The Hubble Space Telescope went into orbit in April 1990 with a primary mirror polished to the wrong shape by 2.2 micrometers, a figure smaller than a fiftieth of a human hair and yet large enough to make a $1.5 billion observatory take pictures roughly as blurry as a good ground telescope on a clear night. The defect was called spherical aberration. It meant that light hitting the outer edge of the 2.4-meter mirror focused in a slightly different place than light hitting the center. The result was a halo around every star and a national embarrassment for NASA that lasted three and a half years.
The fix arrived in December 1993, when seven astronauts on space shuttle Endeavour caught the telescope with a robotic arm, opened it up, and installed corrective optics to compensate for the mistake. NASA’s history timeline says the first servicing mission installed two key instruments: the Wide Field and Planetary Camera 2, which had corrective optics built into the camera itself, and COSTAR, the Corrective Optics Space Telescope Axial Replacement, which corrected the flaw for Hubble’s other first-generation instruments. The story of how a mirror that precise ended up that wrong, and how one shuttle mission rescued the telescope, is one of the clearest modern examples of what tiny engineering errors can do.
A mirror polished to the wrong shape, perfectly
The Hubble primary mirror is 2.4 meters across, about the width of a small car. It was ground and polished at Perkin-Elmer’s optics facility in Danbury, Connecticut. The specification was brutal. The mirror had to be smooth to extraordinary tolerances, and in one sense the contractor achieved exactly that. The surface was not sloppy. It was not carelessly made. It was polished with extreme precision.
It was also the wrong shape.
The outer edge of the mirror was too flat by about 2.2 micrometers. NASA describes the error as roughly one-fiftieth the width of a human hair. That is the kind of number that sounds almost meaningless until it is multiplied across the geometry of a space telescope. Light striking the outer part of the mirror did not come to the same focus as light striking nearer the center. The instruments at the back of the telescope received an image with a sharp core surrounded by a fuzzy halo. Faint objects vanished into that halo.
The cause was traced to a test device called a null corrector, used during polishing to check the mirror’s curve. NASA says the device had a lens spacing error of 1.3 millimeters, which led workers to polish the mirror to the wrong figure. In other words, Hubble’s mirror was not roughly made. It was made beautifully to match a flawed measuring device.
First light, and the slow public admission
Hubble was deployed from space shuttle Discovery on April 25, 1990. The first engineering images were better than many ground-based pictures, but visibly worse than the optical models had promised. Engineers adjusted focus, tilted the secondary mirror, ran alignment procedures, and worked through possible explanations one by one.
On June 27, 1990, NASA announced that Hubble’s primary mirror had spherical aberration. The telescope had been sold to the public as the instrument that would sharpen humanity’s view of the universe. Instead, its first images made it look like a fuzzy eye in orbit. The coverage was brutal. Late-night comedians turned Hubble into a punchline. Congress held hearings. NASA, the agency that had landed Apollo astronauts on the Moon, suddenly looked as if it could no longer build a working telescope.
What saved the project was a design decision made long before the mistake became public. Hubble had been engineered to be serviced by astronauts. Its instruments could slide in and out. Its panels could be opened. Its components were reachable in orbit. That did not make the repair easy, but it made the repair possible.
How you correct a mirror you cannot replace
Replacing the primary mirror in orbit was not realistic. The mirror was the heart of the observatory. Removing it would have meant rebuilding Hubble in space. The fix had to compensate for the error without touching the mirror itself.
The solution was conceptually simple and mechanically unforgiving. If the primary mirror had the wrong curve, engineers could place smaller mirrors with an equal and opposite optical correction into the light path. Hubble would still collect light with the flawed main mirror, but the instruments would receive corrected light before the image was recorded.
Two pieces of hardware made that possible. The first was the Wide Field and Planetary Camera 2, a replacement camera with corrective optics built into its own design. The second was COSTAR. The Smithsonian describes COSTAR as a deployable optical bench with five pairs of small corrective mirrors on arms, designed to send corrected light to Hubble’s Faint Object Camera, Faint Object Spectrograph, and Goddard High Resolution Spectrograph.
NASA has described COSTAR as about the size of a large refrigerator. It was not a science instrument in the usual sense. It was a set of eyeglasses for the instruments Hubble already had. To make room for it, astronauts had to remove the High Speed Photometer, one of Hubble’s original instruments.
Eleven days, five spacewalks, one agency on trial
STS-61 launched on December 2, 1993, with seven astronauts aboard Endeavour. The mission commander was Richard Covey. Kenneth Bowersox was the pilot. Story Musgrave, Jeffrey Hoffman, Kathryn Thornton, Thomas Akers, and Claude Nicollier served as mission specialists. Nicollier operated the robotic arm that grappled Hubble and held it in the shuttle’s payload bay during the repair.
NASA’s STS-61 mission record says the astronauts performed five back-to-back spacewalks totaling 35 hours and 28 minutes. They replaced solar arrays, changed gyroscope hardware, installed the new camera, removed the High Speed Photometer, inserted COSTAR, and added a new computer co-processor.
Every step had been rehearsed. The spacewalking teams alternated so one pair could rest while the other worked. Tools were tethered because anything dropped in orbit could become a dangerous projectile. Bolts, doors, handholds, and instrument swaps had been practiced underwater in a Houston training tank with a full-scale Hubble mockup.
Endeavour redeployed Hubble on December 13. One month later, on January 13, 1994, NASA announced that the new optics had successfully corrected the spherical aberration. The before-and-after images of galaxy M100 told the story instantly. What had looked soft and smeared now resolved into crisp structure. The repair worked.
What 2.2 micrometers cost, and what it bought
The Hubble repair is often used as a parable about quality control. That is fair, but incomplete. A single mis-set test device helped produce a flawed mirror. Warning signs in testing were not acted on as they should have been. Years of science time were compromised. Hundreds of millions of dollars of repair planning, replacement hardware, astronaut training, and shuttle time followed.
The deeper lesson is that Hubble survived because it had been built with failure in mind. Its designers did not assume every component would work forever. They gave astronauts access points, replaceable instruments, standardized interfaces, and a way to hold the telescope steady from the shuttle. Without that serviceability, the mirror flaw might have permanently limited the mission.
The James Webb Space Telescope, by contrast, operates near the Earth-Sun L2 point, far beyond the reach of any current crewed repair mission. Webb’s mirrors were tested and retested under conditions Hubble’s original mirror never faced. The Hubble experience became part of NASA’s institutional memory: when there is no second chance to touch the optics, the ground testing has to carry all of the burden.
The afterlife of a rescued telescope
Servicing Mission 1 was followed by four more Hubble servicing missions, in 1997, 1999, 2002, and 2009. Newer instruments were built with their own corrective optics, so COSTAR eventually became unnecessary. ESA says COSTAR was removed during Servicing Mission 4 and its instrument bay was reused by the Cosmic Origins Spectrograph. The flown COSTAR unit is now part of the Smithsonian National Air and Space Museum collection, though the Smithsonian’s current object page lists it as not on display.
Hubble’s repaired life became far more important than its flawed beginning. NASA says Hubble has made more than 1.7 million observations over its lifetime, with more than 22,000 peer-reviewed science papers published from its discoveries. Its work helped refine the age of the universe, revealed supermassive black holes at the centers of galaxies, studied planets inside and beyond the solar system, and produced deep-field images that changed how the public imagined cosmic time.
The telescope is still active more than 35 years after launch. NASA lists Hubble in low-Earth orbit at about 300 miles altitude, traveling at about 17,000 miles per hour. Even in 2026, it continues to produce fresh science, including observations connected to a wake generated by a companion star orbiting Betelgeuse.
What the number actually means
A 2.2 micrometer error is hard to visualize because human intuition runs out below the width of a fingernail. A human hair is often around 100 micrometers wide. A red blood cell is about 7 micrometers across. A typical bacterium can be around 1 micrometer. The Hubble mirror error sits in that almost invisible world, spread across a curved glass surface 2.4 meters wide.
And yet that distance was enough to send ancient light to the wrong place. Photons that had crossed space for millions or billions of years entered Hubble’s mirror, reflected through the telescope, and missed their intended focus by enough to blur the image. The light was not lost. It was just smeared.
Three and a half years later, a set of small corrective mirrors unfolded into the light path and put that light back where it belonged. Hubble’s defining story is not simply that NASA made a mistake. It is that the telescope was designed in a way that allowed the mistake to be reached, understood, and corrected. The next time a Hubble image appears impossibly sharp, it is worth remembering that its clarity came from optics ground to match an error no human eye could see.
