How I Recovered Half of my Battery's Lost Capacity

[Bubbatech]

What made you decide to never run it below 50% to begin with?

Convenience. I always plug the car in when it is the garage. I have never seen a provision for setting charge limits below 50%, and the vast majority of my trips are less than 20 miles. I live in a relatively small city.

 

[Electric700]

I've seen the service manual excerpt for the model 3 posted a few times. The BMS starts balancing at voltage above 4.0

Which is about 75% currently, I think. So if everyone sets their vehicle to 75 or 80%, battery balancing should automatically start (unless this has recently changed and balancing occurs at states of charge lower than 75%). Thanks.

 

[KenC]

Voltage above 4.0? Isn't that 90+% SOC?

 

[AAKEE]

Voltage above 4.0? Isn't that 90+% SOC?

Nope, rules of thumb:
4.2V = 100%
4.1V = 90%
4.0V = 80%
3.9V = 70%

Its not 100% accurate, and Tesla hides the buffer wich offsets the displayed SOC a little.
We will probably read about 75% SOC at 4.00V
We can not read the OCV so we will se the voltage with a slight load.

 

[KenC]

Nope, rules of thumb:
4.2V = 100%
4.1V = 90%
4.0V = 80%
3.9V = 70%

Its not 100% accurate, and Tesla hides the buffer wich offsets the displayed SOC a little.
We will probably read about 75% SOC at 4.00V
We can not read the OCV so we will se the voltage with a slight load.

Anyhow, I don't know at what level the BMS rebalances cells, but mine typically shows 6mV imbalance, and as you know, I keep my SOC fairly low.

 

[AAKEE]

Anyhow, I don't know at what level the BMS rebalances cells, but mine typically shows 6mV imbalance, and as you know, I keep my SOC fairly low.

I’ve been at 4mV since new for more or less anytime the car has been parked for a while.

Balancing is no hokus pokus, just evening out the differences.
When running low-middle SOC the battery probably wont need balancing.

Top balancing ensures that the battery is full with no cells not fully charged and pushes the difference between cells down in the buffer.

A complete balancing actually put the “perfect balance” upwards, rendering slightly more imbalance at low SOC.

Ive heard a lot of theories when the car is balancing the cells.
Balancing in Teslas way means burning electrical energy up.
Shouldnt be done more than needed as it use energy.
I havent seen any evidence that it balances at low SOC.
Once I charged in steps and had the same imbalance from 55 to 85, after passing 90% and two hours the imbalance was lower.

 

[Hiline]

Thank you for sharing the experience. My 2019 Model 3 with 50k miles has a 280-mile range while the EPA range is 325, representing a 14% range loss which is quite an anomaly for the mileage. I'm gonna need to try the varying state of charge method to see if this is a problem of degradation or of range estimate.

 

[AAKEE]

Thank you for sharing the experience. My 2019 Model 3 with 50k miles has a 280-mile range while the EPA range is 325, representing a 14% range loss which is quite an anomaly for the mileage.

The degradation per mile is sort of a myth.

The degradation from battery cycles is mostly very moderate. Cycle tests of Tesla model 3 2170 cells show about 5% degradation the first about 250 cycles, this despite quite large cycles. Also, the calendar aging is included so the actual cyclic aging is less than 5%.

These 5% means about 5% for 100K km or 60k miles. Or 0.5% per 10k km /6 k miles. (But actually less if we count the cyclic aging alone.)

The main part of the degradation is calendar aging. You live in a warm climate?
What SOC do you charge to, in general?
And at what time on the day do you start the charging?
What is the normal SOC when you arrive home before charging?

For a warm climate we can look at the chart below: 10 months at 25C and 70-90% SOC causes about 5% degradation. This means one year costs 5.5%, and four years costs 11%

The battery will be warmer than the ambient due to driving and charging, and sometimes standing in the sun also.

You also have a few percent cyclic degradation so of we add 3% cyclic aging to the 11% calendar aging we are at your numbers you are looking about.

I'm gonna need to try the varying state of charge method to see if this is a problem of degradation or of range estimate.

It might be worth a try, but as it is not that improbable that you actually have 14% degradation, it might not work or it can temporary increase the range and then fall
back to the numbers you see.

Also: remember that a BMS calibration doesnt increase the battery capacity. It just increase the numbers on the screen.

 

[Hiline]

The degradation per mile is sort of a myth.

The degradation from battery cycles is mostly very moderate. Cycle tests of Tesla model 3 2170 cells show about 5% degradation the first about 250 cycles, this despite quite large cycles. Also, the calendar aging is included so the actual cyclic aging is less than 5%.
View attachment 942761


These 5% means about 5% for 100K km or 60k miles. Or 0.5% per 10k km /6 k miles. (But actually less if we count the cyclic aging alone.)

The main part of the degradation is calendar aging. You live in a warm climate?
What SOC do you charge to, in general?
And at what time on the day do you start the charging?
What is the normal SOC when you arrive home before charging?

For a warm climate we can look at the chart below: 10 months at 25C and 70-90% SOC causes about 5% degradation. This means one year costs 5.5%, and four years costs 11% View attachment 942756

The battery will be warmer than the ambient due to driving and charging, and sometimes standing in the sun also.

You also have a few percent cyclic degradation so of we add 3% cyclic aging to the 11% calendar aging we are at your numbers you are looking about.


It might be worth a try, but as it is not that improbable that you actually have 14% degradation, it might not work or it can temporary increase the range and then fall
back to the numbers you see.

Also: remember that a BMS calibration doesnt increase the battery capacity. It just increase the numbers on the screen.

Everything you said makes sense. I bought this car used and was very surprised to find out how low the range was when I took delivery. My new Model Y degraded by 1% after 9 months so I had assumed the used Model 3 should have something like 10% degradation. The Model 3 degradation is indeed very high compared to other cars of similar mileage according to TeslaFi - the average car with 50k miles has 12 miles more range.

The car was based in Nevada before it came into my possession, so I do assume a high storage temperature. It's also likely that the previous owner used Tesla's recommended state of charge of 90%. 40C at 90% SoC corresponds to 8% degradation in 10 months according to your chart, so I guess this explains the 14% degradation over 4 years. You're probably right that BMS calibration won't move the needle - I know the range is what it is, and I wanted to know what the actual degradation should look like, but given what we know now, it probably doesn't deviate much from 14%.

Hopefully the lower temperature in LA would help slow down the battery degradation. 280 mile is ok, but if the range ever dips below 260, the car will be more like a Standard Range than a Long Range.

 

[AAKEE]

It's also likely that the previous owner used Tesla's recommended state of charge of 90%.

Well, they do not. Thats a myth.

Tesla says ”to under 90%” for daily use.

Which when I read i means any number below 90%. Not recommend 90%.

This doesnt mean it wasnt charged to 90% most times, as the myths have been very wide spread.

The car was based in Nevada before it came into my possession, so I do assume a high storage temperature.

Yes, that pretty much guarantee a higher degradation.

 

[afty]

I observe that quite frequently. If you keep your SOC for a long time between 40-60% and then charge to 70% or to discharge below 30% there will be a large update of the SOC after sleeping. (i.e. up to 5%)

Interesting. For daily driving I keep the car between 50-70%. I took a trip and got back yesterday. Pulled into the garage with 13%, but a couple of hours later it was at 8%. Charged overnight to 70%, which drifted upward to 74%. Interesting seeing this battery adjustment in action.

 

[PCMc]

Interesting. For daily driving I keep the car between 50-70%. I took a trip and got back yesterday. Pulled into the garage with 13%, but a couple of hours later it was at 8%. Charged overnight to 70%, which drifted upward to 74%. Interesting seeing this battery adjustment in action.

So the drop from 13% to 8% when you got back is result of battery cooling. You see the impact much more at low SOC, on my MS90D say below about 20% when I park after a long drive with battery fully up to temp. The upper end shift after charging may be more of an evidence of the BMS calibration adjusting from my experience.

 

[dwatty]

Great detail. Thanks for taking the time.

What a great post. I have a 2018 M3PD, full charge estimate is 286. 32k miles. I will test this theory and report back.

 

[AAKEE]

So the drop from 13% to 8% when you got back is result of battery cooling.

Its not probably battery cooling.
First of all, these days we can see this difference as a blue part of the SOC.

Other, when the battery is very very cold, lile -15C I had a drop of about 4% from the battery being cold.

If we have summer, there is mostly no drop at all for most people. For me in the very north part of the world I might see 1% drop if the car was outside during the night and we had a very cold summer night.


The difference in this case almost certainly comes from the BMS recalculating (measuring) the true SOC ny measuring the cell voltage.
The true SOC can not be measured during a charging session or during driving.
This is because the battery will show a higher voltage during charging and a lower voltage during driving compared to when the battery is at rest.

During drives and charges, the SOC is calculated by counting energy in- and out.

In a perfect world, with a perfect BMS calibration the estimated SOC matches the true SOC at any time.
In the example below we disregard the buffer to make it easier:
Having a 80 kWh capacity fully charged and using 40kWh during a drive means 40/80kWh used = 50%. So the estimated SOC will be 50%. After the car was parked the Bms measured the cell voltages either with the car awake or when sleeping. As the BMS estimation of the capacity is spot on, the estimated SOC matches the true SOC.
We will not se a change in SOC after the car was parked.
(Sleeping will have the HV battery didconnected so this will be better/more exact)

If we had a BMS that overestimated the capacity by 2.5% (true capacity 80 kWh but the estimated capacity was 82kWh), this will happen in the same drive:

Starting value is 100% SOC, used was 40kWh/82 kWh = 48.8%, so estimated SOC after the drive was 51.2%.
After thw car was parked and the SOC has updated it would be 40/80= 50% used, so 50% remaining.
The car would show 51.2% when parking and 50% adter the BMS has read the resl SOC and updated it.

These calculated/expected SOC can be seen with Scan My Tesla and we can also see it update the SOC.

We can also see the displayed SOC reducing or increasing after a drive when the SOC is updated.

I have a lot logged data which covers this.
One of the best examples was a complete 100-0% drive during one day, at the same time as my BMS was fairly off, underestimating the capacity with about 3kWh.
After a 240km drive I parked with 52% and about eight hours later I stsrted the return drive, then showing 54% SOC. The math from the updated SOC actually matched the true capacity measured on the total 100-0% drive.
I was able to calculate how much the BMS was off by using the updated numbers.
To do this, SMT or similar showing the SOC in decimals is needed, and a longer drive to actually ”force a difference” between estimated and real SOC.

You see the impact much more at low SOC, on my MS90D say below about 20% when I park after a long drive with battery fully up to temp. The upper end shift after charging may be more of an evidence of the BMS calibration adjusting from my experience.

I see the updated SOC to follow the same pattern both with charging and driving.

Charging:
The SOC overshoots the set charging target - capacity overestimated.
The SOC undershoots the set charging target - capacity underestimated

Driving:
The SOC adjust down after the car has been parked for a while - capacity overestimated
The SOC adjusts up after being parked- capacity underestimated.

Both these need a true SOC to begin with, like sleeping before the drive (or fully charged). A long drive with lot of charges and no sleep will most probably set the initial SOC value a little off, so in that case the up/down adjustments can not be taken as true for the capacity.

 

[PCMc]

Its not probably battery cooling.
First of all, these days we can see this difference as a blue part of the SOC.

Other, when the battery is very very cold, lile -15C I had a drop of about 4% from the battery being cold.

If we have summer, there is mostly no drop at all for most people. For me in the very north part of the world I might see 1% drop if the car was outside during the night and we had a very cold summer night.


The difference in this case almost certainly comes from the BMS recalculating (measuring) the true SOC ny measuring the cell voltage.
The true SOC can not be measured during a charging session or during driving.
This is because the battery will show a higher voltage during charging and a lower voltage during driving compared to when the battery is at rest.

During drives and charges, the SOC is calculated by counting energy in- and out.

In a perfect world, with a perfect BMS calibration the estimated SOC matches the true SOC at any time.
In the example below we disregard the buffer to make it easier:
Having a 80 kWh capacity fully charged and using 40kWh during a drive means 40/80kWh used = 50%. So the estimated SOC will be 50%. After the car was parked the Bms measured the cell voltages either with the car awake or when sleeping. As the BMS estimation of the capacity is spot on, the estimated SOC matches the true SOC.
We will not se a change in SOC after the car was parked.
(Sleeping will have the HV battery didconnected so this will be better/more exact)

If we had a BMS that overestimated the capacity by 2.5% (true capacity 80 kWh but the estimated capacity was 82kWh), this will happen in the same drive:

Starting value is 100% SOC, used was 40kWh/82 kWh = 48.8%, so estimated SOC after the drive was 51.2%.
After thw car was parked and the SOC has updated it would be 40/80= 50% used, so 50% remaining.
The car would show 51.2% when parking and 50% adter the BMS has read the resl SOC and updated it.

These calculated/expected SOC can be seen with Scan My Tesla and we can also see it update the SOC.

We can also see the displayed SOC reducing or increasing after a drive when the SOC is updated.

I have a lot logged data which covers this.
One of the best examples was a complete 100-0% drive during one day, at the same time as my BMS was fairly off, underestimating the capacity with about 3kWh.
After a 240km drive I parked with 52% and about eight hours later I stsrted the return drive, then showing 54% SOC. The math from the updated SOC actually matched the true capacity measured on the total 100-0% drive.
I was able to calculate how much the BMS was off by using the updated numbers.
To do this, SMT or similar showing the SOC in decimals is needed, and a longer drive to actually ”force a difference” between estimated and real SOC.



I see the updated SOC to follow the same pattern both with charging and driving.

Charging:
The SOC overshoots the set charging target - capacity overestimated.
The SOC undershoots the set charging target - capacity underestimated

Driving:
The SOC adjust down after the car has been parked for a while - capacity overestimated
The SOC adjusts up after being parked- capacity underestimated.

Both these need a true SOC to begin with, like sleeping before the drive (or fully charged). A long drive with lot of charges and no sleep will most probably set the initial SOC value a little off, so in that case the up/down adjustments can not be taken as true for the capacity.

I'm sorry, but I stand by my comment that the initial drop is due to the battery cooling.

It's not the cold ambient issue which you reference and showing with a range of the battery being blue, but the battery cooling from fully warm, working temperatures to ambient.

Fully warm battery may be running 40 deg C or even at little higher, so it's cooling from the fully warm operating state to ambient temp, even moderate ambient temps.

I've seen this multiple times on my MS90D over the years with ambient temp in the 70-80 F (25 deg C) range. I've captured it happening in the datalogs which I collect using my own logger program. As I've said it only tends to be seen at relatively low SOCs, somewhere below 20% SOC in my experience.

This is why you may see if you park your car with the SOC being low that there will actually be an alert pop up recommending that you charge soon.

 

[AAKEE]

I'm sorry, but I stand by my comment that the initial drop is due to the battery cooling.

It's not the cold ambient issue which you reference and showing with a range of the battery being blue, but the battery cooling from fully warm, working temperatures to ambient.

I have many logs about this.

First thing, lithium ion batteries do not change the OCV noticeble when cooling down, even to very cold values.

In the very fine level, like single mV or psrts of, the voltage might drop or increase very little, depending on the chemistry and SOC region.

I took a few Panasonic 2170 NCA cells the otjer winter, measured the voltage in my house at +25C.
Then I did put them outside the house at -26C for about three hours.
My medium class Voltage meter did not detect any difference at all, measures down to hundreds of a volt. ( 1/100 of a volt = about 1%)

The BMS wont experience a drop in SOC due to the cooling of the battery at all.
When using scan my tesla checking the BMS (or looking at teslafi numbers) we can se the true SOC when parking, after the battery has cooled the true SOC is the same but the displayed SOC is lower.
The displayed SOC is recalculated to reflect the usable energy but the true SOC still is the same.

For my (just sold) M3P 2021, I could park it at more or less anytime and then catch it 10hrs later showing the same displayed SOC.
I have a lot of logs, just need to check how to find the teslafi logs when I started a new session on the new car.

This is parking after a 240km drive to my job:

This is driving a short drive two days later, car was parked with sentry off and not connected. OAT was between minus 2 and +10C during these days (historical METAR data from the airport I work at)

For my car this is the normal thing, at many times parked one week at the time and it do not loose more than 1% from the number when setting the car to park.

I’ll continue to state it the other way
Loosing or gaining after park or a charge is due to a BMS wrong capacity estimate.
(Unless really low batt temp like sub freezing, then the largest delta I read in the BMS was 5C when having cell temp -15C
(Take more than a full day at sub -20C to reach that).

 

[awokeinthelbc]

Speaking fresh from memory, my X just errored out last night with BMS_u008 among other messages. It limited my speed and after dinner at restaurant it refused to start up. Internet searching says it’s likely a battery coolant failure.

That likely accounts for in past year it would precondition haphazardly half the time and correlate with battery state dropping over time. I saw various BMS messages in service mode past 6 mos but it didn’t appear while driving so I drove on.

So there is a slow degradation process it tolerates which may mess with your results.

I’ve a 2018 X with 76500 mi. Enough through desert driving activating jet engine fan mode. It should be covered under our extended warranty; google says they charge ~$700 to fix and will discount to $200 if without. One would imagine it covered under battery warranty but it’s not.

Meanwhile my wife’s 3P bought same month actually gained range back; we’ve similar mileage. We now just [free for life] supercharge when low <20% and it reset the BMS too.

 

[Sti-g]

New member (and new Model 3 owner) here. Amazing thread, I have been reading it for the past several days. I lost track of which posts I saw at what page, but I would like to contribute by adding some information.

Regarding balancing, I think there was a final post stating that the shunt (parallel) resistors are not always connected. This is correct. This is how passive balancing works. It can depend on OEM, but there will be a relay or mosfet switch on the board (PCB) connecting a resistor between positive and negative terminals of a given cell to discharge it. Not sure if any OEM is really using board elements for the switch/relay since the same can be done with a mosfet switch in the BMS chip (which saves PCB area & cost). The resistor will have to be on the board for thermal management. Having the mosfet in the chip is reducing balancing current to keep die heating within reason. I have seen typical values around 200-300mA. This small amount of current is the reason why it takes so long to balance cells, considering group of cells (which are connected in parallel) may have capacity of 10s of Ah.

Automotive chips are designed for -40C to 150C junction temperature. Given that environment/battery pack temperature can reach ~60C, combined with die heating of the chip due to its own power consumption can put the die temperature easily around 80-90C, before any balancing current. When you add balancing, say the on chip mosfet switch with 2ohm resistance @ 300mA will generate 0.18W, which can cause die heating of 5C easily. Each BMS chip usually connects to 12-18 cells, so depending on how many cells need balancing, power dissipated during balancing can be pretty high. Some duty cycling can also be used to reduce the average value of the balancing current for thermal management. 300mA balancing current means 14ohm (total) resistance, so the resistor on board will be 12ohms (in this example), which will have to handle 1.08W (12*0.3^2). Typical resistor can only dissipate 0.25W, so handling 1.08W = more expensive resistor. This is another reason why balancing current is low...

There was a question on why Tesla is not doing active balancing. Active balancing is expensive in terms of implementation. There are different ways of doing it but the main problem is you need to be able to take excess charge from ANY cell and move it to ANY cell with lower charge. This causes many combinations. I don't think it is worth the effort. I believe Mercedes EQXX was the first car to use active balancing. EQXX press release

Regarding when the balancing happens. I can't speak how each OEM does this, but to my knowledge, it is done when charging switches to constant-voltage mode, ie >80% SOC. I believe most people commented here that balancing happens around 85-90% SOC.

I also saw some discussion regarding charging at lower currents (in L2 charging) to reduce stress in cells. Level 2 charging is very low power. In Model 3 SR, there are 2976 cells. Charging at 7.8kW means each cell is getting ~2.6W. Compare this to early iPhones, which shipped with 5W chargers... So this is really low power and won't stress the cells. Supercharging at 170kW puts ~57W/cell which is pretty significant These numbers are same for LR version since the number of cells and charging power are both about 50% more.

 

[Fiver]

I believe most people commented here that balancing happens around 85-90% SOC.

Great post. Granted it was a long time ago (eg, many software versions in the past) but Jason Hughes mentioned that balancing in S/X seemed to happen when you got above either 93% or 95%. (I can't remember specifically which percentage).

This, again, was many years ago so who knows what they are using these days.

 

[PCMc]

Great post. Granted it was a long time ago (eg, many software versions in the past) but Jason Hughes mentioned that balancing in S/X seemed to happen when you got above either 93% or 95%. (I can't remember specifically which percentage).

This, again, was many years ago so who knows what they are using these days.

But he also posted later on that with software changes (I think maybe in the 2017 era??) that there was balancing occuring at virtually any SOC once the car was fully powered down.