Tesla "logs" vs reality... potential problem?

[mongo]

Certainly reasonable... but the whole point here is questioning if it is factual.

Which will never be stated on the forum due to NDA of anyone who could actually answer the questions definitively.

With that known, the best possibility for information exchange this venue can offer is to examine the data available and see where the inconsistencies may be.

My interpretation was that the set of circumstances brought into question logging accuracy. So I tried to add into the discussion with what may be logged, what may be processed, and why the different points of view may be able to coexist. Along with the behavior of TPMS systems.

If I have misunderstood, please elaborate, as these things intrigue me.

 

[bcsteeve]

Since the details can't be known, as you correctly say, I don't see any point in guessing at them. I think the bigger picture is far more important - that there must be some skepticism with regard to Tesla's statements of logs as proof of this or that.

 

[mongo]

Since the details can't be known, as you correctly say, I don't see any point in guessing at them. I think the bigger picture is far more important - that there must be some skepticism with regard to Tesla's statements of logs as proof of this or that.

Indeed, Garbage In, Garbage Out.

But I do think it is nice to be able to wrap a plausible, if unproven, scenario around why soemthing is doing what it is.

 

[luckyj]

Understand that you don't prove a negative by saying there's no log of it. All that proves is...there's no log of it. Some events or types of events are just not logged. Don't misinterpret that as "it didn't happen". If someone says as much, they're falling into the same trap. Also understand there will ALWAYS be a disconnect between logs and reality. Logs can't record everything, so there's always something missing from the picture.

If Tesla asserted something didn't happen because there was no log of it, sure...be skeptical. But sometimes Tesla has said that logs indicate a positive assertion of something completely opposite of what the driver said. In those cases, I'd be more inclined to believe the logs.

 

[bcsteeve]

Why? Me pressing the accelerator pedal and the car's system glitch "pressing" the pedal (as it does with cruise control, etc) is going to result in the same "positive" logs, no? Combined with the absence of a report on a reason for it to do so - a logical trap as you say - and they really can't have any idea what happened.

One can assume the driver is embarrassed or doesn't want their insurance to go up so they try to falsely blame the car. But conversely, one can assume that Tesla has far more to lose (public trust, consumer confidence, stock valuation) and they're in a perfect position to manufacture "logs" to say whatever they want. Why do we choose to believe them and not the person that was in the car? In my case, I think its clear I have no reason to lie. Nothing to gain, nothing to lose, etc... because nothing really happened other than evidence of a disconnect between reality and their knowledge of reality. They didn't tell me, "oh sorry, we don't have logs for that" they told me the logs indicated there was not a problem (which isn't the same as not indicating that there is a problem). Careless wording by a low level rep? Perhaps. Or a systemic corporate culture where they want to reinforce that their product is perfect?

Anyway, maybe its not a big deal. Maybe they have different levels of logs that are unlocked by different levels of reps that are triggered by different levels of severity/PR concerns. I just know for me - personally - I have to take whatever they say with a grain of salt because to me, they've demonstrated that they're willing to fudge and/or manufacture facts to avoid scrutiny.

 

[Skotty]

Logs won't record everything. Tesla is saying they have no evidence of any problem in the logs. And a lack of evidence is not evidence of absence. They simply can't confirm what you are saying from the log information they have available. This is most likely different from the unintended acceleration issues where they apparently have actual evidence of throttle position. That's not to say that log information is infallible, but as a programmer, it's a very good bet that if a log and a user disagree, the log was right and the user was wrong. Of course, trying to prove it with more certainty requires additional evidence to further the case of either the logs or the user. Such evidence may or may not be obtainable.

 

[mongo]

OK, you just said:

I have to take whatever they say with a grain of salt because to me, they've demonstrated that they're willing to fudge and/or manufacture facts to avoid scrutiny.

I'm going to step through things, please correct me if I get something wrong.

From your original post:

What was odd is that all four tires were indicating 42psi on the screen, but the two front ones were in orange and the alert at the bottom was on constantly.

All the tires were reading normally at that point in time.

The representative assured me there was nothing wrong and nothing in the logs to indicate a problem and there was no indication from the car's tire pressure sensors that anything was wrong.

Tesla said all the tires were reading normally.
So at that point in time, data was indicating things were hunky dorey.

The issue you were having was that the display showed two orange tires, the reason for which you provided:

Perhaps I should also note that the warning was at one time correct. A couple of days ago the temperature dropped significantly and my pressures were 38 psi in the front and 39 in the back. The two front ones were complaining. So I figured 38 was the threshold. I topped up to 42 all round and it didn't clear for 2 days so that's why I called them this morning.

So the front tires dropped low enough to trigger a warning, and the rears didn't. The warning did not clear until you filled the tires and drove the car.

While driving around the block, about 4 minutes later that same right front one went normal and then a minute later so did the front left and the warning went away.

If the car does not log while off, and you did not drive it when the pressure was below 40, there was no way for Tesla to see that low pressure. (unless it also logged derived errors). At the time you called the values were in a decent range, and since they were good then, they may not have looked back in time to see if they had been low previously. Where is the fudge?

 

[bcsteeve]

@Skotty

I'm sorry, but you weren't on my conversation and you really have no way of saying what "Tesla is saying". As I just stated above, she didn't say there wasn't a log. She said that the logs show that there isn't a problem. To me, that says my car's TPMS system is actively sending logs with some sort of "ok" status. Your assumptions (but that's all they are) may be correct and she may have mispoke, but just as my point about the driver that crashed into the building... the only thing we have to go on are the facts at hand, not made up logs and certainly not guesses that logs don't exist despite what she said.

 

[bcsteeve]

and since they were good then, they may not have looked back in time to see if they had been low previously. Where is the fudge?

she told me straight out that there is no current alert about low pressure, nor had there been for the past couple of days.

 

[bcsteeve]

If the car does not log while off, and you did not drive it when the pressure was below 40, there was no way for Tesla to see that low pressure

And once again, that's entirely your assumptions/imaginations/whatever. You have no idea when/what it does or doesn't log. From what she told me, it *sounded* like it is continually sending all information, not just abnormal ones. Again, perhaps she misspoke. I can't know that.

 

[mongo]

Ah, missed that line. Thank you for pointing it out.

 

[MarcusMaximus]

@Skotty

I'm sorry, but you weren't on my conversation and you really have no way of saying what "Tesla is saying". As I just stated above, she didn't say there wasn't a log. She said that the logs show that there isn't a problem. To me, that says my car's TPMS system is actively sending logs with some sort of "ok" status. Your assumptions (but that's all they are) may be correct and she may have mispoke, but just as my point about the driver that crashed into the building... the only thing we have to go on are the facts at hand, not made up logs and certainly not guesses that logs don't exist despite what she said.

I can kind of see where you’re going with this, but it doesn’t really follow. She said the logs showed there wasn’t a problem. From the info you gave, those logs were correct. All tires were properly inflated, and that was reflected correctly in the logs.

The only piece of information here that’s inaccurate is the warning that appeared on the dash. If anything, this all backs up the accuracy of the logs vs anything else.

 

[Skotty]

@Skotty

As I just stated above, she didn't say there wasn't a log. She said that the logs show that there isn't a problem. To me, that says my car's TPMS system is actively sending logs with some sort of "ok" status.

Here is part of what I'm trying to say. While perhaps misleading, if there isn't a log, it would still be mostly accurate to say the logs show that there isn't a problem. (a more precise response would be "the logs don't show that there is a problem"). Some speculation here, but Tesla probably has 2 resources to look at -- actual historical logs, and present data readout. Chances are, the car didn't log anything about the tires, and the present read out showed pressure ok. In response to that, the lady said what she said. The remaining question would then be why did the display still show the orange tires, and there is a reason for that, and that reason may or may not indicate a bug in the display software. Honestly, at that point, you need a developer involved to explain what the software display is doing. It may very well be operating as intended, if temporarily inconsistent with tire pressure read outs. Or not. Hard to say.

To add to this, it may even be slightly more convoluted. It's possible it logged low pressure, then logged ok again. The software the phone lady was using may have simply ignored those logs because they went back to ok. The lady may or may not be aware of this depending on how her software works.

Here's another nugget that plays into this. The tire pressure thing, talking to the lady on the phone at Tesla, is really a different situation than the unintended acceleration incident. When you have an actual serious legal matter, a different scenario plays out. Actual developers get involved (and possibly engineers too, if information on the hardware is needed), and they skip past any front end software and look directly at data to try to determine what exactly happened for managers and the legal department.

 

[mongo]

Why? Me pressing the accelerator pedal and the car's system glitch "pressing" the pedal (as it does with cruise control, etc) is going to result in the same "positive" logs, no?

No.
Or at least it depends on what is logged.
If they log the raw analog inputs, then they know what the pedal was most likely doing.
If they log the output of the function that maps the analog inputs to a pedal position, they sort of know what the pedal is doing.
If they only log the power command to the drive unit, they know very little about anything.

It the old days, cruise control would move the pedal. For these cars, that is not the case, so from a SW POV, it's totally separate.
DTC checks require monitoring and comparing both analog signals, so I would expect that is the level it gets logged at.

Question for my own curiosity, how long were the tires below 39 psi before you topped them off? Some things may be logged less often than others. I've done embedded systems that logged all the sensors every control cycle. From that, you could replay any event to find SW level issues or sensor faults.

 

[bcsteeve]

A little off-topic:

I've done extensive work on drive-by-wire cruise control systems. I've never seen any system where there is any "monitoring and comparing both analog signals". The TPS consists of two potentiometers and not much more. One fails and the other is used. The idea being that it is unlikely for both to fail during the life of the car. But they do. Like many things, it comes down to reasonable costs vs expected outcomes. I developed an aftermarket cruise system for a Mercedes product, and we simply intercepted the primary signal and output a modified one. It would be completely different (other than at zero) from the secondary. The car was perfectly happy with this, triggering no TPS fault. So I can say with certainty that there was no monitoring and comparing of signals.

Does Tesla? I can't say. Though I kind of doubt it. Perhaps I'll take my pedal apart and see. Doing exactly what I did with the MB kit would prove that one way or the other. However, if it does throw a fault I'm dis-inclined to do this because I have no way of clearing it myself (unless someone's documented that for Tesla?).

If the systems are anything like Mercedes, Renault or Toyota systems that I have experience with then there simply *is no way* to say for sure that the guy pressed the pedal physically vs. a signal generated internally (failed sensor or computer glitch). And that's if their logging system was as robust and continuous as they seem to say, which I have now experienced - I believe - it is not.

Beyond that, Tesla's argument that the pedal was pressed to 100% and the car was just doing what the driver wanted... is complete nonsense. I've sent a few messages to them about this, but of course it gets ignored. They're opening themselves up for legal action even if the guy did press the pedal. The car can detect the wall. Everyone would agree that hitting the wall is a bad outcome. Press the pedal or no, the car should not launch into the wall... ever. If the car is parked and an object is in front of it and a person hits the accelerator pedal, the assumption should be that they've made a mistake. Lives and property saved. Someone somewhere is going to argue that if the car has the ability to prevent the crash, it should. If it doesn't, it is liable.

 

[bcsteeve]

Question for my own curiosity, how long were the tires below 39 psi before you topped them off?.

I didn't notice the warning until I left the garage. I'm not certain if it was on when I got in the car, or if it took moving before they faulted. But at that point, I was late and I saw it wasn't *that* low, so I did my driving for the day and topped up when I got home. I'm going to guess that we're talking 4 to 6 individual trips for a total of 20 to 30 kilometers over the course of about 4 hours.

 

[MarcusMaximus]

The car can detect the wall. Everyone would agree that hitting the wall is a bad outcome.

IF Tesla can guarantee, 100%, that it is, in fact, a wall that’s being detected. If there’s anything wrong with the sensors causing them to think a collision is imminent when the driver is actually, say, on an open freeway, then slamming on the brakes and stopping the car is a VERY bad outcome, likely resulting in death, with Tesla fully at fault.

That’s why Tesla’s design here is to trust the driver. The car will try to mitigate an imminent collision to the extent it can safely do so, but if the driver REALLY tells it to keep going, it has to do so.

 

[mongo]

I didn't notice the warning until I left the garage. I'm not certain if it was on when I got in the car, or if it took moving before they faulted. But at that point, I was late and I saw it wasn't *that* low, so I did my driving for the day and topped up when I got home. I'm going to guess that we're talking 4 to 6 individual trips for a total of 20 to 30 kilometers over the course of about 4 hours.

Definitely seems long enough to log.

When you say TPS are you referring to the Throttle Potion Sensor on the throttle body? Did those have stepper motor drive? (antoehr layer of redundancy)
For accelerator pedal on drive by wire, there are two potentiometers, one increases and one decreases as the throttle is pressed (can also have different supply voltages). If they don't track, the ECU throws a DTC (P2135, P2138 are related), and may go into a limp mode (had to replace wife's when it went bad).
Electronic Throttle: Where's The Cable?

I expect Tesla's system is different than the OEM's mentioned, since there is no throttle body or external cruise control. It can log the accelerator position and log the cruise control command separately. Especially following the Toyota unintended acceleration issue.

The "car shouldn't let me ram things" has been discussed on other threads. I'm of the opinion that the driver should be the final say on going and stopping. Too many edge cases of reason why ramming something was the better option to leave it to a binary GO/NO-GO decision. Is it a solid wall?, is it a snow drift?

 

[bcsteeve]

IF Tesla can guarantee, 100%, that it is, in fact, a wall that’s being detected. If there’s anything wrong with the sensors causing them to think a collision is imminent when the driver is actually, say, on an open freeway, then slamming on the brakes and stopping the car is a VERY bad outcome, likely resulting in death, with Tesla fully at fault.

That’s why Tesla’s design here is to trust the driver. The car will try to mitigate an imminent collision to the extent it can safely do so, but if the driver REALLY tells it to keep going, it has to do so.

Not at all true. I'm [mostly] with you on the freeway scenario, but that's not what we're talking about here. There's NO CASE anyone can conceivably come up with, I'm sure, where sitting parked in front of a wall warrants the full use of an accelerator. In fact, it shouldn't matter parked or otherwise... 100% throttle should be FIRST read as a fault/mistake and only accepted IF sensors show all clear. Even on the freeway, if you're cruising along and the sensors detect an object in front of you and your input is to go 100% throttle, the assumption should be driver error or sensor fault, not "okie dokey"

 

[bcsteeve]

Too many edge cases of reason why ramming something was the better option to leave it to a binary GO/NO-GO decision. Is it a solid wall?, is it a snow drift?

Doesn't matter.. why in either case would there ever be a reason to ram it at full throttle? I agree that the driver should be in control, when it is reasonable to assume the driver is, in fact, in control. Sudden 100% throttle is not a reason to assume that. I'd say the ONLY time anyone ever plants their foot (in a Tesla) full to the floor is when they are launching. Even passing there's no reason to put your pedal to the metal. In those cases, it is perfectly sane for the car to be given first reading and only allow it if the sensors show clear ahead. That's just plain common sense stuff. What "edge case" could you possibly come up with otherwise?