they overlap in time just fine. This is because even remotely good cells have Coulombic efficiency over 99.9% (typically 99.97%), so if the same cell is discharged by x mAh twice, it's (exactly enough) at the same SoC, regardless of the current it was discharged at - as long as you stay within limits of no damage occuring to the cell.

I mean that in a sense that we are measuring more than just a resistor.

Yes, it's kind of a variable resistor that someone turns in interesting ways depending on their mood . Of course, resistance is only a number that models the actual phenomenon, which is a voltage drop. Voltage drop is caused by ions going to their places too slowly. It's like a traffic jam; road seems full, and voltage indicates "fuller" (or emptier) battery, whereas in fact the cars are still on their way, not at home yet. Voltage "updates slowly".

With Nickel, you don't "charge" a battery, you feed the current into it that makes some chemical reactions go at a rate almost independent of the charge level.
The only way to know you are done is to look for tiny variations in voltage or charge fast enough for the battery to heat up rapidly when the active material gets depleted.

While Li-Ion is pretty much a capacitor - current in is dependent on the state of charge. Limit it and done.
Just WTF.

Yes, about 99.97-99.99% of the charge fed into li-ion battery does exactly the right chemical reaction which is kind of transparent to the user, and the rest slowly degrades the battery, as long as the simple constraints are met - no huge overcurrents, no huge overvoltages, no charging at very low temperatures. Li-ion also stores quite well at any SoC, so no huge constraints here. Fully charged? No problem. Battery at 50%? No problem. Empty battery? No problem. Just don't overdischarge or overcharge.

Also, nowadays even the cheapest laptop cells can provide quite nice discharge currents at good efficiency. Compare the 1C of the cheapest energy-type li-ion cell to the 0.2C of lead-acid which is sometimes considered "quick discharge" or otherwise the Peukert effect messes everything up; no Peukert at all on li-ion!

I had at one point (2009-ish) crashed a quadcopter enough to puncture the LiPo. One tiny little hole, and yet it was smoking in 10 seconds and ablaze in 15.
Teaches you to respect these...

Puncture reactions have been getting much better (on quality cells only, of course; all RC LiPo stuff is shit anyway) lately, due to advanced separator design. PTC in 18650 cap can only protect against external shorts. But when the battery is internally shorted (because of the puncture), the electrodes heat up but before thermal runaway, so-called "shutdown separator" melts and shuts down any ion transfer, stopping current generation. If this all happens quickly and evenly enough inside the battery, it can be made so that no spot reaches the thermal runaway temperature (which for LCO or NCA is about 160 deg C). But these mechanisms are still somewhat unreliable. Much better than nothing, but never design a product so that the cell could be punctured, even a high quality one, even though it is unlikely that it goes all flamey. But it happened to Tesla when enough nearby cells were physically crushed at the same time.

Kinda makes you wonder if maybe the technology changed over the years?

Indeed!

But never trust random noname Hobbyking batteries. From the safety viewpoint, they range from "extremely dangerous" to "slightly horrible". (And anything else than Sony, Sanyo, Panasonic, Samsung, LG and a few others are "nonames".)

[ralphd]

I call BS on claims that lithium ion (Co, Mn) are the simplest to charge.  LiFePO4 are much easier.  You can even use 4 in series to replace 12V lead acid battieries in cars, motorcycles, etc.