3.0 Battery Longevity

[hcsharp]

Main takeaways (based on single 18650 cells) you may see in the data:

1. Recharge often and avoid large depth of discharges

Thanks for posting this great study. It does indeed show that shallow discharges are better for battery life. I'd love to see this study performed on the new 3.0 cells.

2. If you do not use full capacity, charge and discharge your cell as much in the mid range as possible. Lower range is good also, but not very practical and can lead to higher currents with voltage depression.

You cannot draw this conclusion from the study. Cars are different from generic studies in that vehicles generate more heat in the cells at lower SOC. You don't know what the results would have been, for example, if 1 SOC had been allowed to get warmer than 3 SOC which it would in an automotive application. Or 3 SOC vs 5 SOC, in which the results were very close, but what if 5 SOC didn't generate as much heat in the cells as 3 SOC? Would the results have been closer? Or even more cycles for 5 SOC than 3? Since Tesla never charges any cells to 1 (4.2v) even in range mode you really don't know what the results would have been for a Roadster.

Tesla has performed A LOT of research in this area as it applies to cars. I suspect they've found the magic balance between mid-range (promoting more cycles) vs high-range of SOC (promoting less heat and therefore more cycles) to provide the greatest cycle life.

 

[ion_1]

You cannot draw this conclusion from the study. Cars are different from generic studies in that vehicles generate more heat in the cells at lower SOC. You don't know what the results would have been, for example, if 1 SOC had been allowed to get warmer than 3 SOC which it would in an automotive application. Or 3 SOC vs 5 SOC, in which the results were very close, but what if 5 SOC didn't generate as much heat in the cells as 3 SOC? Would the results have been closer? Or even more cycles for 5 SOC than 3? Since Tesla never charges any cells to 1 (4.2v) even in range mode you really don't know what the results would have been for a Roadster.

Tesla has performed A LOT of research in this area as it applies to cars. I suspect they've found the magic balance between mid-range (promoting more cycles) vs high-range of SOC (promoting less heat and therefore more cycles) to provide the greatest cycle life.

The takeaways that I listed were only related to the study data that I posted, and in that case, I believe it is a correct and obvious conclusion. A conclusion which the authors also concluded, and of course the reviewers who allowed the paper to be published.
I was VERY careful to state that this was for single 18650 ONLY not in a battery pack, in car, etc. and I agree and did state that you cannot necessarily extrapolate. I even mentioned you specifically when I did. I also did mention that lower voltages could generate more heat. This is obvious as you are running at constant power thus require more current for lower voltages.

30% discharge at 2C is quite extreme, but could be relevant if you are tracking the car.

These are also for different (possibly) chemistry than the roadster 3.0. I tried to be careful.
However, these type of conclusions are long know to be correct for most Li-ion cells and I thought some here would be interested in this study. Faster discharge rates within each SOC 30% segment lead to a faster increase in impedance for the mid and upper SOCs. The effect is not as great as SOC. The effect could be heat related or chemistry related due to overpotential inducing new reactions or both.

Inside a battery pack, as you correctly state, is another story. I am even more concerned with extended times at high temp in ambient heat in the parking lot in summer near fully charged, etc.

One thing to also note is that the impedance of the battery changes as a state of charge. The impedance of the electrode materials change as a function of state of charge. The two ends is where it gets nasty, especially towards the end of discharge.

I agree Tesla has found a balance. Range when you need it, but sparingly. Standard which gives the best return on useable mileage balanced with retaining life. One question that pops up from owners is whether you are commuting only a few tens of miles a day, is there a better window to operate in (if you could), and if so how could this relate to the 3.0 battery? The standard window is optimized to give you a decent amount of range, but what if you don't need all that range...Running at a slightly lower SOC with a LiCoO2 based battery doesn't change the voltage that much as the voltage profile is quite flat. As such, running at a lower state of charge would not induce a higher current leading to more heat if you are in a narrow window of driving range. 3.0 may be a very different story.

Tesla and especially Panasonic (and LG) has performed research which I very deeply respect, but so has others. What is given to the consumer may not be the best solution but a compromised solution to keep consumers on the average extremely happy. One great thing about this forum is it allows owners to think more deeply, possibly beyond what Tesla has brought to the table, whether it be Brakes, Bushings for control arms (Wiztecy) or battery data analysis (Bolosky).

The most important item would be to have a solid confirmation of what LG cells are inside the 3.0 roadster. Rumors were HG2, but I am not sure that is correct.

Until then, Bolosky and the contributors are king around here as they are the only one with real relevant data.

 

[dpeilow]

The most important item would be to have a solid confirmation of what LG cells are inside the 3.0 roadster. Rumors were HG2, but I am not sure that is correct.

I'm really not sure what advantage Tesla has by keeping this proprietary. It's a cell type that will be a one-off for the Roadster.

 

[bolosky]

30% discharge at 2C is quite extreme, but could be relevant if you are tracking the car.

I've always wondered what a study like would look like with an actual car profile for charge/discharge rates. While we might have a few seconds at ~3.5C (me getting on the freeway with a full charge is about that), most of the time it's much lower, typically 0.3C for freeway driving and less (plus lots of 0 and regen at ~0.2C) for traffic, then charging at just over 0.1C (40A @220v minus fans and pumps and whatnot). I don't have a really good feel for how different it would be, but my guess is it's much better than 30% steady discharge/recharge at 1C/2C for a given SOC.

I suppose I could set it up and actually do the experiment, it's not all that hard.

 

[bolosky]

Here's a 3D version of both charts with axes for days, miles and CAC. When I ran a two-variable linear regression, I unsurprisingly got a much better fit, with R squared of 0.82 (meaning roughly that time and miles combine to explain 82% of the variation in CAC). The line has an intercept of 214 (good, that's a new car), and slopes of -0.0271 Ah/day and -0.001 Ah/mile. I didn't plot the regression line on the 3D graph (yet), I'm still learning R.

The big outlier at the top is #660, which is at just under 413 miles in 282 days (and < 3K miles lifetime). I think that with that little driving the algorithm isn't doing much of anything at all.

Let me know if you like this format instead of or in addition to the 2 2D charts, or if I should just skip it in the future.

 

[sethr]

I'm really not sure what advantage Tesla has by keeping this proprietary. It's a cell type that will be a one-off for the Roadster.

I have never really understood why Tesla is so stingy with information. They won't even tell you your CAC, even though they know most Roadster owners know how to access it (I look at the "secret" screen of the VDS). If they gave us data about the 3.0 cells I might, maybe consider spending the $29K. Given the data you all are presenting, I'm rather dubious. By the time the CAC levels off on the current crop of 3.0's, they may well not be offering it anymore. Who knows? Probably Tesla, but I'm sure they will never say.

Right now, though, I'm waiting for them to address my current situation ("Powertrain Problem, fault # 287"). Two weeks before they'll even take my Roadster in for service. (Sorry - off topic, I shouldn't rant here about service.)

 

[DeedWest]

...Right now, though, I'm waiting for them to address my current situation ("Powertrain Problem, fault # 287"). Two weeks before they'll even take my Roadster in for service. (Sorry - off topic, I shouldn't rant here about service.)

It may be your 400V Controller. $1,566.62 installed.

 

[sethr]

It may be your 400V Controller. $1,566.62 installed.

Happily, I am still under an extended warranty. From reading the many helpful posts on this forum, I believe you are correct. I just hope Tesla has parts!

 

[hcsharp]

The takeaways that I listed were only related to the study data that I posted, and in that case, I believe it is a correct and obvious conclusion. A conclusion which the authors also concluded, and of course the reviewers who allowed the paper to be published.
I was VERY careful to state that this was for single 18650 ONLY not in a battery pack, in car, etc. and I agree and did state that you cannot necessarily extrapolate. I even mentioned you specifically when I did. I also did mention that lower voltages could generate more heat. This is obvious as you are running at constant power thus require more current for lower voltages.

OK, I'm sorry I misunderstood you. Since this is the 3.0 Longevity thread I thought you were drawing a parallel to the 3.0 pack with your takeaways. My bad. Over the years I've seen so many people conclude that they should run their EVs at a lower SOC window based on studies that don't simulate an automotive application that I've probably become a little reactive...

Tesla told us when they first announced the 3.0 upgrade that they had to modify a circuit board in the PEM to allow it to charge the cells to slightly higher voltage because the new chemistry would not suffer from it. They must have meant range mode because it looks to me like the std mode is the same voltage as the OEM cells.

 

[hcsharp]

Let me know if you like this format instead of or in addition to the 2 2D charts, or if I should just skip it in the future.

I like the new format although it would be nice to display the colors to show the z axis better. It's hard to tell how deep some of those points are.

Thanks again for your efforts.

 

[bolosky]

I just got 6 months' more data for #425 (thanks, ecarfan!). It's interesting and a bit of an outlier. It's now the car with the largest number of days (460) and it's far above the CAC of comparably old cars, even #537, which has fewer miles. On the per-mile chart, it's been trending up (!) for the last roughly 2K miles. Even more interesting is that it's curve is remarkably similar to mine (#670) at the same mileage range. In fact, from about 5200 to the end, it's got all the same bumps as 670, although the climb at the end is much less. It's probably a coincidence, but it's remarkable nevertheless, and now I'm curious to see what happens to other cars in that range. I attributed that bump in my car to range charging it every day for a week as an experiment (and there's some evidence that 425 was also range charged when it started up because of longer drives and at least one time where the daily log shows 89% SOC.) Still, now I wonder if there was also something else going on. We'll find out with more data over the next few months, I assume.

I'll keep tweaking the 3D graph to see if I can make it look better. In the interim, here are the two older-style ones.

(BTW, I looked and I'm still at about 191 with several hundred miles more than is shown here. Maybe I'll update when I cross 20K.)

 

[ion_1]

OK, I'm sorry I misunderstood you. Since this is the 3.0 Longevity thread I thought you were drawing a parallel to the 3.0 pack with your takeaways. My bad. Over the years I've seen so many people conclude that they should run their EVs at a lower SOC window based on studies that don't simulate an automotive application that I've probably become a little reactive...

Tesla told us when they first announced the 3.0 upgrade that they had to modify a circuit board in the PEM to allow it to charge the cells to slightly higher voltage because the new chemistry would not suffer from it. They must have meant range mode because it looks to me like the std mode is the same voltage as the OEM cells.

No problem at all.

As you mentioned, in the context of the thread it could have been taken the wrong way.
Just trying to present some info that may be related to the health of the core foundation of the battery, the single cell 18650.
From there, there are all the deleterious factors associated being built into a functioning battery with real life conditions.

I completely agree with your concerns regarding battery heat, one must be sensitive to heat. Lower SOC would require one to carefully factor in what % of capacity they burn through so to avoid voltage depression and intrinsic impedance increase towards the end of discharge and especially their driving habits (high rate discharge). If not carefully thought through, due to heat and polarization effects, it could be more detrimental than helpful. I think we are on the same page, we are just approaching the discussion from opposite ends of scale.

Interesting re your comment on PEM. This is in line that NMC based cells and to some degree NCA cells are more voltage tolerant, but age differently than LCO. However, my confidence in a robust 3.0 Range charge is not great because as you and others state, there is a larger capacity difference between range and standard for the 3.0 cells vs 1-2.x cells. Range charge is a good show/marketing for mileage, but Tesla is playing it conservative for standard charge. This is encouraging, but telling...

 

[ion_1]

I attributed that bump in my car to range charging it every day for a week as an experiment (and there's some evidence that 425 was also range charged when it started up because of longer drives and at least one time where the daily log shows 89% SOC.) Still, now I wonder if there was also something else going on. We'll find out with more data over the next few months, I assume.

Interesting comment.
With all our (and likely Tesla's) concerns re range charging, the few range charges you did did not seem to have a very significant negative impact on life at this point. It would be great to somehow to tally the approx # of range charges and say put in parentheses next to the car number in the legend. Range charge gives a lot of mileage benefit in the 3.0, it would be great to see if it has a major detrimental effect in the long run.

Looks like there are a number of batts giving evidence they are leveling off (a bit early to conclude), just waiting for 181 to turn the corner...

 

[ecarfan]

I just got 6 months' more data for #425 (thanks, ecarfan!). It's interesting and a bit of an outlier.

Sigh. Story of my life...and I wouldn't have it any other way.

Thanks for all your work. Very interesting data!

 

[bolosky]

Updating with another month's data for my car, which is now at nearly 21K since the upgrade. My CAC is about the same as it was 2K miles ago, at 192. This is now the longest period without a net decline that I've seen, which is encouraging but by no means conclusive that I've gotten to some kind of knee in the curve.

If you look closely around 20K, you'll see a day with a long drive (two dots far apart on the mileage chart) with a corresponding increase in CAC. That day, I range charged, drove the car ~170 miles, range charged again (though not to full this time), and drove it back, hitting the charger with 29 ideal miles (not counting the bottom 10%), so SOC probably about 18%. It had a notable jump up in CAC the day of the long drive (after the range charge) from 190.8 to 191.3. The data point was most likely taken while the car was recharging the second time. The next day, it dropped a whole amp hour to 190.3. I've seen that pattern before: range charges make the CAC (and here the emphasis is on "calculated," since I seriously doubt this reflects the real state of the cells) go up, and deep discharges make it go down.

A week later, around 20.5K miles I did another range charge for a long-ish drive. This time, however, I drove about 180 miles and didn't get the battery to such a low SOC. This time, there was a big jump (191.0 to 192.3) but with no corresponding drop the next day.

You can see a similar thing for #33 around 3500-4000 miles: there's a couple of days with long-ish drives and increases in CAC, followed by a third day with a very long drive (nearly 300 miles) and a big decrease in CAC (199.9 to 196.2). I don't know if the car was range charged, or how low the SOC went on the third day, but it seems likely that he did about the same thing as I did with similar (though more dramatic) results.

On the other hand, the big drop in #537 at 2873 miles (nearly 3 Ah) was without any driving at all, so that was probably something else. Likewise with the second big drop at 3975 miles.

So, it looks like range charging the car (which is showing the BMS the top of the capacity) makes its estimate go up, while deep discharges (showing it the bottom) makes it go down. Interesting. And, on a larger scale, for #670 things seem to have been pretty level for an unusually long time at about 90% of initial CAC, which in the end may be more interesting if it continues.

Without further ado, the charts.

 

[ecarfan]

I've seen that pattern before: range charges make the CAC (and here the emphasis is on "calculated," since I seriously doubt this reflects the real state of the cells) go up, and deep discharges make it go down.

Excellent point, and a persuasive reminder not to take sudden CAC changes too seriously.

My car (#425 in the graphs) had I think 3 range charges near the end of September which appears to have bumped up the estimated CAC. During that time I never drove it below 50 miles remaining (in Range Mode) if I recall correctly.

it looks like range charging the car (which is showing the BMS the top of the capacity) makes its estimate go up, while deep discharges (showing it the bottom) makes it go down.

Great analysis, thank you!

 

[gregd]

So, it looks like range charging the car (which is showing the BMS the top of the capacity) makes its estimate go up, while deep discharges (showing it the bottom) makes it go down.

The take-away here may be that the bottom of the battery's capacity is not so well known to the car, and is likely closer than it appears. So, don't count on those last few 10s of miles! This supports the recent report of someone who had their car shut down "early" during a long trip.

 

[bolosky]

The take-away here may be that the bottom of the battery's capacity is not so well known to the car, and is likely closer than it appears. So, don't count on those last few 10s of miles! This supports the recent report of someone who had their car shut down "early" during a long trip.

I agree. I think that it's probably best never to plan to use the last 10% that you unlock with range mode. If something unexpected happens, then maybe it's there, but I wouldn't count on it. We now have two different pieces of evidence to suggest that there might be a problem (CAC and the Roadster that ran out of battery unexpectedly).

 

[slcasner]

You can see a similar thing for #33 around 3500-4000 miles: there's a couple of days with long-ish drives and increases in CAC, followed by a third day with a very long drive (nearly 300 miles) and a big decrease in CAC (199.9 to 196.2). I don't know if the car was range charged, or how low the SOC went on the third day, but it seems likely that he did about the same thing as I did with similar (though more dramatic) results.

This was a trip to Santa Barbara on 4/21/2017 and return on 4/24/2017 with a side trip to Los Angeles on 4/23/2017. Both the trip down and back were about 290 miles. Both began with a range mode charge and included a partial standard mode partial charge along the way. There was a range mode charge also before the side trip. On the way home the SOC went down to 8% and 22 ideal miles in standard mode.

 

[gregd]

On the way home the SOC went down to 8% and 22 ideal miles in standard mode.

Standard mode? Standard mode hides the bottom 12%-ish of the battery capacity from the ideal miles calculation, so 8% SoC should already be underwater in standard mode, no? Those numbers are about right for Range mode.