Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Unintended Acceleration - Find the Cause

SylviaSylvia Posts: 1,636
edited December 2013 in Toyota
Every car company has received complaints from consumers relating to vehicles that suffered unintended acceleration. Many incidents are not fully addressed by recalls. NHTSA is responding to the challenge with more of what they have already done: additional investigations.

Isn’t it time to try a different approach? We at think so.


  • steverstever Posts: 52,683
    That's a nice teaser lead-in Sylvia. :shades:

    And the rest of the story ....

    Auto Web site Edmunds offers $1 million for cause of runaway Toyotas (Washington Post)

    Here's the official press release from Edmunds.
  • graphicguygraphicguy SW OhioPosts: 10,290, that's one way to get inundated....offer money....lots of it, to show UA in Toyotas.

    Me thinks you'll be hearing from Dr.Gilbert very soon.

    For that kind of dough, I may go out and buy a Corolla and do it myself.
    2018 Acura TLX SH AWD ASpec
  • xluxlu Posts: 457
    edited March 2010
    Many people think it's good to go with digital on everything. It's not always the case.

    Let's start with a very simple control in every electronics, the volume control. In the old products, the volume control is a simple potentiometer. If you turn it slowly, you can change volume in indefinitely small steps (fine adjustment). If you turn it really fast, you can bring the volume from min to max in one revolution and in less than half a second. That's the beauty of analog (mechanical) controls.

    Now let's see how a digital volume control works. The volume knob turns a digital encoder which has certain contacts at certain position. Let's say the encoder has 8 contact positions in a revolution (at 12, 1:30, 3, 4:30, 6, 7:30, 9, 10:30 o'clock position). The CPU would check periodically where the current contact position is and compare with the past position to determine which direction the volume knob is turning and how many steps; the CPU would change volume accordingly.

    You can see two major problems of digital volume already:

    1) Fine adjustment vs. fast change: if you want to make very fine changes in volume, say 8000 steps from min to max. Then the knob needs to be turned 1000 revolutions for that! If you want to make less total turns to one revolution like the analog volume, then you only have 8 steps!

    2) Rotation direction and speed detection: let's say the CPU checks the encoder position 4 times a second (CPU needs to multi-tasking with many controls). The 1st time the CPU found the contact is at 12 o'clock; 2nd time is at 9 o'clock. Then the CPU assumes the volume is turned down by 2 clicks. But the user may have turned the knob up (clockwise) very fast from 12 o'clock to 9 o'clock between the 1/4 second CPU checks the encoder! You can see that the CPU could easily and totally miss-interprets the user action in the opposite way and wrong number of steps!

    These are the 2 exact challenges when designing the car's pedal, steering and brake into digital (electronics) controls.

    Camry unintended acceleration is probably due to the CPU miss-interpret the driver pedal action in the wrong direction.

    Prius brake hesitation is probably due to the CPU not checking the brake encoder (sensor) fast enough.

    Corolla steering problem is probably due to not enough fine steps in the steering encoder (sensor).

    I challenge Toyota engineers to post their solution to the problems I listed above in their car design.
  • imidazol97imidazol97 Crossroads of America I70 & I75 Posts: 21,504
    >I may go out and buy a Corolla

    But think how bad the resale value is likely to be on a toyota now. :P

    2015 Cruze 2LT, 2014 Malibu 2LT, 2008 Cobalt 2LT

  • steverstever Posts: 52,683
    edited March 2010
    That analog knob fails when you are trying to tune a radio station however, especially if you are using a short wave radio. Or if you are trying to dial in a planetary or star position with your telescope.

    And if the potentiometer jams on high, you get unintended acceleration "really fast."

    In the good old days, a relatively common complaint was a "flat spot" in the accelerator. Drive by wire can avoid that issue. A mechanic device can kink or break or have an object dislodge and jam the throttle cable. So there's cons to heading back in that direction too.
  • wwestwwest Posts: 10,706
    I strongly suspect that at the end of the day it will be found that there is no "single" causative factor. Confluence of factors would be my guess. For instance, simultaneous opening of the A/C compressor clutch relay and startup of the ABS pumpmotor. Lots of RFI/EMI "spewed" about within the engine compartment and within the 12 volt wiring due to the inductive "kick" of the electromagnetic A/C clutch coil, while at the same time a HUGE load on the 12 volt system dragging it down to a minimum voltage.

    Personally I would start by adding voltage "snubber" devices to the various inductive devices, A/C clutch, transaxle solenoids, DC motors, etc. I would also provide SOLID ground/supply wires directly from the battery post to the engine/transaxle control module, fused directly at the battery.
  • eliaselias Posts: 2,147
    edited March 2010
    xlu, it seems to me that you are proposing ideas in the right direction and that you give examples of things that can cause "glitches".
    However I think the root cause issues are substantially more complex than you have described and involve both race-conditions (software bugs relating to unexpected timing of separate processes) and multiple simultaneous electronic failures or "glitches".
    That is, I don't think the problem occurs due to "single point of failure" of either software or hardware, else we would be seeing way more catastrophic failures, and failures that were easily repeatable.
    Instead I think the catastrophic failures are something like a "cascade" of worst-case/corner-case race-conditions.
    That professor dude seems to be demonstrating this empirically.
    BTW, I think Toyota engineers are doing (and redoing) the exact same kind of tests and that they are diligent, skilled, and ethical engineers.
  • wwestwwest Posts: 10,706
    Wrong on both counts.

    1. Fine adjustment vs fast change.

    Check you PC mouse setup. You can have it set for very fine resolution AND fast change. The more distance you move the cursor the more rapid the cursor accelerates....

    2. Rotational encoders, optical and mechanical, ALWAYS have two outputs, 90 degrees in phase separation, so direction of rotation can be discerned..

    Proper hardware and software design is almost magical.

    The gas pedal sensors are not "encoders" in your sense of the word, but hall effect position sensors. The two sensor outputs are intentionally displaced in voltage output by ~0.80 volts but otherwise track each other linearly as the pedal moves from idle to full depression. So it is not possible for the firmware, correctly written firmware, to miss-interpret the gas pedal position.
  • xluxlu Posts: 457
    edited March 2010
    Your response seems to be too absolute.

    In the 1st point, yes, I checked the mouse, both mechanical and optical. In the mechanical mouse, the movement in each axle is sensed by a small bar which drives a wheel with many holes. The optical sensor would read the hole counts. The resolution for one rotation is a fixed number (say 16); so no matter how you set it up, there's only 16 steps in one resolution. When you make the mouse faster, each step just jumps a bigger distance. The optical mouse works with the fixed amount of steps as well. The ball has many small dots which reflect lights. Again, fixed number of steps. You won't get more total steps no matter what you do.

    In the 2nd point, I do not agree that the encoder ALWAYS have 2 outputs. I designed hundreds of products with the encoders of 1 output.

    I'm not an auto engineer so I do not know if they use hall sensor in car design. If they do, there could be even bigger problems because Hall sensor is very sensitive to the EMI interferences.
  • xluxlu Posts: 457
    edited March 2010
    The analog dial in a short wave radio is not easy to tune because you don't know the frequency you are at. The digital tuner displays the frequency or wave length so you can even direct key in the the numbers. Also the radio station frequencies are not continues spread out on the spectrum; they are either 10 kHz or 9 kHz apart; so it's a discrete or digital setup which naturally works better with digital tuner.

    When potentiometer jams, it jams. When a digital pedal jams, it also jams. Toyota already discovered that. We are trying to find the cause by electronics beyond the mechanical jam.
  • wwestwwest Posts: 10,706
    No, what I said was that the PC software that "watches" the mouse output can be set to automatically move the cursor at a faster and faster rate the longer you keep the mouse in motion. So, via the "magic" of software you can have "fine" resolution and FAST tracking.

    Only an encoder that has constant motion, does not reverse direction, or direction is of no interest, can be viable with a single output. ABS wheel speed sensors come immediately to mind.

    VSC stearing wheel rotational position sensors CANNOT be subject to failures that would undoubtedly result from a single output encoder.
  • xluxlu Posts: 457
    edited March 2010
    There are all kinds of software tricks to sense if the user wants fast change and jump big steps. This adds another layer of software interpretation and thus more chance for miss-interpretation. One example is the automatic transmission lag. No matter how well it's designed, it still needs a split second to see what driver's foot is really doing, before jump in action. The manual transmission never has such hesitation. You press the pedal, the car goes immediately.

    Are you saying that a good encoder should have 2 outputs or you know for sure Toyota uses ones with 2 outputs? The 1 output encoder is widely used in volume control which is a bi-directional device.
  • wwestwwest Posts: 10,706
    I suspect that my '01 Porsche C4 with manual transmission also has the same long lag as today's automatics, at least from the standpoint of the elapsed time between my mental decision to "go" vs when the C4 begins moving. From the time I decide to change gears I need to depress the clutch, move the shifter, and then release the clutch, and hopefully coordinate the two functions correctly, clutch release and gas pedal depression.

    The only difference is that my "delay" occurs before that gas pedal depression whereas with the automatic the delay occurs after.

    "..The manual transmission never has such hesitation.."

    Yes, it does, EVERY time.

    The rotary encoder in the stearing wheel of my '01 F/awd RX300 is most definity dual phase.

    I'm also quite sure the volume, tuning, cabin temperature setting controls are also dual phase outputs. I will check tomorrow to see.

    Think of a swinging door, opens inward or outward. Start out with door closed, no sunlight, and now we open it to sunlight...

    Did it swing inward or outward...??

    All the single sensor "knows" is that the signal went from dark to light.

    And by the by the hesitation you feel, the historical 1-2 second downshift delay/hesitation, is a "fix" for the design flaw incorporated in the U160E/F transaxle during the initial design phase back in ~97-98.
  • 210delray210delray Posts: 4,722
    Man, this topic has gotten boring in a hurry!
  • steverstever Posts: 52,683
    edited March 2010
    We need to wait until midnight, gulp some Mt. Dew and a Red Bull and open a coding window in our other monitor. We'll have this thing whipped by 5 am. :D
  • kc_nkkc_nk Posts: 1
    can it be the cruise control problem (like cruise control computer system/software, wire or anything linked to cruise control)?
  • Let me first say, I don’t know what is causing the problems. Either way you’re barking up the wrong tree. 1. Rotary encoders are not used in gas pedals. They wouldn’t work in this case. A gas pedal needs absolute position not relative, otherwise the pedal position would have to be read even with the battery disconnected. It is most likely a potentiometer connected to an ADC. I'm going to tear into my Mazda peddle tomorrow to verify it, but just thinking about it for 3 seconds it would take an idiot not to design in at least 1 backup/redundant pot and reading instrument to verify the pedal’s position.
  • I doubt this is the case - in my experience, throttle position sensors (TPS) have an analogue output which is fed into the PCM which has an onboard A/D converter to digitise it.

    The scenario you describe with rotary encoding is very common and basically precludes this configuration from being used in this type of application.

    Generally, the PCM will expect a TPS voltage range between X and Y where 0 < X < Y < Vmax - let's say X = 1.0V, Y = 4.5V, and Vmax = 5V. Any failure of the TPS should either go open-circuit (which should have the appropriate circuitry to pull this down to 0V), or Vmax. Any of these failure conditions should log a TPS range fault in the PCM and the PCM will drive the throttle to 0% output.

    The only way the TPS could be at fault would be if it somehow failed AND maintaned an a healthy output voltage. Possible? Can't say without intimate knowledge of the electronics, but certainly cannot be ruled out.
  • flaviusflavius Posts: 3
    Here's what I know for shure about unintended acceleration on a Renault Logan car:

    - they come from the factory without cruise control option. If you install the one from a Renault Megane II , on a Logan powered by a 1.5 dCi 85bhp diesel engine, around 90km/h almost 100% of times, you will get this acceleration thing, with cruise control enabled. Well, the reason for this, is the air friction force, wich is stopping the vehicle. The sensor detects a slight and constant decrease in speed, and tells the computer to give it more gas to compensate. But it keeps accelerating untill you disable it.
    - 2nd common issue, is that when you are cruising at say 60 - 70 km/h , and you get just a small gust of wind from the front, again, the vehicle is accelerating past the set speed of the cruise control.

    So, in my case, I blame the cruise control thingy for the unintended acceleration. Maybe this is the same reason for the cars affected by it, in this article.

    What's really strange is that these diesel engine Logan's have drive by wire acceleration, and they don't have issues like these other cars, wich are way more expensive and "intelligent". A Logan is around 8000Euro's...

    PS: Please excuse my bad english, I am not a native english speaker.
  • plektoplekto Posts: 3,738
    edited March 2010
    Even if you had the idea, though, it basically is impossible for a normal person to claim the prize due to this part:

    The rules, which will be announced shortly, will challenge participants to demonstrate in a controlled environment a repeatable factor that will cause an unmodified new vehicle to accelerate suddenly and unexpectedly.

    This basically means a test lab or similar setup with monitoring and so on. But the problem is likely to be a worn part somewhere which can't BE reliably tested for unless one can get their hands on pretty much an entire car that has had the problem(and survived intact, plus hasn't been "fixed" or altered by Toyota).

    So even if I knew the cause, which I have my own theory about, as do most of us here, there's no practical way to claim the prize unless we have a lot of free time and money to buy a known defective vehicle and do repeated testing without destroying the car in the process (chain it down to a dyno or something - who has the ability to rig up something like that?)

    There should be part of the prize for the first person to figure out the cause and post it here even if they can't prove it themselves. We'll know the cause in a year or less, most likely, and it would be easy enough to verify.(It's likely to be already discussed or will be in the next two weeks)

    BTW - Here's my post from Jan 28 about it and my theory. GM(as an example) normally uses potentiometers instead of Hall Effect sensors. Potentiometers tend to exhibit noticeable signs when they are wearing out. Hall Effect sensors have hardly any warning.!keywords=allin%3Amsgtex- - - - - - - - - t%20date%3A3m%20throttle%20plekto#MSG86

    Here is an article that sheds light on this as well: - - - - - - - ls/
    This is from Jan 30, two days after my post.

    The Toyota parts in question use a Hall Effect type sensor(the above article mentions "magnet" multiple times). When it fails, the sensor reads either 0 or 100 depending upon how it is wired into the rest of the system. This isn't an easy to fix problem since it requires a total redesign of the cars in question back to either a potentiometer(less accurate, more prone to wear and dirt, but still minimally functional when worn) design or a normal throttle cable. The article above, though, like every one that I could find online - at least before my Jan 28 post, (and most since) don't target the TYPE of sensor as the potential problem - just the manufacturing differences.

    In a nutshell: It's the use of magnets and Hall Effect sensors that is causing the problems in the entire industry.

    EDIT - potentiometers also fail(and can also cause full-throttle acceleration), but give more warning as a rule and complaints tend to be along the lines of "the throttle surges" or "it won't maintain a constant speed" as opposed to "what just happened?". One gets repaired and the other bites you without warning.
  • flaviusflavius Posts: 3
    edited March 2010
    You are right in your statement Plekto... they should give a prize for a theoretical idea and one for a practical one. I can't afford one of those cars to test my idea. If they give me one for free, just for testing, then we are in money :)

    I gave it some thinking... a more powerful car (mine it's only 85bhp) suddenly accelerating towards a 90 degrees corner (left, right, doesn't matter) would give some intense sensations. BAD sensations. :sick:
  • eliaselias Posts: 2,147
    edited March 2010
    xlu, have you ever staged/raced a car at a drag race? try it some time, it's fun. :)
    you would see that the 'reaction-time' and other times measured indicate the opposite of what you are saying about manual vs automatic, agreeing with wwest on that point. for reasons other than 'lag time', usually the dragracers prefer the automatics.

    i'm not aware of anyone dragracing any recent toyota but hey, it could have happened somewhere - maybe Prius does reduce the measured reaction time compared to other slow and lousy-handling cars.

    but i'll disagree on another wwest point:
    the hall effect sensor can certainly malfunction such as via its connections/wiring, so i disagree with the idea that it is impossible for a computer to misread a hall effect sensor. and it's certainly possible for someone one to write software that misreads a hall effect sensor reading, or that has any arbitrary bug.
  • plektoplekto Posts: 3,738
    edited March 2010
    True, but a potentiometer is dead-simple to read positioning data from by comparison. There's almost nothing to go wrong in the software unless the programmers are horrendously daft. "read resistance from sensor A" is a lot easier to program for than "find position of moving object in sensor A". That's the issue, IMO, and until they stop using these sensors, you'll get problems like these(or programmers who can't figure out how to adequately deal with it).

    BTW, GM's method of dealing with faulty potentiometers is to notice that the impedance is missing or jumping around and it triggers the famous "limp home" mode almost all of the time. A Hall Effect sensor won't be able to tell unless it's a very sophisticated design and the programming is as well. The issue is in a nutshell what happens when the system thinks the magnet went missing/disappeared or got physically stuck.(note a potentiometer will almost never physically get stuck all of a sudden, either, barring enormous abuse or again, some warning)

    EDIT - as further evidence(since it's getting late here and I need *some sleep), consider what happens if we were to replace the Hall Effect sensor in the TPS unit with a potentiometer. The type and manner in which it would fail would be different. The end result would be few if any crashes.

    That logically points to the underlying technology itself as the root cause. Not how it's interpreted or being dealt with(even IF the software is part of the problem). One will fail suddenly and without any warning and the other type of sensor almost never does. (or certainly far far less often does it fail suddenly and without any warning beforehand)

    To bring it around to cars(since we're all gear-heads here :) ), it's like comparing a car with carbs to one with fuel injection. The real issue isn't why does my car get vapor lock and is hard to start in the winter and gets such miserable performance in the mountains. It's that the same engine with fuel injection added exhibits none of these issues. Perfect example - the old Toyota 22R/RE engines. Same exact design other than fuel injection added to the later models. One has all of the issues of carbs and also stalls on steep uphill climbs(floods). And the other does not. Using a different technology on the same engine solved the problems of the former design.
  • judoaljudoal Posts: 1
    The 2008 subaru impreza sport with manual transmission and an electronic throttle, has a significant lag in the accelerator response that is most notable in 1st and 2nd gear. This reminds me of a problem I had in a college engineering control systems course regarding stability. This problem concerned remotely controlling an object over a great distance that is moving away. At what point does it become unstable? While Subaru says that the vehicle is operating normally it makes me wonder if the imposed lag is intentional and if they decreased it, if the system would be unstable. I have no way of testing it, but perhaps someone with access to the microcontroller can play with the lag in the toyota and other vehicles to see if that changes things. Perhaps there is an interaction with the vehicle operator that some combination of operator action and the system, because of the timing and phase considerations, the system becomes unstable.
  • imidazol97imidazol97 Crossroads of America I70 & I75 Posts: 21,504
    >unmodified new vehicle

    First the vehicle has to be "new." If you've owned it, it's no longer new.

    Second the "unmodified." To be able to test and effect the problem, some modification may be needed. A vehicle that has the problem that's a year or more old may not have had modifications in the problem area made to it when built, without fanfare by toyota, and is a better testbed for finding one or more of the causes for unintended runaway acceleration.

    2015 Cruze 2LT, 2014 Malibu 2LT, 2008 Cobalt 2LT

  • botiabotia Posts: 1
    Here are some of the obvious causes:
    * Alcohol
    * Old age
    * Not paying attention
    * Excuse to get out of accident
    * Floor mat
    * Kids toys
    * Leg cramps after exercise
    * Bad feedback from sensor
  • I dont know what the car has already, i presume there's multiple redundancy on accelerator and brake sensors etc

    A simple safety measure could be that if a user tries to brake while the car is accelerating that the brake overrides the accelerator. This should not be the result of the normal accelerator brake processing in the emu. Instead something physical via a seperate control channel, such as gas flow into pistons progressively reduced the longer / harder that the brakes are applied.

    An additional measure, if the car notices that it's not slowing down after a few seconds of brakes applied then that's a strong indication that there's a major safety problem and engine power output should be progressively reduced. Again this should be implemented on a seperate control channel to normal braking / acceleration processing

    Downside is reduced driving experience for people who want to build up revs while braking, but the upside is that braking will always slow down the car even if the accelerator is floored
  • enphorcer600enphorcer600 Posts: 1
    edited March 2010
    I have a funny feeling it could be RF interference from transmitters. After blowing out my UPS with only 5 watts of radio power… it can really be a problem to susceptible circuits. Different circuits would become resonant at specific frequencies. If there was a strong enough transmitter near a car, on the frequency that is resonant with critical speed control circuit, boom. Could also be illegal CB radio operators using illegal linear amplifiers. Some of these mobile amps go up to 1000W, and produce broad spectrally unclean signals. Could just be Russian roulette that&#146;s catching some of these cars. Could be toyota&#146;s standard electronic control package is susceptible to this.

    Auto makers are supposed to test their cars to make sure EMI/RF isn&#146;t a problem, but who knows if something slipped through the cracks, and who knows what kind of tests dot they perform.
  • I had same problem with 1996 Chevy Blazer, where if you got the pedal close to the floor it would get trapped under the floor mat. Heard of an 2005 F150 that would do the same thing.

    Either having a raised platform under the gas pedal so you could avoid getting it stuck under the floor mat. I have seen that in a few models I suppose that is why it is there, but every auto maker should take that precaution.

    Also with electronics if the car is under heavy breaking while the car is unattended acceleration the car should shift into neutral to avoid any power getting to the wheels. Then you could safely come to a stop and with todays engines and rev limiters the engines should be fine if they are stuck revving in neutral.
  • houdini1houdini1 Kansas City areaPosts: 7,642
    Good list, I will add a couple more:

    1. Stupidity

    2. Panic

    3. Big feet

    2013 LX 570 2016 LS 460

Sign In or Register to comment.