It's arbitrary saying 100 divisions is as accurate as it goes, and anything finer is 'empty resolution'.
No, it isn't. If you quantize something to 1 unit you introduce an error with standard deviation of 1/sqrt(12). Look up "quantizing"
If that was the case, the car would not be able to calculate single digit range numbers, yet it does.
Nonsense 0.41*295 = 120.95.
When you look at the CAN bus data you can see the car calculates remaining capacity much more accurately. I have done that for 4 years and I can see it adding and subtracting energy at 0.01 kWh increments. I can see it accurately measure single digit Watts.
When I calculate pH from an electrodes temperature and voltage readings I use floating point and get an answer to floating point resolution. But as I know pH expressed to more than two decimal points is meaningless I only display two. Or put another way, if my ADC has resolution of 0.1 mV then I know the accuracy of any reading is at best 0.1/sqrt(12). That number may get converted to floating point (very precise) but it is still not accurate to better than 0.1/sqrt(12). That error gets propagated throughout the rest of the arithmetic, mixed in with the error from the temperature reading and magnified by the "geometry" of the problem and the final answer that comes out while it may have 12 digits beyond the decimal point is only accurate to 2 so I only display 2.
It is perfectly capable of measuring and calculating with very small amounts of power and energy. Very small amount of power draw are accurately measured and shown.
The problem is that coulomb counting does not give very good estimates of SoC. To determine SoC by coulomb counting you have to know where you are counting from. That is if you want to know what the SoC is after adding Q coulombs you have to know what the state was before you added the Q. In order to do that you need to estimate the the OC voltage of the cells and have the detailed charge/discharge curve for the temperature at which the measurement is made. This is pretty complicated and the "geometry" of the problem thus pretty hairy. Thus coulomb counting only gives a relatively rough estimate and even if you do the arithmetic in floating point (which I gather they do) the answer is not accurate (though it may be precise). Do you understand the difference between accuracy and precision? I suspect not. I've run into lots of engineers who don't let alone lay people.
Definitely not 'empty resolution'.
Thus precision in an SoC estimate beyond two significant digits or at most perhaps 3, even though it come from coulomb counting and voltage measurement combined in a Kalman filter, is definitely empty resolution. And guess what? The engineers at Tesla know that and that's why they show the driver two significant digits.
When you look at the CAN bus data you can see the car calculates remaining capacity much more accurately.
No, it calculates it with more precision. You really need to understand this before trying to draw conclusions about the displays.
Percent is a random number that doesn't tell you how far you can go.
Whaaaa? The displayed value of how far you can go is calculated from the estimated SoC. How far you can go is the available charge (really energy) divided by the energy estimated to be required to go one mile (or km). The problem is, of course, that that range is an estimate and its numerical value depends on how it was estimated. Thus a range estimate contains both the error incurred in trying to estimate SoC
and the error inherent in the Wh/mi estimate.
What % tells you is how much usable battery you have left. That is the most important piece of information to you. If I see battery dipping below 5% I know I'd better find charge tout suite. I know that that nominally means I have 15 miles left and I know that I can stretch that out if I hypermile. I'm not really trying to push % displays because I think people should use whatever they are comfortable with. But evidently people get lulled into thinking that they have 14 miles range left when their SoC is 0 % available.
There is no user error. 14 miles left is far far more than misjudged energy usage can explain.
I don't think there was ever a suggestion that user's were making calculation errors. The observation was that people who use the % display rather than the miles display don't seem to be reporting this pickle.
Even at 50% higher energy usage the actual remaining range would be 9 miles instead of 14. Still far away from shutting down.
Well I have no idea as I never use the miles range setting but I think it might be the other way round. IOW you get down to 5% usable but as you have been rolling downhill with a tail wind your car is measuring 100 Wh/mi (I actually saw -146 on a drive the other day). Assuming your car has the 100 kW battery that means about 5000/100 = 50 miles estimated range when in fact your rated range is more like 15 miles which is what you will experience when you come to the bottom of the hill or 7 if you then have to go back up the next hill. The problem with this thesis is that I don't know that the little battery gauge displays estimated miles or just rated. If it's just rated then there is, in effect, no difference in what the display tells you be it in miles or km or %.
So if the car does shut down at 14 miles left, something dramatic has gone wrong because it is supposed to have 14 miles left plus some emergency buffer. Usually its individual cells/bricks that have degraded more than the rest and show a sudden voltage drop causing the BMS to do a premature shutdown. Again, there have been several cases exactly like this and Tesla has replaced the battery under warranty.
I am not saying that there cannot be battery or BMS failures. If I see a utilization graph that shows the actual utilization tracking the predicted nicely to some point at which the actual curve plummets below the 0 line then I agree that there is something awry in the battery system. That's why I keep asking "what did the utilization curve show?" But I never get an answer. That tells me that either people don't understand what I am asking or that they don't monitor that graph. If there is any good to come out of this thread it would be that people would have gotten the idea that they should monitor that graph.