Hubble mirror flaw 1990 traces back to a single piece of measuring equipment, roughly the size of a coin, that was off by 1.3 millimeters. That’s it. That’s the whole root cause. A $1.5 billion space telescope, over a decade in development, launched into orbit with a primary mirror ground with extraordinary, meticulous precision — to exactly the wrong shape — because one small measuring device had been assembled slightly incorrectly, and nobody caught it before launch.
I’ve spent a career debugging systems, and I have a very specific, very uncomfortable relationship with stories like this one. Not because they’re rare. Because they’re not. The failure mode here — a tiny, easily-overlooked error propagating undetected through months of otherwise flawless, rigorous work — is a pattern I recognize immediately, the way you recognize a stranger’s face because it reminds you of someone you actually know.
The mirror itself wasn’t sloppy. It wasn’t rushed. It was, in a very real sense, one of the most precisely manufactured objects human beings had built up to that point. It was also completely wrong, and it took NASA two months in orbit to figure out why, and three and a half years to fix it, in a repair mission that remains one of the most audacious pieces of engineering ever performed in space.
An Instrument Sold as a Promise
To understand why the flaw hurt as much as it did, you have to understand what Hubble had been promised to be, publicly, for years before it ever launched.
The Hubble Space Telescope had been in development since the 1970s, survived the 1986 Challenger disaster that grounded the entire Space Shuttle program for more than two years, and finally launched aboard Space Shuttle Discovery on April 24, 1990. It carried a 2.4-meter primary mirror, a price tag north of $1.5 billion, and roughly a decade of public expectation that this instrument would fundamentally sharpen humanity’s view of the universe — sharper than anything achievable through Earth’s atmosphere, which blurs and distorts starlight before it ever reaches a ground-based telescope.
NASA had spent years building public anticipation around Hubble specifically as a leap forward, not an incremental improvement. It was framed, repeatedly, as the instrument that would let us finally see clearly.
I find something worth sitting with in how much of the eventual embarrassment came from that framing itself. A flawed mirror on an ordinary satellite is a technical problem. A flawed mirror on an instrument that had been sold to Congress, to the press, and to the public as humanity’s clearest window into the cosmos is a very different kind of failure — one measured not just in dollars but in credibility, at an agency that had buried three astronauts on the launchpad in 1967 and lost seven more just four years before Hubble’s launch.
The First Weeks: A False Dawn
In May 1990, weeks after reaching orbit, Hubble captured images of a double star system in the Carina constellation. Early reports described the images as strikingly clear. For a brief window, it looked like the telescope was performing exactly as promised.
I think about that false dawn more than I probably should, because I recognize the shape of it from my own work. Early positive signals, in any complex system, have a way of quieting the specific kind of scrutiny that would actually catch a deeper problem. When something appears to be working in the first pass, the instinct is to relax slightly, to move attention toward the next task, rather than to keep testing the thing that just seemed fine. Hubble’s early images weren’t fabricated or misleading exactly — they were genuinely an improvement over ground-based telescopes in some respects. But they weren’t the full picture, and the full picture took a few more weeks to surface.
In late June 1990, Hubble failed a focusing test. The images that should have resolved into sharp points of starlight instead showed a soft, smeared halo around every object — as if every star in the universe had been photographed slightly out of focus, uniformly, no matter how the telescope adjusted.
The Two Months of Denial
What happened next is, I think, the part of the story that gets skipped over most often, and the part I find most genuinely instructive.
According to Ed Weiler, who worked on the Hubble program at the time, NASA engineers and scientists spent roughly a month in what he later described plainly as denial — from May into early June — before accepting what the data was actually telling them. Weiler recalled the specific moment his own resistance broke: astronomer Sandy Faber, at a meeting reviewing the optical problems, leaned over and told him she was now convinced the telescope had spherical aberration. Weiler said that when Faber told him that, he finally believed it — not because the data had changed, but because he trusted the person delivering the conclusion.
I find that detail almost uncomfortably relatable. As an engineer, I like to think that conclusions land because of the evidence, cleanly, the moment the evidence is sufficient. In practice, that’s not quite how belief actually works, even among rigorous, credentialed scientists staring directly at their own data. Sometimes you need to hear the right person say the hard thing out loud before you’re willing to fully accept what you already, technically, know.
Once the diagnosis was accepted, NASA formed the Hubble Space Telescope Optical Systems Board of Investigation on July 2, 1990. Investigators reviewed fabrication documentation, tested the equipment originally used to build and verify the mirror, and interviewed the people who had worked on its construction. What they found was almost absurdly small in physical terms and almost unbelievably large in consequence.
The Washer
The primary mirror had been ground and polished by the Perkin-Elmer Corporation using a device called a null corrector — a precision instrument designed to verify, during fabrication, that the mirror’s surface matched the exact curve required. The null corrector itself contained a small metering rod with a cap on one end, positioned using a specific measuring device. During assembly of that measuring device, technicians had used a washer to space out a lens, and that washer had been installed roughly 1.3 millimeters out of position — later attributed to a chip of paint that had come off the device’s own metering rod during handling, throwing off a laser measurement by that fraction.
That 1.3-millimeter error in the measuring device translated into a primary mirror ground with its outer edge approximately 2.2 micrometers too flat — a discrepancy NASA later described as roughly one-fiftieth the width of a single human hair.
I want to sit with those two numbers side by side for a moment, because the gap between them is the entire story. A paint chip, dislodged during handling, caused a laser to mismeasure by roughly a millimeter and a quarter. That millimeter and a quarter, propagated through months of otherwise genuinely excellent, careful polishing work, produced a mirror error of two-millionths of a meter — a distance so small it’s difficult to even hold in your head as a physical quantity. And that essentially invisible error was large enough to blur every single image the most expensive space telescope in history would ever take, for the first three and a half years of its operational life.
As someone who has spent a fair amount of time debugging problems that traced back to some small, easily overlooked assumption baked in early and never questioned again, I find this almost too on-the-nose to be real. The mirror itself was polished with genuinely extraordinary precision — precise enough, in fact, that its flaw was itself remarkably consistent and smooth, rather than rough or uneven. Investigators later noted that the aberration was so textbook-perfect it almost certainly originated during fabrication rather than from damage after the fact. NASA had built a mirror with world-class precision to exactly the wrong specification, and nobody caught it, because the null corrector used to verify the mirror’s shape was the very instrument that had been subtly wrong all along. You cannot catch an error using the same flawed tool that produced it. That’s not a mirror problem. That’s a fundamentally epistemological one, and I recognize it instantly from every code review I’ve ever sat through where the test suite itself contained the same wrong assumption as the code it was supposed to be checking.
A 1991 US General Accounting Office report later noted that Kodak had also submitted a competing bid to build Hubble’s mirror, using an independent testing approach that likely would have caught the error. Perkin-Elmer’s bid had been selected instead. I don’t know what to do with that fact except note it, the way you note any detail that makes a story more uncomfortable rather than more satisfying. There was, apparently, a version of this project where a different testing methodology would have caught the flaw before launch. It simply wasn’t the version that got built.
Giving a Space Telescope Glasses
Here’s the part of the story that I think deserves to be far more famous than it is, because the actual fix was genuinely elegant, and genuinely difficult, in roughly equal measure.
NASA did not replace Hubble’s primary mirror. They couldn’t — the telescope was already in orbit, and swapping out a 2.4-meter mirror in space was never a realistic option. Instead, once the flaw had been precisely measured, engineers realized something useful: the error was consistent and predictable. If you know exactly how a lens is wrong, you can design a second lens with the precisely opposite error, and the two flaws will cancel out.
That’s essentially what happened. During Servicing Mission 1 in December 1993, seven astronauts aboard Space Shuttle Endeavour captured Hubble with the shuttle’s robotic arm and installed two key pieces of hardware. The Wide Field and Planetary Camera 2 had corrective optics built directly into the camera itself. And COSTAR — the Corrective Optics Space Telescope Axial Replacement, a device roughly the size of a phone booth packed with small motorized mirrors — was installed to correct the light feeding into Hubble’s other instruments before it ever reached them.
The popular description of this fix as “giving Hubble glasses” is, genuinely, close to exactly accurate. Nobody reshaped the flawed mirror. Nobody replaced it. Engineers built a second set of optics with a precisely calculated, opposite imperfection, and let the two errors cancel each other out through corrected light paths — which is very close to the literal mechanism behind a pair of prescription eyeglasses correcting a misshapen eye.
I find something genuinely satisfying about that solution, beyond its cleverness. It didn’t require pretending the original error hadn’t happened, or somehow erasing it. It required precisely characterizing the mistake and building something specifically shaped to compensate for it. As someone who has occasionally had to ship a fix that works around a flaw rather than eliminating it at the source — because the source is buried too deep in a system to safely touch — I recognize this as a genuinely respectable form of engineering, not a lesser one. Sometimes the correct fix isn’t undoing the mistake. It’s building something that cancels it out.
Endeavour redeployed Hubble on December 13, 1993. One month later, on January 13, 1994, NASA announced the correction had worked. The before-and-after images of the galaxy M100 told the story almost instantly — a soft, smeared blur resolving into crisp, defined structure, the same object, the same telescope, the difference entirely attributable to a small package of corrective optics installed by astronauts in orbit.
What the Flaw Actually Cost
It’s worth being honest about what those three and a half years actually meant, rather than skipping straight to the triumphant repair.
Hubble wasn’t useless during that period — it still outperformed ground-based telescopes in some respects, even with the flaw. But an enormous amount of the observatory’s planned science, the very reason it had been built at such expense, was compromised or delayed. Years of research time, telescope allocation, and scientific momentum were spent working around a limitation that shouldn’t have existed. For an agency still recovering from the Challenger disaster just four years earlier, the very public discovery of a “billion-dollar mistake” — the Shuttle fleet grounded by unrelated fuel leaks that same year, the Space Station program over budget and behind schedule — landed at close to the worst possible moment for NASA’s credibility.
I think there’s a real cost to acknowledge here that the eventual triumphant repair sometimes lets us skip past too quickly. The fix was genuinely brilliant. It doesn’t erase the years of degraded science, or the researchers who had planned observations around an instrument that couldn’t yet deliver what it had promised, or the simple, unglamorous fact that a paint chip on a measuring device cost the field of astronomy three and a half years it never got back.
A Thought to Leave You With
What I keep returning to, in this story, isn’t really the repair — as elegant as it was. It’s the null corrector.
The instrument built specifically to catch this exact kind of error was itself the source of the error. Every check performed during fabrication used that same subtly miscalibrated tool, which meant every check came back looking correct, because the reference point itself was wrong. It wasn’t a case of nobody checking. It was a case of checking thoroughly, rigorously, repeatedly — against a standard that was quietly, invisibly broken from the very beginning.
I think about this every time I trust a test suite a little too completely, or lean on a monitoring system without asking what assumptions are baked into how it measures success in the first place. The most dangerous errors, in my experience, aren’t the ones nobody looked for. They’re the ones that were checked for diligently, using a reference point that turned out to be flawed in exactly the same way as the thing it was supposed to catch. Hubble’s engineers weren’t careless. They were meticulous, and precise, and thorough, and it still wasn’t enough, because the ruler itself was bent.
The telescope launched nearly blind despite years of careful work. It flew for three and a half years producing images nobody was fully satisfied with. And then seven astronauts, three hundred miles up, gave a billion-dollar mistake a very precisely calculated pair of glasses — and Hubble spent the next three decades showing us some of the clearest images of the universe humanity has ever seen, built on top of a flaw that never actually went away. It’s still there, in the original mirror, exactly as wrong as it always was. We just learned to see around it.
More Stories Like This
← Previous: Apollo 11 Moon Landing: The Real Story Behind One Giant Leap in 1969
→ Next: The Wrong Map That Accidentally Led to the New World