QUOTE (helvick @ Dec 16 2006, 04:46 AM)
![*](http://www.unmannedspaceflight.com/style_images/ip.boardpr/post_snapback.gif)
That's a bit of a sweeping generalization there oDoug. Of all the failures only the MCO english/metric units problem and the commanding error that led to the loss of Phobos 1 could really be called "simple". Out of 21 failures by the various participants in the Mars race I'd say that 2 cases caused by simple mistakes hardly deserves to be described as "most".
Perhaps... though, when you look at the numbers, it seems that the majority of Mars probe failures fall into distinct categories:
Decent engineering, bad workmanship: This plagued a lot of the early Soviet Mars probes.
Launch failure: This is responsible for 40% of the American failures -- Mariners 3 and 8.
Simple mistakes: In addition to MCO and Phobos 1, I would add the failure to fully evaluate how the MPL software would react when the landing gear deployed as a "simple" mistake, making this category responsible for another 40% of American failures. Granted, this probably doesn't qualiify as "most," but it's only equaled by lauinch failures, at least for the American program. (And, hey, wasn't one of the Viking spacecraft accidentally shut down for good by a bad command load? That sort of falls in here, too...)
Plain old bad luck: I put a few failures in this category, including the Mars 6 lander, Beagle 2, the DS2 penetrators, and even Mars Observer. In any complex mechanism, you will always have mechanical failures, and these missions tended to run into them at critical points in the missions. Either that, or had the bad luck of hitting the ground at the wrong angle, or onto a badly placed rock, or onto the side of a hill, or during a global dust storm... in other words, just getting on Mr. Murphy's bad side.
So, OK, maybe "most" isn't appropriate. But you gotta admit, of the various categories, it ain't insignificant, either...
QUOTE (helvick @ Dec 16 2006, 04:46 AM)
![*](http://www.unmannedspaceflight.com/style_images/ip.boardpr/post_snapback.gif)
As to your other point they can and should test, test, test and then test some more where they can but at the same time we have to accept that there is a point where you have to stop chasing perfection and run with what your best engineering tells you is "good enough". That answer will sometimes be wrong and we will lose probes in the future but if we insisted on chasing zero risk we would end up with a robotic program that rarely launched anything.
I totally agree with the old engineering maxim that "Better is the mortal enemy of good enough." I just think that, in some cases, you have to raise the bar a bit in your definition of "good enough." I would say that, for example, a full simulated run of the EDL software, with all expected events represented accurately in the simulation environment, ought to be an absolute requirement for all Mars landers. The failure of this box being checked off (or even existing on the checklist, for all I know) in the MPL development cycle most likely caused its failure. It's this kind of thing -- flying the mission with a fatal flaw in its software that could have been caught with a single full-sim run of the EDL software -- that I think you just have to commit yourself to achieving, regardless of its impact on development costs.
Of course, as with anything I write here, that's just my own $.02's worth...
-the other Doug