Diagnosing Lithium cells?

[Artlav]

I got a box full of used 18650 cells with unknown history, and i'd like to find out what the history was.
To that end i built a battery analyzer that can charge and discharge the cells while logging the current and voltage curves.

The question is - now what?
Specifically, i remember seeing several papers that described in detail what sorts of abuse or wear leave what types of signs on the discharge curves.
I.e. being charged below zero would make the cell linger around 4V, then drop fast is the only one i remember.

Anyone knows where to find info like that?

[dmwahl]

The main things you'll see are reduced capacity and increased internal resistance, which will show up as greater droop in the voltage curve at a given load.

Nice tester though, do you have the design posted anywhere?

[doobedoobedo]

I agree totally with dmwahl

Do an internal resistance test, it'll tell you much more than discharge curves alone.

Mind you if you're only going to use them for low current applications I don't suppose it much matters.

[crazy horse]

http://batteryuniversity.com/learn/article/testing_lithium_based_batteries

Lots of good battery information on this site.

[amyk]

http://batteryuniversity.com/learn/article/testing_lithium_based_batteries

Lots of goodmisleading battery information on this site.

https://en.wikipedia.org/wiki/Talk%3ALithium-ion_battery#Dubious_information_originating_from_discredited_source

[Artlav]

Interesting.
Conversely, that talk page is the only reference i could find about the "battery university" being discredited.
Do anyone have any details?
I thought it was a good and useful site, and there is a lot of valid info there.

[eas]

http://batteryuniversity.com/learn/article/testing_lithium_based_batteries

Lots of goodmisleading battery information on this site.

https://en.wikipedia.org/wiki/Talk%3ALithium-ion_battery#Dubious_information_originating_from_discredited_source

What a clusterf*ck that whole discussion is. It reads like a cargo-cult version of intellectual rigor.

[crazy horse]

It does seem if it is a bit personal.

[ez24]

Interesting

[Jay_Diddy_B]

Hi,

I can contribute this information. I recently measured an 18650 Cell made by PK Cell, nominal capacity 2200 mAh.
This is my measurement setup:




The equipment was controlled using HPIB.

These are the results:






The horizontal axis is depth of discharge, 0 is fully charged, 1 is fully discharged.

The ESR was measured by interrupting the discharge current and measuring the change in voltage.

I also measured a Ultracell Rose 4200mAh cell obtained from eBay. This graph show the raw data. The blips shows the voltage when the load was turned off to measure ESR.



This battery had a measured capacity of 2250 mAh.


Regards,

Jay_Diddy_B

[Artlav]

I can contribute this information.
...
The ESR was measured by interrupting the discharge current and measuring the change in voltage.

That's interesting.
What i plan on doing, is to add ESR logging during discharge.
Measure voltage, turn the load off, wait 100us, measure voltage, subtract, do the math, turn the load back on.
Would that be a valid method?
The descriptions i google up mention the inverse being done - that is, load the cell for a millisecond, measure the voltage drop.

More importantly, does the length of the delay matter?
When the discharge switches off at the end, the cell quickly gains voltage, from 3 to 3.6 in a few seconds.
So, how long is long enough, and how long is too long?

Nice tester though, do you have the design posted anywhere?

No, i don't. It's a pretty straightforward thing, but if there is interest i can describe it.
One thing is, it's only a few days old and i've already fixed a few big bugs, so there won't be any quality in hurry.

[Jay_Diddy_B]

Artlav,

I think the best way to measure the ESR would be to reduce (or increase) the load current and calculate R= delta V / delta I.

I would use a scope to determine if the timing matters.

I had to use relatively long timing, because I had the DMM free-running, I could have shortened the time if I knew when the DMMs are going to read.

Being consistent is important, if you want to compare cells.

Regards,

Jay_Diddy_B

[ralphd]

http://batteryuniversity.com/learn/article/testing_lithium_based_batteries

Lots of goodmisleading battery information on this site.

https://en.wikipedia.org/wiki/Talk%3ALithium-ion_battery#Dubious_information_originating_from_discredited_source

The wiki post is dubious.  Battery University published verifiable details.  The SEM photos might be hard to do on your own, but charge/discharge and IR measurements are easy to verify.
http://batteryuniversity.com/learn/article/how_to_restore_nickel_based_batteries

[dmwahl]

However you end up measuring internal resistance, be sure to use a 4 wire connection so that the resistance of your test leads doesn't affect the result. You may have thought of that already, but just wanted to mention it.

[Artlav]

I would use a scope to determine if the timing matters.

Ok, made it and checked - the voltage settles within about 1us, i suspect more due to the FET switching delay than anything else.
After that it stays flat for hundreds of us. So 100us is good.
Haven't tried to find the upper limit, however.

How valid the measurements are is a big question.

I'm getting 90mV (+-5mV, 3.9 Ohm load to open circuit+10k sense divider) of difference on both near-dead and half-decent cells, and it's not changing much at all from full (4.1V) to empty (3.2V).
Starting at 90mV at 4V, it dropped to 80mV at 3.2V, or 87 mOhm to 97 mOhm increase near the end.

"Near-dead" as in they started at about 1000mAh, and are down to 400mAh after 3 charge-discharge cycles.
"Half-decent" as in they are at 1200mAh out of the rated 2200mAh, and it changed little over the same 3 cycles.
2 of each, all at once.

Sounds like the ESR is not really that indicative of anything.
Confirmed on a scope, so it's not a rig error.

Should i look at absolute values, or am i doing something wrong?

be sure to use a 4 wire connection so that the resistance of your test leads doesn't affect the result.

Naturally.

[doobedoobedo]

A Li-ion that's lost half it's capacity and has 87-97mOhm internal resistance is not half-decent. It's mostly dead.

It'd still work for something that doesn't draw much current (less than an amp) and only requires a single cell if you're determined to use it.

[eas]

Member HJK has published curves on a huge number of LiIon cells on his site. Seeing how new cells of the type you are testing perform should be useful in evaluating old cells.

From the reading I've done,  the internal resistance of "modern" (ie, those made in the last decade or more) lithium ion cells is of limited use on its own for determining general cell health because most cells are now formulated to maintain a relatively low IR over their target lifespan. The preliminary testing I did with a half-dozen or so used and new-old-stock 18650 cells bore this out. There was a correlation between cell capacity/cycle count, IR and calendar age, but just by eyeballing things I could tell it didn't have great predictive power.

Furthermore, I found in limited tests that cells that were at less than 2/3rds their original capacity still could have a long useful life in front of them. After 50+ cycles, the cells I tested only had a further 1-2% reduction in capacity. This was at relatively high discharge/charge rates too. I don't recall the exact values, but my discharges were at over 1.5A. Not enough for some applications, but not a trickle to an 8-bit MCU either.

Oh, also, yes, the amount of rest time when doing IR tests is probably significant, but as long as you are consistent in your methodology, it should be useful for relative measurements. Keep in mind that the IR numbers from battery manufacturers are usually obtained using an AC impedance measurement technique. It's probably not the most useful measurement for a lot of purposes since it bears little resemblance to the DC workloads cells are subject to, but it's quick, and it apparently works well enough for matching cells for packs.

[Artlav]

A Li-ion that's lost half it's capacity and has 87-97mOhm internal resistance is not half-decent. It's mostly dead.

Curiously enough, the graphs in eas' link show that 100mOhm is about average IR for these cells.

Also, i'm charging them to 4.1V, so the comparable capacity would be about 1500mAh... Which is still close to half of 2200mAh.
Well, it would go into the weak bin, i guess.

I've got a few genuine Sanyos int the rig at the moment, should be a nice reference point on what a good cell looks like.

Furthermore, I found in limited tests that cells that were at less than 2/3rds their original capacity still could have a long useful life in front of them. After 50+ cycles, the cells I tested only had a further 1-2% reduction in capacity.

Is there a dependence on the voltage they were found at (presumably after spending a while at it)?
The bad cells were found at 2.85V, and over 6 cycles they went from 600mAh to 200-300mAh.
The better cells were found at 3.75V, and they kept at around 1200-1300mAh over the same 6 cycles.

Here are the curves. Voltage on the top, truncated to 3V, current on the bottom.
There is a bit of inconsistency as i tweaked the rig along the way, but the patterns are still here.

Good cell, charges:


Good cell, discharges:


Bad cell, charges.
The spikes on the first curve is me probing the current and freezing components to check thermal stability.
Interesting how it goes to 4.1V almost at once.


Bad cell, discharges.
It looks almost as if the cell settled at some value, after the stress burned out whatever unstable capacity was in there.

[Siwastaja]

No official standard to measure DC resistance. Dropping the load off for several seconds every now and then and calculating R = dU/dI gives good approximations. Keep the method fixed between the test for proper comparability.

Or if you do full cycles - both charge and discharge - you can just calculate the DCR from the graph when the graphs are aligned on the common axis (mAh instead of time, unless the charge and discharge current is same). For example, if you charged at 1A and discharged at 2A, look at voltages at the same aligned point (for example, 500 mAh before the cell was full; and 500 mAh after the discharge started. If it was 4.00V while charging and 3.70V while discharging, you'll get R = dU/dI = 0.30V/3A = 100 mOhm.

Often, official "end-of-life" is either 30% drop in capacity or 100% increase in DCR, whichever comes first. Some typical laptop 18650 cells show a DCR of about 50 mOhm when new, so 100 mOhm would be end-of-life. But increased DCR may not be significant in your application at all! Cells with increased resistance can still offer good efficiency and good amount energy when discharged and charged more slowly. Similarly, if you are fine with capacity slightly less than the rated 70%, why toss them.

In addition to these two mechanisms, leakage current a.k.a. self-discharge can also increase significantly in some cases. Leave a rather fully charged cell alone. Measure the voltage after 24 hours, and again after a week or so. You will see if some cells leak significantly more than the others.

For safety reasons, toss any cells that show below about 2V without even trying to charge them. For cells reading between about 2V and 3.0V, precharge them at very low current (C/50 to C/100) to normal empty voltage levels (to about 3.3V) before normal charging.

Never charge li-ion cell, new or used, below 0 degC. Very low currents may be acceptable, but don't risk it. It will at least ruin the life and might cause safety issues in the long run, or in substandard cells. Keeping the temperature higher (30-40 deg C instead of room temp) while charging is good for the life.

[doobedoobedo]

Lithium cells don't like to be stored discharged, the closer they are to around 3.8V the less they dislike it, so yes that could explain the good v bad cells especially if they've been left like that for quite a while.

I'll fess up and say I don't really use round cells much, I use a lot of LiPos though and the 2S2P 1500mAh 35C pack I have in front of me has DCR of 15.1mOhms (tested using voltage drop/current calculation) so 100mOhms for a single cell sounds crazy high in comparison.

[eas]

Initial voltage of recovered cells can be useful for screening, but I think some people put too much stock in it and pitch cells that could have a long useful second life. Low voltage could be a sign of cells with high self-discharge, or the time since last charge, or a flaw in the pack design. Cells may be compromised, or not.

I like Siwastaja's approach. Pitch anything below 2v. Precharge anything between 2 and 3. Charge, check self-discharge after ~12 hours. Any that have dropped by more than about 0.1v below termination voltage could be put through a gentle charge/discharge cycle and checked again.

Not surprised that cells that only showed 600mAh capacity quickly deteriorated. As for their capacity loss leveling off, my understanding is that capacity loss decline is generally due to the Li-ions in the electrolyte reacting with electrode or other electrolyte materials, typically during charging.

[NiHaoMike]

Keeping the temperature higher (30-40 deg C instead of room temp) while charging is good for the life.

Doesn't high temperature age the cells faster?

I recommend limiting charge voltage to 4.1V/cell and charge rate to 1/2C if that is OK in your application.

[Siwastaja]

Lithium cells don't like to be stored discharged, the closer they are to around 3.8V the less they dislike it, so yes that could explain the good v bad cells especially if they've been left like that for quite a while.

Source on this? I have heard the opposite. Of course avoid storing at 0% SoC, but something like 20% could be very good for the life. My own research is still going on, but I'll get exact results later.

I'll fess up and say I don't really use round cells much, I use a lot of LiPos though and the 2S2P 1500mAh 35C pack I have in front of me has DCR of 15.1mOhms (tested using voltage drop/current calculation) so 100mOhms for a single cell sounds crazy high in comparison.

Lithium ion cells are made for many different applications. They can optimize the internal structure (basically, active surface area to active mass ratio) to produce so called energy cells or so called power cells (and anything between these). Both offer near 100% efficiency at extremely slow charge/discharge. Power cells can provide higher currents without overheating. But this is a compromise that lowers the energy density (Wh/kg) and increases price per energy ($/kWh). So you always try to use energy cells in slow discharge, slow charge applications, like laptops, electric vehicles etc. Many 18650 energy cells work best when fully discharged during more than 1.5 hours (0.7C), even though they are often specified 3C peak, 2C continous discharge. If end-of-life is considered 100% increase in DCR, then the rating goes down even more. Power cells are used in power tools; many 18650 power cells are specified to 8C to 12C, i.e. full discharge in less than 6 minutes.

100mOhm DCR is perfectly normal for an aged 18650 laptop-type "energy" cell which has doubled its internal resistance, but can still be great if the application discharges it slowly (below 0.5C). World is full of low-power applications for these recycled cells - the leakage is more important.

[Siwastaja]

Keeping the temperature higher (30-40 deg C instead of room temp) while charging is good for the life.

Doesn't high temperature age the cells faster?

In storage, higher temperature causes the cells to age faster due to electrolyte reaction.
But in actual cycling, lithium plating during charging is the most prominent reaction.

I came across a really good research paper some time ago where they tested this on Panasonic production cells, and it was surprising how much charging at 45degC instead of 25degC helped with the life. Charging is where most of the cycling damage occurs, and as we all (hopefully) know, at 0 deg C it is completely forbidden because it can destroy the cell almost immediately, so this is a very logical finding indeed.

But when in storage, keep the cells as cool as possible.

Another interesting finding in their paper was that storing at 80-90% SoC didn't provide any benefit compared to 100%; you really need to store below 60% (preferably below 50%) to get the benefit. OTOH, my findings show that it's the charging that does the damage between (80%)-90%-100%, so the traditional "only charge to 4.1V" tip is a good one anyway.

Summary for best life:
* Charge at slightly elevated temperatures to slow down lithium plating - but don't go to extremes, over 45degC is often forbidden by the manufacturer and may cause electrolyte reactions.
* Discharge temperature doesn't matter so much, but higher temperature reduces energy losses.
* Store as cool as possible. Store at 40-50% SoC.
* Limit charge rate. 0.5C is the real maximum with many cells! If you can, further limit it slightly after 4.0V - kind of pre-CV-phase.

[doobedoobedo]

Lithium cells don't like to be stored discharged, the closer they are to around 3.8V the less they dislike it, so yes that could explain the good v bad cells especially if they've been left like that for quite a while.

Source on this? I have heard the opposite. Of course avoid storing at 0% SoC, but something like 20% could be very good for the life. My own research is still going on, but I'll get exact results later.

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.

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.
The discharged cell will have puffed considerably and is only fit for the bin, the fully charged cell may well have puffed a bit, but it won't be like the 3V balloon and the 3.8V cell won't have changed.

I'll fess up and say I don't really use round cells much, I use a lot of LiPos though and the 2S2P 1500mAh 35C pack I have in front of me has DCR of 15.1mOhms (tested using voltage drop/current calculation) so 100mOhms for a single cell sounds crazy high in comparison.

Lithium ion cells are made for many different applications. They can optimize the internal structure (basically, active surface area to active mass ratio) to produce so called energy cells or so called power cells (and anything between these). Both offer near 100% efficiency at extremely slow charge/discharge. Power cells can provide higher currents without overheating. But this is a compromise that lowers the energy density (Wh/kg) and increases price per energy ($/kWh). So you always try to use energy cells in slow discharge, slow charge applications, like laptops, electric vehicles etc. Many 18650 energy cells work best when fully discharged during more than 1.5 hours (0.7C), even though they are often specified 3C peak, 2C continous discharge. If end-of-life is considered 100% increase in DCR, then the rating goes down even more. Power cells are used in power tools; many 18650 power cells are specified to 8C to 12C, i.e. full discharge in less than 6 minutes.

100mOhm DCR is perfectly normal for an aged 18650 laptop-type "energy" cell which has doubled its internal resistance, but can still be great if the application discharges it slowly (below 0.5C). World is full of low-power applications for these recycled cells - the leakage is more important.

No argument with this.