Just to make sure I am not missing anything, your research shows that according to BMS software the following to be true, (based on MIN buffer size of 1 kWh) right?
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
jgs
Active Member
If there's one thing this thread convinces me of, it's that Tesla made the right call in ditching the kWh designators on car models.
FlatSix911
Porsche 918 Hybrid
FlatSix911
Porsche 918 Hybrid
This is one thing that convinces me that Tesla was dishonest with the original kWh numbers...
This makes a lot more sense and seems to align with what we were speculating about in this thread about why displayed rated range can be confusing. As always, thanks for the insight on what you find/learn.
FlatSix911
Porsche 918 Hybrid
Not exactly... here is the original data from wk057 as published on Electrek...
Tesla’s hacked Battery Management System exposes the real usable capacity of its battery packs
- Original 60 – ~61 kWh total capacity, ~58.5 kWh usable.
- 85/P85/85D/P85D – ~81.5 kWh total capacity, ~77.5 kWh usable
- 90D/P90D – ~85.8 kWh total capacity, 81.8 kWh usable
- Original 70 – ~71.2 kWh total capacity, 68.8 kWh usable
- 75/75D – 75 kWh total capacity, 72.6 kWh usable
- Software limited 60/60D – 62.4 kWh usable
- Software limited 70/70D – 65.9 kWh usable
Exactly, hence my request to @wk057 for clarification in the post you are responding to, to make sure I am not missing anything: results of his current research do not match results of his previous research published by Electrek.
And neither match Tesla actual physical charge depleting testing performed by Tesla for CARB as documented in their EPA Application dated 12/02/2014. Note that this actual charge depleting testing was then duplicated and matched by EPA official confirmatory testing performed at the "Volkswagen Test Center Oxnard" on 12/23/2014. I provided links to and excerpts from these actual charge depleting testing data published by the EPA. Tesla also tested 85 kWh pack using continuous C/3 rate of discharge, resulting in 83 kWh of energy extracted from the pack.
So this is not about Tesla "lying". This claim doesn't cut it. The only way it can be explained, if one wants to take this route, is that Tesla equipped the car it provided to EPA for independent testing with the battery, larger than the cars sold to customers. I see zero chance of this.
Last edited:
Have you found anything new about balancing? Back then you wrote the BMS triggers the balancing once constant voltage charging is reached. It then bleeds off the highest cell blocks. You mentioned back then that this point is reached aprox at 93% state of charge. I have done a lot of data logging and noticed in my 2014 Model S 85, the point at which it switches to constant voltage charging is always just under 90%.
I basically have two questions. Have you found that the BMS uses other measuring points or methods to determine an imbalance and triggers the bleeding resistors? Was the 93% a guess or are you sure about it? I just want to confirm if it's really necessary to charge to 93% to trigger balancing or if it's enough to 90% based my findings that the CVC is reached just under the 90% level.
There is just so much misinformation out there about balancing. Many people claim, without any evidence, that balancing happens when the battery is charged to 100% and that's why it takes so long to charge to 100%. If you could shine a little more light on the balancing Tesla is using, that would be awesome!
Looks about right for the max values possible with a new pack that has seen no real world use calibration yet, yes.
Will explain more below.
It's been a while, but I'm pretty sure the numbers above were calculated based on EEPROM values pulled from the BMS systems of vehicles in my fleet, and took an average of the various types. I did this a few different ways back then, but the results would be super close either way. Suffice it to say, data from the massive leak I was given goes along with the original data pretty well, so I stand by it for the most part except that the buffer portion would be off by ~2 kWh on the total capacity numbers based on newer knowledge.
Anyway, because this data was from vehicles that had some use, even minimal use, the data has some amount of real world calibration in it, and this was before I had reverse engineered enough to locate the initial factory values, and before I understood exactly what the "buffer" was.
In the end, the 60/70/75 packs would almost always calibrate upwards of their factory setting, since the physical setup was slightly under spec'd by a little bit, while the vast majority of "85"/"90" packs would calibrate downward from their initial factory values. With the 100 packs, they stopped the shenanigans, however there are a lot of outliers with the 100 after calibration (I've seen 100 packs calibrated to 103 kWh, as well as some at 96, both relatively new).
In summary, the original data was actually pretty generous. The refined data puts the total capacities of the 60/70/75s a bit tighter to their advertised sizes, and the total values of the "85" and "90" even lower than I'd said previously.
Oh! And for those wondering...
The Model 3 packs have these initial factory values (total pack size):
"Long Range" - 74
"Mid Range" - 62
"Short Range" - 50
They also incorporate the same "buffer" system, except they must have got wise to people sniffing this data on CAN and artificially inflate the "total" reported values with the buffer included already, unlike S/X. Yeah, have gotten most of the 3 BMS re'd as well ... including some interesting S3 code they left in there (Note: Pre-refresh is S, refresh is S2, upcoming model is S3, internally).
Worth an update here.
Tesla has changed the balancing algorithm many many times over the years. Originally this was a very dumb setup that would only kick in once a cell group reached a threshold voltage, usually around 90-93% SoC. This is no longer the case.
First, let me point out that Tesla's BMS software has come a LONG way... I'd consider it a work of art now. Lots of genius in there. It's absolutely amazing and full kudos where kudos are due here.
One thing they're now able to do is to calculate out the capacity of individual bricks of cells (96 in the 85/90/100, 84 in the rest) based on a ton of factors and compute this in near real time, in a full range of conditions, with almost magical accuracy. They're basically running physics simulations (similar to how they calculate out unmeasurable metrics in the inverter firmware, like rotor temperature) of the entire pack based on measured power usage/charge, balancer usage, temperature, temperature delta based on coolant flow and coolant temp, predicting temperature gradients, and probably 100 more variables. This is the holy grail of proper balancing for safety and longevity for a battery pack. This is not a dumb system anymore by any means. Knowing the actual capacity of the individual bricks allows them to know exactly which ones need cell bleeders enabled, and for exactly how long. With this data, they can balance on the fly at any SoC, and just use top and bottom SoC windows for fine tuning, validation, and calibration.
The car balances all the time whenever its needed. It knows when a cell group will need balancing before it's even out of balance... which is really freaking weird when you think about it, especially if you're watching a playback of the pack balancing and voltages and see it engage a balancer on a cell group that doesn't look out of balance at all, and watch it fall completely in line still at the end of a charge or discharge cycle. It keeps track of which groups will need it, which wont, how long they'll need it, how much they've been balanced, etc.
It really is an epic setup now.
The short answer to the balancing question: It balances any time it needs to balance.
As for SoC shenanigans, yes getting closer to 100% or 0% will give it a chance to tune things better... but it's not needed anymore. Just charge like you need to, and drive.
Presumably above capacities are derived from BMS software data.
The actual physical EPA depleting discharge testing data (based on SAE J1634 Multi-Cycle Procedure), don't match the above data. Below are USABLE capacities.
LR: 78,270 Wh
LR AWD: 79,213 Wh
MR: 63,180 Wh
SR+ 54,523 Wh
Tesla, TSLA & the Investment World: the 2019 Investors' Roundtable
Didn't we already establish somewhere that the EPA numbers are BS? As in, they don't actual measure this data and instead infer it from other measurements? Don't feel like digging for this, but I clearly recall this being the case.
Besides, I've done some capacity tests on the LR 3 pack... there is no way in hell it's > 74 kWh new.
If you find me even a photo of a Model 3 trip meter that shows anywhere close to 78 kWh since last charge on a 3, I'll be impressed. Until then, let's just agree that the EPA kWh data is useless.
Last edited:
Nope.
There is clearly a consistent difference between the charge depleting testing done by independent professional labs and data reported by you. I don't quite understand your desire to dismiss work by others instead of trying to understand the reason for differences and learn something from it.
Had put this in an edit, but guess it needs its own post:
Edit: Nm, found it.
Battery Electric Vehicle Test Standards
SAE J1634
Chassis dynamometer testing similar to conventional vehicles
Range must be found for each test cycle
Final Results
***** AC Wh / km (energy going into the charger) ******
So, accounting for charger (in)efficiency, the EPA numbers you posted match pretty well within a reasonable error margin.
I've clocked the Model 3 charger at between 92 and 96% efficient, depending on a bunch of factors.
TL;DR: The EPA numbers you posted are derived from power into the charger, as this is the result of SAEJ1634... NOT PACK CAPACITY.
There, understand the reason for the differences? lol
Edit: Anyway, I'm done with this thread again. I put the facts out there, take them or leave them. Arguing with @vgrinshpun is like talking to a wall anyway, and I'm not going to waste my time. If anyone has anything interesting or relevant besides "But Tesla says" or "but EPA test data that I'm not properly reading says".... tag me.
Last edited:
The testing includes data for both Vehicle DC Wh and Total (System AC) Wh, both are shown in the post I linked. The numbers I listed above are DC energy extracted out of the battery. The energy going into the charger is higher due to charging losses.
Wow. Didn't expect to see this thread resurrected.
Been sitting on this, mostly, but since the last time I posted anything relevant here, a Tesla insider leaked a TON of data to me about the entire fleet of Tesla vehicles. That data quite clearly shows that no 85 was ever 85, no 90 was ever 90. In fact, the vast majority of 90 packs have even degraded to be below the capacity of the vast majority of 85 packs at the time that data was given to me. This is their own internal data. 60, 70, 75, and 100s get a pass, as most of those meet spec. 85 and 90 folks got mislead, and they should be labeled 80 and 85 respectively. Since Tesla charges.. what, $3k to upgrade from a 70-75? ... I feel like we should all get $3k in credit or something. Not going to happen, but it is what it is.
Also since that time I've run capacity tests on loads of 85-type modules... and, you guessed it, not even the best come out to 85 kWh total capacity.
Next, I actually spoke with an ex-Tesla engineer who actually worked on the BMS software early on. He explained that the 4 kWh "buffer" doesn't mean what we think it means. It was actually a kludge built in to smooth the range calculation, and to make sure you actually could hit zero miles consistently without getting stranded. On average, the actual capacity "buffer" is 2 kWh, so the code was written so that a 4 kWh window was used and scaled along with the SoC as the car discharged, adjusted and calibrated as possible based on other measurements. This was to ensure that the range calculation would never adjust abruptly, and should never (rarely) run out of capacity while rated miles were > 0. TL;DR: The actual capacity left on the table by the value of the "buffer" is targeted by the BMS to be half that much.
Finally, in the past 2 years since my last substance post in this thread, I've almost completely reverse engineered the Tesla BMS software and hardware. Guess what's in there? Oh, right the factory "new" profiles for every battery type. More interestingly is that the other values (like those actually reported by the BMS) are directly derived from the initial factory capacity number. Anyone want to take a stab at what the new-from-the-factory values are for the "85"? Spoiler: Not 85. If you guessed 80, you win. (And for those playing along at home, the remaining initial values: 60 for the 60, 70 for the 70, 75 for the 75, 85 for the 90, and 100 for the 100.... one of these things is not like the other...). (Note: My reverse engineering also confirms what the engineer told me above.)
Let's stop beating this dead horse and get over it. Tesla lies, and they lied about the 85 capacity. Still love the products, hate the company.
What would say about the 90kw packs adjusting on the fly with regards to remaining range so that by the time a driver reaches zero, they would have only been able to go their EPA consumption * USABLE_CAPACITY when the starting range remaining number was calculated with NOMINAL_REMAINING * EPA consumption?
I would say that is called lying.
"Calculated" != "Measured"
That data is based on incorrect assumptions and is simply not correct.
Look, I'll give you my Model 3 *and* a Model S, if you get me a Model 3 built between release and now that 78 kWh can be discharged from for driving on a single charge. Like, I will sign the titles over after a properly calibrated shunt installed on the pack shows net 78 or more kWh discharged after a 100% to dead drive.
I'm willing to put real money (ie two Teslas) on the line to prove this is complete BS. Give me something equally compelling to back up your claims, and we'll let the community weigh accordingly... or don't, and that speaks just as clearly.
Last edited:
Similar threads
- Question
