Diagnosing Lithium cells?

[Siwastaja]

I call BS

I can see so many shining, elaborate arguments right here, in reply to my long expert posts!

LiFePO4 are much easier.

BTW. Your sentences don't even make any sense. LiFePO4 is lithium ion. My post included LiFePO4. LiFePO4 is not easier than LiFePO4.

... even tough LFP is a rarely used niche product mostly manufactured by substandard companies, which was one of the reasons I decided to ditch it. The energy density is horribly poor, especially the Chinese cells are now lacking so much behind -- 95 Wh/kg versus 270 Wh/kg of the state-of-the-art NCA cells. LFP is also too expensive because of the Chinese price fixing strategy. 30-40% higher price for some 60-70% worse product just doesn't cut it.

And, LiFePO4 is more difficult than LCO/NCA/NCM in one regard -- it has very flat curve with no indication of SoC. It is harder to use in some applications (like my bidirectional battery cycling workstation which needs a "storage" battery kept at 50%). OTOH, flatter curve can be nice in some cases. It's even closer to the "ideal" battery model, or would be if there was no knee sections in the curve.

You can even use 4 in series to replace 12V lead acid battieries in cars, motorcycles, etc.

The only reason you cannot use 4 LCO/NCA/NMC in series to replace a 12V battery is because the voltage is higher and your car bulbs will burn out with too much voltage. Nothing to do with difficulty of charging.     

You can replace a 24V car battery with a 7s NCA. I'd do it without cell-by-cell monitoring, but it's not recommended. It's exactly the same for LFP! The "LFP never requires a BMS" internet forum bullshit is just based on fact that no one professionally uses LFP, and no quality manufacturer makes it, so nothing matters the slightest. You can claim anything and do anything. And of course it works. NCA is sometimes professionally used without BMS, for example by BOSCH.

Once again: li-ion (inluding LFP, d'oh ) charging is easiest to understand. Simple chemical reactions, practically no side reactions messing things up. No Peukert effect, and charge in = charge out. Near to ideal battery model. No extremely complex side reaction stuff from lead acid or NiCd/NiMh. No specialized charger IC needed at all for good results, because there is no magic. Li-ion charger ICs are just CCCV power supplies.

[amyk]

You cannot use 4 LCO/NCA/NMC in series to replace a 12V battery, because the voltage is higher and your car bulbs will burn out with too much voltage.

He is talking about 4S LFP, which is around 12V nominal.

[Siwastaja]

You cannot use 4 LCO/NCA/NMC in series to replace a 12V battery, because the voltage is higher and your car bulbs will burn out with too much voltage.

He is talking about 4S LFP, which is around 12V nominal.

I know, you probably missed his point, and hence my reply too.

[KL27x]

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,

The truth is that li ions cells do not even need CC-CV. This is just one fancy way to describe it. To distill this algorithm even further, all you need is current and voltage limiting. This is why a 4.2V current source with a series resistor works perfectly fine and is not the cause of cells igniting. As long as the value of the resistor limits the maximum charge current to within acceptable limit, this method is perfectly safe and perfectly capable. Since we know the minimum float voltage of a li ion cell in the usable range, and in fact your application and/or battery circuit must define that lower limit if you want your batteries to last, ohms law is all it takes to choose the right resistor. It will take longer to charge the cell, particularly the last 20%, but there is no safety concern. I suggested this as a valid and capable method of charging cells 7-8 years ago. I never purchased or reverse engineered any li ion charging circuit, although I have ripped several out and thrown them in the trash. I read the datasheet of a li ion charging IC and understood what the graphs meant on a basic level, rather than reading more into it than what it was. Good thing there are engineers in China that don't get their education from random commercial websites with ties to commercial industry.

[Siwastaja]

The truth is that li ions cells do not even need CC-CV. This is just one fancy way to describe it. To distill this algorithm even further, all you need is current and voltage limiting.

Exactly!

Li-ion is easy, because there are only limits. It's easier to follow limits that are completely stateless than to think about steps. There are no steps. You just follow a few very easy rules all the time.

Another good thing is that you don't need to do anything to keep the cell in shape. Don't want to fully charge? Then don't charge fully. Ditch the "CV" phase, make a CC circuit and a voltage latch that disables the charger at 4.2V. Charge as much or as little as you want to, as long as you follow the simple limiting values.

Constant voltage supply, such as an LM317 regulator set at 4.2V, with a series resistor is perfectly fine, if you confirm manually that you don't start charging overdischarged cells (there is a minimum voltage limit for charging!).

So we have:
Voltage range for safe charging & discharging (typically, 2.5V to 4.2V)
Maximum charging current (typically at least 0.5C) and maximum discharging current (typically at least 1C)
Temperature range for safe charging (typically, 0 degC to 45 degC)
Temperature range for safe discharging (typically, -20 degC to 60 degC)

All of these make perfect sense and are easy to obey with simple circuitry.

And every time someone asks: "Can I do it my way?" The only question is: "Are you following the limits?" If yes, then it simply doesn't matter. Li-ion is a simple charge sponge for the user. For the manufacturer, it's full of expensive, complex high-tech manufacturing processes.

Optimizing for maximum life requires some extra knowledge, but it is not hugely important. Just following the limits gives better lifetime than any other widely used battery technology there has been. The same goes for bending the limits; some are not hugely important, and if you know what you are doing, you can do more.

[djQUAN]

I fully agree with the "Li-ION charging is easiest" statement.

You put current in and limit voltage that's it. Although the voltage limit should be pretty accurate but that is not a problem these days with the available devices.

Lead acid needs a four step charge process. which has bulk, absorb, float and optionally, terminate and needs to be done with timers and different voltages.

And it doesn't really like staying in a partially discharged state which occurs in quite a bit of applications.

[Siwastaja]

Although the voltage limit should be pretty accurate

Relatively accurate, yes, but not overly so. Any half-decent voltage reference will do, as long as it's not complete shit. And the voltage isn't  temperature-dependent like lead acid termination voltage is!

I tested Samsung ICR18650-26H cycled with different end voltages. The highest was 4.30. I terminated the test after 100 cycles - there was approx. 0% capacity fade on all samples, including the one charged to 4.30V on every cycle.

Datasheets often specify maximum at 4.25V.

And, if you are wise, you implement normal charging to only 4.15V, or even 4.10V. That way, you have quite a lot of margin before 4.30V, after which the risk of severe damage really starts going up.

I did some 10 cycles (maybe 8 or 9) charged to 4.45V and didn't see any capacity fade, either, but I'm quite sure I would have seen problems if I had carried on cycling. It can happen rather suddenly; severe capacity drop could happen during a few cycles after 20-50 cycles. And there is significant safety issue with charging to such high voltages. Charging to 4.45V brought about 20% more usable energy storage capacity ! Don't do this at home, or at least use proper precautions such as closed steel casing without flammable materials nearby, even though the risk of quality cells blowing up from such "slight" abuse is smallish because of their advanced separator design (among other things). For example, I haven't been able to blow up a Sony 18650 cell by applying 30V from a 10A power supply for an hour or so, not even venting, but that doesn't mean it couldn't go up in flames. The protections are good, but not perfect and not meant to be actually tried in actual products.

But if you are even remotely interested about safety (as everyone probably should), the most solid tip is: only use genuine cells from the world-class major manufacturers, which are Panasonic, Sanyo, Sony, Samsung, LG and possibly a few Chinese ones like Lishen, but none of the ones praised by hobbyists are any good. You are at a risk, albeit small, when using such products, even when obeying the limits, but especially in the case of abuse.

[Artlav]

Meanwhile, i tried several ways of measuring DCR.
0.1 second wait every 10 seconds, and 1 second wait every minute.
Of the four methods (curves, 0.0001s, 0.1s, 1s) none give the same results.

-100us one does not differentiate anything, and basically always reports around 90mOhms.
-0.1s method gives 170-200mOhm curve for the good cells.
-1s method gives a 200mOhm to 300mOhm curve for the same good cells, and 600-700mOhm curve for bad cells. Also, it gives significant artifacts on the discharge curve, the voltage takes about 5 seconds to settle back to the pre-measure level.
-The curves method give a 50mOhm to a peak of several Ohms near the end more or less straight curve for a good cell, and 100mOhm up to a similar peak for a bad cell.

All in all, the 0.1s method seem to be the best, as it differentiate good from bad without interfering with the rest of the process.
The curves method might be a close second, since it also seem differentiating, but that need some more work to clean it up.

even tough LFP is a rarely used niche product mostly manufactured by substandard companies, which was one of the reasons I decided to ditch it.

Hm, i heard differently. Specifically, that there were many industrial applications, car manufacturers using them to replace lead-acid starter batteries in the newer cars, various hobby-grade electric vehicles, go carts and so on.
The advantages are supposed to be good cycle life and abuse resistance (no thermal runaway), even if the energy density is less.

I got a bank of 40152 LiFePO4 cells working as a UPS for the 24V lights&trinkets wiring i have at home (you can see the the rig plugged into it on the first post's picture).
However, all i can say is that they work. Haven't done any stress tests on them.

I did a table of $/KWh for various cells on on sale, and they came out the cheapest at about $600 per KWh, above only the crap chineese and noname cells.

Now, used 18650 can be as low as $100 per KWh, but they come in a variety of condition and quality, and by the time you filter out the good ones you're back into $400-$500 range.

[djQUAN]

I have a bunch of recycled 18650's from old laptops and tested each one (charge then discharge at 1A to see remaining capacity) and most are still pretty good. I use a hobby charger for testing and charging packs.

Edit: I have experienced overheating of some cells when charging but it occurred even when the cells have been discharged only down to 3.2V. Charging current is reasonable for a single string of 18650's (1A) It got pretty hot and I was able to pull it out on time so I thought those cells were toast. Left it to cool and waited for fireworks but nothing happened so I charged them again a day later but at a lower rate (300mA) and it terminated and is now back in service as if nothing had happened. Anyone know what was at play here? Cells were laptop pulls (genuine Sanyo cells) but still have some life in them.



I also have a couple packs of LiFePO4, the big ones 40152 also. I used a 4S one as a portable battery pack and another 4S3P pack as a huge power bank.

From my limited experience with those, LiFePO4 seems pretty good to me compared to a lead acid of similar size for the application.

[Siwastaja]

Of the four methods (curves, 0.0001s, 0.1s, 1s) none give the same results.

This is not surprising, since you are not measuring DC resistance, but AC impedance. You have already found the answer by yourself, you just need to realize what you are seeing:

the voltage takes about 5 seconds to settle back to the pre-measure level

This is what DC resistance is all about! It's the voltage sag between two different current levels. To actually see the DC component, you need to use a long enough pulse that all AC effects are removed, i.e., the voltage needs to settle back fully.

I haven't even tried going much below 5 seconds, because I don't want to see AC effects. Cell chemistry is "sluggish". You'll need a rather long pulse. IIRC, I use 3 or 4 seconds in one system, a bulk cycling setup, and 10 seconds in another, a more precision setup.

Also, it gives significant artifacts on the discharge curve

Yes, you have to choose between perfectly smooth discharge curve (no data lacking), or a good DCR measurement curve. You can't get both at once. Or you can do two different discharges and get both perfect discharge curves, and a full DCR curve calculated from the two curves.

I like the pulse method because it gives almost perfect discharge curve and a slightly quantized but good enough DCR curve, on one go.

However, I don't see much use for "perfect" discharge curve. You can measure the DCR once every two minutes, for example, and use a 10-second pulse. Then, you'll have about 20 seconds of missing data, which you can fill in with linear interpolation. The error in discharge curve during the interpolated sections will be rather minimal; not much happens during 20 seconds when discharging at 0.5C or maybe 1C. Just don't try to get the DCR from the very last maybe 0 to 5% SoC.

All in all, the 0.1s method seem to be the best

It's probably good enough to differentiate between good and bad, it gives idea of the real DC resistance, and you'll get clean curves, so go for it.

[Siwastaja]

even tough LFP is a rarely used niche product mostly manufactured by substandard companies, which was one of the reasons I decided to ditch it.

Hm, i heard differently. Specifically, that there were many industrial applications, car manufacturers using them to replace lead-acid starter batteries in the newer cars, various hobby-grade electric vehicles, go carts and so on

Partially true, yes.

The problem with forum hearsay is the it's never questioned or checked, the same "facts" circulate year to year.

There was significant academic (and probably industrial) interest in the LFP chemistry in early 2000's, more than 10 years ago. It was a promising technology for a few reasons:

(1) It is cobalt-free - cobalt is somewhat significant cost factor in the traditional LCO chemistry, also slightly toxic.
(2) Significantly higher onset temperature for thermal runaway (compared to LCO), so, safer
(3) Lower voltage helps slow down electrolyte reactions (overgrowing SEI layer) - longer calendar life
(4) Lower voltage helps slow down lithium plating - longer cycle life, higher charging C rate
(5) The cathode lattice structure allows it to be fully charged - LCO cathode can only be charged to about 50%
(6) The energy density tradeoff is only about 20-30% compared to LCO - a good tradeoff!
(7) Finally, the price could be much lower than LCO! Li-ion is hugely expensive - this could be the revolution!

But the problem, as always, is that the ideas and practice do not always meet. In fact, everything went the opposite way:
(1) They got the manganese and nickel work better in the cathode (the problem of Mn in cathode was its dissolution in electrolyte, not sure about what was holding nickel back). The most state-of-the art NCA chemistry has gotten rid of 80% of the cobalt, and NMC has lost 67% of it. These new chemistries are both better than LCO in all respects, not compromises.

(2) This holds true; thermal runaway is a lot less likely on LFP cathode. It is not impossible, however. Anyone who claims that, is lying. However, thermal runaway onset temperature on cathode is only a small part of complete li-ion safety. LFP still has thermal runaway, it has exactly the same burning electrolyte composition which can shoot out of the cell and ignite. The DIY EV community has had several LFP related fires that resulted in total loss of the property. I'm not saying that LFP is extremely dangerous, but I'm just comparing. It may not be safest at all. We don't have enough real safety data to say one way or the other, but my gut feeling says that in practice, the real LFP may be considerably worse in safety.

While LCO, NCA and NMC are at the risk of thermal runaway at lower onset temperatures (at about 160 to 200 deg C, instead of about 350 deg C for LFP, IIRC, feel free to check the numbers), the safety can be brought by advanced design ideas, mostly in the separator film. Such R&D is only done by the Big Players. There was this paper (again, can't remember the title or the authors, sorry) where they tested short-circuiting and overcharging (12V supply, IIRC) commercial 18650 LCO and 25650 LFP cells, and all LFP samples reacted much more badly because of the lack of protections and safety design.

The problem with LFP safety is that the manufacturers are small players with lack of advanced safety measures, and they are trusting too much in "safe" chemisty. And while the cathode chemistry is a lot safer, the resulting complete product may be less safe.

(3&4) Electrolyte and electrode design has come a long way. Nowadays, 4.20V is just fine. Modern cells still do increase their DCR in storage, but the fully charged calendar life has at least doubled from a few years to nearly 10 years. The modern cells also exhibit better structure on the electrodes, causing better ion transfer to the lattice, causing less lithium plating even at higher voltages.

They could even go as far as go and release first 4.30V then 4.35V LCO cells to increase the energy density. Now the best products are back to 4.20V, but they do have leeway in the reactions, and very good life.

So, the problems that LFP promised to solve, were solved in another ways, without the huge tradeoffs of LFP.

(5) LFP cathode can be fully charged, but so what? It's gravimetric capacity is hugely poor, so even when it can be fully utilized, it's still horribly poor compared to LCO, NCA or NCM.

(6) Before 2005-2007, when the LFP was still a "thing", a typical state-of-the-art LCO cell was about 150-160 Wh/kg. LFP was expected to be at about 120-130 Wh/kg, only a 20% tradeoff in energy density. But LFP was rather "fully developed" from the beginning, and large companies with proper R&D didn't see it could work out. LCO went to about 220 Wh/kg, and then became NCA which is now at about 270 Wh/kg on the state of the art products, whereas LFP is still at 120-130 Wh/kg. The Tesla's battery pack, which gives the car a 500km driving range, weighs something like 500 kg which is already a lot. If done in LFP, it would weigh more than 1000 kg, which would be totally impossible.

(7) Price. So finally, most of the LFP, which is a tiny market share on li-ion, probably less than a percent, are produced by smallish Chinese producers with no or substandard quality control. They clearly have a price fixing agreement going on. The prices have gone down much slower than on any other li-ion chemistry. 6-7 years ago it was different, but in 2015, NCA cells from Samsung or LG or maybe even Panasonic are so much cheaper than the Chinese LFP, per energy - about $200 / kWh vs. $300/kWh.

DIY EV conversions done in LFP are very limited in range because of the poor energy density, both gravimetric and volumetric, and high price of the LFP. They are mostly of sub 100-mile design.

LFP still has one advantage, and it's the voltage range which is compatible with 12V lead acid system, so a 4s LFP pack can be used as a direct replacement for a 6s lead acid pack. People love to think that there is more to this, and attribute this to the "safety" of LFP or "no BMS" stuff, but really, it's basically only about the voltage (both average voltage and the curve shape). LFP has lower voltage (which partly attributes to the low energy density), and while it's generally a bad thing, it happens to help with this particular scenario, by pure luck.

I did a table of $/KWh for various cells on on sale, and they came out the cheapest at about $600 per KWh, above only the crap chineese and noname cells.

You can get proper Samsung NCA cells, not the most modern stuff, but still almost 2x better energy density than any LFP, for $200/kWh once you find the right distributors and can buy a large enough batch. I have purchased from a local Finnish distributor.

Now, used 18650 can be as low as $100 per KWh, but they come in a variety of condition and quality, and by the time you filter out the good ones you're back into $400-$500 range.

Sounds horrible. I would buy used 18650 for $100/kWh only if they are properly characterized and still show at least 80% of original capacity and <150% of original DCR.

I have considered making a good multichannel test rig for quickly characterizing a lot of 18650's, but then I would pay something like $20-30/kWh from "raw" cells that were only selected with a multimeter (voltage at least 2.0V), because I would expect that at least half of them are rejected, and testing also takes time and energy, even when fully automatized.

I like to try to answer any questions regarding this or anything else li-ion related.