In June 1990 the freshly launched Hubble Space Telescope returned images no sharper than ground telescopes: its 2.4-metre primary mirror had been ground to the wrong shape, introducing severe spherical aberration. The mirror was judged too difficult to return to Earth, and engineers considered desperate options — sending a spacewalking astronaut inside the optical tube to swap the secondary mirror, or fitting an aperture shade to mask the mirror's flawed outer regions.

The winning idea inverted the problem. Ball Aerospace optical engineer James Crocker, at an ESA crisis meeting in Germany, noticed his hotel showerhead clamped to a vertical rod at adjustable heights and angles; a similar sliding mount could carry small corrective mirrors into each instrument's light path. The result, COSTAR, cost a budgeted US$50 million, was the size of a telephone booth, and replaced the least-used instrument, the High Speed Photometer. The new Wide Field and Planetary Camera 2, installed in the same 1993 mission, carried its own built-in correction.

Astronauts installed COSTAR on the December 1993 STS-61 mission in eleven days of spacewalks, and the instrument's mirror arms were deployed on 28 December 1993. On the corrected Faint Object Camera, the fraction of a star's light concentrated within 0.1 arcseconds jumped from 18% to 85% — a 1.6-magnitude sensitivity gain — and effective resolution improved from 66 to 43 milliarcseconds, at the modest price of 20% lower throughput and a smaller field of view.

Correct the light path downstream, not the mirror

The mirror's error was stable and fully characterized from star images, so its opposite could be manufactured on the ground and trusted to cancel it in orbit.

Bringing Hubble home or operating inside its optical tube were impossible or reckless; the instrument bay was the one interface already designed to be serviced by astronauts.

Two small mirrors per instrument cost a few grams of glass versus recoating or replacing a 2.4-metre primary — the fix scaled with the error, not the aperture.

Giving up the High Speed Photometer, the least capable instrument, converted the trade into a near-free slot on the telescope.

Designing new instruments with built-in correction (WFPC2) let the same mission fix the system from two ends at once.

When a defect sits in a component you cannot replace, push the correction downstream to where you can reach. A known deterministic error can be cancelled at the cheapest boundary.

The repair worked as calculated: Hubble delivered the deep fields and distance measurements that recalibrated cosmology, and the mission became NASA's template for on-orbit servicing. Every instrument installed afterwards carried its own corrective optics, which made COSTAR obsolete — it was removed in the 2009 servicing mission to free its bay for the Cosmic Origins Spectrograph and returned to Earth, where it is now a museum piece. Crocker's showerhead story became NASA's standing parable for where good engineering ideas come from.

FOLLOW THE EVIDENCE

The sources

  1. In-orbit performance of the COSTAR-corrected Faint Object Camera ntrs.nasa.gov
  2. Corrective Optics Space Telescope Axial Replacement en.wikipedia.org