Off Their Rocker? Detroit Free Press Postulates Cosmic Radio Waves Are Causing ECU Issues

Interesting theory, now it would be interesting to know how often such events occur in Toyota's, and compare that to other makers.

Right. This postulate is like saying, "A butterfly fart in Brazil is the underlining cause of a hurricane in the Gulf of Mexico."

Well they really don't know what is the cause of the acceleration on Toyota cars as I see. Whatever the problems is it, I hope the keep searching deeply to avoid more issues on customers.

I building a Tin foil hat for my Prius.

Were this true, then we would expect to see more incidents of unintended acceleration at higher altitudes.

I didn't think the planet, let alone the atmosphere was thick enough to even bend cosmic radiation.

What they are talking about when they refer to "cosmic radiation" is the influx of particles into the atmosphere from space. In this case neutrons, which are progressively filtered out of the atmosphere. The ground neutron spectrum is around 1/300 of that at 40,000 feet.

This is why the two groups of people who know a lot about SEU are people who design and run very large computers and aircraft manufacturers.

Funny, 204E60. Of course, not exactly what I meant...

cdoke, what's the deal with your glamour-shot avatar photo

Its not a glamour-shot (it isn't even photoshopped). It was taken by a NY fashion photographer. Why have a modeling photo as a profile image? Well, maybe I am incredibly vain, maybe I'm incredibly insecure, or maybe there is no reason.

The most probable cause is multi-input processing error.

ECU's in today's autos work like a big "look up table", i.e., for a given air temperature, throttle command, engine temp, engine load, RPM, oxygen sensor measurement, anti-knock sensor, etc, then set the fuel injectors, valve timing, and ignition timing to predetermined values. The true number of inputs is very large and the range of the sensor inputs is even larger.

For example, for a single output such as ignition timing and two inputs, air temp and RPM, the calculations look like; the air temperature range is from -20F to +120F = 140 inputs. Combined with RPM from 1000 to 6000 in resolution of 100 RPM = 50 inputs, resulting in 140 x 50 = 7,000 data points that must be individually tested.

However, if we consider adding the oxygen sensor inputs to the equation, 200 in all, then the total number of calculations goes to 7,000 x 200 = 1,400,000.

When all of the inputs are considered over their full expected operating range, the total number of calculations that must be validated is in the billions.

Add into this situation the possibility and uncertainty of any sensor sending slightly erroneous data when the sensor is starting to fail intermittently and THE PROBLEM is that the extent of these calculations has now grown too large for testing, either in software or on the road, to ever test each combination.

And thus, it is highly likely that there is an error in the ECU calculations for some very obscure combination of inputs (like momentary full throttle on a 72F day with the knock sensor slightly out of spec and the engine at 3250 RPM).

Then, and only then, does the calculation result in a error that commands full throttle and will not read the gas pedal position any longer correctly, resulting in a full throttle run away.

You can test all you want, but until you hit those exact conditions, you will never repeat or replicate the error.

Nothing beats an absolute simplification of a problem that at its root is user mistakes.
So you figure all of Toyota cant test variables that hit a billion calculations. A billion calculation ?
Well a BILLION calculations a SECOND is called GigaFlop aka GFLOPS(10^9). (GIGA Floating Point Operations Per Second)
A simply Cell processor (in the PS3) is capable of 32 GFLOPS,
Amd 5970 graphics CPU reaches 4640 GFLOPS in single precision and 928 GFLOPS in double precision.
A standard general purpose CPU from Intel is about 76 GFLOPS in double precision (i7-965 XE)

so yea, either the NHTSA or the various testers who write ECU software, are too cheap to buy an off the shelf computer or graphics card to
run simulations... and use calculators from the 1970's by your observations.
ECU software or dedicated processor software is written to wholly different standards then the crap you find on your computer. Its because it functions as a critical engineering component. Parameter testing is one of the most basic tests any hardware needs to pass at its very least.



Se yeah a simple computer you can buy ..