PDA

View Full Version : Why you need cell level monitoring



podolefsky
25 February 2014, 1117
There’s been some discussion of different balancing methods over here (http://www.elmoto.net/showthread.php?1347-Cell-balancing-top-bottom-static-and-dynamic). One thing that came up was pack vs cell level monitoring, so I wanted to start a new thread on that. Some people believe you can just use pack voltage. I’m going to try and explain why that’s a risky proposition.

To be sure, there are lots of reasons cell level monitoring is a good idea. I'm avoiding drift or self-discharge of a healthy pack (which seem to be controversial) and focusing on the case of a catastrophic failure.

Part of this is because of the recent discussion, and part is because I recently lost a cell in my pack – I believe due to an internal short that caused it to completely self-discharge overnight. Some folks might say that my BMS killed the cell. It did not. My BMS told me I have a bad cell and is probably saving me from losing the entire pack. (Full disclosure, it also warned me while I was killing it, but I ignored it so I could get home. Oops.)

Just to be clear, nothing in my argument depends on whether you top or bottom balance. I’m going to explain using top balancing, just to remove that controversy, then I’ll explain why you have basically the same problem with bottom balancing.

podolefsky
25 February 2014, 1118
This is a little involved, so bear with me.

Suppose you have a pack of 24 cells in series. The cells are top balanced. Say they’re LiFePO4, so max charge voltage is 3.65 x 24 = 87.6V and min voltage is 2.8 x 24 = 67.2V.

If all the cells are healthy and the pack stays balanced, then you can charge to 87.6 and every cell will be at 3.65V. Let’s say you know that at 70V, one cell hits 2.8V and the rest are higher. So, 70V cutoff protects that one low cell and all the others.

Now suppose one of your cells starts to go bad by developing an internal short. The way you model this is a resistor in parallel with the cell. As it charges, some of the current shunts through the resistor, so that cell charges less, while the other cells charge more. At the end, your pack voltage will read 87.6, but that bad cell might only be at 3V while the others hit 3.68V. They’re getting overcharged.

With an internal short, the cell usually self-discharges very quickly. The next day it might be empty. You would see this as the pack rest voltage being about 3.3V lower than normal. Hopefully you notice, because what happens next is bad.

When you charge again that bad cell isn’t just starting low, it’s starting at zero and charging slowly. So, the other cells get overcharged even more. When you discharge again, that bad cell starts low and it goes way below the 2.8V limit you thought you had. It gets driven even harder, making the short worse, making the other cells go even higher on the next charge.

If you keep that up, eventually that cell will be a dead short, taken out of the system. When you charge again, now you are charging 23 cells to 87.6V. Since they’re top balanced, they’ll all hit 3.81V. They’re now getting very overcharged. It just so happens that overcharging can lead to internal shorts, meaning you could cause more of the same problem. All of this happens while you continue to see 87.6V high and 70V low.

What if you bottom balance? Consider the same process, leading to the worst case - you only have 23 cells working and one shorted. Your low voltage cutoff is 2.8V x 24 = 67.2V. Say that when the pack was good, you knew one cell hit 3.65V when the pack is at 85V. That’s your max charge voltage. Now you have 23 cells, so they all charge higher, and that high cell gets pushed up. Since it’s at the knee of the charge curve, its voltage goes up much faster than the other cells which haven’t gotten there yet. Now you’re killing that one high cell very quickly. If you kill it off, you make the situation even worse, because now you have two cells out of the system.

Basically whether you top or bottom balance, if you only pay attention to pack voltage, and you only really pay attention at the top and bottom, you run the risk of losing cells. The reason is that the cells could be doing something unexpected, and it’s masked by the fact that the pack will keep hitting your high and low voltage limits. It’s fine if everything behaves, but if you have one bad cell you can have a cascade that takes down the whole pack.

If you have cell level monitoring, it will stop the charge before you drive the cells over the edge. Without it, you’re flying blind.

chipper6
25 February 2014, 1140
If you check the voltage after the charge is complete with one cell dead, it will not be at the expected voltage. So you wouldn't have a cascading loss of cells unless you ignore your gauges, same that happened when you ignored your bms.

Sent from my HTC One using Tapatalk

frodus
25 February 2014, 1157
You can't really ignore your monitoring if it's wired correctly... It would fault and shut down the pack before it cascaded. Orion, Elithion. Reap, etc are wired this way. As soon as that once cell acts weird it would tell you and then shut down either charger or controller until you fix the issue. That's the real point of a BMS... Not balancing. It's there to look for overvoltage, undervoltage and over temp and to shut the pack off to protect it. People lose cells with a bms because they're bad to begin with or they don't allow it to shut off the pack.

So... If you have no cell monitoring and you check your pack voltage after the charge is complete you'll never notice because the cells are all overcharged while that one is low or shorted. Pack voltage looks the same because the charger went to its programmed voltage.

I'm not for or against a balancing BMS... I mean I like a balanced pack... Who doesn't.

I'm all for a monitoring system that protects the pack when **** hits the fan. Using pack level monitoring is only looking at big picture and relies on the rider to judge if there's an error or not. Are you sitting there while the bike is charging... Looking at voltages? No.

Even a simple monitor like the cell-log8's is a great way to shut things down when a cell goes wonky. More advanced systems give you more flexibility.... But the function is the same.... To protect the pack from damage. If a cell goes tits up, it ensures that cell doesn't take down the entire pack.

podolefsky
25 February 2014, 1242
Yeah, I really should have paid attention to my cell monitors (let's not even call it a BMS). I actually don't have it hooked up to take control because I want the freedom to push it a bit - which I know is bad practice. I'm learning from that mistake.

There are ways that pack voltage will let you know something is wrong, but it's really iffy. Like Travis said, if the 23 remaining cells are overcharged, then you won't see a full cell voltage difference in pack voltage. The other cells will sit higher and tend to mask it. Also, in my experience, resting pack voltage varies from charge to charge. If I worried every time is was slightly different I'd be investigating false positives all the time.

I'm not trying to convince anyone they HAVE TO do this. Whatever you decide is up to you - I'm just putting the info out there so it's a well informed decision. My choice is to have big red flashing lights letting me know something is wrong the moment it happens, and a system that stops me from killing the pack if I'm not completely paying attention...which is a lot, since I ride my bike almost every day (weather permitting). I come home from work, plug it in, get up in the morning and ride it. The initial work I put in setting up the system means every ride is that much simpler.

Warren
25 February 2014, 1256
podolefsky,

(Full disclosure, it also warned me while I was killing it, but I ignored it so I could get home. Oops.)

Please explain how it warned you. What did you see?

podolefsky
25 February 2014, 1322
I have a screen built into my tank that reads out various data. There is text that gives you warnings, for this one it says in red text "Alarm: cell voltage is too low".

chipper6
25 February 2014, 1323
I think I will be running a test this week to fill in my thoughts. I don't like making comments I can not back up. If conclusive I'll provide a video with my thoughts for the discussion.

I feel the title of the post makes it seam you do want to convince people to cell level monitor. I would also like to throw up the idea of partial pack level monitoring (e.g. groups of 6 cells in the 24 cell pack) as another option.

Sent from my HTC One using Tapatalk

Nuts & Volts
25 February 2014, 1324
If you check the voltage after the charge is complete with one cell dead, it will not be at the expected voltage. So you wouldn't have a cascading loss of cells unless you ignore your gauges, same that happened when you ignored your bms.

Sent from my HTC One using Tapatalk

False, your bulk charger is set for a max pack voltage. Lets say 24V with 6cells would be 4V per cell. One dead cell then so you have 24V with 5cells which gives 4.8V per cell. They will probably fail and go into thermal runaway. You have no killed all 6 cells.

With more cells in series the problem is mitigated, but still there.

For reference I do not run a BMS on a 26s4p EIG pack and I top balance the packs every 20 cycles or so.


I think I will be running a test this week to fill in my thoughts. I don't like making comments I can not back up. If conclusive I'll provide a video with my thoughts for the discussion.

I feel the title of the post makes it seam you do want to convince people to cell level monitor. I would also like to throw up the idea of partial pack level monitoring (e.g. groups of 6 cells in the 24 cell pack) as another option.

Sent from my HTC One using Tapatalk

Honda actually did the latter option of monitoring voltage of every 6 cell group (1.2V NiMH cells) in the first generation Insight (I currently own one). I kind of like this idea, but still has the same problem as my above comment

chipper6
25 February 2014, 1331
Thanks for pointing out the small pack problem I was responding to the example he presented of 24s pack. This problem is even more so mitigated for car size packs.

Sent from my HTC One using Tapatalk

podolefsky
25 February 2014, 1334
I think I will be running a test this week to fill in my thoughts. I don't like making comments I can not back up. If conclusive I'll provide a video with my thoughts for the discussion.

I feel the title of the post makes it seam you do want to convince people to cell level monitor. I would also like to throw up the idea of partial pack level monitoring (e.g. groups of 6 cells in the 24 cell pack) as another option.

Sent from my HTC One using Tapatalk


Cool, tests are always a great idea, love to see what you come up with.

I would say yes, I am very strongly suggesting people cell level monitor, but like I said it's up to you. If it's your own bike that you're willing to maintain judiciously, you can get away with pack level. Just know the risks.

frodus
25 February 2014, 1349
Exactly. The more granular the voltage monitoring, the better you can protect the pack. Monitoring smaller subpacks will give you a better view than measuring at the pack level only.... And a cell level monitored pack would give you a better view than measuring at the subpack level.

If you monitor at the subpack level, at least only a subpack can go wonky. If you wire such that it disconnects a load... Then you save the other subpacks from damage.

And yes.... All of these issues with monitoring are far less visible when using lower voltage packs. For car size packs and high voltage packs, a little unbalance can mean a lot more trouble.

Nuts & Volts
25 February 2014, 1513
Thanks for pointing out the small pack problem I was responding to the example he presented of 24s pack. This problem is even more so mitigated for car size packs.

Sent from my HTC One using Tapatalk

Yea I realize my example wasn't that great, was trying to just do quick math. On my 26s pack if I have a bulk charger go to (4.15V*26) 107.9V then I will get over 4.3V per cell if 1 cell dies. At 96s it would be 4.19V which is safe, but not healthy and now I have ~4x as many failure points.

Anyways I think we get the idea

Warren
25 February 2014, 1623
podolefsky,

I have a screen built into my tank that reads out various data. There is text that gives you warnings, for this one it says in red text "Alarm: cell voltage is too low".

Was this with the pack was under load? About how many C? Did it give you a voltage for that cell?

podolefsky
25 February 2014, 1641
podolefsky,

I have a screen built into my tank that reads out various data. There is text that gives you warnings, for this one it says in red text "Alarm: cell voltage is too low".

Was this with the pack was under load? About how many C? Did it give you a voltage for that cell?


Yes. I was trying to nurse it home through city streets. It's hard to remember, but I'm sure I was pulling less than C/2 (30A) once I got the alarm. For the last few hundred feet I was staying below C/6 (10A), creeping along at a walking pace.

It does give me individual cell voltage. I hit a button and it flips to a screen with all the cell voltages. All the cells were low, this one especially. I don't remember exactly how low, but for sure <2.0V.

When I went to charge it again, that cell came up much more slowly than the others. I stopped the charge after a while and let it sit overnight. Now that cell is at 0.0V, all others are about 3.3V. Slow charge and fast self-discharge are both signs of an internal short.

podolefsky
26 February 2014, 0925
If you check the voltage after the charge is complete with one cell dead, it will not be at the expected voltage. So you wouldn't have a cascading loss of cells unless you ignore your gauges, same that happened when you ignored your bms.

This is a good point, and it's had me thinking...and doing some testing, just so I have some data to back up what I'm saying.

It really depends on the cell type. LiFePO4 tend to return to 3.3-3.4V regardless of how high they're charged. Higher charge, they'll sit a bit higher, but not much. LiCo (like hobby lipo) and NMC (like EIG) tend to stay right where you charged them.

Here's the data. I took 2 each of LiFePO4, LiCo, and NMC. Charged one a little higher than the other, exactly enough so that a pack made of 24 of the low cell would charge to the same voltage as 23 of the high cell. Then I let them sit for 10 hours and measured the voltage.

Pack 1 is 24 x cell 1, pack 2 is 23 x cell 2. Tells you would you would see if you had a dead cell in pack 2.

________ Time __ Cell 1 __ Cell 2 _ Pack 1 __ Pack 2 __ Pack Delta
LiFePO4 __ 0 ____ 3.65 ___ 3.81 __ 87.60 ___ 87.60 _____ 0
______ __ 10hr __ 3.36 ___ 3.38 __ 80.64 ___ 77.74 ____ 2.9

LiCo _____ 0 ____ 4.00 ___ 4.17 ___ 96.0 ____ 95.9 ____ 0.1
________ 10hr __ 4.00 ___ 4.16 ___ 96.0 ____ 95.7 ____ 0.3

NMC _____ 0 ____ 4.00 ___ 4.17 ___ 96.0 ____ 95.9 ____ 0.1
________ 10hr __ 4.00 ___ 4.16 ___ 96.0 ____ 95.7 ____ 0.3


The 0.01V difference is within the accuracy of my voltmeter, so basically LiCo and NMC don't drop at all. LiFe does. If you were looking at a LiFe pack, you might notice 2.9V. If you had the other two, you'd never see a difference.

chipper6
26 February 2014, 0936
Beat me to it, good to know about the other chemistries. Nice work

Sent from my HTC One using Tapatalk

podolefsky
26 February 2014, 1024
I'd still be interested in what you come up with. Not all LiFePO4 are the same. I did this test with Headways. My GBS cells have a slightly different discharge curve.

chipper6
26 February 2014, 1037
My question was going to be, how long do I have to wait for to see the difference? 1 min, 5 min, 10 min? Basicly fill in the decay curve until it showed. If you unplug the charger and hop on, you may not catch it, but if you wait x time you would (for life). I may have this data all ready save, will check tonight

Sent from my HTC One using Tapatalk

podolefsky
26 February 2014, 1047
That's a good question. Great to see if you have it.

frodus
26 February 2014, 1051
Why not switch a dc-dc converter or some other small load on first.... It would bleed that surface charge a bit.

podolefsky
26 February 2014, 1055
Why not switch a dc-dc converter or some other small load on first.... It would bleed that surface charge a bit.

Do you mean in the tests I ran? Sure, that would speed things up. I wanted to simulate, as closely as possible, charging a pack and then letting it sit with no load.

frodus
26 February 2014, 1111
No... I mean for general operation without bms. After charging... Just load the pack a bit. The voltage will fall down to normal levels and you'd see a lower than normal voltage.

podolefsky
26 February 2014, 1145
Ah, I see. Yeah, that should work. Just need to figure out how much load and for how long before you know you're at normal level without taking out too much.

That's part of the trouble I see with using pack voltage. You have to know how your pack behaves really thoroughly, and pay attention to all kinds of other conditions so you can pick out deviations of a couple volts here or there and decide if it's normal or a problem. You can see those changes from all sorts of things - how long it sat since charge, temperature... It's a big balancing act.

chipper6
26 February 2014, 1203
It's a good point, when you turn the key on, it should take that surface charge off relatively quickly. If you had a bms, it would probably need that dc to dc on so it would be able to see it quickly.


Sent from my HTC One using Tapatalk

chipper6
27 February 2014, 0226
Here is a zoom in on a Charge Discharge Cycle I did last year on some A123 20 A-hr cells.

Data was collected every 10 seconds on a Powerlab8

The cell was charged at 25 amps to 3.6 volts and held until current dropped to 2amps. This is where the chart begins (time = 0).

As you can see the voltage falls nearly 1.5 volts in just over 5 minutes when the discharge cycle began and 25 amps was applied.

This is just from my initial cycling of the cells, I still have to run a cycle to put the scenario together.

5488

podolefsky
27 February 2014, 0938
Nice data, thanks for posting.

Do you know about where those A123 cells sit at rest (like after a day)?

chipper6
27 February 2014, 1017
Probably same as your headways ~3.35 ish I'll check the files I have or just charge the pack I have.

Sent from my HTC One using Tapatalk