Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

This site may earn commission on affiliate links.
I know I’ve seen Superchargers that charge an increasing rate per kWh as you approach full. I think in the car if you click on a Supercharger it will show you if a tiered rate is in effect.

In fact, here’s the nearest one to me, albeit in minutes rather than kWh:

52937327102_2736d9161e_z.jpg
Eddie a lot of people up voted your post so let me address the confusion you have started here.

You replied to this quote

I wonder if Tesla would increase the cost to charge from 90% to 100% as it takes the longest even with our Teslas.


you said
I know I’ve seen Superchargers that charge an increasing rate per kWh as you approach full.

now lets look at the key concept in the screenshot you used

1685500664852.png



0-60 kW is the slower charging that happens at the end of your charge and notice it's cheaper per minute, the opposite of what you suggested.

60-100 kW is a medium speed so not affected by your confusion

100-150 kW is a high speed charging that happens earlier in the charge curve when your battery is almost empty. It's more expensive per minute on the table, the opposite of what you suggested.

There is one situation that is an exception (the very beginning of a cold soaked battery in extreme cold conditions we won't see here in TN) where slow charging happens at the start of a charge. But the vast majority of the time the slow charging is at the end of the charge.

The tiers here aren't increasing cost as the battery approaches full, they are just trying to use tiers to approximate a per kWh cost in a state that doesn't allow them to price it per kWh.
 
Last edited:
While it's true that Tesla has multiple iterations of (different) hardware they support, I'm positive each requires specific tuning and configuration of the system.

For instance, just in cameras/imaging sensors, you have different characteristics such as:

-Sensitivity
-Resolution
-Bayer-pattern implementation
-Chroma filter wavelengths
-Noise envelope
-Fixed patter noise
-Saturation behavior
-Gamma response
-Aspect ratio
-Field angle
-Temperature sensitivity
-Tolerances
-etc....

And then there's placement around the vehicle (no 2 models are same size/shape), height, angle, windshield rake, etc...

Repeat for each discrete FSD component, and you have a multitude of parameters to account for. It's certainly possible to "generalize" the system such that other components can be used in a different vehicle implementation, but there's a large amount of work to do so, especially for a safety-critical system.
I don't know if Tesla can adjust their simulated training data to account for some of these differences.

The software could build up a 3D image map in a common format, for all models, regardless of the differences?

If there is a smart way to allow for this, I'm sure they have found it..
 
  • Like
Reactions: oldTAVguy
I don't know if Tesla can adjust their simulated training data to account for some of these differences.

The software could build up a 3D image map in a common format, for all models, regardless of the differences?

If there is a smart way to allow for this, I'm sure they have found it..
We humans don't have our eyes in the exactly same position in two different cars either. Yet we can drive different cars reasonably well. Our brain somehow figures out some approximation of where the car is and how the environment around it looks like. An artificial neural network can do the same. Give it a few million videos from a few hundred thousands of cars, autolabel the videos correctly and train the neural network to output where everything is and it will learn to do the camera calibration or whatever helps it solve the problem most efficiently...

I think Tesla intends to keep the camera suite, add some semis to the mix and the neural network will just learn to identify if it's being run in a semi or roadster(not a very hard problem), but learn from Y and 3 how other cars behave and use this for Model 25k. For HW3/HW4 mix of data they can probably use autolabel and simulation to simulate how HW3 would have seen the HW4 video and vice versa.
 
Last edited:
Eddie a lot of people up voted your post so let me address the confusion you have started here….

The tiers here aren't increasing cost as the battery approaches full, they are just trying to use tiers to approximate a per kWh cost in a state that doesn't allow them to price it per kWh.

Thank you for the clarification. I was, apparently, interpreting it wrong.
 
Thank you for the clarification. I was, apparently, interpreting it wrong.
Easy to do, I also read it wrong when I first saw it. The units of measurements aren’t intuitive for non-EE types like me, so I think of it as: kW = how fast electricity flows, kWh is storage (capacity or how “full” your car battery is).
 
I've been holding and buying since 2017 and will continue to hold till I die and my children get the stock

I agree the day-to-day price doesn't matter in the long run, but when Tesla was over $400 I tipped a lot better...
I used to give out what I referred to as "Tesla Twenties" to street musicians. Now they have to be satisfied with ten.
 
I don't know if Tesla can adjust their simulated training data to account for some of these differences.

The software could build up a 3D image map in a common format, for all models, regardless of the differences?

If there is a smart way to allow for this, I'm sure they have found it..
As the training data attempts to portray "actual" events, I think it more likely that in order to generalize FSD to the point that it can be implemented by other manufacturers using other/new hardware sensors, Tesla would have to specify the needed parameters for a device (minimum camera resolution, FOV, etc...) as well as a method to supply a profile the devices for FSD to ingest. Once that happens a calibration process for the FSD system for that specific model implmentation....
 
As the training data attempts to portray "actual" events, I think it more likely that in order to generalize FSD to the point that it can be implemented by other manufacturers using other/new hardware sensors, Tesla would have to specify the needed parameters for a device (minimum camera resolution, FOV, etc...) as well as a method to supply a profile the devices for FSD to ingest. Once that happens a calibration process for the FSD system for that specific model implmentation....
I don’t see why any manufacturer would want that, let alone Tesla. Imagine licensing iOS for a Samsung made phone. You couldn’t use any differentiating features on the Samsung and you run the risk that something in iOS doesn’t work right. In a car, that could very well be dangerous. There is really no incentive for Tesla to license what they consider the eventual sole value of the company.
 
I don’t see why any manufacturer would want that, let alone Tesla. Imagine licensing iOS for a Samsung made phone. You couldn’t use any differentiating features on the Samsung and you run the risk that something in iOS doesn’t work right. In a car, that could very well be dangerous. There is really no incentive for Tesla to license what they consider the eventual sole value of the company.

I tend to agree. I was responding to the initial question posed by @Captkerosene:

The idea of Tesla licensing their FSD tech has been hypothesized many times here. My question is, is it practical? (I don't have the technical knowledge to have an informed opinion on the subject.) So, I'm asking the bored (SIC) whether it's feasible or not.


And the following assertion by @dhanson865 that:

They can use whatever camera supplier they want, brakes, drive train, batteries can all vary.

Just with Tesla, FSD is already working with dozens of different battery pack and dozens of different motor configurations (hundreds of combinations of those). Doesn't really matter to FSD.

It's not really that simple.