Model 3

Temps, Precondition, Regen data with GRAPHS

2456710

Comments

  • edited November 2019
    @Bighorn: you are referencing Tesla's statement right?
  • edited November 2019
    I was referencing the poster who said battery preconditioning does less than driving to heat up the battery in another thread. She may have referenced a Tesla employee, I forget.
  • edited November 2019
    Yes, ok just wanted to clarify.
  • edited November 2019
    Bumping because of a few threads and posts that would be served by seeing this thread.
  • edited November -1
    Added new Full vs limited Regen overlay graph to the OP, 70-0 full regen, 60-0mph limited Regen
  • edited November 2019
    Added Heater power vs Time to OP
  • edited November 2019
    Whoah derotam!

    What will FISHEV do now confronted with actual data versus his terribly wrong statements and equally wrong Maths?
  • edited November -1
    First regen 70-0 shows multiple modes the car is going through. Almost looks temporal, around 4-4.5 seconds. Note the periodic velocity changes. I don't think they are perceivable by driver. But different slopes on power curve for each mode.

    Wondering why heater power has significant noise until something happens and noise disappears.
  • edited November 2019
    @jallred: Yeah so I don't necessarily believe that those really exist. Probably something with the recorded data. I do have to pull out occasional errors, like a speed reading that shows I am going 287 miles an hour.

    I started to investigate those but I wanted to just get the full graph done. If they keep showing up I will investigate further.

    With the heater data, yeah, again I am not going to really try and analyze that right now...I think it would be a very low return on investment of my time.
  • edited November 2019
    @jallred: your killing me... so on that 70-0 vs time, the Torque values show the same bumps... anyway....moving on for now, haha.
  • edited November 2019
    Wow. How refreshing! Turns out that Trolls are allergic to fact based commentary and discussion. Whodathunkit? Excellent!
  • edited November 2019
    @derotam - Interestingly, your graphs look very similar to the ones that come out of TeslaFi if you click into the drive data:
    https://imgur.com/a/Z4Hh069
  • edited November 2019
    @hokiegirl: Yes, there is a lot of metrics that you can actually get using the API. I have never used TeslaFi and don't want to pay a subscription. Even though TeslaFi is always recording, that is A LOT of data to sift through to get what you want to look at. I don't need everything record. Just the applicable metrics for what I want to look at.
  • edited November 2019
    They are definitely real. The show up in each parameter.

    For cabin heat, I'm going to make a guess that they PWM the resistive heat. Makes power draw happen gradually. But maybe most important if they slowly heat up that heater they maybe can prevent it from making pinging noises as the metal expands. I heard that last winter but haven't heard it yet this fall.

    Not sure why I'm killing you. As in funny, annoying? Just making observations that hoping someone can find useful.
  • edited November -1
    ...for making me go back and look at those anomalies. haha.
  • edited November 2019
    I don't think it is PWM, the ramp is still "steep" and I would like to know how that would reduce metal expansion noise. It looks like random noise, not a purposefully generated signal.
  • edited November 2019
    @derotam - I'm lazy but like numbers, so for me, the subscription is worth it. :) I don't have the patience to do what you are...but I do like to look at my various trends.
  • edited November 2019
    Ramp isn't steep. Goes from around 0 to 6kW in about 30 seconds.

    You don't think that heating the metal slow could reduce expansion noise? You might be right, but I feel like the faster it is heated the more noise it will make.

    Assuming that the heater is the dominant draw on the power (might not be), I would think that a current controlled heating element would ramp with less noise.

    With PWM there would be high frequency alterations to the power and if the sampling rate is too low to see them they can start to look like noise in the time domain. Especially if there is no clock synchronization between PWM and data recording.

    There is probably a ton of sampling jitter in CANbus data, so under sampled phenomena are definitely going to look noisy.

    A ramped current level wouldn't be under sampled and wouldn't look as noisy.
  • edited November 2019
    Sorry, starting with 3rd paragraph noise is sampled data noise, not audible noise.
  • edited November 2019
    @hokiegir1: Yeah, I am mostly lazy because I don't have a lot of experience manipulating large data sets. My wanting the flexibility in manipulating the data overrode Tesla'Fi's subscription.

    l use Excel for the most part and I don't code so I'm not using VB code, just Excel built in functions. I have learned a lot in the last few days by using my friend Google.
  • edited November 2019
    @derotam - I use excel almost exclusively at work (pivot tables and linked sheet lookup formulas are my friends), so if you get stuck and google isn't helpful, let me know. :) I generally work with financial data, but some of the methods to identify trends would be the same regardless of the source.
  • edited November 2019
    In case anyone is interested in what the raw data looks like...Below is Time and Battery Power. Time is in milliseconds.

    32102 5.981508
    32102
    32103 6.174189
    32103
    32103
    32115 6.13552
    32115
    32115 6.173545
    32125
    32133
    32133 6.36444
    32142
    32142 6.17274
  • edited November 2019
    @hokiegir1: I'll keep that in mind. I am also looking at a program called UDPLogger(windows) that graphs everything right as you load the raw data. It has a simple interface for selecting/deselecting the fields to show. The good thing is that it automatically normalizes the time scale. This issue with UDPLogger is that I cant change what is on the x-axis from in the program.

    So my biggest issue right now with Excel is how do you make a graph with whatever data on the X axis but have the scale be linear between the low and high value in the raw data? Right now I am adding in the missing time values to make that linear.
  • edited November -1
    Ah. So the sampling jitter is enormous. That's good to know, we have to be sure that we don't look for trends in data that we are under sampling.

    Why do you get samples that have a time of the sample but no data? And the time of sample doesn't have sufficient resolution as seen by multiple samples with the same time.

    For plots to look at by human eye the only issue will be that sampling jitter will appear as noisy (y-axis) data.

    Thanks for posting all of this and taking the time to acquire the data. This is like the best thing ever.
  • edited November 2019
    @jallred: There is other data in that set, they were just not relevant, like Battery Temp, coolant flow rates...etc. First think I usually do is compress the data in excel so that anything with the same time stamp gets put in one Row.
Sign In or Register to comment.