OK, so the modules, not counting the piping for the coolant loop or the small edges they rest on the sides, are 26 1/2" long and 11 1/4" wide. With the support edges they're about 12" wide. The coolant loop connectors add another ~2".
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
JRP3
Hyperactive Member
That is correct, it is not possible that a single cell failure will disable the pack.
Thanks for the photos wk057. It is great to see inside packs like this. My colleagues and I have had a closer look at the cell monitoring PCB and made a few notes about what we could see. Apologies if this is repeating anything that has already been said:
The PCB will allows discharging of cells, or more accurately, groups of parallel cells. The charging & discharging will likely be at pack level, and any cells that are over charged can have the extra charge dissipated as heat, preventing the overall pack voltage being dictated by one rouge cell. There is also a secondary system to prevent cells going out of their safe operating region (either voltage or temperature). This is the basic premise behind EV battery management.
The board is controlled by a Silicon Labs MCU C8051F530 which looks to be powered from the monitored HV battery directly.
The host controls the 6 cell Texas Instruments BMS chip BQ76PL536A-Q1, passively balancing the cells. The BMS controls each channels 39.5Ω balance resistors giving 106mA discharging, just under 0.05% per hour for 74 off 2900mA cells. The BMS chip safety outputs look like they are only heading to the MCU.
The BMS ADCs are connected to the cells via inductive filtering. The cell connections are fused, although fusing isn’t consistent across the PCBs.
The host has communication to elsewhere on the vehicle across an isolation barrier that keeps the HV system separate from the LV system with two in and two out channels at 5kV isolation, with the rest of the PCB tracked with an isolation gap of what looks to be, using EN60950, roughly 400V of reinforced insulation.
The PCB testing capability seems reasonably extensive and looks to have been tested once after manufacture, suggesting a first time pass.
The PCB will allows discharging of cells, or more accurately, groups of parallel cells. The charging & discharging will likely be at pack level, and any cells that are over charged can have the extra charge dissipated as heat, preventing the overall pack voltage being dictated by one rouge cell. There is also a secondary system to prevent cells going out of their safe operating region (either voltage or temperature). This is the basic premise behind EV battery management.
The board is controlled by a Silicon Labs MCU C8051F530 which looks to be powered from the monitored HV battery directly.
The host controls the 6 cell Texas Instruments BMS chip BQ76PL536A-Q1, passively balancing the cells. The BMS controls each channels 39.5Ω balance resistors giving 106mA discharging, just under 0.05% per hour for 74 off 2900mA cells. The BMS chip safety outputs look like they are only heading to the MCU.
The BMS ADCs are connected to the cells via inductive filtering. The cell connections are fused, although fusing isn’t consistent across the PCBs.
The host has communication to elsewhere on the vehicle across an isolation barrier that keeps the HV system separate from the LV system with two in and two out channels at 5kV isolation, with the rest of the PCB tracked with an isolation gap of what looks to be, using EN60950, roughly 400V of reinforced insulation.
The PCB testing capability seems reasonably extensive and looks to have been tested once after manufacture, suggesting a first time pass.
Last edited by a moderator:
lolachampcar
Well-Known Member
I like to think of it in terms of hard disk technology where the device works around known acceptable failure rates.
Except I think it's worth pointing out there's a significant difference: A hard disk initially maintains spare unused sectors that it can map in to use when an existing sector fails, thus "working around"" a failure. As far as we know the Tesla pack maintains no such spare capacity, however it can continue working despite a failure.
The reason why this is important is that while a disk will have enough spare sectors to remap over it's expected life thus maintaining original capacity, the Tesla pack will degrade in overall capacity with a cell failure.
By my calculations, a single cell failure results in a 1.35% loss of overall pack capacity, despite there being over 7,000 cells in the pack. (note however, that subsequent cell failures don't necessarily degrade an additional 1.35% unless it's in the same module... and with 96 modules that's much less likely).
Last edited:
I could see the potential point of failure being the dual sided cell level fuses.
As I noted earlier in the thread there are indications that a handful of them were manually repaired at the factory prior to final assembly, meaning there certainly is not 100% success with these connections.
I did remove the protective plastic from a module to get a better look at these and, well, they're pretty weak little wires. A tiny touch moves them. However nothing can touch them inside the pack, so, they're safe from that.
My concern would be over time vibration rattling a weak one loose and causing an intermittent connection.
Not saying this is happening anywhere since I assume Tesla has accounted for this, but if there were a point of failure that would be it... 1 of 14,208 connections failing would not seem statistically improbable.
As I noted earlier in the thread there are indications that a handful of them were manually repaired at the factory prior to final assembly, meaning there certainly is not 100% success with these connections.
I did remove the protective plastic from a module to get a better look at these and, well, they're pretty weak little wires. A tiny touch moves them. However nothing can touch them inside the pack, so, they're safe from that.
My concern would be over time vibration rattling a weak one loose and causing an intermittent connection.
Not saying this is happening anywhere since I assume Tesla has accounted for this, but if there were a point of failure that would be it... 1 of 14,208 connections failing would not seem statistically improbable.
SeminoleFSU
Voluntaryist
I've often wondered how many points of failure exist in the battery pack using so many (thousands) of cells... I know Tesla touts having fewer moving parts, etc and therefore should be more reliable.... but.... 
I do realize many of those points of failures wouldn't mean an outright failure, just a diminished capacity.... but...
Since owners don't get to see the actual diagnostics on the pack to identify failed cells/connections, etc... 1.35% loss by failure of a cell here or there could easily be written of as "normal" degradation, and not a failure inside the pack.... and since the owners have no way to verify or refute- they'll just have to live with whatever Tesla decides.
Am I missing anything here?
- - - Updated - - -
Thinking out loud some more.... I wonder if Tesla's pack degradation prediction graphs, etc assume individual cell failures at certain points, or just assume all cells continue to operate normally, but just degrade?
I do realize many of those points of failures wouldn't mean an outright failure, just a diminished capacity.... but...
Since owners don't get to see the actual diagnostics on the pack to identify failed cells/connections, etc... 1.35% loss by failure of a cell here or there could easily be written of as "normal" degradation, and not a failure inside the pack.... and since the owners have no way to verify or refute- they'll just have to live with whatever Tesla decides.
Am I missing anything here?
- - - Updated - - -
Thinking out loud some more.... I wonder if Tesla's pack degradation prediction graphs, etc assume individual cell failures at certain points, or just assume all cells continue to operate normally, but just degrade?
I've often wondered how many points of failure exist in the battery pack using so many (thousands) of cells... I know Tesla touts having fewer moving parts, etc and therefore should be more reliable.... but....
I do realize many of those points of failures wouldn't mean an outright failure, just a diminished capacity.... but...
Since owners don't get to see the actual diagnostics on the pack to identify failed cells/connections, etc... 1.35% loss by failure of a cell here or there could easily be written of as "normal" degradation, and not a failure inside the pack.... and since the owners have no way to verify or refute- they'll just have to live with whatever Tesla decides.
Am I missing anything here?
- - - Updated - - -
Thinking out loud some more.... I wonder if Tesla's pack degradation prediction graphs, etc assume individual cell failures at certain points, or just assume all cells continue to operate normally, but just degrade?
I was just getting ready to reply to your email when your edit showed up... I'd bet you are correct. That the overall diminished pack capacity probably allows for a percentage of cell-failure-caused failure as well. I'm not sure if it's more likely that an intercoonect would fail or the cell itself would.
Given that you could lose a cell in each of the 96 modules and still only be degraded "one cell's worth" of capacity (1.35%), then as long as you don't expect to lose too many then I'm assuming it's not going to be a terribly significant factor in the overall degradation.
I'm no statistician, but (all else being equal) isn't the chance of losing 2 cells in the same pack (1/96) ^2 = 0.011% ? (thanks to hcsharp for catching my error)
Last edited:
No. It's .011% (all else being equal). But all else is not equal. Once you lose one cell in a brick then the other cells in that brick are stressed that much harder. The degradation from use becomes exponentially faster in that brick. With time, those other cells and fuses are more likely to fail. They will get hotter, discharged lower, and always have more current flowing through them.
Ah, right you are.. my numerical answer was correct, I forgot to then recognize that percentage account's for two of those decimal points.
But I suspect those fuses are sized for to prevent thermal runaway of a battery in case of a short, massive module failure, etc... not the failure of a cell (or two or three), is such that they are going to be affected by that small of a failure.
SeminoleFSU
Voluntaryist
Ah this makes sense- so I wonder if their models account for that as well? I'm assuming yes- since they probably come up with the data by testing several to dozens of completely built packs, not just the individual components that comprise the pack.... *we hope*
- - - Updated - - -
The .011% number is hard for me to wrap my mind around. Does that mean over the total life of the pack? 8yrs? per year?
This is the chance of losing two cells and having them both be in the same module of one of the 96 individual modules in the pack.
So, although a single cell loss immediately brings the entire pack capacity down by 1.35%... you could lose a cell in all the remaining 95 modules without losing any more capacity. It would take another cell loss in the same module that had already lost one previously to bring down the pack another 1.35%...
So from that standpoint, it would be at any point over the lifetime of the pack.
lolachampcar
Well-Known Member
SeminoleFSU
Voluntaryist
glhs272
Unnamed plug faced villian
What could be much worse than a cell short is a bad cell that has excessive internal resistance and won't hold a charge/drains down quickly. This would bring down it's sister cells wired in parallel along with it. Not sure if the BMS would be able to compensate for this. This could possibly lead to complete pack failure if it drained down a whole group of 74 cells in parallel.
Not holding a charge would seem to be the same as a cell that failed open or had a blown fuse. That's essentially infinite resistance.
An internal short, or very low resistance would drag down the rest of the module, no?
Last edited:
glhs272
Unnamed plug faced villian
I incorrectly worded that.
So in one scenario where a cell suddenly failed and became very low resistance or completely internally shorted, the cell would draw a lot of current from the rest of the cells in parallel and would blow the cell level fuse and take the bad cell out of the system, thus self healing more or less.
I am thinking of a cell failure mode where the cell does not sustain voltage over time as it's brothers wired in parallel do. This would be some kind of higher resistance failure in that it is not drawing as much current from it's neighbor cells, thus not blowing the fuse but draining down the pack quicker than the balancing system can compensate. A slower failure. From a pack perspective it would appear that the whole pack is not taking much of a charge anymore and the pack rapidly decays. I guess what I am trying to describe is a cascade failure. I wonder if the pack is susceptible to this?
So in one scenario where a cell suddenly failed and became very low resistance or completely internally shorted, the cell would draw a lot of current from the rest of the cells in parallel and would blow the cell level fuse and take the bad cell out of the system, thus self healing more or less.
I am thinking of a cell failure mode where the cell does not sustain voltage over time as it's brothers wired in parallel do. This would be some kind of higher resistance failure in that it is not drawing as much current from it's neighbor cells, thus not blowing the fuse but draining down the pack quicker than the balancing system can compensate. A slower failure. From a pack perspective it would appear that the whole pack is not taking much of a charge anymore and the pack rapidly decays. I guess what I am trying to describe is a cascade failure. I wonder if the pack is susceptible to this?
Isn't it obvious that it is? This has been discussed previously in the Roadster threads. I think it's the Achilles heel of a Tesla battery pack. Most of our batteries will die that way - with most of the cells still in good shape and a few bricks in bad condition. For 2 years I've been trying to think of a good engineering solution to this but haven't come up with anything affordable or cheap enough.
spaceballs
Member
It's obvious to people who have read into the details of this pack and the roadster pack threads. If I was Tesla the next gen pack step would be to hire custom hardware guys to build custom ASIC IC per for each cell. This IC would take existing cell monitor features (i.e. OV/OC/UV cutoffs) but add in the ability to individually "addressable" each cell (there are many ways to do this) to the ESS/BMS/BMB/whatever. So that if any brick/groups that has accelerated self-drainage it can do an binary search until the problem cell is found and disconnect it from the group (or some other technique). In the volumes that Tesla would be needing them these ASICs it would be the winner for cost. Though other side of the coin, now the problem for people like EV DIY's, is then Tesla now has an level of control (at the cell level) where even if you took the pack apart, the custom bricks/groups could still be unusable unless you software hack it, or remove the IC's.
If Tesla does go down this path make sure they keep the wire fuse! As for a prototype I built a custom roadster brick that had each cell had it's own battery monitor IC on it, everything was fine and dandy. The problem was the IC's had a feature there it would disconnect the cell from the group if more than 1.75Amp is drawn. Which at first glance doesn't seem like a problem, but it becomes one when say you have an undersized pack, and a neighbor that didn't understand "keep it under X Amps", they thought it was X miles per hour. 99% of the cells did what was predicted and disconnect them self's from the group, problem was ~2 cells didn't and all the current went through those, lucky my cells had PTC fuses on them so no cells went up in smoke.. but many of the IC didn't survive and 1-2 of them actually became shorts, so fuse is still required!
Last edited:
Sounds like interesting research, for how many cells was this done--do you have a thread or blog describing what you did and how it worked, etc.?
Similar threads
