Electronics > Projects, Designs, and Technical Stuff
Panasonic NKY467B2 36V 15AH 540Wh. Ebike Li ion upgrade, burning my father'ass?
Zucca:
Siwastaja
Once I started I could not stop reading all of it, I was addicted to it. I thank you so much.
You not only inspired me to go deeper in my chinese dead horse bike project, but also now I fell like one day I will/can/should develop a Li Ion energy storage pack for my future home.
There are so many questions and comments I want to write now, but I prefer to digest it first, do some experiments and report back.
I will ask Dave to create a Battery section in the forum. Probably better not to have it, most of the know how should be in the section "Renewable Energy"
--- Quote ---Renewable Energy: Solar, wind, thermal, nuclear, energy storage, Electric Vehicles etc.
--- End quote ---
Ops I should have posted this there....
nctnico:
@Siwastaja: I'm wondering: is any of your BMS designs used in large production runs of battery packs? From what I've seen the TI chips are pretty common to use as an analog front-end for a microcontroller. If the chips from TI where as bad as you describe nobody would use them.
Siwastaja:
--- Quote from: nctnico on October 11, 2018, 01:09:21 pm ---@Siwastaja: I'm wondering: is any of your BMS designs used in large production runs of battery packs? From what I've seen the TI chips are pretty common to use as an analog front-end for a microcontroller. If the chips from TI where as bad as you describe nobody would use them.
--- End quote ---
Oh, the classic "million flies" argument :)! OK, thanks for your interest really, let me explain my point of view a bit more.
I never dared to commercialize the "full home brew" BMS in large scale, as I explained above, so I kept it small scale. In total, I had 500 and later another 500 cell modules made. I still have some. I want to be careful and avoid making the same mistakes others did. And while my design reduced the complexity and BOM cost by about 3-10x compared to the competitors, this doesn't prove ultimate reliability.
So, if my manifesto says that the existing li-ion products are too unreliable, I would need to prove A) that they really are; and B) that mine are considerably better, and proving or disproving reliability takes a lot of time and resources! I would also be fighting against industry giants, with hard-to-really-prove claims. Not proving every claim may be OK on forum discussions (no one expects it from themselves, only from the others who disagree), but for really professional work on safety and reliability, this marketing speech right here wouldn't cut it.
And I'm not actually a reliablity-nut at all. I'm mostly just fine with the reliability of the TI chip. I'm also fine with an occasional lithium ion fire happening, if the thing is sorted out properly, root cause investigated and fixed.
What I'm not fine with is all the hypocrisy and lack of information around these solutions - let alone the absolute classic: the false sense of security for those who are more concerned about the safety than I am. What I'm also not fine with is the BMS cargo cult, which manifests itself very well in the way how BMS manufacturers need to document in big red letters that yes, you indeed need to connect your BMS shutdown signals to actually shut something down. (For example, see the very first words of Elithion manual: http://lithiumate.elithion.com/php/index.php ) This usage pattern which really exists is not too far from using a wooden BMS unit with bamboo wires that looks real. Even on these forums, I have heard this argument multiple times:
"BMS is absolutely critical for safety".
Then, when questioned, they go on:
"Geez, just put in some BMS chip and you are done!"
They put quite a lot of trust on that random, unnamed chip, and quite a lot of trust on the application engineer deploying it properly, for such "safety critical" thing!
BTW, AFAIK, and correct me if I'm wrong, but most of the "TI chips" and alike do not have any kind of safety qualifications done (they wouldn't probably pass), and if the product is in relevant group so that you need the safety stuff done, you as a system integrator take the full risk, and you need to understand the big picture completely, all the small details included.
Only the most expensive chips very few people end up using have the paperwork done (which doesn't prove much, as explained earlier).
. . .
So I went on to make case-by-case custom solutions where I need li-ion batteries as a part of something larger, because I like large problems, often done in "good enough" manner. I always try to prefer a single-cell solution or max 2s-3s (without cell-level management) and step up the voltage, if possible, and "manage" the cell by the MCU that sits there anyway. If necessary, I employ completely redundant analog backups with comparators and voltage references. Completely redundant also means redundant power switch transistors. Makes me sleep better. Yeah, I don't trust TI - too amateurish, too many products in too quick cycles, with too little attention to details that matter. Also, tend to need some babysitting.
Yes, I have needed to make a small-scale recall and bodge fix some devices because of a compounded failure of:
1) Me failing SOA calculations in hurry, so that one level of security is wiped out by a shorted MOSFET,
2) TI failing their li-ion chip design, probably in hurry as well, so that the protection which should exist doesn't, and they happily overcharge the cell by actively driving current to it while their internal non-connected signal tells to stop.
This didn't have cell-level measurement - which wouldn't have helped, since the controller was screaming "overvoltage" anyway, cells were in perfect balance; the signal was just ignored. It's the #1 n00b mistake as explained in the Elithion manual, but when it's inside a TI's chip, you can't do anything but add your own external protection. Did I say something about false sense of security? Or babysitting these chips?
--
For your questioning of "nobody would use them" -- it's as I explained before; the chips I'm complaining about are probably "good enough" so they they don't fail all the time, or in much bigger numbers than the devices they are designed in would fail anyway, and when they fail, it's not catastrophic, because the cells handle the abuse because the li-ion R&D business - think about Sony, Panasonic, Samsung, etc. have been really responsible. It's the high tech I'd be proud of, but which gets little credit - only the extremely rare li-ion "explosions" are reported on media!
Now, the funny thing is, when a BMS fails, the whole product fails. When it's (hopefully) examined by the designer for forensics, who is not a li-ion BMS expert, but "just designed in a nice & easy chip", the first thing they do, they measure the cell voltages. A cell is at 0V - so must have been a bad cell! Probably a wrong conclusion. In fact, it was a failed BMS, which killed the cell. The reason gets classified wrong. I'm 99% positive that TI BMS chips fail at 10x or 100x the rate compared to failures originating from the cell. Even 1000x wouldn't surprise me - cell failures are such rare incidences. If cells would randomly short out or start leaking, massive paralleled packs of 18650's, widely used not only by TESLA, would show massive numbers of problems.
The safety is built in the cells. If the cells were really dangerous as people expect, the game would be totally different, and we would have very reliable BMS solutions available out of necessity.
Please understand that TI's just a typical representative example. Others are similar. TI is most widely used and seems to have the largest portfolio. And I think they can handle my critique >:D. And maybe I'm still a bit angry about my 4.63V cells and their partial responsibility about it, and want to vent off?
My point? It isn't to claim these products are total unreliable crap which automatically blow up your battery pack instantly.
It is, people tend to represent these BMS products as completely irreplaceable and extremely robust, extremely well designed safety devices. This is really not the case. Many proper battery systems, typically less than 6s, do not use any cell-level BMS at all.
Yes, I have designed in a TI li-ion I2C MCU AFE once - maybe I was eating the wrong mushrooms while doing that decision? I didn't even go on testing it after getting the prototypes, but desoldered it from the first prototype. It has an I2C interface through which you turn balancing resistors on, then use another message to turn them off. There is no timeout. If this message never comes through - for example, due to I2C bus lockup, which is surprisingly common - you probably have a bricked product. Yeah, implement all the watchdogs. Just to be sure, inject the "typical" I2C reset pattern of clock transitions. How do you test it is effective in actual lockup condition? Any such event during the whole product lifetime could brick it for good. MTBF must be extremely high for no field returns. Really doing this properly on the MCU is a lot of work. So it's not a properly integrated solution - you need considerable work to babysit it. If they just had implemented an utterly trivial timeout counter... But no, every integrated solution seems to have at least one such showstopper-class deficiency.
This feature alone prevents me from using said product. These products are designed for low-quality cheap crap. The performance is similar to what I'd expect from a $0.02 Shenzen special available at LSCS, meant to be used in a toy. What I don't like is the mental image we are given.
But dead products happen! Some percentage of field returns is usually accepted on consumer electronics. In industrial, having a maintenance agreement, and having things break down every now and then (but not too often) can be a good milking cow, depends on how you play the game.
I'm also not proposing any simple answer on "what you should use". Sadly, I'd like to, but I don't know. I don't have a really reliable, good BMS system in my mind. Mine is probably not such; nor is TI's. I revisited the COTS BMS chip offerings this year, spending full working week (60 hours) going through all available BMS products for a 6s battery. None of them made me feel confident.
I expect and hope the safety in battery technology itself to continue increasing so that the BMS would become even less critical. Now, it's already noncritical enough so that the market doesn't need to produce really safe and reliable BMS products.
People want easy answers and quick solutions. Application notes and application field engineers exist for this purpose. This time, I can't give any easy answer - I haven't found it myself. Sorry for that.
TLDR: People just use a TI chip since that's what they are recommended to do. People don't know the failure modes of the batteries. People are happy if a typical product last a typical average lifetime, and returns are within some typical levels.
nctnico:
Sorry for being a bit obnoxious but I'm getting a bit wary when people say 'the whole world has it the wrong way around'. You do make good points. One of the reasons my customer produces battery packs in the NL is because there are almost no decent Li-ion packs available on the market. Most of what comes out of China is too crappy (unreliable) for high-end commercial / industrial use. They do quite a bit of research themselves as well.
When it comes to safety the BMSes I have been dealing with (for high volume production battery packs) ultimately have a fuse which interrupts the circuit when the battery gets shorted and the BMS doesn't cut the power. Other than that there is a big reliance on software and cleverly designed hardware avoiding single points of failure (can't explain further due to confidentiality) where it comes to protection. Other than that battery packs are also required to pass CE and UN38.3 testing. The UN38.3 safety testing is particulary interesting because these involve testing multiple packs for vibration, temperature, charging, discharging. If a pack survives the UN38.3 torture testing you have some degree of a assurance the battery pack won't fall apart or catch fire by itself.
Siwastaja:
--- Quote from: nctnico on October 11, 2018, 06:05:30 pm ---Sorry for being a bit obnoxious but I'm getting a bit wary when people say 'the whole world has it the wrong way around'.
--- End quote ---
I agree; OTOH I'm not saying "the whole world has it the wrong way around". It's not that black and white. I'd say, the whole world has it in a non-optimal way, and people design things without thinking. Which I think you'd agree with. It's not limited to li-ion battery management. People do the separate analog and digital ground planes as well as taught by application notes, without giving it a thought. Now, there is one difference: no one is saying that your things catches fire and explodes by using the wrong kind of ground plane. This kind of simplified argumentation is true when discussing about li-ion management.
--- Quote ---there are almost no decent Li-ion packs available on the market. Most of what comes out of China is too crappy (unreliable) for high-end commercial / industrial use.
--- End quote ---
We noticed exactly the same. Spawning our own cell testing system to verify/prove/disprove assumptions and, foremost, evaluate real-life cycle lifes in different conditions. Trying to find "optimum way", I built this in 2014-2015: . It's still in use, I now use it for building battery packs for mobile robots in a related startup I now design for... Not using nickel strip but direct copper interfaces is both cost and performance optimization.
--- Quote ---When it comes to safety the BMSes I have been dealing with (for high volume production battery packs) ultimately have a fuse which interrupts the circuit when the battery gets shorted and the BMS doesn't cut the power.
--- End quote ---
Yes. The bog standard fuse. Rated correctly, it's the most important protection. (Note that some BMS's have some peculiar limitations when adding fuses or contactors mid-pack. Always be sure to understand these limitations.)
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version