Please be careful with ones accusations of unintended acceleration. I worked for VW/Audi in the US in Purchasing in the late 70's early 80's and saw what effect these unfounded accusations can do to a company. 60 minutes TV show did a hatchet job on Audi and devastated the company for over 10 years. They nearly didn't make it back. This is 99.9% always driver error. The computers in the car don't lie. I don't think any of us want TESLA to go under because of rumors and false accusations.
I feel like those that know how to drive a stick shift car, that are then driving a Tesla with creep off, are not going to have this problem.
I came from a 6-speed manual car immediately prior to my Tesla, and I keep creep off because it’s what I’m used to. Granted it’s closer to an SMG transmission than a true manual because you lack the clutch which gives added control in low speed situations, but it’s sinilar enough I feel like I’m in a native environment.
I can definitely see someone that’s only ever driven automatics getting into a Tesla with creep off and totally screwing up
I came from a 6-speed manual car immediately prior to my Tesla, and I keep creep off because it’s what I’m used to. Granted it’s closer to an SMG transmission than a true manual because you lack the clutch which gives added control in low speed situations, but it’s sinilar enough I feel like I’m in a native environment.
I can definitely see someone that’s only ever driven automatics getting into a Tesla with creep off and totally screwing up
I have driven both manual and auto for many years now and move seamlessly between them.
When you shift from manual to auto you tend to need to keep your left foot out of the way to avoid the urge to hit the clutch, but I'm wondering if in fact with Tesla, and creep turned off, it would be better to just continue with two foot driving...
When you shift from manual to auto you tend to need to keep your left foot out of the way to avoid the urge to hit the clutch, but I'm wondering if in fact with Tesla, and creep turned off, it would be better to just continue with two foot driving...
Wifey nailed our garage door with her Camry. She backed out of the drive, daughter ran out and yelled she wanted to go, Wifey pulled back into the driveway, daughter got in, Wifey looked back and went forward, startled she "hit the brakes" and slammed into the garage door. No amount of logic and eyewitness account by daughter would convince her she hit the wrong pedal. $700 for garage door and all is forgotten.
Unintended acceleration does happen. But the only time it has been the vehicle's fault that I can recall was BMW and their idle setting on stick shifts. They would go all the way to full throttle to keep the engine from dying, not good in a panic stop when someone forgets to put in the clutch. I had one of these and tried it out there was a lurch at about 10 MPH as the brakes fought the engine. This flaw was short lived on one model in 2006.
Unintended acceleration does happen. But the only time it has been the vehicle's fault that I can recall was BMW and their idle setting on stick shifts. They would go all the way to full throttle to keep the engine from dying, not good in a panic stop when someone forgets to put in the clutch. I had one of these and tried it out there was a lurch at about 10 MPH as the brakes fought the engine. This flaw was short lived on one model in 2006.
Have any of you read this study?
https://www.autosafety.org/sites/default/files/1989 NHTSA SA Study Report & Appendices A-D(1).pdf
Yes, I know it is old. Read the factors on Page X. I think one of them still applies to Tesla vehicles, and do not see a reasonable way to mitigate them without the possibility of introducing artificial intelligence failures in place of the current behaviors. We have already found some possible faults with such a system just in this thread.
https://www.autosafety.org/sites/default/files/1989 NHTSA SA Study Report & Appendices A-D(1).pdf
Yes, I know it is old. Read the factors on Page X. I think one of them still applies to Tesla vehicles, and do not see a reasonable way to mitigate them without the possibility of introducing artificial intelligence failures in place of the current behaviors. We have already found some possible faults with such a system just in this thread.
Jeep also had a problem and it was tied to the computer compensating for a carboned or misadjusted EGR pintle in the throttle body that affected air intake. It could easily rev to 2000 + RPM and if shifted into gear would suddenly lurch forward and the brakes had to be stood on to bring it to a stop. It would sometimes do it and sometimes not but more often after the vehicle was warmed up and the vehicle was shut off then restarted. People were definitely run over and accidents occurred and attributed to the driver. Yes, they shouldn't have put it into gear at that high of revs but drivers didn't know or pay attention. However, the drivers vehicle shouldn't have had that problem in the first place. Jeep never admitted to anything being wrong on their end.
Last edited:
There is a class action lawsuit underway for "spontaneous acceleration." Tesla filed a motion to dismiss but the court denied all motions and the case is going into discovery at the end of the summer. You can find the case and the lawyer on the internet if you haven't already. Far too many cases of this, especially when the car is "creeping" that Tesla always denies. Otherwise the safest car on the highway. A big otherwise.
bonnie
I play a nice person on twitter.
I hope Tesla can recover costs for defending a frivolous suit.
Perhaps the AutoPilot knows when you are parking, and decides to embarrass the driver? This could be something. Elon did say something about machines taking over eventually.
I, for one, welcome our new automotive overlords.
Okay it happened to me too.
This morning wee hours very little to no traffic. I was stopped at a red light with no one in front of me.
And then it happened.
All I could remember was the light turning green. The car lurched suddenly and before I knew I was past 50mph. Looking in the rear mirror it appeared all the cars behind me were racing backwards.
“But Officer, I swear I didn’t touch the accelerator.”
Can I join the Class action ?
This morning wee hours very little to no traffic. I was stopped at a red light with no one in front of me.
And then it happened.
All I could remember was the light turning green. The car lurched suddenly and before I knew I was past 50mph. Looking in the rear mirror it appeared all the cars behind me were racing backwards.
“But Officer, I swear I didn’t touch the accelerator.”
Can I join the Class action ?
Last edited:
Are you sure that this isn't the one that was settled before it became a class action? If it is a different one can you point us to the details?
As a result of the previous discussion, I have come up with a possible explanation for sudden unintended acceleration in Tesla vehicles that also explains how the accelerator pedal sensors can have large outputs without the driver pressing on the accelerator pedal. You can read this explanation by going to Dr. Ronald A. Belt’s Sudden Acceleration Papers - Center for Autosafety and reading the paper entitled “Tesla’s Sudden Acceleration Log Data – What it Shows – 5/1/18”. This explanation provides a testable theory of sudden unintended acceleration in all Tesla Vehicles.
Tough read, but it makes sense. Unfortunately I still have questions.
Could the 0.1 sec exact timing be the sampling rate of the data recording allowing a human 0.05 sec variation to be masked? I notice the data points do not ramp supporting the low data recording rate idea.
Temperature is cited as a factor, more data is needed as to when "unintended acceleration" happens. If temperature is not a factor due to car being just started or in cold weather this analysis is in question.
Could the 0.1 sec exact timing be the sampling rate of the data recording allowing a human 0.05 sec variation to be masked? I notice the data points do not ramp supporting the low data recording rate idea.
Temperature is cited as a factor, more data is needed as to when "unintended acceleration" happens. If temperature is not a factor due to car being just started or in cold weather this analysis is in question.
[The Duke said: "Tough read, but it makes sense. Unfortunately I still have questions.
Could the 0.1 sec exact timing be the sampling rate of the data recording allowing a human 0.05 sec variation to be masked? I notice the data points do not ramp supporting the low data recording rate idea.
Temperature is cited as a factor, more data is needed as to when "unintended acceleration" happens. If temperature is not a factor due to car being just started or in cold weather this analysis is in question".][/QUOTE]
I don't completely understand your question about a 0.1 sec sampling rate allowing a human 0.05 sec variation to be masked. I was originally curious about the 0.1 second accuracy of the time durations in the figure, and this curiosity led me to a possible explanation for not only these time durations (i.e., a common 1.0 sec recovery time), but also the non-zero values of the sensor outputs. After coming up with an explanation for the non-zero values of the sensor outputs (i.e., that they were caused by a ground drop in the two sensor ground lines caused by some leakage current), it occurred to me that my explanation for the non-zero sensor outputs is independent of the fact that the sensor outputs have peaks and valleys with rise times accurate to 0.1 sec. In other words, the same explanation applies even if no peaks and valleys are present at all, as long as the sensor outputs become non-zero without the driver stepping on the accelerator pedal.
Your comment that if (motor) temperature is not a factor, then this analysis is in question is absolutely correct. Without a high (motor) temperature, there is no speed sensor leakage current and therefore, no voltage drop in the accelerator pedal sensor ground lines to cause a higher sensor output. It is interesting, however, that Tesla uses several temperature sensors to monitor the motor stator, and several printed circuit board functions inside the inverter housing.
I am still trying to understand how Tesla engineers can pinpoint transitions in the pedal sensor log data with an accuracy of 0.1 sec. The log data that Tesla has provided to NHTSA shows that the accelerator pedal sensor is sampled at a data rate of only 1 Hz in the logs downloaded from a vehicle. This log data is obtained from an SD card in the Gateway CPU between the Ethernet and the CAN buses. The accelerator pedal sensors are sampled at a much faster 100 Hz rate data rate on the CAN bus, however. So it should be possible to see fast changes in the depression of the accelerator pedal if one is able to get this high speed CAN bus data into the log data being read from the vehicle. How this is done is still a mystery to me. When writing my paper, I assumed that the analog output of one of the pedal sensors was converted into a digital form by amplifying it using a Schmitt trigger, after which the digital data was used to start and stop a clock,giving very accurate time durations. But I am also aware that some researchers believe that this same accuracy can be obtained solely from the pedal sensor amplitude data.
Could the 0.1 sec exact timing be the sampling rate of the data recording allowing a human 0.05 sec variation to be masked? I notice the data points do not ramp supporting the low data recording rate idea.
Temperature is cited as a factor, more data is needed as to when "unintended acceleration" happens. If temperature is not a factor due to car being just started or in cold weather this analysis is in question".][/QUOTE]
I don't completely understand your question about a 0.1 sec sampling rate allowing a human 0.05 sec variation to be masked. I was originally curious about the 0.1 second accuracy of the time durations in the figure, and this curiosity led me to a possible explanation for not only these time durations (i.e., a common 1.0 sec recovery time), but also the non-zero values of the sensor outputs. After coming up with an explanation for the non-zero values of the sensor outputs (i.e., that they were caused by a ground drop in the two sensor ground lines caused by some leakage current), it occurred to me that my explanation for the non-zero sensor outputs is independent of the fact that the sensor outputs have peaks and valleys with rise times accurate to 0.1 sec. In other words, the same explanation applies even if no peaks and valleys are present at all, as long as the sensor outputs become non-zero without the driver stepping on the accelerator pedal.
Your comment that if (motor) temperature is not a factor, then this analysis is in question is absolutely correct. Without a high (motor) temperature, there is no speed sensor leakage current and therefore, no voltage drop in the accelerator pedal sensor ground lines to cause a higher sensor output. It is interesting, however, that Tesla uses several temperature sensors to monitor the motor stator, and several printed circuit board functions inside the inverter housing.
I am still trying to understand how Tesla engineers can pinpoint transitions in the pedal sensor log data with an accuracy of 0.1 sec. The log data that Tesla has provided to NHTSA shows that the accelerator pedal sensor is sampled at a data rate of only 1 Hz in the logs downloaded from a vehicle. This log data is obtained from an SD card in the Gateway CPU between the Ethernet and the CAN buses. The accelerator pedal sensors are sampled at a much faster 100 Hz rate data rate on the CAN bus, however. So it should be possible to see fast changes in the depression of the accelerator pedal if one is able to get this high speed CAN bus data into the log data being read from the vehicle. How this is done is still a mystery to me. When writing my paper, I assumed that the analog output of one of the pedal sensors was converted into a digital form by amplifying it using a Schmitt trigger, after which the digital data was used to start and stop a clock,giving very accurate time durations. But I am also aware that some researchers believe that this same accuracy can be obtained solely from the pedal sensor amplitude data.
Dr. J
Active Member
This is clearly over my head, so quite likely this is a stupid observation. I sense a contradiction (all quotations from page 2).
On the one hand: "But this does not explain the zero-amplitude periods of one second duration between the periods of higher amplitude. These zero-amplitude periods were explained by Tesla engineers as the driver alternately pressing on the accelerator pedal and releasing the accelerator pedal; i.e., “stabbing” at the accelerator pedal. Note that the time periods of the peaks and valleys are accurate to 0.1 second. There is no way that a human driver can produce four time periods of exactly one second duration accurate to 0.1 second by alternately pressing and releasing the accelerator pedal."
On the other hand, the chart says: "Amplitudes are logged once per second." If amplitudes are logged once per second, doesn't this by definition explain why the chart shows the exactly one second duration of amplitudes? I don't get the relationship between the 0.1 second something-or-other and the 1 second logging of amplitude.
Ok, I'll give a stab at this,
1. It would be highly unusual to use the exact same 5V regulator for a analog sensor as your main 5V, if for nothing else then noise immunity.
2. Even if the 5V regulator dropped out, there should be voltage monitors which would have flagged this.
3. In addition, even if the voltage to the pedal sensor drooped/dropped wouldn't the redundant and opposing sensors either continue to work, or flag a fault as not matching?
4. Wouldn't you expect to see strange log data after an impact with a solid object, such as accelerator being stabbed at (as if it was the brake) but with 0 speed because the car has no velocity
5. And finally, if we are discussing a normal parking operation, why was the brake not pressed prior to any of this happening, as would be expected during a parking operation, rather only 5 seconds after impact. If one's foot were covering the brake, if there was an impact, that impact would tend to press ones foot down onto the brake in a sudden stabbing motion.
This seems like quite a lot of work. Since you are willing to put so much effort into this, I'm shocked that you haven't taken WK up on his offer to pull the low level logs from your car. I'd even bet that you could talk him into giving you a drive unit circuit board which you could check the power regulation on.
Peter
1. It would be highly unusual to use the exact same 5V regulator for a analog sensor as your main 5V, if for nothing else then noise immunity.
2. Even if the 5V regulator dropped out, there should be voltage monitors which would have flagged this.
3. In addition, even if the voltage to the pedal sensor drooped/dropped wouldn't the redundant and opposing sensors either continue to work, or flag a fault as not matching?
4. Wouldn't you expect to see strange log data after an impact with a solid object, such as accelerator being stabbed at (as if it was the brake) but with 0 speed because the car has no velocity
5. And finally, if we are discussing a normal parking operation, why was the brake not pressed prior to any of this happening, as would be expected during a parking operation, rather only 5 seconds after impact. If one's foot were covering the brake, if there was an impact, that impact would tend to press ones foot down onto the brake in a sudden stabbing motion.
This seems like quite a lot of work. Since you are willing to put so much effort into this, I'm shocked that you haven't taken WK up on his offer to pull the low level logs from your car. I'd even bet that you could talk him into giving you a drive unit circuit board which you could check the power regulation on.
Peter
No, because the four “peaks” have durations of 1.3 sec, 1.0 sec, 1.0 sec, and 1.5 sec – durations which the Tesla engineer was very specific about. So, if the amplitude sample times determine the peak durations, then peak durations of 1.3 and 1.5 sec would be impossible unless one assumes that the amplitude sampling times vary somehow. This would be very unusual and difficult to design. The only way to explain peak durations of exactly 1.3 and 1.5 sec is to assume that the amplitude samples are taken at a fixed 1 sec rate, but that the transitions occur at different times than the amplitude samples are taken, with the transition times measured to an accuracy of 0.1 sec and the amplitude samples being taken in between the transitions at times not tied to the transitions. This assumption also explains how the last two peaks can be detected without having associated amplitude samples. It is likely that these two amplitude samples really were taken, but they were not given to the driver by Tesla because they were unusual in some respect. Maybe they showed an amplitude greater than 100%. (I know that greater than 100% sounds impossible, but wk057 has shown in his document entitled “Tesla Model S CAN Bus Deciphering” that the range of values for the pedal position sensors on the CAN bus goes from 0% to 102%. This may be to allow some tolerance for the variation in maximum pedal position from one vehicle to another).
As I mentioned above, I am still trying to understand how Tesla engineers can pinpoint transitions in the pedal sensor log data with an accuracy of 0.1 sec. I tried to explain this by assuming that the analog output of one of the pedal sensors is converted into a digital form by amplifying it using a Schmitt trigger, after which the digital data is used to start and stop a clock, giving very accurate time for transitions in the accelerator pedal sensor data, in a simple manner similar to the brake data. If someone else can provide a better explanation for this 0.1 sec accuracy, then I would welcome their explanation.
Here are my answers to each of your questions above:
1. It would be highly unusual to use the exact same 5V regulator for a analog sensor as your main 5V, if for nothing else then noise immunity. Throughout the entire auto industry, 5V is the standard for powering all sensors on the automobile. During the Toyota crisis on sudden acceleration, many researchers noted that Toyota used the same 5V regulator for both accelerator pedal sensors, and that both pedal sensors used the same ground connection. This seems to be a widespread industry practice, even though it is not good design practice according to fault tolerance experts.
2. Even if the 5V regulator dropped out, there should be voltage monitors which would have flagged this. I agree with you on this. In the case I mentioned, there was no discussion by the Tesla engineer of other DTC’s or voltage monitors on the vehicle.
3. In addition, even if the voltage to the pedal sensor drooped/dropped wouldn't the redundant and opposing sensors either continue to work, or flag a fault as not matching? If the supply voltage to both pedal sensors dropped out, then both pedal sensors would have been disabled during the assumed one second regulator recovery time. With both sensors disabled, redundancy would not have helped to detect the dropout because redundancy was technically absent. It may be possible to detect such a voltage dropout by other means, but one can only design such means if one first knows what he is trying to detect such a voltage dropout.
4. Wouldn't you expect to see strange log data after an impact with a solid object, such as accelerator being stabbed at (as if it was the brake) but with 0 speed because the car has no velocity? I don’t understand why the accelerator pedal sensor would behave any differently before or after an impact, or any differently whether the vehicle is moving or not. I can understand how a difference can occur if the pedal sensors are damaged, but that does not seem to be the case here.
5. And finally, if we are discussing a normal parking operation, why was the brake not pressed prior to any of this happening, as would be expected during a parking operation, rather only 5 seconds after impact. If one's foot were covering the brake, if there was an impact. That impact would tend to press one’s foot down onto the brake in a sudden stabbing motion. The driver claims that she was slowly entering a perpendicular parking space on level ground in front of a building. The vehicle’s motor was quiet and her foot was hovering over the brake pedal, but not in contact with it. She then heard the motor rev up by itself, but the vehicle did not accelerate immediately. About a second later (in her perception), the motor noise revved up to maximum amount by itself and the vehicle accelerated over a concrete curb and crashed into the building wall. She maintains the brakes were never applied by her, although Tesla logs indicate she did do so briefly prior to the crash and once after the crash. The crash caused the air bags to deploy, but there was only minor property damage done to the building. There was substantial front end damage to the vehicle and damage to one of rear wheels.
This seems like quite a lot of work. Since you are willing to put so much effort into this, I'm shocked that you haven't taken WK up on his offer to pull the low level logs from your car. I'd even bet that you could talk him into giving you a drive unit circuit board which you could check the power regulation on. The incident did not happen to me, but to another driver. I only tried to find an explanation for the incident. I did contact wk057 to ask if this data looked like other data he had taken on ten other vehicles, but he refused to answer my question saying that he agreed not to share the actual data from the vehicles he's recovered information from. When I responded that drivers who asked him to download data from their vehicles might appreciate the sharing of their generic data if it led to a better understanding of the cause of their sudden acceleration incidents, as long as their identity remains anonymous, he did not respond back. Therefore, I ask any drivers who have such data to send it to me directly in a personal message, and I will make the same non-disclosure offer as wk057.
Similar threads
