Hubble Space Telescope mirror defect

Six Sigma's limits — for a single-unit precision instrument, statistical quality control doesn't apply. The Lincoln Lab adaptive optics back-application is the secondary teaching point.

If you're trying to make bricks for a barn or a house, four Sigma may be okay. I started saying this back in the '90s when they had the problem with the Hubble telescope. Anybody know what happened with the Hubble telescope? Two-billion-dollar device to look at stars and probe back into the earliest ages of the universe, and they screwed up when they ground the mirror. It wasn't as nice and parabolic as it should be. The problem is, if you're making Hubble telescopes, it's not easy to go fix it when it's up there. You really needed process controls that were much better than even Six Sigma.

Think of the Hubble telescope. The Hubble had a few design problems on the mirrors, and they had a two-billion-dollar project that all of a sudden could have been junked if it hadn't been for Lincoln Lab developing a military technology to compute away the distortions of the atmosphere. One of the reasons for putting the Hubble telescope up in space was to get rid of atmospheric distortions. Well, they put it up there and found they had distortions in their mirror. We'll never build another Hubble telescope to do optical imaging in space, because now, because of places like MIT Lincoln Lab, they can shoot a laser beam through the atmosphere, measure the distortion, and back-calculate. There's a satellite up there measuring the distortion of the beam going through the atmosphere, and they can now take pictures from Earth through the atmosphere that are just as accurate as the Hubble telescope, because they do computer correction for the distortions. So we will never need to build another Hubble telescope.

Risk = hazard × consequence framing. One-in-a-billion quality required because the system cannot be repaired (with a parenthetical acknowledgment that fixes were ultimately done).

The risk-benefit here — you know what risk is defined as? Risk is defined nowadays by psychologists as the hazard times the consequence. If I'm manufacturing bricks for a building, I can probably tolerate one bad brick in a thousand, because if a brick is bad, the guy will probably throw it out, or even if he does mold it into the building, the building's not gonna collapse because of one bad brick with a crack in it. But if I'm building Hubble telescopes, my quality should be more like one chance in a billion of failure, because the consequence is you can't fix it. Actually, they did do some fixes, but they're very difficult and very expensive.