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

FSDb improved speedometer accuracy (still ~1 mph high)

This site may earn commission on affiliate links.
Here's one interesting thing I've noticed since activating the FSDb option last month:

My Tesla's speed readout used to register significantly high, like 6% or more. If the speed readout claimed 45 mph, the real speed was typically 42 mph. I noted this based first on my own judgment, then confirmed it with a couple of roadside speed-radar displays, and further verified it using one of the commonly available GPS-based Android apps. I wondered (and still do) why Tesla didn't run a background calibration routine using its own built-in GPS.

After switching to FSDb, I noticed that the car now feels faster for the same speed setting, traveling the same familiar roads. Again I broke out the GPS speedometer app and sure enough, 45 mph on the screen now typically reads 44 mph. This may not seem like a big deal, but it's actually pretty noticeable when you commute the same route every day.

The next question, of course, is why does it still indicate just a bit high, and not right on true speed? - which again, the car is perfectly capable of calibrating within the integer-mph readout resolution.

I just watched Dirty Tesla's 2020 vs 2023 FSDb comparison video*. At 10:07, he mentions the NHTSA-mandated elimination of rolling stops, but with a detail I hadn't recognized before: Even though the car on FSD does achieve a complete zero-mph stop, the speed readout shows 1 mph at that moment. He says NHTSA also mandates that the speedo should not indicate a lower speed than actual; the inevitable result of such a policy is that it's set with a slight ~1mph bias to err on the high side.

I did a quick search and couldn't find a clear reference to such a rule, but I found this webpage on the topic:
Are speedometers designed to read high?

Anyway, the bottom line is that the FSDb stack definitely improved my car's previously significant speedometer error.

* I'd recommend this video, not only for tthis arcane speedometer discussion, but generally for its perspective on the FSD beta progress since 2020. Chris also takes the last 5 minutes of this ~19 minute video to give a nice verbal summary of the often-discussed missing i.e. unimplemented basic capabilities and common foibles of the present FSDb system. This reinforces my contention that, while he's certainly enthusiastic and good-natured in his presentation, the "shill" accusation is undeserved. He doesn't shy away from honest assessment of the problems.​
48629702898_644d91ba51_b.jpg
Admin note: Image added for Blog Feed thumbnail
 
  • Like
Reactions: n.one.one
I've also noticed this change, though it never occured to me that it might be associated with FSDb. When I first got my MYLR, the speed reported by the car was up to 3-4 MPH higher than the speed reported by my radar detector (which uses GPS). More recently, the difference has decreased to 1 MPH or less. I'm not sure whether it is because of FSDb or a general correction in Tesla's software. I think the only way to determine that would be to get some reports from Tesla owners who are not in the beta, but those owners are not likely to be reading this forum :)

Note that speed measurement by the car is dependent on an accurate measurement of the tire circumference, which can change because of tire wear, tire pressure, and other factors. OEM vs replacement tires of another brand could also change the calculation.

Tesla is in a unique position to be able to get this right, though. Every Tesla has a GPS system for navigation. It should be an easy matter to use that system to calculate speed just as other devices do it. I'm not suggesting that Tesla should report GPS-based speed as its speedometer., But it could easily maintain a long-term comparison of measured speed vs GPS-calculated speed and use this data to adjust the actual speedometer measurement. Effectively it would be using the GPS data to periodically recalculate the tire circumference.
 
  • Like
Reactions: JHCCAZ
Every Tesla has a GPS system for navigation. It should be an easy matter to use that system to calculate speed just as other devices do it. I'm not suggesting that Tesla should report GPS-based speed as its speedometer., But it could easily maintain a long-term comparison of measured speed vs GPS-calculated speed and use this data to adjust the actual speedometer measurement. Effectively it would be using the GPS data to periodically recalculate the tire circumference.
Agreed. That's exactly what I meant in referring to speedometer GPS calibration.

And yes it's possible that the timing of the fix is a coincidence, but my guess is that it's due to the FSDb stack. I had the same 2022.44 branch for a while before requesting FSDb, and that's what I noticed the change. Also, I don't believe there's been any mention of this issue in the general branch release notes.
 
Interesting stuff. I wonder if they discovered that the error in the speedometer was causing the car to plan to to drive a specific velocity profile for FSD, but the execution was off (by however much the speedometer is off). That could cause some issues in tight situations, close merges, shooting-the-gap UPLs, over or under braking for cars crossing in-front etc. On production autopilot it's probably significantly less important to drive a precise velocity profile. Wild speculation on the reasoning here, of course.