3.0 Battery Longevity

[bolosky]

If we use 300 miles as one complete cycle (it's probably a little less, but roughly) then I'm a little over 100 cycles, and certainly nowhere close to 200. You would think that the shallow, middle-of-the-range cycling that we do would be nicer to the cells than the deep cycles usually used for tests like that.

I've seen a similar capacity vs. cycle graph for what someone thought was one of the Model S cells, and it had a very different shape than the original Roadster cell. It had a very steep dropoff until it was at about 10% loss, followed by a very flat section for many hundreds of cycles. I've been hoping for a long time that this is what we're dealing with in the 3.0 battery. My car's performance is somewhat encouraging, but I still feel like we need much more data to really be sure.

I wonder how hard it would be to run a set of cycle tests on cells ourselves. You'd just need a rig that measured voltage and current and had enough smarts to be able to run the proper cycles. Then we could buy a few of the putative 3.0 cells, measure them under varying conditions (deep fast cycles, middle slower cycles like we expect, whatever else we think of) and generate our own graphs. If those end up looking like what we're seeing for the car and they do indeed have a long plateau, I'd feel a lot better. The tricky part might be managing temperatures to make it look like it's in the car.

 

[ion_1]

I wonder how hard it would be to run a set of cycle tests on cells ourselves. You'd just need a rig that measured voltage and current and had enough smarts to be able to run the proper cycles. Then we could buy a few of the putative 3.0 cells, measure them under varying conditions (deep fast cycles, middle slower cycles like we expect, whatever else we think of) and generate our own graphs. If those end up looking like what we're seeing for the car and they do indeed have a long plateau, I'd feel a lot better. The tricky part might be managing temperatures to make it look like it's in the car.

This test is decently straight forward. A model protocol can be extracted from a "typical" drive scenario and programmed into a battery cycler and run in repetition. The only factor that would be missing is calendar life degradation (cells can be run at 40C max to accelerate).. We just need to know for certain what these cells are...Ideally run these in parallel with a couple of Panasonic LCOs and maybe a NCA cell representative of an S. This way we could establish a known benchmark to validate the protocol.

But the fundamental question after a couple of years still remains, what LG cells are these? If we know this for certain than many of our questions will be answered immediately even without testing.This is the most important question by far. Much extensive characterization data can be extracted from the manufacturer.

 

[bolosky]

I just ordered an el-cheapo cell tester from China. I'll get some LH HG2 cells and do some cycle lifetime tests. It'll take a month for the tester to show up, and then quite a long time to run the tests (if you run at 0.25C then one cycle is 8 hours because you have both charge and discharge, so 500 cycles = ~6 months). So don't hold your breath on this...

 

[dpeilow]

But the fundamental question after a couple of years still remains, what LG cells are these?

Something no doubt Tesla could answer in a flash, along with confirmation as to whether they are behaving as expected and will plateau. The fact that it is radio silence is what concerns me.

 

[Kerios]

I wonder how hard it would be to run a set of cycle tests on cells ourselves. You'd just need a rig that measured voltage and current and had enough smarts to be able to run the proper cycles. Then we could buy a few of the putative 3.0 cells, measure them under varying conditions (deep fast cycles, middle slower cycles like we expect, whatever else we think of) and generate our own graphs. If those end up looking like what we're seeing for the car and they do indeed have a long plateau, I'd feel a lot better. The tricky part might be managing temperatures to make it look like it's in the car.

Its a good point, but I suspect we'd need to get identical cells to be certain and without cracking a 3.0 pack to get them that be difficiult. I wonder if anyone has a 3.0 Roadster which is offline / out of commission even written off perhaps from where we could get an actual example; that would be ideal for subsequent testing.

I'd ask my SC but...

 

[Kerios]

Did a range charges last night. Range shows now 302, that's a drop of 45 from the original 347, or 13% in 14 months. This sucks.

 

[slcasner]

So it's an interesting question why the BMS thinks the capacity of the bricks varies so much. It would also be interesting to see the voltages after a range rather than standard charge.

I have recently done some range charges so I can answer that question.

On 8/18, I started a range charge ahead of a long drive. The charge ran from midnight to about 7 am, but there was not enough time for balancing to finish before I needed to leave shortly after 9 am. Here's the balancing matrix at 7:48 with Vave=4.32, Vmin=4.30, Vmax=4.33 and CAC=184.06:

Here is the balancing matrix at 9:05 just before driving away, Vave=4.32, Vmin=4.29, Vmax=4.32:

On 8/21 I planned another long drive. This time I did a full standard charge the day before so that the range mode charge would finish earlier and leave more time for balancing, but still it did not quite finish. The charge ran from 11 pm to 3 am. At 7:02 with Vave=4.32, Vmin=4.30, Vmax=4.32, CAC=185.07the balancing matrix was:

At 8:11, with Vave=4.32, Vmin=4.29, Vmax=4.32, the balancing matrix appeared to be almost finished:

However, just before leaving at 11:27 the balancing was worse again, with Vave=4.31, Vmin=4.29, Vmax=4.31:

On 8/25 I needed to do another range charge, and this time I arranged for the charge to be finished at 1 am and by 7:37 the balancing matrix was cleared with Vave=4.31, Vmin=4.29 and Vmax=4.32 and CAC=185.96:

The voltage spread was always within 0.03V, so per what @MLAUTO said the pack is not out of balance. It does appear that my pack may have a weak brick, though, because each time I looked the Vmin value was read from brick #20.

I will send a PM with the latest log info.

 

[CharityDreams]

I got my 3.0 upgrade in early October, so it's been about 4 months. In that time, I've put roughly 6K miles on the battery.

When I first got it, I did a couple of range charges both to see what it was showing and also because I had to do a long drive (and then back). The best that I saw was 344 ideal miles with a CAC of 215.04.

Since then, I hadn't needed to do a long drive, so I didn't range charge it. I noticed that the ideal miles in a standard mode charge declined somewhat, from maybe 225 to 217 or so. I looked for the first time in a long while, and the CAC had declined to 201.45, a little more than a 6% loss.

I thought that maybe the problem was that the battery couldn't balance properly with the roughly 75% SOC in ideal mode, so I range charged it, let it sit plugged in for a long time to balance, topped off, and repeated a few times.

The result of this is that the CAC improved to 202.24 (total capacity at full charge 74.36 kWh as opposed to 77.26 kWh new) and the standard range charge is back to 221. However, the best range mode charge I saw was 326, a full 18 miles less than when the battery was new (again, 4 months and 6K miles ago, so not very long ago).

I figure that one of several things may be going on, rated from best to worst in their implications:
1) The 3.0 battery has cells that have a quick initial dropoff in capacity, but then level out for a long while. I know that some of the Model S cells are like this (and I think in that case that Tesla just sells them as if they're at the plateau level). So, I should expect to stay near this capacity for a very long time, and the R80 is really more of an R74.
2) There's something wrong with my particular battery, and Tesla will just replace it for me since it's a dud.
3) The 3.0 batteries lose 30 ideal miles/10k of range (as opposed to ~3.7 in Tom Saxton's study for the 2.0 battery). So at 33K miles a 3.0 battery will have the same capacity as a new 2.0, and quickly descend after that. That is, the 3.0 batteries don't work and I spent $30K making my car be useless earlier than if I'd have kept the 2.0 battery.

I really hope that Tesla didn't screw up in cell selection and it's not #3. Or that if it is that they'll suck it up and fix all of our quickly decaying batteries.

Have any of the other 3.0 owners seen a decrease like this? How much mileage do you have on your new battery?

I would like to talk to you on the telephone. I have a 2010 Tesla Roadster. Lawrence in Hawaii. (808)636-3333

 

[S-2000 Roadster]

So here's a hypothesis. The CAC algorithm doesn't get any data unless you range charge the car or run it to very low SOC. Otherwise, it's a complete guess based on what the engineers thought when they wrote the code. When you DO range charge or run to low SOC, it gets some data, but it averages that in with what it already thinks (which may be based on old data, or may be based on lots of guessing).

The guesses that it makes are essentially just drawing a line from when it was new to the current time. However, when it gets new data from a range charge, it not only adjusts the CAC, it adjusts its estimate of the slope of the line. So, range charging not only increases the CAC, it decreases its rate of decline. Again, this isn't saying that it helps the cells (it almost certainly doesn't), it's that it helps the estimate.

I just found this thread and was going to ask how we know that we can trust the first range estimates, so I'm glad to see your hypothesis. My understanding is that predicting actual battery charge is quite difficult, and prone to error. Considering the small number of Roadsters, the lack of focus on this model, and the difficulty of the problem, it seems very likely that we're not going to be able to expect very accurate numbers. Perhaps the firmware just needs more work, but Tesla Motors hasn't dedicated the engineering resources to this particular problem.

The good news is that lots of R80 owners are seeing something like 345 miles in Range mode right after the upgrade, so that must be a reasonable estimate of the improvement.

 

[dhrivnak]

The good news is that lots of R80 owners are seeing something like 345 miles in Range mode right after the upgrade, so that must be a reasonable estimate of the improvement.

While the above is true, I have seen my range charge change from 347 when new, to 319 fourteen months later. A loss of 2 miles a month. That seems quite steep for most of us. I am still trying to see if that loss is real (best scenario) or if the loss tapers off after a steep decline in the first year (less bad scenario). But if the 2 miles a month continues, then our 345 mile pack is below 200 miles in just 6 years making no better than the LEAF with no thermal management.

 

[Kerios]

While the above is true, I have seen my range charge change from 347 when new, to 319 fourteen months later. A loss of 2 miles a month. That seems quite steep for most of us. I am still trying to see if that loss is real (best scenario) or if the loss tapers off after a steep decline in the first year (less bad scenario). But if the 2 miles a month continues, then our 345 mile pack is below 200 miles in just 6 years making no better than the LEAF with no thermal management.

Yup, and it gets worse, I started off with 347 also last June and I’m now seeing range charging sub 302miles, a drop of 13% in 14 months. That sucks, and it drops further every time I drive the car; this isn’t about firmware, it’s about battery chemistry and sloppy engineering.

 

[bolosky]

Yup, and it gets worse, I started off with 347 also last June and I’m now seeing range charging sub 302miles, a drop of 13% in 14 months. That sucks, and it drops further every time I drive the car; this isn’t about firmware, it’s about battery chemistry and sloppy engineering.

I think it's all three. And I suspect that there's really a plateau on the battery. My car has been holding at roughly the same CAC (~189) for over 10 months now. I did the experiment of doing a full range charge and then running the battery all the way down to see what happened two weeks ago and have been procrastinating writing it up (and will continue to procrastinate for at least one more day), but the bottom line is that it delivered roughly what it claimed, and while the CAC went down a little after doing it, it didn't go down much or stay down. So, at the very least I'm sure that my battery capacity hasn't continued down on the steep original curve.

 

[dpeilow]

Having just driven across Europe doing several standard and range mode charges in succession I've seen an interesting phenomenon. A full range mode charge at the start was 325 miles but has gone up to 338 miles, however CAC has gone up from just under 201 to just under 203 Ah. That suggests the algorithm is not linear.

Furthermore, when the battery is warm (around 40°C) I have seen real estimated ranges of well over 350 miles cruising at 68-70 mph.

 

[slcasner]

@dpeilow remember that the range numbers you get immediately after charging stops are not realistic. It takes a little while for the range number to settle down to a stable value. I have not observed whether the CAC number similarly varies, but if not, that might explain the non-linearity.

 

[dpeilow]

The CAC has gone up and stayed at that value for about a week now. I did see a little bit of settling, but it can be topped off again afterwards, and when I say genuine range I mean on track to hit that after 200+ miles of driving, not immediately after a charge.

 

[bolosky]

The CAC has gone up and stayed at that value for about a week now. I did see a little bit of settling, but it can be topped off again afterwards, and when I say genuine range I mean on track to hit that after 200+ miles of driving, not immediately after a charge.

CAC going up after range charging is consistent with what I've seen both in my car and in the data. I suspect that it's just because the estimation algorithm gets information that it has previously been too pessimistic.

Not only that, but I wouldn't be surprised if your future rate of decline is now less than it was as well. That clearly happened to my car (I wrote a post about this a few months ago).

Estimated depends so much on current driving conditions that I just ignore it. I live in Washington State where there are lots of hills, so it mostly tells me if I've been going up or down recently.

I'll process your data and post an update soon.

 

[bolosky]

Here's the latest version of the graphs, with updated data from four cars: mine (670), and 2, 33, and 330(EU). Thanks for the data, slcasner and dpeilow!

2 really hasn't been driven, and its CAC hasn't been updated since the last data point for it. It's got 132 miles on the 3.0 battery (and 4171 on the car). I think all we really learn here is that if you don't drive at least a little, the CAC algorithm doesn't do anything at all. That's why it's such an outlier on the by-days graph.

#33 still shows average Ah 201, min (CAC) 186, so it's off by 15Ah. Not sure why this is happening given that slcasner said the car thinks its balanced. That said, the CAC has improved somewhat after some range charges.

#330(EU) similarly did some range charges/long drives and had its CAC go up. Interestingly, its newest data point on the by-miles graph is right on top of mine, and tied for the best CAC at that mileage (~8K) of any car in the study.

My car has shown a little decline after a long

steady ​

period. This is almost certainly because I did a drive designed to run the battery all the way from full to empty. It's possible that I actually damaged my battery a little by letting it get a tad too empty, but it also might be the algorithm reacting to seeing the real bottom of the battery. I still owe you a write-up of that, which I hope to get done soon.

 

[dpeilow]

So I take back what I said above: After briefly going over 203 Ah, the CAC is now down to 199.3 Ah. I noticed a big drop on the first charge immediately after I went north of the Alps on my trip, where ambient temperature dropped from 30°C to around 10°C.

With the warm battery temperature range gains I saw, I wonder if the algorithm cannot cope with the way this chemistry seems to perform when warm and then drop back when cold?

 

[hcsharp]

It's possible that I actually damaged my battery a little by letting it get a tad too empty,

That's unlikely with the new cells. One of their attributes compared to the OEM cells is that they have a lower minimum voltage. The Roadster is unable to take advantage of this because the charging system can't start that low and still control the amperage through the cells. The new cells also have a slightly higher maximum voltage which is the main reason the PEM needs to be modified.

 

[bolosky]

That's unlikely with the new cells. One of their attributes compared to the OEM cells is that they have a lower minimum voltage. The Roadster is unable to take advantage of this because the charging system can't start that low and still control the amperage through the cells. The new cells also have a slightly higher maximum voltage which is the main reason the PEM needs to be modified.

I managed to get Vmin down to 2.6V (and VMax 3.5V). I did this by parking the car and letting the AC/heat run. As soon as I noticed this, I plugged it back in and it went into "Recovery Charging" where it drew 7A for a while before coming back up to a more normal level.

BTW, do you know for sure what the new cells are? I've gotten conflicting reports.