NASA lost a $125 million spacecraft at Mars because one engineering team used English units while another expected metric — the mismatch quietly pushed the orbiter onto the wrong trajectory after nine months in space

A little after 2am Pacific time on 23 September 1999, a spacecraft slipped behind Mars and never came out the other side.

Mars Climate Orbiter was supposed to skim about 226 kilometres above the surface, burn its main engine for 16 minutes and settle into a long looping orbit as the first weather satellite stationed at another planet. It arrived roughly 170 kilometres lower than that. Everything looked normal until the carrier signal dropped out 49 seconds early and stayed gone. NPR called the loss of the $125 million orbiter human error, a tidy label for something that took eight more causes to explain.

Pound-seconds where newton-seconds belonged

Reaction wheels inside the orbiter spun up as sunlight pushed against its solar array, and small thrusters had to fire periodically to bleed that momentum off. Each firing gave the craft a tiny shove. Ground software called SM_FORCES worked out how big each shove was and wrote the figure into a file that navigators at the Jet Propulsion Laboratory fed into their trajectory model.

The specification governing that file required metric units, newton-seconds. What the code delivered was pound-force-seconds. One pound of force equals 4.45 newtons, so every nudge the navigators modelled came in at less than a quarter of its real strength, as documented by the Mishap Investigation Board in its first report that November. Meanwhile the version of the same calculation running on board the spacecraft used metric and was correct. Only the copy on Earth was wrong, which left the orbiter with a better idea of where it was headed than the people flying it.

Nine months of tiny nudges

How does a factor-of-four discrepancy survive 286 days of scrutiny?

Partly by hiding inside something the flight team was not expecting. Mars Climate Orbiter carried a single asymmetric solar array, unlike the symmetrical wings on Mars Global Surveyor, so solar pressure built momentum far faster and desaturation firings happened 10 to 14 times more often than the navigators had planned for. A daily 180 degree flip designed to cancel that build-up had been dropped from the flight plan after engineering trade studies, and nobody passed the decision along to navigation.

Geometry helped conceal it too. The desaturation thrusters fired mostly sideways to the line between Earth and the spacecraft, and Doppler tracking can only sense motion along that line. The drift stayed small day to day and sat squarely in the one blind spot of the instrument meant to catch it.

Format errors and bad attitude data made the thruster files unusable for the first four months of the cruise, so the navigators worked from emailed notifications and modelled the nudges themselves. Once clean files arrived in April 1999, the numbers looked wrong within a week. Through that spring and summer, Doppler tracking kept implying a closer approach than the other solutions, and those concerns went up informally and were never resolved. Staff traded email about it rather than filing the formal anomaly reports that would have forced an answer.

Twenty-four hours out

With a day left before insertion, the number on the screen kept getting worse.

A final course correction on 15 September was aimed at that 226 kilometre approach. Within the week, orbit estimates had slid to between 150 and 170 kilometres. About an hour before the engine burn, fresh tracking data put the closest approach at 110 kilometres. Anything below 80 was considered unsurvivable.

A fifth correction burn existed as a contingency and could have lifted the path. It was raised out loud, briefly, and never flown. No criteria had been written for calling it, no procedure tested, no room left in the onboard sequence to upload one.

Nobody had built the machinery for making that decision, so nobody made it.

Reconstruction afterwards placed the true low point at 57 kilometres, deep enough into the atmosphere that NASA now lists the mission as lost on arrival.

Eight other causes

“A major error in our understanding of the spacecraft’s path” is how board chairman Arthur Stephenson summed it up when NASA announced the findings. Stephenson, director of Marshall Space Flight Center, led a review that named one root cause and eight contributing ones, and the measurement mismatch was only the first.

Staffing drew particular criticism. Three Mars missions were running simultaneously out of one operations office, and in the weeks before arrival a single navigator was assigned to this one. That navigation team came aboard shortly before launch, missed the design reviews, and assumed the new spacecraft would behave like Mars Global Surveyor.

Beneath the technical findings sat something closer to an attitude problem: a sense, widely shared among the flight teams, that reaching Mars safely had become routine. The board’s advice to the team readying the next landing was blunt: don’t assume anything is safe simply because it has always worked before.

Ten weeks later, Mars Polar Lander went quiet during descent and was lost on arrival as well, most likely because a false touchdown signal shut its engines off early. By 2002 the General Accounting Office was reporting that the agency still had no dependable way of gathering or circulating what its projects learned, and that managers named a perceived intolerance for mistakes among the reasons they stayed quiet. That first report on the orbiter has sat in NASA’s Lessons Learned Information System ever since, public and indexed and one search away. Finding it was never the difficult part.