Re: BristolBachelor
"More on that story please!"
Well let's see. From where I was working at the time, it was sometime in the 90's. Some details are deliberately left out, like the country, airport, aircraft etc. As I understand it, I could use the investigation in general terms as an example when talking about investigations, failures or systems engineering, but certain details that aren't relevant to that use might make people blush. So here is an example of a system level failure and a small part about the investigation that is worth baring in mind when designing equipment that might be expected to undergo post mortem if something goes wrong.
Pilot flying a small regional jet came in to land and set the engines to idle (oops 1). During final approach, he decided he didn't like something, and decided to do a go-around, so he throttled the engines to 100%. The engines now being below nominal operating temp because of the time spent at idle, throttled up a bit, but not to 100% (if you go from 0% to 100% in a sudden jump, the turbine blades heat-up, expand quicker than the outer of the engine and things get really noisy really quickly).
Realising that he wasn't getting 100% thrust, the pilot decided that he was going to have to land anyway, so changed his mind in a hurry, and got the plane onto the runway. Sometime around now, with the engines up to nominal temp and with the throttle still at 100% (Oops 2), ramped up the trust to the max. Plane shot down the runway like a bat out of hell, overshot and went into a forest.
They scrambled a helicopter to locate the plane as it was a bit hard to follow the random path through the trees. The investigation could not interview the flight crew and reconstructed evens from the flight recorders, radar records, positions of controls, etc. We were asked to retrieve whatever we could from the built-in records in our unit. The unit does a built-in-test (BIT) every time at start-up, a fuller BIT when commanded, and some periodic on-going checking during normal use. Also any anomalies are recorded - all done to identify possible faults for maintenance.
The box was a bit of a mess as I said earlier. The E² (effectively flash that is erased byte at a time instead of page at a time) dies were taken to the manufacturer who had a nice test set-up that could probe directly on the bond-pads. Can't remember which manufacturer it was, but it's likely that they don't exist by that name anymore and have been bought and assimilated so many times it would be difficult to find out. Anyway, the data was read-out of the die, and for speed was copied into a new chip and plugged into an engineering unit to read-out the logs. Separately, a manual search was done through the raw data to confirm that nothing was missed by the log read-out.