It's the "unknowns" not accounted for in the models that are the (potential) issue. If you have time read through this Ars Technica thread where a bunch of these questions are hashed out: http://arstechnica.com/civis/viewtopic.php?f=23&t=116333...

For those without the time, I'll highlight one post from page 5:

"Unintentional radiators.

The reason they ask you to power down ALL electronics, and not just turn off radios, is that the oscillators (the clock that runs all digital devices) on the circuit board can act as miniature radios, in and of themselves, via clock signals on circuit traces. They emit at the clock frequency, and in some cases many higher level harmonics of those frequencies.

As I've stated several times, it's actually pretty easy to mitigate against known frequencies and signalling techniques...it's the ones you don't know about that are the problem. EMI can be downright spooky.

As an example: I was once testing a medical device (for the Medical Device Directive) that was required to failsafe, since it would literally be touching a patient (it was a combination pulseoximeter and a few other things all rolled into one.

There were numerous tests for both emissions and immunity, and things were going along ducky until we noticed two separate failure modes that weren't considered failsafe. At the time, we were testing ESD immunity up to 20kV. It would pass one time, and not another. We thought maybe we had a bad unit, so we got a few more from the manufacturer. That entire week we kept trying to figure out how to make the failure repeatable, without luck. My coworker and I went in over the weekend, and could simply not make any of the units fail at all, with the exact same test.

That's what triggered our thought process...what else could be causing the issue during the work week but not on a weekend?

Other immunity tests! Turns out, a dozen or so meters away, a different device was undergoing a different test...that wasn't required for the MDD. It was a conducted immunity test (it may have been Electric Fast Transients, can't recall) but the actual test signal was leaking out of that lab, and into our lab, via AC lines in the building. Our chief engineer submitted a proposal to add whatever test it was to the MDD, but I don't know whatever became of that.

I've seen simple clocks inside electronic gear cause CPUs to go haywire...in effect a single system interfering with itself. I've also seen extremely low power, yet very high frequency harmonics invade and corrupt function of another device, several meters away."

what spa942 said.

From the linked thread:

"The low power in a digital watch is certainly a factor but the main thing that keeps them from interfering is very low operating frequency. In older, simpler watches this was 32 kHz! Only one octave above the audible range. In more modern multifunction watches with stopwatches that go to 0.01 seconds (as if anyone can push a button that precisely, but never mind that) they do have to be clocked faster, but they are still clocked as slowly as possible, to save power.... typically 1 MHz or less.

The RF from these things is all but undetectable unless you put a pickup directly against the watch. "

What confuses me is why they don't ensure that planes' electronics are properly shielded rather than testing a bunch of electronics that emit very weak to weak EM. If a cell phone could create problems for a flight during take-off, what about a lunatic pointing a HERF weapon at the plane?

Nothing can be perfectly shielded in the real world and instruments or controllers may need very long cables to connect to their respective sensors and actuators making shielding difficult.

Now add in that many of these instruments actually need to process an RF signal and so won't work if shielded, and it becomes a difficult problem to solve.

But whatever it is, you don’t want to do all your tests in a “simulation,” then have some scientists say “Oops, our bad, we forgot to simulate lower pressure or increased gamma rays or what-have-you,” when we can simply fly the device and all of our testing equipment.

Now excuse me, I have to return to preparing for the Invasion of Moghron from the Frozen Seas of Neptune

cost of plane itself + cost of all the human lives on it = many millions, or "priceless", right?

seems cheap in comparison. plus they don't need to perform this test on every plane before every single flight. only once or a few times, per model permutation. If you spread the testing cost across all the operational instances where they'd be de-risked, I bet it would be pennies per flight, at most.

In reality, there almost always is."