amazing balancing after 145k miles

[hacer]

I understand that most Lithium Ion chargers do balancing at the end the of the charge process and they do it by charging up the lower ones to bring them up to the same level as the higher ones. The main difference is that in most applications, the whole balancing controller and most of the BMS is done in the charger so the battery pack itself is can be made very simple and less expensive. In an EV this doesn't apply.

There are good reasons to bleed off higher cells rather than charging lower ones. If you want to charge up the lower ones, you need to be connected to a power source for much longer. The charger would run at a very low (=inefficient) power level for hours. It also means it would only work if the car is connected to power. ANd worst of all, it would only work if the battery is fully charged and then you have to keep it at that level for an extended time. We all know that's not good for a Lithium battery.

Actually there is no requirement for a high state of charge to perform by-pass balancing during charging. The only thing necessary is that the charging current be small, which is always possible to achieve if you just want to balance and aren't really trying to charge the batteries much. Since the car is presumably often charged overnight and usually doesn't need the full night to charge it would be very easy to run a balancing step after charge completion (at any state of charge) using only a trickle charge for balancing. Given that Tesla says the car is happy when plugged in whenever not being driven that hints at such a strategy. But I have no evidence for it.

Bleeding off the higher cells has the advantage that is can be done independent of a power source and independent of any other battery operation. It also means you can do it at a slow rate as long as you want. In other words, it only needs a relatively short time to determine how much difference there is, then it runs independently. The amount of energy that is bled off to balance is insignificant. It's a very small price to pay for a much bigger advantage compared to the alternative.

look here:
Pics/Info: Inside the battery pack

Pics/Info: Inside the battery pack
Pics/Info: Inside the battery pack
Pics/Info: Inside the battery pack
Pics/Info: Inside the battery pack
Pics/Info: Inside the battery pack
Pics/Info: Inside the battery pack

Thank you. Somehow I missed that whole thread of Jason's battery tear down. The photograph of the BMS module is particularly interesting. It shows an array of inductors near the fuses going to each cell tap within the modules. This suggest to me two possible purposes:
(1) They allow a "bleed" type cell current to be unaffected by the transient voltages caused by large power draw or regen during driving which would indicate that the bleeding is indeed done over long periods regardless of what the car is doing.
(2) The BMS doesn't generally "bleed" energy, but instead uses a (low power) switch mode conversion to transfer energy from the weak cells into the stronger ones, at least within the same module. That would be an efficient means to balance the battery when lightly loaded. But it would be difficult to do with significant dynamic loads and doesn't solve inter-module balancing so I think this is less likely.

Either way, there is still the question of why "trigger" this based on a high charge state? The cell voltages are always available and a timer-based bleed program could be re-programmed every single time the car goes to sleep. So why not do that since it would result in the most consistent balancing across all states of charge?

I'd like to ask you to show your cell voltages at some lower states of charge to see how balanced it is then. If those number are also stellar, then for sure Tesla does an amazing job of keeping the cells balanced.

 

[BBone]

looking at the CAN bus data. I use 'Scan My Tesla' for Android.

View attachment 271331

Nice! I didn't know this app existed. Reminds me of LEAFSpy for the LEAF. I'm trying the Demo now on my KeyONE. Will buy if its good.

 

[JasonA-EV]

Balancing is ALWAYS going on. I wish this whole "you need to drop to single digits and then go to 90something+%" just to balance would just stop!

The BMS is always active and working, the only thing going down into single digits and then going to 95+% is doing is degrading your pack more and more

No wonder I have the best 70d capacity around and the best Rav4EV pack around. And yes, I have plenty of mileage on my cars, I just know how to treat batteries the correct way!

 

[David99]

Balancing is ALWAYS going on. I wish this whole "you need to drop to single digits and then go to 90something+%" just to balance would just stop!

The BMS is always active and working, the only thing going down into single digits and then going to 95+% is doing is degrading your pack more and more

I agree, the misinformation about balancing that gets repeated here over and over is not helpful. Doing a full cycle here and there does help though! Read here
A senior tesla executive's comforting answer to concerns re: "loss of range"

 

[ran349]

I agree, the misinformation about balancing that gets repeated here over and over is not helpful. Doing a full cycle here and there does help though! Read here
A senior tesla executive's comforting answer to concerns re: "loss of range"

Sounds like you're contradicting yourself, because you say balancing talk is misinformation but doing a full cycle once and a while helps. Which is it? Also, that link is going on 4 years old, and I would say based on the threads here, range loss ( at least apparent loss) is still a pretty big concern for a lot of people. Do you know of any updated discussion from Tesla on this?

 

[David99]

Sounds like you're contradicting yourself, because you say balancing talk is misinformation but doing a full cycle once and a while helps. Which is it? Also, that link is going on 4 years old, and I would say based on the threads here, range loss ( at least apparent loss) is still a pretty big concern for a lot of people. Do you know of any updated discussion from Tesla on this?

I'm not sure you read the posts I wrote here. I'm not contradicting myself. Balancing is happening all the time. A full charge/discharge cycle is recalibrating the capacity/range estimate. Those are two separate things and have nothing to do with each other.

Based on my observation of my battery over almost 100k miles I found that the pack never goes out of balance enough that it would lose any significant amount of range. Even when only partially charging for a long time. I believe that what Jason wrote is sure correct (balancing is initiated at 93% charge level or higher) but that might not be the only time it is. I have not charged my battery to even 90 for well over two moths and still the cells were very close together. So either they don't drift apart much or balancing is more sophisticated than just being triggered at 93%. Either way, there no reason to charge your car up all the way to trigger balancing. So yes I totally agree that the recommendation to charge your battery all the way up to balance it is nonsense.

But for a whole different reason, namely the calibration of the range estimating, it does make sense to do a full cycle here and there. Yes that email is old, but the fundamental issue remains. You can't directly measure a battery's exact state of charge and capacity at any given point without fully charging and then discharging it. The error will always be there. It really doesn't hurt a battery to do an almost full cycle once every few months. Chances are you are going to do so anyways when using your car normally. There will be a situation where you are going through the entire battery.

 

[ran349]

I'm not sure you read the posts I wrote here. I'm not contradicting myself. Balancing is happening all the time. A full charge/discharge cycle is recalibrating the capacity/range estimate. Those are two separate things and have nothing to do with each other.

Based on my observation of my battery over almost 100k miles I found that the pack never goes out of balance enough that it would lose any significant amount of range. Even when only partially charging for a long time. I believe that what Jason wrote is sure correct (balancing is initiated at 93% charge level or higher) but that might not be the only time it is. I have not charged my battery to even 90 for well over two moths and still the cells were very close together. So either they don't drift apart much or balancing is more sophisticated than just being triggered at 93%. Either way, there no reason to charge your car up all the way to trigger balancing. So yes I totally agree that the recommendation to charge your battery all the way up to balance it is nonsense.

But for a whole different reason, namely the calibration of the range estimating, it does make sense to do a full cycle here and there. Yes that email is old, but the fundamental issue remains. You can't directly measure a battery's exact state of charge and capacity at any given point without fully charging and then discharging it. The error will always be there. It really doesn't hurt a battery to do an almost full cycle once every few months. Chances are you are going to do so anyways when using your car normally. There will be a situation where you are going through the entire battery.

But I think most people only care about balancing in the sense of getting more range, or at least more perceived range. That was what I was assuming balancing meant. My battery shows about 6% range loss at 45k miles, but most of that came almost immediately after I got the car. The question is how much real capacity has my battery lost? But I usually charge between 40- 80%, maybe 1 time I got close to 100%. But no matter how low or high my battery charge % shows, it always calculates to the same 100% value, so it doesn't appear to have any hidden range available. But I am not going to drive to zero miles or below to find out for sure.

 

[apacheguy]

6% loss is extremely unlikely to be due to balancing with Tesla’s BMS. I’ve gone 4 months before without a range charge and the cells are still within 4 mV.

Just because it calculates the same 100% value does not mean there isn’t range below 0. My vehicle has about 4% below 0 and it has charged to 258 rated miles for the past 2 years. Hasn’t gone up or down.

 

[ran349]

6% loss is extremely unlikely to be due to balancing with Tesla’s BMS. I’ve gone 4 months before without a range charge and the cells are still within 4 mV.

Just because it calculates the same 100% value does not mean there isn’t range below 0. My vehicle has about 4% below 0 and it has charged to 258 rated miles for the past 2 years. Hasn’t gone up or down.

I would think that if you actually drove to 4% below zero, then when you charged to 100% it would recalibrate and show extra rated miles and that 4% below zero wouldn't be there the next time. At least that is the way I would interpret the letter from the Tesla executive link above from David99:
" They are correct that it will mitigate the inaccuracy of the capacity algorithm, resulting in a higher displayed range, but it is misleading as the actual amount of stored energy does not change."

 

[Naonak]

I would agree with that as well @ran349, are people not seeing this happening?

I need to recalibrate one of my batteries... maybe I will go through a cycle and see what happens.

 

[apacheguy]

I have not driven past 0 displayed rated miles. However, others have and have not noticed any added range on the next top up. This actually makes sense because rated miles are based on nominal full pack energy. The algorithm shifts some of those kWh below 0 on some vehicles, so it is accounted for.

As an example, about two weeks ago I rolled in with 5% displayed on the IC. I checked the values on CAN and it was still showing 7.4 kWh remaining. 7.4/76 = 9.7%. Thus, the vehicle is aware of this residual capacity but for whatever reason decides to shove it below 0.

 

[Olds442]

243 miles at 100%.

that's awesome.

 

[snd92]

Balancing happens at all SOCs, but the calibration of BMS happens at 90-100% SOC.

 

[David99]

Balancing happens at all SOCs, but the calibration of BMS happens at 90-100% SOC.

That is the theory based on Jason's experiments. But it doesn't mean there are other times or methods that the BMS uses to measure cell balance. As I said before, the reason I believe that is because I have not charged my car higher than 80 % for several months and the balance was still as good as always. But I have no actual evidence.

 

[snd92]

By calibration of BMS, I meant the guess of mile/kilometers/battery capacity (guess of remaining kWh). I do fully agree with you @David99. My pack is balanced well and I have not charged above 70 % in the past year or so.

 

[FlyingCookie]

According to Jason Hughes, who has opened many Tesla batteries and figured out a lot of what's going on inside the battery, balancing is triggered at a high state of charge (93% or above). Once it is triggered the BMS boards controlling each module work independent of state of charge, driving or charging or idling. Basically, the BMS figures out how much the modules are apart and then sends a command to bleed off x amount of energy. The BMS boards then do that completely independent of what else is going on with the battery. That means it's completely irrelevant what state of charge the battery is at. The resistors that bleed off the excess energy of the highest modules are rather small, so this process can take hours, maybe even days. That is what Jason has been finding. If that is true, and given his level of expertise I have no doubt about it, then it makes no difference what level of charge the battery is. Seeing a 2 mV difference is a meaningful measure.

It's very easy to balance at a high SOC, so a 2 mV difference at 90 or 100% isn't meaningful. The BMS does start the process of balancing the battery above 93% SOC, but when you start pulling energy from the cells the voltage on the weaker cells will fall faster, even if they were "balanced" at a high SOC.

If you really want to see how balanced your pack is, pull it down to 4% and then check the cell voltages. I'd bet you'll find much larger differences.

If a single cell in a single brick has failed, that brick will hit 4.06 volts before the rest of the bricks do, but the rest will eventually catch up and make the pack appear to be balanced. That same brick will also hit the cutoff voltage before the rest do. That's why a single cell failure can cause a 1.5% loss in range on an 85KWh pack.

 

[apacheguy]

Yes, the cell spread increases at lower SOC but even a healthy battery can have a 10 mV or greater cell spread in the single digit range. Doesn’t mean there is anything wrong.

 

[mongo]

you really want to see how balanced your pack is, pull it down to 4% and then check the cell voltages. I'd bet you'll find much larger differences.

I'd call that sub module (parallel group) uniformity as opposed to balanced.

If a single cell in a single brick has failed, that brick will hit 4.06 volts before the rest of the bricks do, but the rest will eventually catch up and make the pack appear to be balanced. That same brick will also hit the cutoff voltage before the rest do. That's why a single cell failure can cause a 1.5% loss in range on an 85KWh pack.

This is why they equalize at high SOC. If all the groups hit the high end of their voltage at the same time, that gives the maximum available power from the pack. If a group is weaker or missing a cell, its voltage will drop faster and will set the low end of the pack's capacity. However, on the next charge cycle, all groups will again hit the max voltage together (Assuming a charge to 100%, if the typical charge is 90% and the BMS sets them equal at that point, the weak group will also set the top end).

It shows an array of inductors near the fuses going to each cell tap within the modules.

I believe those are filters to reduce high frequency noise (part of your option 1). Linear Tech and others make controllers to perform charge transfer between groups, but the arrangement would be different.
For well matched cells in a long term use case, the bleed resistor only needs to dissipate power equal to the worst case difference in module self discharge plus a little extra for initial balancing.

 

[hacer]

[QUOTE="
I believe those are filters to reduce high frequency noise (part of your option 1). Linear Tech and others make controllers to perform charge transfer between groups, but the arrangement would be different.
For well matched cells in a long term use case, the bleed resistor only needs to dissipate power equal to the worst case difference in module self discharge plus a little extra for initial balancing.[/QUOTE]
Yes, after reading the whole thread on wk057's battery tear down it is evident that the balance circuit can in fact only shunt a group of cells with 39 Ohms, or ~ 0.1A current. Having thought about this, along with the description of these shunt elements being left on for long periods to bleed down, rather than being used for long periods during high-end charging and reached a sad conclusion for the case of failed cells:

If a cell fails, and blows it's cell-level fuse (or fails open), then one of the 96 series elements has 73/74ths (or 1.4% less) the capacity of all the other series elements. When you go to charge to a high capacity, this segment's will increase faster than all others, and eventually will reach it's upper limit and shut off the charging entirely. So approximately the whole battery is now at 73/74th of what it was before the cell failure which is unavoidable. But now the balancing circuit kicks in and slowly discharges the weak segment until it's voltage matches that of the others, after which it as only (73/74)^2 (or 2.7% less) energy it could have stored before the cell failure. When you discharge the battery then this segment's voltage will fall faster than the others, since it has less energy stored and it will reach the low-voltage limit before all of the others. The effect is to (slightly worse than) double the effective loss of capacity of the battery.

If instead, the balancing was done during charging, the weak segment would have been given as much energy as it could store, as would the other segments, and so the capacity loss is not doubled. Unfortunately that is essentially impossible if only 0.1A of bypass current is available because the balancing portion of the charging cycle would have to run for more than 24 hours which is totally impractical.

I've suspected that my car might have a failed cell because it seemed like my 90% suffered a rapid drop back when mine was a 70 and I had no way to 100% charge to see if balancing was the culprit. So I'm slightly worse than 6% range loss after 21k miles; now I think that this could be 3% aging loss + 3% loss due to a single cell failure.

Bleeding is a good strategy if the segments mostly have similar capacity, but simply become unbalanced due to impedance differences or thermal gradients, etc. But it's not so great in the case of cell failures or differential capacity degredation.

 

[mongo]

[QUOTE="
I believe those are filters to reduce high frequency noise (part of your option 1). Linear Tech and others make controllers to perform charge transfer between groups, but the arrangement would be different.
For well matched cells in a long term use case, the bleed resistor only needs to dissipate power equal to the worst case difference in module self discharge plus a little extra for initial balancing.

Yes, after reading the whole thread on wk057's battery tear down it is evident that the balance circuit can in fact only shunt a group of cells with 39 Ohms, or ~ 0.1A current. Having thought about this, along with the description of these shunt elements being left on for long periods to bleed down, rather than being used for long periods during high-end charging and reached a sad conclusion for the case of failed cells:

If a cell fails, and blows it's cell-level fuse (or fails open), then one of the 96 series elements has 73/74ths (or 1.4% less) the capacity of all the other series elements. When you go to charge to a high capacity, this segment's will increase faster than all others, and eventually will reach it's upper limit and shut off the charging entirely. So approximately the whole battery is now at 73/74th of what it was before the cell failure which is unavoidable. But now the balancing circuit kicks in and slowly discharges the weak segment until it's voltage matches that of the others, after which it as only (73/74)^2 (or 2.7% less) energy it could have stored before the cell failure. When you discharge the battery then this segment's voltage will fall faster than the others, since it has less energy stored and it will reach the low-voltage limit before all of the others. The effect is to (slightly worse than) double the effective loss of capacity of the battery.

If instead, the balancing was done during charging, the weak segment would have been given as much energy as it could store, as would the other segments, and so the capacity loss is not doubled. Unfortunately that is essentially impossible if only 0.1A of bypass current is available because the balancing portion of the charging cycle would have to run for more than 24 hours which is totally impractical.

I've suspected that my car might have a failed cell because it seemed like my 90% suffered a rapid drop back when mine was a 70 and I had no way to 100% charge to see if balancing was the culprit. So I'm slightly worse than 6% range loss after 21k miles; now I think that this could be 3% aging loss + 3% loss due to a single cell failure.

Bleeding is a good strategy if the segments mostly have similar capacity, but simply become unbalanced due to impedance differences or thermal gradients, etc. But it's not so great in the case of cell failures or differential capacity degredation.[/QUOTE]

It's not as bad as you think. Yes, the group with a missing cell will hit the max voltage first. Not good, and the BMS will drain it , also not looking good. But in the overall system, the small group is now more aligned with the normal ones (when charged). After a few cycles of this, the small group will hit the max V at the same time as the normal groups. Ideally the BMS will recognize this condition and not drain all thee other modules when the smaller group V drops below the normal.

If there were 10 cells normally, and the failure occurred at 50%
Small norm
Failure 50% 50%
charge 100% 95% (50% × 9/10 swing)
Balance 95% 95%
Discharge: 50% 54.5% (45*.9 for norm)
charge 100% 99.5% (50*.9 for norm)


Balancing by shunting current is problematic because you either need to take the high group out of the circuit and short past it (complexity and added resistive losses during normal operation) or else add a shunt path in parallel that matches the cell voltage and handles a large portion (all) of the charge current. So for a 35kW 100A charge, it would need to dissipate 400W or so.
Active balancing is possible too, but is overkill most of the time. The small bleed approach is low cost, low complexity, and over a small number of cycles can get the small group high V aligned with the rest of the pack. (A really smart BMS can adjust it regardless of current SOC)