Pics/Info: Inside the battery pack

[Johan]

The BMS really doesn't seem involved in the actual charging in the Model S and battery. The Model S charger itself, as visible in some photos others have posted, just feeds the battery high voltage to the main connector. The same with supercharging. So, sanely applying a charge voltage to the pack should charge it fine.

So basically as long as you can control the voltage being fed you can make sure the battery doesn't overcharge. As for making you own "crude" BMS I guess you could monitor the voltage of the modules individually, since they each have their own wiring, and perhaps instead of having a system that "bleeds off" modules to balance all the time instead you could manually,a couple of times per year (?), charge the pack up, then bleed off module per module as required to balance the pack, using something that loads/draws appropriate current? My guess is this would be good enough for stationary use, since you won't be neither supercharging nor drawing hundreds of Amps? Also perhaps active cooling will be unnescessary for your intended use and climate?

 

[wk057]

Really great pictures and insights. I like that you compare it to a tank gives me a good feeling driving a MS.

One little thing to note: The serial number you blurred out in the images is also included in the data matrix code next to it. So you might want to blur that as well. You also have more serial numbers and data matrix codes visible on parts of other photos - just in case you want to anonymize the source of the pack here, you could consider blurring these as well.

Again, great job and please keep us updated. This is a really interesting project! :smile:

Yeah, I actually don't even know why I bothered blurring the serial number in the first place... so no worries. Thanks

Thank you very much not just for the photos, but for being the vanguard for the battery-bank-from-Tesla crowd. I am very jealous: my own battery bank consists of 9,000 lbs of Absolyte IIP Pb-Ca absorbed glass mat, 1150Ah at 48V. That's 16 550-lb 6V batteries (3*nominal2V each cell). Now, each one of those was heavy, esp. hoisting them on top of each other (2 columns of 8 each. Are you planning to have your system neatly up against a wall? That would be fantastic.

But, although my system is better* than yours in one important way...I'd much rather be able to have yours. Far more elegant, far more compact - even if/when you get a second skateboard to mangle.

*Better: these Absolytes have extraordinary keeping characteristics in cold weather. That is, -40º and lower. Tesla's system, I have it on unimpeachable authority, are not appropriate for environments far less frigid than that. My bank has endured many days at temps ranging from -30ºF to -50ºF (their first season, seven years ago), and since then I keep them for weeks at a time at temps from 0ºF to about -15ºF. This system is what AT&T and other telecom companies use in remote Alaskan mountaintops; in my more controlled environment it should be good for another 50 years.

But back to your system: it's not clear to me what part the coolant bath will play for you. But perhaps my questions should be on your other thread?

I plan to keep the pack in a climate controlled environment (room temp). So, shouldn't be much of an issue.
But yeah, I'll try to keep this about the pack, and my thread about my project details.

Too bad about the pack not wanting to accept a charge as-is... I suspect unless the BMS is alive to allow it the contractors might not close?

It's not that it doesn't want to accept a charge... I'm sure it will just fine, I just have to get something capable of doing so. The equipment I have currently isn't capable and the appropriate equipment is lagging behind the pack on delivery...

So that's an interesting cell layout in each module. There are four groups of cells in each module, with each group having a common battery contact plate visible on the top. The two larger quadrants each have 148 cells, and the two smaller 74 cells. Can you tell if the batteries in the larger groups are oriented with half the cells anode-up, and half cathode-up? That's the only way I can seemingly make the numbers work to come up with a nominal pack voltage of ~390V.

With (rouugh numbers) a cell voltage of ~4v, they cant be series groups of 74 or 148 cells per module, as the voltages would be either too high or too low (not to mention that the common battery plates dictate there is a parallel connection in each group).
Thus, it appears that the 16 modules in the entire pack must be series connected. That dictates each cell module must be ~390/16=24V.

The only way to get 24 volts out of each module with 4v cells would be to have 6 groups of parallel cell sets connected in series. But there are only 4 visible groups.

Given that you would not want unequal groups (so as to keep current capacity even), I'm guessing that each of the larger visible groups is actually two groups of 74 cells, using that common top plate to connect them in series. That means half the cells in those larger groups have to be physically inverted.

It's hard to see clearly through the plastic top, but it appears that some of the battery contact areas slightly triangular (cathode?), whereas others are circular (anode?). The fact that there are "+" signs stamped in to each plate made me wonder if that implied cathode connections only...

If so, then I'd guess that those two larger groups have two smaller contact plates on the bottom of each of those larger groups. That allows 6 cells sets in each module at ~24v@230A, which is the neighborhood of the pack rating.

Interesting...

There are six groups of cells in series with each group having all cells in parallel. They are inverted as needed to facilitate this. I did not count them exactly (yet).

Each module would come to a nominal voltage of 22.2V, and a fully charged voltage of 25.2V based on 3.7V per cell, which seems to be correct from what I've gathered thus far. All modules are connected in series, and through a main fuse. This yields a nominal voltage of 355.2V and a fully charged voltage of 403.2V.

The pack appears to be able to be discharged at around 900A on a P85 for short periods...

There will be some way for the BMS to bleed off the cells as needed. This will be done continuously during charging, as part of the pack balancing procedure. As the cells are probably very well matched, they may not need to bleed off much energy, so the load dump resistors may be quite small. Is there any chance that if you take the pack to bits, you could get a photo of the BMS board for us? That would be very interesting.

I haven't been able to figure out if it is able to do this or not... I will try to examine some of the BMS boards a little more closely later, however.




Thanks everyone for your comments and encouragement.

 

[scaesare]

Y
There are six groups of cells in series with each group having all cells in parallel. They are inverted as needed to facilitate this. I did not count them exactly (yet).

That's what I had been wondering... appreciate the confirmation.

Good stuff... thanks again for sharing...

 

[hockeythug]

What's your background? Are you a electrical engineer?

 

[djp]

Each module would come to a nominal voltage of 22.2V, and a fully charged voltage of 25.2V based on 3.7V per cell, which seems to be correct from what I've gathered thus far.

I think the fully charged resting voltage is closer to 4.1V. This diagnostic screen for the BMS has the cells at 3.957V.

View attachment 42080

Active cooling doesn't kick in until 55C, so you should be fine without a temperature management system at the rate you're charging/discharging.

 

[stopcrazypp]

From what I can tell the BMS is only capable of monitoring the modules. They do not seem to have the ability to charge or discharge individual modules or module sections.

This does not make a whole lot of sense to me. Previously people with out of balance packs were able to trigger a balance cycle and get measurable results. So obviously the pack must have some way to balance the modules (either through the main BMS or internally). The simplest way is dumping energy from modules into resistors.

 

[wk057]

What's your background? Are you a electrical engineer?

Not professionally, but I've been a pretty advanced hobbyist in the field for years now.

This does not make a whole lot of sense to me. Previously people with out of balance packs were able to trigger a balance cycle and get measurable results. So obviously the pack must have some way to balance the modules (either through the main BMS or internally). The simplest way is dumping energy from modules into resistors.

It is possible... I haven't had a chance to fully investigate this as of yet, but at a glance I didn't see any resistors or circuitry that would immediately catch as something that would perform this function.

I'm curious where you saw that people are able to tell that they "were able to trigger a balance cycle". From what I understand people just fully charged their packs and discharged them until their rated range updated according to the latest data. This doesn't necessarily imply any active balancing, just that the computer was able to gather data and recalculate the potential range and range of SoC use better after the cycle.

The BMS on the modules also do not appear to have any thermal connectivity to the cooling loop or anything else that would be capable of dissipating any significant amount of heat. Keep in mind that these are ~5.3kWh modules. So each of the six groups in the module is ~0.89kWh. Even a 0.1V difference could mean quite a bit of power to be dissipated before making a change. From what I can see so far there is nothing thermally capable of sinking that much heat in any reasonable time frame on the BMS modules.

Also, the wiring used on the modules is very small gauge (maybe 22AWG)... and low voltage (single cell type voltages 3-4V). I doubt more than a a couple of amps (maybe 10-15W at these voltages) could safely be pushed through them for any amount of time. Just 1% out of balance would mean sinking about 9Wh of energy somewhere... or lighting up the small BMS lead to the module's section at a couple of amps for an hour or more... those leads can't be larger than maybe 22 AWG (something like 1A) so I can't see Tesla thinking this would be a good idea.

So, just from a physical standpoint, I don't see how active balancing could be done by the BMS here in any useful form. Again, maybe it is possible and haven't discovered it yet, but, so far I don't see it.

 

[Johan]

Not professionally, but I've been a pretty advanced hobbyist in the field for years now.



It is possible... I haven't had a chance to fully investigate this as of yet, but at a glance I didn't see any resistors or circuitry that would immediately catch as something that would perform this function.

I'm curious where you saw that people are able to tell that they "were able to trigger a balance cycle". From what I understand people just fully charged their packs and discharged them until their rated range updated according to the latest data. This doesn't necessarily imply any active balancing, just that the computer was able to gather data and recalculate the potential range and range of SoC use better after the cycle.

The BMS on the modules also do not appear to have any thermal connectivity to the cooling loop or anything else that would be capable of dissipating any significant amount of heat. Keep in mind that these are ~5.3kWh modules. So each of the six groups in the module is ~0.89kWh. Even a 0.1V difference could mean quite a bit of power to be dissipated before making a change. From what I can see so far there is nothing thermally capable of sinking that much heat in any reasonable time frame on the BMS modules.

Also, the wiring used on the modules is very small gauge (maybe 22AWG)... and low voltage (single cell type voltages 3-4V). I doubt more than a a couple of amps (maybe 10-15W at these voltages) could safely be pushed through them for any amount of time. Just 1% out of balance would mean sinking about 9Wh of energy somewhere... or lighting up the small BMS lead to the module's section at a couple of amps for an hour or more... those leads can't be larger than maybe 22 AWG (something like 1A) so I can't see Tesla thinking this would be a good idea.

So, just from a physical standpoint, I don't see how active balancing could be done by the BMS here in any useful form. Again, maybe it is possible and haven't discovered it yet, but, so far I don't see it.

This is very interesting indeed. The very basic question arises: how does the pack balance? We must assume it does balance, if it was just the BMS doing a better calculation I really do believe pack longevity would be much poorer than what real world reports are saying. Fundamentally the pack can balance either through selective charging or selective discharging of modules or individual cells. There is nothing in what you have shown that would suggest that this is possible on the individual cell level, but you also seem to think it impossible on the module level???

From my understanding there is nothing in the physical/chemical properties of a battery composed of many cells that would make it "autobalance"??? For example that as you are applying charging voltage the cells with relatively lower voltage would somehow "take" more of the electrical energy than those with higher voltage because of a larger voltage "gradient" (with regards to the constant external charging voltage being applied). If this was the case unbalacing wouldn't arise, right?

 

[arg]

I'm curious where you saw that people are able to tell that they "were able to trigger a balance cycle". From what I understand people just fully charged their packs and discharged them until their rated range updated according to the latest data. This doesn't necessarily imply any active balancing, just that the computer was able to gather data and recalculate the potential range and range of SoC use better after the cycle.

The main evidence of active balancing is the behaviour at the end of a charge where people have reported it sitting for many minutes at 99% done with not much obviously going on.

Also, the wiring used on the modules is very small gauge (maybe 22AWG)... and low voltage (single cell type voltages 3-4V). I doubt more than a a couple of amps (maybe 10-15W at these voltages) could safely be pushed through them for any amount of time. Just 1% out of balance would mean sinking about 9Wh of energy somewhere... or lighting up the small BMS lead to the module's section at a couple of amps for an hour or more... those leads can't be larger than maybe 22 AWG (something like 1A) so I can't see Tesla thinking this would be a good idea.

Balancing does appear to be very slow (particularly for packs that have got measurably out of balance), so these sort of numbers are not impossible. In designs I have used (for much smaller packs, but using 18650 cells and per-datasheet values from the charge controller chips), the balancing was at about 100mA per cell. To balance at an equivalent rate, that would mean 3 or 4 amps for the blocks of paralleled cells, though it wouldn't be unreasonable for Tesla to do the balancing more slowly.

 

[FLDarren]

wk057, this is one of the coolest things I've seen on these forums. I really hope you find a way to make it all work. I'll be following your work. Very very cool.

 

[wk057]

The main evidence of active balancing is the behaviour at the end of a charge where people have reported it sitting for many minutes at 99% done with not much obviously going on.
Balancing does appear to be very slow (particularly for packs that have got measurably out of balance), so these sort of numbers are not impossible. In designs I have used (for much smaller packs, but using 18650 cells and per-datasheet values from the charge controller chips), the balancing was at about 100mA per cell. To balance at an equivalent rate, that would mean 3 or 4 amps for the blocks of paralleled cells, though it wouldn't be unreasonable for Tesla to do the balancing more slowly.

I'm not sure the power taper curve is necessarily evidence of balancing... just a slow reduction in the amount of power given to/taken by the pack under charge approaching an asymptote. Again, I could be wrong.

I didn't even see anything even the size of a 1W resistor, let alone 10W+ on the BMS board. Keep in mind I have not removed them and can so far only see what is easily visible from the top and on the side of the oddball module.

- - - Updated - - -

Come to think of it, the reduction in rated range on some packs could be attributed to slowly imparted imbalance, since the pack can only charge until one section reaches max voltage, and discharge until one section reaches low voltage. This is perhaps evidence of the lack of self/auto balancing... (combined with actual cell degradation)

Again, definitely could be wrong. I may examine the BMS in more detail soon.

 

[apacheguy]

Come to think of it, the reduction in rated range on some packs could be attributed to slowly imparted imbalance, since the pack can only charge until one section reaches max voltage, and discharge until one section reaches low voltage. This is perhaps evidence of the lack of self/auto balancing... (combined with actual cell degradation)

That would be very interesting, if true. I find it hard to believe that "the most advanced BMS" does not auto balance the cells. That would imply accelerated degradation as certain cells fell out of balance and cycled deeper into lower voltage territory than the rest of the cells that are at higher voltage.

 

[wk057]

That would be very interesting, if true. I find it hard to believe that "the most advanced BMS" does not auto balance the cells. That would imply accelerated degradation as certain cells fell out of balance and cycled deeper into lower voltage territory than the rest of the cells that are at higher voltage.

Was doing some more research and calculations on the topic and I've come to the conclusion that the ~0.9kWh 3.7V nominal module sub-sections could likely be kept reasonably balanced with something like a ~500mA load, which is much less than I had originally estimated. It would still be a decent amount of waste heat to put somewhere in that sealed pack potentially, but probably doable. If the BMS is advanced enough it could potentially dump this current into an adjacent module sub-section instead of completely wasting it.

I have not examined the actual BMS PCBs more closely yet, but that should give more insight later.

- - - Updated - - -

On a side note, does anyone happen to have a working Model S that isn't under the new owner purchase agreement (ie, not subject to the "no reverse engineering" portion) and would be brave enough to try and figure out how to interface with the BMS? Should be pretty simple actually, after removing the rear seat. Most likely some type of CAN comms...

... or am I going to have to keep an eye out for a decently priced salvage

 

[NuclearPowered]

Even the most basic hobby R/C lithium battery chargers take the battery packs up to max voltage and then bleed off the highest cells. The more advanced ones will actively charge the lower voltage cells. This all happens at the end of the charge when the pack is full. This is consistent with Tesla balancing at 99% rated range. I wouldn't be surprised, since you aren't seeing any resistors on the BMS boards, that the BMS is connecting the highest voltage and lowest voltage modules in parallel to actively balance without wasting energy. That type of setup would be a pretty advanced BMS, maybe even the 'most advanced'.

 

[islandbayy]

I've been using BMS systems for about 10 years that actively balance like that. I have built 26 EV Conversions, and the BMS system that did that was one a guy was "home building" originally for the Zap Cars. Obviously much more simple being Lead acid and only 6-10 drive train batteries in a conversion, however, it was extremely small even for a home brew system. It actively equalized all the batteries. It shuffled power between all batteries until the voltage was equal. And it did it all the time, not just during charging. Doesnt take much to equalize a battery pack. Even 1 amp, or 500mAh, or even 100 mAh can be enough. Remember, it's not like the car is taking fully charged battery, and trying to charge a dead one. It is taking a fully charged battery, and bleeding that power into a slightly lower voltage battery.
Example, Say, Brick "A" is fully charged at 4.2v, and Brick "B" is at 4.15. The difference between the two is very minor. Even at 100mAh, that charge can be shuffled fairly quickly.
And as the software is currently in-accessible to us at this time, it is quite possible that it is ALWAYS trying to balance, or at least always trying while plugged in (Possibly why Tesla says plug in whenever possible when not in use). Charging to 100% can help speed the process up, but its much more simple then people realize. Or simple at least, in terms of where and when to shuffle a charge, the actual design of one of these systems, thats a wee bit beyond my skill level.

 

[NuclearPowered]

I've been using BMS systems for about 10 years that actively balance like that. I have built 26 EV Conversions, and the BMS system that did that was one a guy was "home building" originally for the Zap Cars. Obviously much more simple being Lead acid and only 6-10 drive train batteries in a conversion, however, it was extremely small even for a home brew system. It actively equalized all the batteries. It shuffled power between all batteries until the voltage was equal. And it did it all the time, not just during charging. Doesnt take much to equalize a battery pack. Even 1 amp, or 500mAh, or even 100 mAh can be enough. Remember, it's not like the car is taking fully charged battery, and trying to charge a dead one. It is taking a fully charged battery, and bleeding that power into a slightly lower voltage battery.
Example, Say, Brick "A" is fully charged at 4.2v, and Brick "B" is at 4.15. The difference between the two is very minor. Even at 100mAh, that charge can be shuffled fairly quickly.
And as the software is currently in-accessible to us at this time, it is quite possible that it is ALWAYS trying to balance, or at least always trying while plugged in (Possibly why Tesla says plug in whenever possible when not in use). Charging to 100% can help speed the process up, but its much more simple then people realize. Or simple at least, in terms of where and when to shuffle a charge, the actual design of one of these systems, thats a wee bit beyond my skill level.

It wouldn't surprise me if the BMS could put all modules in parallel simultaneously, limiting current flow where appropriate. It could attempt to keep all modules in balance while otherwise the pack behaves normally. Suffice to say that if a relatively advanced hobby grade charger for R/C models can do balancing by cell to cell energy transfer and a home built BMS was doing the same 10 years ago, Tesla's engineers have implemented something better.

 

[scaesare]

This is an interesting overview of continuous pack balancing:

SWE Li-ion Cell Balancing Overview

It specifically discusses doing so in a multiple cell (or parallel module) configuration.

 

[Johan]

... Also the smallest basic unit if the pack - six cells in serial - would be the only unit that can't be balanced. The BMS, would see these 6 cells as one and they could never balance between themselves. Anything above that level, i.e. within a module and between modules can balance if they are somehow connected in parallell.

When operating it seems the large modules are in serial, but perhaps as suggested by others, the BMS can open and close parallell connections between them for balancing?

 

[Teo]

On a side note, does anyone happen to have a working Model S that isn't under the new owner purchase agreement (ie, not subject to the "no reverse engineering" portion) and would be brave enough to try and figure out how to interface with the BMS?

Otmar has a working salvage Model S. Here is his topic:
Electric of a VW Westfalia using a Model S platform

Here is a contact form on his website.
About | Stretchla Blog

 

[omarsultan]

Very cool thread - thanks for bringing us along for the ride.