Chevy Bolt EV Forum banner

121 - 140 of 175 Posts

·
Registered
Bolt EV, 2019, Premier
Joined
·
75 Posts
One other thought, a lot of BMS controllers have the ability to test for opens or shorts on the sense leads. This process could also lead to anomalous voltage readings that are not really relevant. This might have something to do with the groups of high voltages? Supposition. Needless to say I was reminded that the whole subject is quite complex. lol
 

·
Registered
Joined
·
6,256 Posts
Ahhh okay so this might explain the variance if you’re using a different tool. You may be right in terms of how the app extracts or handles the data.
My graph is an average of two samples for each cell. When I sent the CSV file from Torque Pro to Microsoft Ultra Office Spreadsheet, it did exactly as 2019EVLT described. I had to renumber all the columns to get them in the right order too.
 

·
Registered
Joined
·
1,091 Posts
My graph is an average of two samples for each cell. When I sent the CSV file from Torque Pro to Microsoft Ultra Office Spreadsheet, it did exactly as 2019EVLT described. I had to renumber all the columns to get them in the right order too.
Same with me on EngineLink, the cells on the CSV file were a mess initially, had to fix it on Excel. What I was referring to was the specific reader / software that 2019EVLT was using; I was speculating that perhaps it was numbering the cells differently. But yes this is purely speculative, without seeing the PIDs used this is all conjecture, but at least one way of potentially explaining the seemingly random arrangement of the “bad” cells compared to the consistent outcomes of the ones we’ve seen thus far.
 

·
Registered
Joined
·
890 Posts
I know there is one person (sorry, don't remember the name) who bought the tools to do dealer updates himself. I wonder if he has the GM tools to get more "official" cell voltage readings. Ultimately it would be good to compare those to Torque Pro. I suspect they will probably be the same but on the off chance maybe we are using the wrong ones or there is some problem, it would be good to know Torque Pro is giving us the true values. I don't know how the PID list that most of us downloaded came to be, but sometimes it's someone just finding some things in the data stream and using them. For all we know, there's a "raw" voltage value and a "corrected" value and we may not be using the "corrected" voltage values. I do think that unlikely but it might be worth at least checking.

Mike
 

·
Registered
Joined
·
890 Posts
I went back and watched a portion of Prof. Kelly's video and noticed that the BECM seems to monitor in 6 groups of 16 cells. I overlaid the cells from each group on top of my graph and found something that might be interesting (not sure).

bgbgbg.jpg

First, I think prof. Kelly might have two mislabelled: I think 16-33 should be labeled 16-32 and 33-48 should be 32-48. But be that as it may...

He did mention that each cell at the endpoints of each group of 6 are duplicated because you need a reference voltage to measure the cells in the next group (paraphrasing). I take that to mean that given all the wiring and internals, when the BECM monitors 1-16, it monitors 16-32 next and "compares" what "channel 1" saw for 16 and compares it to what "channel 2" saw for that same cell 16. The question then becomes, what if there is a small difference? I don't know how/why this is done but maybe they have the equivalent of six 16 channel A/D converters to read the voltage. So if channel 1 reads cell 16 as 4.00V and channel 2 sees cell 16 as 4.01V, the BECM might use a -.01V offset for the raw voltages read in channel 2 (for cells 16-32). And so on. Or it may try to "smooth" the readings around the reference cells in some "prorated" fashion.

Beyond that, I found it interesting that if you look at my graph with what I'm calling the 6 channels, you notice that if we name the channels 1-6 from left to right, all the odd channels are "bad" and all the even ones are "good". I also notice that for channels 1, 3, and 5 which I'm calling "bad", exactly half the channel is bad and the other half is fine. In fact, the variance in the even channels (good ones) is about .008V. The variance in the odd ones is .02V.

What does all this mean? I have no idea. Maybe someone more familiar can fill in the blanks. It could be that the BECM has an issue related to some of this. Could be that there are some correction factors (by channel) or "offsets" that Torque Pro isn't considering and is instead displaying the raw values. It kinda makes sense because if you used those endpoint cells to get some type of reference, you'd think you'd actually USE that reference for something, otherwise why even have it. So the system may read raw values and then apply some correction factor. Depending on how the software tries to "spread out" the corrections, I could see the corrections being applied in "prorated" sections which end up affecting about half the channel: 8 to the left and 8 to the right of the reference, moving down the line thus smoothing out the final readings.

I'm not a big believer in coincidences when things seem to align with something. Maybe this is all bunk but I thought I'd post it because someone may be better at interpretting this. Sometimes I miss the forest for the trees. :)

Mike
 

·
Registered
Joined
·
1,091 Posts
I know there is one person (sorry, don't remember the name) who bought the tools to do dealer updates himself. I wonder if he has the GM tools to get more "official" cell voltage readings. Ultimately it would be good to compare those to Torque Pro. I suspect they will probably be the same but on the off chance maybe we are using the wrong ones or there is some problem, it would be good to know Torque Pro is giving us the true values. I don't know how the PID list that most of us downloaded came to be, but sometimes it's someone just finding some things in the data stream and using them. For all we know, there's a "raw" voltage value and a "corrected" value and we may not be using the "corrected" voltage values. I do think that unlikely but it might be worth at least checking.

Mike
I like this train of thinking / troubleshooting. On the raw vs actual or corrected values, whatever the case may be we’re seeing pretty consistent results here, which, statistically speaking, is less and less likely due to be from chance alone.
 

·
Registered
Bolt EV, 2019, Premier
Joined
·
75 Posts
Well I was all wet when I stated that my Bolt did not show this sort of variation. My eyeball filter was defective. This is from my 2019 that I have been level 1 charging to about 75%. It is also plugged nearly 24/7.

31760


This data was from averaging 10,000 readings from Torque Pro. The data really does not look that noisy to me. It appears to me that the resolution is on the order of 12 bits. The minimum change per channel seems to be about .001221 V. If you assume a 5V reference dividing by this you get 4095. This is what 12 bits would give you.

It certainly looks like there are similarities with some of the other deviation profiles. I think I see a pattern every 8 channels. This is a nice "binary" number that might be attributed to some symmetry of the measurement system. For example, possibly groups of 8 are using the same A/D converter or reference?
 

·
Registered
Joined
·
200 Posts
I disagree that "tight" voltage tolerances couldn't cause fires. There's a reason the first thing GM engineers pointed to was 100% charging. While below 100%, a 0.02 V variance isn't anything more than a nuisance that might slightly affect capacity. At 100%, however, it could cause a serious issue, especially because the Bolt EV's battery provides almost no buffer on the upper end, and the buffer does have can be overridden by "burping" the battery.

The Bolt EV's cells should settle on a post-charging voltage of no more than 4.17 V. A 0.02 V variance at that point could cause a lot of damage to cells, and if the Bolt EVs BMS is attempting to balance cells at that point, a 0.02 V variance could create a lot of heat both in the electronics and the pack.
You're misextrapolating though. The voltage variance is 4.15..4.17, not 4.17..4.19.

Also this chemistry is "fine" in charging up to 4.4V, in the sense that there's just a lot more degradation, but it's not going to cause a fire.

I would argue that it's not that there's next to no buffer, there is no buffer. The rest voltage of a fully (and slowly) charged Bolt cell is the rest voltage of a properly charged cell that cuts off at 0.05C.
 

·
Registered
Joined
·
200 Posts
The charger is pumping juice into every cell during charging, because they are in series. There are no relays between cells, allowing them to be bypassed, to prevent this. As long as the charger is on, current is flowing through every cell. The voltage of each "cell" will be capped at the BMS designed limit, however, by bleeding off the excess current from each full cell to its bleeder resistor.

At the end of a 100% charge, using the onboard charger, the charger is putting out less than 2000 watts, at 400 volts, so under 5 amps flowing through every cell. Bleeding off 5 amps x 4.165 volts = 20.8 watts per cell, During balancing, for some length of time, every one of the 96 cells is bleeding off that much power...the whole 2000 watts.

This is easy to see here where module #6, which the BMS sits on, is 9 F warmer than the others, and module #5 which is permanently attached to it is almost as warm.

View attachment 31720 View attachment 31721

However, I agree that they need the BMS to be more sophisticated that that. If there is a failing cell that can't reach or maintain the target voltage, that failure is exhibited as heat in the cell. Sooner or later it will go into thermal runaway.
Does anyone actually have proof of balancing working? I'm fairly convinced, at least based on what I've seen, that it will only trigger if it's 0.03V out from min to max. Mine are 0.025V from min to max and I have never seen balancing.

I've done a number of tests where I've charged to full (L1 and L2, and restarted to top-up), and the voltages don't change or balance either being plugged in, or at any time unplugged up to 24hr later.
 

·
Registered
Joined
·
200 Posts
Interesting indeed. I think this is the first we have seen that doesn't have the low and high cells clumped together in the same spots. It's crazy how we all have a .02V spread though, regardless of which cells we are talking about.
Mine have always had a 0.025V spread, FWIW.
 

·
Registered
Joined
·
200 Posts
Yeah, just a theory. The alternative is, though, that LG's quality control at that point was just so poor that cells coming off the same assembly line couldn't be balanced. I don't see how that's plausible, though, as that would likely require an issue with the chemistry itself. Given that GM can add newer modules to older packs and still align their voltages makes me think that's not right.

Maybe another interesting bit of data to gather... Do we have voltage spreads for Bolt EV owners whose batteries were replaced under recall? Both the entire battery and/or individual modules (for the second round).
When I toured LG's Holland, MI factory, they were producing Bolt cells. I wasn't allowed to take pictures, but they do test every cell coming off the line automatically, before and after aging, and some fail and are removed.

Do we know for a fact that they add new modules to old batteries? I assumed that they simply used old modules from recalled batteries.
 

·
Registered
Joined
·
200 Posts
That case is not exactly proof that the car needs to be charging. IIRC, the car was charged to full or near full and then immediately driven to the destination. The problem could've begun while finishing the charge and then slowly evolved while the car was being driven, and then escalated once it was parked. From that video of the Bolt in Germany going up in flames, it appears that at least the rest of the car's electronics on the 12V system were working for a while after a catastrophic incident.
These sorts of problems (lithium ion fires) are almost always from dendrites forming which pierce the membrane between layers, causing a short. This is not something that just starts and happens within a matter of a few hours. It's a long process.
 

·
Registered
Joined
·
200 Posts
The single most important function of a BMS is to protect the battery. That means disconnecting the battery or discontinuing the charge if the correct voltage can not be achieved. A well designed BMS will not just sit there and allow electrons to be poured into a cells if in a certain time frame and compared to the other cells it does not come up to snuff. This is exactly the opposite of what is suppose to happen. If this is how the BMS was designed, then I say again, god help us because we are driving time bombs.
If GJetson is correct, and this is a passive system then all the cells are charged a little on the high side and then a resistor bleeds the voltage of the highest cells down to the lowest.
Each cell is 55 amp hours and since we know we can get a voltage reading from all 96 of them, each cell voltage can be individually bleed down. Watts = Volts x amps. Trying to bleed 55 amps down by .02 volts should release 1.1 watts of heat. That is 1/60 of a 60 watt light bulb. Let's say half of the cells need to be bleed down. That is still only 53 watts of heat energy.
Most likely the resistors are in the BMS as I saw on indication of them around the pack in the Weber video. 53 watts is not a lot of heat. I may be wrong with some of my math but you get the general idea, dissipating this much heat should not be a problem and if the BMS is pumping juice into the whole pack in a futile effort to bring all the cell into alignment then it is doing the absolute opposite of what a BMS should be doing and is a total engineering disaster.
You're assuming that a desired end-result 0.02V drop would mean that you cause a 0.02V drop and hold there. That would be glacially slow and very inefficient, and rather difficult to create (you'd need variable resistors).

The system works with a transistor shorting 2x20ohm resistors in series to the cell. So if my math is correct, you're looking at about 104mA and 430mW of drain per cell 3-pack. It can hypothetically drain multiple cells at once if needed. Not very fast, but it doesn't need to be.
 

·
Registered
Joined
·
200 Posts
My graph is an average of two samples for each cell. When I sent the CSV file from Torque Pro to Microsoft Ultra Office Spreadsheet, it did exactly as 2019EVLT described. I had to renumber all the columns to get them in the right order too.
Rather than renumber, just do a search & replace to get rid of the text, leaving only the cell numbers. Then you can transpose paste and sort then graph. Makes it easier.
 

·
Registered
Joined
·
4,176 Posts
Do we know for a fact that they add new modules to old batteries? I assumed that they simply used old modules from recalled batteries.
A fellow Bolt EV owner who had the "cold weather" update done on his 2019 was told that one of his modules needed to be replaced before they could do the update and that they replaced it with a "new" module.
 

·
Registered
Joined
·
200 Posts
A fellow Bolt EV owner who had the "cold weather" update done on his 2019 was told that one of his modules needed to be replaced before they could do the update and that they replaced it with a "new" module.
Right, but "new to him" might not be "actually new". ;)

It would seem pretty prudent to replace modules under warranty with similarly used cells. The warranty even explicitly says this:

the high voltage battery may be replaced with either a new or factory refurbished high voltage battery with an energy capacity (kWh storage) level at or within approximately 10% of that of the original battery at the time of warranty repair.
So if they're willing to replace your entire battery with a refurbished one, I don't see why they wouldn't replace a module with a refurbished one as well - especially since they've gotta have a few hundred from the entire pack replacements that have happened.
 

·
Registered
Joined
·
200 Posts
Another interesting tidbit to keep in mind:

The normal charging process is supposed to be at up to 1C (57kW) with a ceiling of 4.20V/cell and cuts off at 0.05C (~3kW). After a ~15 minute rest, the open cell voltage should be between 4.15 and 4.17V/cell, depending on the specific datasheet being referenced. If we are fast charging, this is roughly what happens - the DCFC stops around 7A to the battery or about 2.8kW.

But our system doesn't do that with L1 or L2. We charge at 4.17V/cell and cut off around 1kW, and wind up with about the same net result - a rest voltage close to 4.17V.

I doubt that this is a problem, as if it was we'd have more than just 5 fires... but it's interesting to note.
 

·
Registered
Joined
·
6,256 Posts
Does anyone actually have proof of balancing working? I'm fairly convinced, at least based on what I've seen, that it will only trigger if it's 0.03V out from min to max. Mine are 0.025V from min to max and I have never seen balancing.

I've done a number of tests where I've charged to full (L1 and L2, and restarted to top-up), and the voltages don't change or balance either being plugged in, or at any time unplugged up to 24hr later.
Thanks for joining this discussion. Our Bolt's battery pack, only shows voltage spreads less than 0.025 volts in the center of a charge. At either end it is always 0.025 volts.

The views on here range all over the map. Some have argued their packs are balancing at hilltop, or 90% for later Bolts. Some believe their packs balance at any state of charge. I have always assumed it only balances at 100%, full, but have always been mystified by the terrible job it appears to do. We rarely charge to 100%, and then never leave it sit for more than an hour, so if it is attempting to balance we may never see the final result.

You may well be right, and none of the packs without failing cells ever balance, as they don't reach 0.030 volts spread.

My only argument/evidence for balancing happening is based on the high temperature of module #6, which has the BMS sitting on it. and the temperature of module # 5 which is often near, or at, the same temperature, since it is permanently attached to #6. But this could simply be due to the electronics in the BMS generating heat from monitoring, and not actually balancing, as I see similar temperature differences at hilltop after preconditioning the cabin, when I don't believe it is balancing.

11-15-20 leaving 2.jpg start at 8.08 am.jpg

If you are correct, and none of the properly functioning Bolts ever need balancing, maybe, instead of trying to balance packs with a 0.03 volts spread, they should alert owners to take it to the dealer immediately.
 

·
Registered
Joined
·
890 Posts
If balancing doesn't occur until .03V, then maybe what we are seeing with the voltage distribution is a result of an intentional manufacturing process. What if they decided that the cells within each pack must be no more than .02V out, and when arranging the cells in the pack, they put the ones that are -.01V in certain places in the pack and the ones that are +.01V in other specific places... and that's why we are seeing a similar distribution of high/low cells?

And what if this distribution-of-cells is what they got wrong at the Korean plant?

Mike
 
121 - 140 of 175 Posts
Top