WitipediaIt's funny because it's true.
Create account Log in

Ariane 5 Flight 501

From Witipedia, the funny encyclopedia
Ariane 5 Flight 501
Date4 June 1996
PlaceKourou
Altitude at breakupabout 3,700 m
Timeabout 37 seconds after the start of the ignition sequence
Causea 64-bit floating-point value converted to a 16-bit integer
VariableBH, the horizontal bias
Backupfailed 72 milliseconds earlier, for the same reason

Ariane 5 Flight 501 was the maiden flight of the Ariane 5 launcher. On 4 June 1996, about 40 seconds after the start of the flight sequence and at an altitude of about 3,700 metres, the vehicle veered off its path, broke up and exploded. An inquiry board chaired by Jacques-Louis Lions traced the failure to a single conversion in the software of the inertial reference system: a 64-bit floating-point number turned into a 16-bit signed integer that could not hold it.[1]

The number was a horizontal-bias value called BH. It was larger than the same calculation had ever been on Ariane 4, because Ariane 5's early trajectory is different. The conversion was not protected. Other conversions next to it were.

The function that was finished[edit]

The overflow happened in the alignment of the inertial platform. That module produces a useful result only before lift-off. Once the rocket is flying, the inquiry board wrote, the function serves no purpose. It kept running for about 40 seconds of flight anyway, because that was how long Ariane 4 had needed it, in case of a hold late in the countdown. Ariane 5 did not need it. The code had been carried over, and the Ariane 5 trajectory had been left out of the requirements for the unit.[1]

Three variables that could overflow had been left unprotected, after a decision that the computer should not spend its time checking things that were physically impossible. For BH, the impossible value was the one the new rocket produced.

The spare[edit]

There were two inertial reference systems, one active and one in hot standby, so that a failure of the first could be survived by switching to the second. They ran the same software. The backup declared the same exception one data cycle earlier, 72 milliseconds before the active unit did. The on-board computer therefore had nothing left to switch to. What it received from the failed unit was a diagnostic bit pattern, which it read as flight data, and it commanded the nozzles to their stops. The angle of attack passed 20 degrees. The boosters separated. The self-destruct, which was working, destroyed the launcher.[1]

The specification said that on any exception the processor should shut down. Both processors did. The inquiry board noted that this switched off two units that were, until the moment of the exception, still healthy.

See also[edit]

References[edit]

  1. ^ Inquiry Board (J. L. Lions, chairman), Ariane 5: Flight 501 Failure, report of 19 July 1996.

Rate this article

Two questions, two separate answers. An article can be useless and hilarious, or accurate and dull; the ratings are kept apart so editors can tell which problem they have. Anonymous ratings are counted per IP address.

Was this helpful?Did you actually learn the thing? No ratings yet
Was this funny?Did the joke land, and was it true? No ratings yet