Battery pack/module replacement (out of main)

[wk057]

Who are you and what have you done with our Karen?

Our Karen would know that all 4 modules are in series, the only parallel wiring is internal to the modules for the individual bricks that are inside the module.

As far as replacing an individual model @wk057 says that it is almost impossible, at least with the S&X packs, because they would have to have almost the identical capacity as the other modules or things get out of balance too quickly and cause problems.

Even worse with a 3 pack, since you have dozens of groups that would go out of balance instead of just 6.

 

[JRP3]

Tesla specifically designed the pack to be modular so that one can just replace individual modules.

They were designed to be easy to configure different pack sizes and for some isolation. If replacing bad modules were important why did Tesla go from smaller modules in the S/X to 4 large modules in the 3? As far as ease of replacement, at least with the S/X packs, it requires destructive removal of the top cover. You might find after market companies offering module replacement services but I'd bet you'll never see that from Tesla. They'd rather just sell you a new pack, or a new car.

Larger sure, but does leakage get that much worse with age?

Let's look at it with an extreme example. Make a "pack" with two cells, one old with 5ah capacity and one with 10ah capacity. The smaller cell will always be operating at a higher relative C rate, (a 2:1 ratio), will have higher heat loads, will discharge more deeply, and since it's older it will have higher internal resistance, exacerbating all the other issues. That's not a setup I'd want to be creating if I were a company who sold new battery packs and cars. As I said above a 3rd party may offer such, maybe with used and more closely matched modules.

 

[KarenRei]

They were designed to be easy to configure different pack sizes and for some isolation. If replacing bad modules were important why did Tesla go from smaller modules in the S/X to 4 large modules in the 3? As far as ease of replacement, at least with the S/X packs, it requires destructive removal of the top cover. You might find after market companies offering module replacement services but I'd bet you'll never see that from Tesla. They'd rather just sell you a new pack, or a new car.

Did a search. Found this. Musk says that at the end of life they'll replace all the modules, but not the pack:

Elon Musk on Twitter

$5-7k to replace all the modules. $1,25k-$1,75k per module (although replacing a single module would be expected to be more expensive than that, due to it bearing all of the labour costs of removing, opening, closing, and reinstalling the pack).

I've seen elsewhere (although I don't remember where) where Tesla and/or Musk talked about replacing individual modules if there's a defect rather than a whole pack. And I also seem to recall people having their car going in for service due to HV battery problems, and the solution was to replace a defective module.

 

[mongo]

Let's look at it with an extreme example. Make a "pack" with two cells, one old with 5ah capacity and one with 10ah capacity. The smaller cell will always be operating at a higher relative C rate, (a 2:1 ratio), will have higher heat loads, will discharge more deeply, and since it's older it will have higher internal resistance, exacerbating all the other issues. That's not a setup I'd want to be creating if I were a company who sold new battery packs and cars. As I said above a 3rd party may offer such, maybe with used and more closely matched modules.

I was talking purely about BMS balancing and not whether mixing modules is a good idea (mostly not due to the valid points you raise, might be worthwhile for gross module failure. I think Elon's price quote

For degredation, yeah replace all the modules($5k-$7k per Elon).

implied a full 4 module replacement and keep the rest of the pack):

Internal series resistance does not impact balance, current in always equals current out and current is what moves the ions around. Internal parallel (or leakage) resistance does imoact balance and is what the BMS has to continuously compensate for (best vs worse group).

If the system limits to a safe C rate for the weakest module, then that is also a non factor in balancing (more relevant to degredation).

In the 5Ah 10Ah case, the BMS could get both modules up to 100% SOC or 4.xx Volts. Then it would cut output when the 5Ah hit the voltage floor. On the next recharge, no additional balancing would be needed as both modules supplied 5Ah.

If you were dealing with lead acid cells where capacity is tightly linked to C rate, then a 5Ah 10Ah setup would be a BMS workout.

 

[JRP3]

Did a search. Found this. Musk says that at the end of life they'll replace all the modules, but not the pack:

If the pack structure and electronics are in good shape then I suppose replacing all the modules makes sense. I don't think it makes sense to mix old and new modules. Once you drop a pack and open it up might as well replace them all.

 

[JRP3]

If you were dealing with lead acid cells where capacity is tightly linked to C rate, then a 5Ah 10Ah setup would be a BMS workout.

Note the capacities at different C rates

 

[mongo]

Note the capacities at different C rates

I see where you are coming from. Varied Ah capacity can indeed cause continuous unbalancing. It appears lithium ion have a Peukert consant range of 1 to 1.09 https://www.researchgate.net/figure...us-lithium-ion-battery-brands_fig12_231169804

I do wonder though how much of that plot's diversion is attributable to effects like series resistance that don't impact balance. (I'm looking at you voltage offset at time 0).

 

[MP3Mike]

$5-7k to replace all the modules. $1,25k-$1,75k per module (although replacing a single module would be expected to be more expensive than that, due to it bearing all of the labour costs of removing, opening, closing, and reinstalling the pack).

I do wonder if his estimate is $5-7k is for the LR Model 3 pack, or is $5k for the SR+ pack and $7k for the LR pack?

 

[KarenRei]

I do wonder if his estimate is $5-7k is for the LR Model 3 pack, or is $5k for the SR+ pack and $7k for the LR pack?

Or it could relate to how far in the future is being discussed. Or the variation on Tesla's estimates for costs. Or really anything.

 

[hacer]

... In the 5Ah 10Ah case, the BMS could get both modules up to 100% SOC or 4.xx Volts. Then it would cut output when the 5Ah hit the voltage floor. On the next recharge, no additional balancing would be needed as both modules supplied 5Ah...

When you first install the module that is possible, but the BMS system will never again allow the 10Ah module to charge to a higher voltage. During the first discharging, the BMS will detect the weak module having lower voltage, then bleed down the strong module voltage to match. When recharging the BMS will stop the charge once the weaker module hits its upper voltage limit which will happen long before the the strong module gets there. So the strong module will also be limited to 5Ah capacity but with a lower voltage. The BMS is probably smart enough not to discharge the weak module to bring it into "balance" because it presumably realizes the modules lower capacity.

So this works out to the total capacity being N times the weakest parallel cell string capacity.

 

[mongo]

When you first install the module that is possible, but the BMS system will never again allow the 10Ah module to charge to a higher voltage. During the first discharging, the BMS will detect the weak module having lower voltage, then bleed down the strong module voltage to match. When recharging the BMS will stop the charge once the weaker module hits its upper voltage limit which will happen long before the the strong module gets there. So the strong module will also be limited to 5Ah capacity but with a lower voltage. The BMS is probably smart enough not to discharge the weak module to bring it into "balance" because it presumably realizes the modules lower capacity.

So this works out to the total capacity being N times the weakest parallel cell string capacity.

Yah no argument from me that the weakest parallel group sets the capacity.
Regarding balancing, I have adding the condition of the BMS targeting one specific SOC/ voltage for the balance point. Ideally in the 90-100% range to get the max pack voltage. That prevents the situation you refer to. No reason to balance at a low SOC.

Some BMS reduce the charge current to the bleed level to allow topping off of all sections (voltage limit shunt style))l.

 

[hacer]

Yah no argument from me that the weakest parallel group sets the capacity.
Regarding balancing, I have adding the condition of the BMS targeting one specific SOC/ voltage for the balance point. Ideally in the 90-100% range to get the max pack voltage. That prevents the situation you refer to. No reason to balance at a low SOC.

Some BMS reduce the charge current to the bleed level to allow topping off of all sections (voltage limit shunt style))l.

My point is that there is no way to charge the higher capacity group once the lower capacity ones reach their voltage limit. If for example, the weaker cells become weaker over time or if the higher capacity cells have greater self-discharge (for example if they were at a higher potential) then over time the higher capacity group will stop charging at a progressively lower voltage because it's in series with cells that must stop charging.

The bleed level is so low that it would require impractically long charging times to raise the SOC of the strong group (if it is much stronger as Karen originally suggested).

There is some advantage to having higher pack voltage (greater top speed, slightly greater peak performance), but there is no means for the BMS to achieve it because inevitably the strong cells will be forced toward the lower voltage where their capacity matches the weak group.

edit: there is a reason to balance at low SOC which is to enable full capacity recharge.

 

[mongo]

My point is that there is no way to charge the higher capacity group once the lower capacity ones reach their voltage limit. If for example, the weaker cells become weaker over time or if the higher capacity cells have greater self-discharge (for example if they were at a higher potential) then over time the higher capacity group will stop charging at a progressively lower voltage because it's in series with cells that must stop charging.

The bleed level is so low that it would require impractically long charging times to raise the SOC of the strong group (if it is much stronger as Karen originally suggested).

There is some advantage to having higher pack voltage (greater top speed, slightly greater peak performance), but there is no means for the BMS to achieve it because inevitably the strong cells will be forced toward the lower voltage where their capacity matches the weak group.

edit: there is a reason to balance at low SOC which is to enable full capacity recharge.

I think we agree, but terminology (edit: or my use of it) is causing confusion.
Yeah, if the leakage rate of the two module types is more than the capacity of the BMS, it is never going to work. At the critical level of leakage == capacity the BMS would be balancing (bleeding) continuously so it would be active/ bleeding at lower SOCs. What I am saying is at that range it should not be trying to get the voltages at those lower SOCs to match, but planning ahead so the voltages match at 100%.

For example, if there were no (or identical) leakage. The BMS would not do any balancing after everything lines up at 100% even though the modules have different voltages at every SOC below that.

If the BMS alway trys to equalize voltages regardless of SOC, then the 10Ah would be using a 5Ah range somewhere between low end of 0% (battery discharged most of the time) and high end of range at 100% (battery fully charged most of the time). The best senario is to get the 100% point lined up so the 10Ah has the highest possible voltage.

 

[JRP3]

There is some advantage to having higher pack voltage (greater top speed, slightly greater peak performance), but there is no means for the BMS to achieve it because inevitably the strong cells will be forced toward the lower voltage where their capacity matches the weak group.

Most BMS use top balancing, when each cell group hits 4.2 volts they start to bleed off charge until all the other cell groups hit 4.2 volts. That way they all get to 100% SOC, (if set to charge to 100%). I think Tesla actually balances at lower SOC's but the idea is the same, all cells get to the same level of SOC near the top, some just have less capacity and are uneven near the bottom.

 

[mongo]

Most BMS use top balancing, when each cell group hits 4.2 volts they start to bleed off charge until all the other cell groups hit 4.2 volts. That way they all get to 100% SOC, (if set to charge to 100%). I think Tesla actually balances at lower SOC's but the idea is the same, all cells get to the same level of SOC near the top, some just have less capacity and are uneven near the bottom.

Yah, that was the original style with voltage clamps that would also signal the charger to cut the charge rate to match the bleed rate.

LT has some nice monitor chips that let you read each group voltage and transfer power to the multi group module instead of just dissipating it. They also have dissipative versions. Either can balance at any point which cuts the peak bleed rate needed and removes the neer for a 'float' type charge rate.

 

[hacer]

Most BMS use top balancing, when each cell group hits 4.2 volts they start to bleed off charge until all the other cell groups hit 4.2 volts. That way they all get to 100% SOC, (if set to charge to 100%). I think Tesla actually balances at lower SOC's but the idea is the same, all cells get to the same level of SOC near the top, some just have less capacity and are uneven near the bottom.

Sorry, these cells are in SERIES with low-resistance heavy connections capable of carrying over 1,000 A. That wired connection simply never goes away. The BMS can only differentially charge series elements in so far as it can clamp the voltage of a group of cells while forwarding some (or all) charge current around them.

Karen's original argument was having a module that had 40% more capacity than the other modules. Let's say the original modules had 16 kWh each when new and had degraded to 14.4 kWh when a new module with 22 kWh capacity was used to replace a bad module. Suppose that this higher capacity module was at 20% SOC when installed (for safety of transportation). So that module had 4.5 kWh when installed. Suppose the non-replaced ones were also at a low 20% SOC for safety during disassembly, so those each have 2.88 kWh SOC. You put it all together and then plug in expecting the BMS to balance this?

The first 46 kWh of charging is easy - all the modules are simply charged in series. At that point you have 3 modules fully charged but the new one still needs another 6 kWh.Tear-down evidence suggests that there is a minimum of 1k Ohm in the bypass path (Tesla Model 3 - NextGen Battery - EVTV Motor Verks) which would limit bypass current to 4.2 mA (or less). It would take about 1.6 years of continuous balancing to bring that module up. Not practical. [6000W-h/(100V * 0.0042A) = 14,285 hours]

If instead we're generous and say that despite evidence to the contrary the BMS can bypass 500 mA around cells as needed then it could be balanced in only 120 hours (~ 5 days) of continuous balancing. Still not practical. If we wanted to do it in say 12 hours of balancing after the 46kWh of charging (still pretty long IMO) then the BMS would need to bypass 5 A. That would mean the bypassing circuits would dissipate 1.5 kW. I see no evidence that the BMS boards can dissipate that. Even if somehow the BMS uses a switch-mode transfer system (per @Artful Dodger's suggestion) it still can't be done because there is no mechanism to do such a transfer across modules! At a minimum the wiring between modules to do this simply doesn't exist.

My conclusion is that should the strong module ever get to a lower SOC (relative to it's own capability) the BMS will never be able to bring it back up.

Edit: of course all of this is predicated on the idea that one module has much higher capacity than the others (The original concept that @KarenRei proposed). The BMS can balance a battery that has modules/parallel cell groups of similar capacity. Unequal capacity is also easily accomodated provided it is due to a different number of series cells (as found in the different modules of the model 3). Note that my math above assumed (for simplicity) 4 identical voltage modules. The conclusion don't change when worked for modules having differing numbers of series cells.

 

[mongo]

My conclusion is that should the strong module ever get to a lower SOC (relative to it's own capability) the BMS will never be able to bring it back up.

Isn't that equally true for identical capacity modules? You're still dealing with the same Ah level of imbalance.

In a module replacement scenario, I'd expect them to precondition the modules to a matching level.

 

[hacer]

Isn't that equally true for identical capacity modules? You're still dealing with the same Ah level of imbalance.

In a module replacement scenario, I'd expect them to precondition the modules to a matching level.

In principal, you are not dealing "with the same level of imbalance". The batteries were designed and manufactured at the same time, with the same materials and operated with the same currents in the same environment so they should, for the most part, degrade the same way. Meaning the capacity imbalances are small and the BMS re-balancing never allows them to have very different SOC.

In the case where one module is markedly different from another, it is a different story. There is no reason to expect that the self-discharge, impedance (thus local temperatures), and other properties would be similar between modules with different materials, different histories and specifically different capacity.

I think that you would want to pre-condition the modules to a common SOC (thus different voltage due to different capacity) upon installation, and the BMS would need to realize one group was markedly stronger than the others and NOT discharge the weaker groups to match the lower voltage of the strong module.

Keeping the SOC at the low end is easy because it can be charged with whatever the other cells are taking because it's in no danger of reaching its upper limit voltage; if for some reason the strong module goes into a lower SOC than the weak modules (unlikely because of it's lower voltage should reduce the self-discharge), the balancing circuit can discharge the weak modules much more easily because they are low capacity and you can do balancing discharge whether or not the car is being driven (in case it takes a long time). That's not true if you are trying to force it to the high end.

 

[JRP3]

Sorry, these cells are in SERIES with low-resistance heavy connections capable of carrying over 1,000 A. That wired connection simply never goes away. The BMS can only differentially charge series elements in so far as it can clamp the voltage of a group of cells while forwarding some (or all) charge current around them.

Not sure what you mean, cells in series can certainly be balanced in relation to each other.

 

[mongo]

Not sure what you mean, cells in series can certainly be balanced in relation to each other.

I think they are pointing out you can only shunt power with the way the cells are set up (and that you can't shut Supercharger level charge current).