3.0 Battery Longevity

[jaanton]

Re Tesla monitoring 3.0 battery issues... I'm doubtful that Tesla would be monitoring this. The 1.x Roadsters were the only car they had for a while so monitoring them was really important. The 2.x Roadsters had telemetry via cell. The telemetry is gone now, no more AT&T 2G service. Roadsters are scattered and rare compared to the rest of the fleet. If there is a problem with CAC measurement, I'm doubtful that Tesla is going to see it easily.

At some point shortly after the Model S was out, Tesla Service changed their software and when I asked about old history they said they had trouble getting to it. I was a bit unhappy because my car had some special service notes and they appeared to be gone.

I hope the community will keep making these graphs and maybe the OVMS people could start tracking.

(That gets me thinking... Another idea... I had an idea of custom OVMS behavior software triggers and CAC tracking would be nice.)

 

[bolosky]

Here's the lifetime CAC vs. Mileage with ecarfan's Roadster #425 added. Two things are interesting: on the original batteries, we had roughly equal CACs at upgrade time (134 for him, 132 for me) even though I drove a lot more (and a lot more after my first battery replacement). That makes it seem like the battery age matters more than the miles driven, which pretty much directly contradicts what Tom Saxton found in the Plug-in-America Roadster battery survey. Tom's got way more data than just two cars and did a more careful analysis than my quick-and-dirty graph here, so I'd be inclined to believe him. Still, I was a little surprised. Second, the slope of decline after the upgrades is essentially equal (he hasn't driven enough to get to the point where I started doing the range charges, so that isn't an issue). So, that's consistent with my hypothesis that the CAC computation algorithm isn't coping correctly with the 3.0 battery and standard charge.

 

[bolosky]

Here's a graph that zooms in on the 3.0 upgrade period, and normalizes the x-axis to miles after upgrade. This shows much better that the decline in CAC for the two cars is roughly the same.

 

[bolosky]

And finally, the data for the old batteries graphed as CAC vs. time since battery install. For my second battery I started the graph at the point where it had fully balanced (more-or-less its highest CAC), since it took a couple of weeks to balance. This graph shows my car dropping faster than ecanfan's (remember, I drive it ~4x as much as he does), so it makes it clear that there's both a calendar age and mileage based effect.

 

[bolosky]

I'd love to add more cars to this. If you want to contribute logs, please PM me. All you have to do is get the log from the car and upload it to the internet, I'll do the rest.

 

[bolosky]

The slopes of the lines for the 3.0 batteries are -.0026 Ah/mile for 425 and -.0024 Ah/mile for 670 (with 670 measured up to where I started range charging the first time around 5800 miles into the new battery). So, they're really close before my little excursion. The slope after the end of the range charging is a little steeper, -.0031 Ah/mile, measured from after it stopped flattening out). So, maybe it is trying to work itself back down to the original line.

In any case, this just can't possibly describe the real capacity of the cells. It's got to be algorithmic problems, which means that we don't really have much idea how the cells are doing.

 

[gregd]

In any case, this just can't possibly describe the real capacity of the cells. It's got to be algorithmic problems, which means that we don't really have much idea how the cells are doing.

Possibly true, but wouldn't the car's under-estimation of the actual capacity limit how far one can drive? When you've got ample capacity, it probably wouldn't make a difference. But if this trend continues, or if you're actually needing all of the extra range, it's got to hurt.

 

[bolosky]

Possibly true, but wouldn't the car's under-estimation of the actual capacity limit how far one can drive? When you've got ample capacity, it probably wouldn't make a difference. But if this trend continues, or if you're actually needing all of the extra range, it's got to hurt.

That's assuming that the CAC really determines when the car quits. I've seen cases where, when the range gets low, the car will adjust the ideal miles remaining (usually down, but I suppose it could be up).

 

[gregd]

That's assuming that the CAC really determines when the car quits. I've seen cases where, when the range gets low, the car will adjust the ideal miles remaining (usually down, but I suppose it could be up).

I expect they're all inter-related. I'm going to (finally) pick up my car this afternoon. I'll see if they can shed some light on this.

 

[dpeilow]

When my 3.0 battery was new it range charged to the mid 350's like others have posted about (if I recall correctly). The last time I range charged was about a month ago and it stopped at 346.

Are you sure about this? The highest I had ever seen claimed previously (and there was photographic evidence) was 346.

 

[ecarfan]

Are you sure about this? The highest I had ever seen claimed previously (and there was photographic evidence) was 346.

You are correct, my error. Sorry!

What I have seen over the past couple of months is that my standard charge Ideal Range number has drifted down from about 236 to 223-224 in the past week. I have not done a range charge for about two months, haven't needed to. But likely will next month.

 

[gregd]

I had my chat with the folks at the Rocklin SC. I gave them a copy of Bolosky's 3.0 battery graph, and explained our puzzle: Is the decline real, or a consequence of the CAC algorithm? If it's real, when will it flatten out, or will it continue?

They didn't have an answer, of course, but had some contacts they could pass the graph and question to. I'll follow up with them from time to time to see what they come up with.

 

[DeedWest]

Are you sure about this? The highest I had ever seen claimed previously (and there was photographic evidence) was 346.

There's actually a photo of 351 in Range Mode on Roadster 242 that was recently for sale.

EDIT: just found it.

 

[ion_1]

Here's a graph that zooms in on the 3.0 upgrade period, and normalizes the x-axis to miles after upgrade. This shows much better that the decline in CAC for the two cars is roughly the same.

View attachment 221495

Great plot. The trend is amazing reproducible. Assuming at least one of the batteries has seen a deep discharge and no self correction of CAC, my gut feeling this is true battery capacity fade. If these are LG HG2 cells as other have reported, these cells have a pretty sharp 10-15% capacity fade initially, then level out beautifully (based on LG published data). Let's see what happens, indeed one battery may be showing evidence of it already. Regardless, I am still amazed at the reproducibility of the data trend. Thank you for posting!

 

[ecarfan]

Here's the lifetime CAC vs. Mileage with ecarfan's Roadster #425 added. Two things are interesting: on the original batteries, we had roughly equal CACs at upgrade time (134 for him, 132 for me) even though I drove a lot more (and a lot more after my first battery replacement). That makes it seem like the battery age matters more than the miles driven

Very interesting, and a surprise to see the age effect.

Here's a graph that zooms in on the 3.0 upgrade period, and normalizes the x-axis to miles after upgrade. This shows much better that the decline in CAC for the two cars is roughly the same.

Yes, remarkably similar! Thanks for posting.

 

[ecarfan]

Just want to provide an explanation as to why the CAC data for my original battery drops so sharply at about the 1,900 day mark. That was several months before I purchased the car (I'm the second owner). I was told by the seller that Tesla had replaced one of the sheets because the CAC declined precipitously, and I have documentation from Tesla that supports that. So I assume that is why the data shows that sharp drop. But what is striking is that after the sheet replacement the CAC didn't improve.

And finally, the data for the old batteries graphed as CAC vs. time since battery install. For my second battery I started the graph at the point where it had fully balanced (more-or-less its highest CAC), since it took a couple of weeks to balance. This graph shows my car dropping faster than ecanfan's (remember, I drive it ~4x as much as he does), so it makes it clear that there's both a calendar age and mileage based effect.
View attachment 221500

 

[bolosky]

The owner of #181 contributed his data to me (thanks, Dave!), and I took another log from my car after doing a long drive and a few range charges, so I have a new graph. I continue to believe that what's happening here is that when the car charges in standard mode that it has essentially no data as to the battery capacity, so the default estimate from Tesla about battery degradation kicks in and just guesses based on mileage. On the other hand, when you charge it fully it does get some data, and so uses observations to affect its estimate. Since it was estimating degradation too much, range charges make the CAC go up (while probably making the actual battery capacity go down).

I noticed that the slope of the line for my car (#670) is less after the range charge around 9000 miles, so maybe I've hit the knee in Tesla's degradation estimate. OTOH, #181 seems to have continued declining past that CAC (though not mileage), so maybe I'm just seeing the effect of more range charges.

I'd love to have more cars for this graph, so if you're willing to donate your log file, please send it to me (or PM me and I'll give you an email address to which to send it).

 

[gregd]

Verrrry interesting... Has #181 ever been range charged?

What sort of ideal range is the car quoting for #670 (standard and range mode)?

 

[DaveD]

Has #181 ever been range charged?

Yes, but I think the last time was in October of last year.

 

[bolosky]

What sort of ideal range is the car quoting for #670 (standard and range mode)?

670 has standard charge ideal range around 215 and IIRC range mode is ~314 (or was it 320?)