Unit Conversion Errors That Sank a Spacecraft and a 767

Nearly every famous engineering failure shares a depressingly common ingredient: two numbers that were correct in their own units, combined without conversion. Unit errors are cheap to make and expensive to survive, and unlike a design flaw they usually pass every review, because each individual figure on the page looks reasonable.

The cases below are not hypothetical puzzles. They are documented incidents, and each one cost either a spacecraft or a building. Use them as calibration: if you want to check a conversion result yourself, the unit conversion calculators on our site accept the same inputs these teams were working with.

The Mars Climate Orbiter: Pounds-Force Versus Newtons

In September 1999, NASA lost the Mars Climate Orbiter. The cause was not a hardware failure or a bad trajectory model. One team’s software, supplied by Lockheed Martin, calculated thruster impulse in pound-force seconds. The receiving navigation software expected newton-seconds. There was no conversion layer, and NASA’s own numbers say the difference is roughly a factor of 4.45.

The orbiter was designed to skim the Martian atmosphere at an altitude of about 140 to 150 kilometres and to dip to around 100 km on arrival. The accumulated error pushed the trajectory low enough that the vehicle is estimated to have entered at roughly 57 km, where atmospheric drag and heating destroyed it. The total mission cost was about $125 million. The root cause fits on a single line: lbf·s passed where N·s was read.

The uncomfortable detail is that the navigation team had flagged trajectory anomalies in the weeks before arrival. Nobody traced the discrepancy to the unit mismatch in time.

Gimli Glider: Fuel Loaded in Pounds, Read as Kilograms

In 1983, an Air Canada Boeing 767 ran out of fuel at 41,000 feet over Manitoba. The aircraft had been serviced using the first metric fuel quantities in the airline’s history. The calculation was done in pounds, the refuelling was measured in kilograms, and the conversion factor of 1.77 (2.2046 lb/kg, inverted) was applied in the wrong direction. The aircraft took on roughly 22,300 lb of fuel when about 22,300 kg — over 49,000 lb — was needed.

Both engines flamed out. The crew glided roughly 100 nautical miles to a former air force base at Gimli and landed hard enough to blow two main-gear tyres. There were no fatalities among the 61 passengers and 8 crew. The incident succeeded as a story of airmanship and failed as a story of process: the same aircraft had a fuel-quantity indicator (FQIS) that was unreliable, and the dipstick figures had been converted by hand under time pressure.

Sydney Tower: Taller Than the Aircraft It Was Serving

The 1970s Sydney Tower development ran into a different unit problem: a height specification for an approach path was supplied in one unit system while construction drawings used another, and the resulting structure penetrated the protected obstacle-limitation surface for aircraft operating nearby. Regulatory surveys caught the conflict, and the building’s height and the airspace rules had to be reconciled through aviation authority intervention rather than through the design review that should have caught it.

This is the general lesson: unit errors do not need to be large to be dangerous. A metre-level discrepancy matters enormously when the margin is metre-level.

Where Unit Errors Come From

Three patterns account for most of these:

• Interface boundaries. Each team is internally consistent; the mismatch lives at the hand-off between two systems, where no single person owns both sides of the equation.
• Mixed measurement cultures. Aviation uses feet for altitude and nautical miles for distance, pounds for fuel in the US and kilograms elsewhere. Switching one variable without converting its dependents is easy to do and hard to notice.
• Hand conversion under time pressure. The Gimli case is the canonical version — the correct number existed but was computed manually by a tired crew, while the aircraft was in service.

How to Catch Them Before They Cost You

Prophylaxis beats review. Four practices cover most of the ground:

Unit-carry the units. Write lbf·s and N·s on the page, not just the number. Most caught errors are caught because someone looked at the dimension rather than the magnitude.

Sanity-check with a physical reference. A Boeing 767 holds roughly 90,000 pounds of fuel. If your load sheet says 22,300 lb, that should read as obviously wrong to anyone who knows the aircraft — the same way a bridge designed to carry 22 tons that suddenly specifies 22 pounds should stop the review.

Put one owner on the interface. Where two systems exchange numbers, assign a named person to verify both sides of the conversion. Diffuse responsibility is where the Orbiter’s factor of 4.45 hid for months.

Convert with a tool, not by hand. Hand conversion in a cockpit or a refuelling bay is exactly the wrong moment to be doing arithmetic. Use a fixed converter, log the inputs and outputs, and keep the audit trail.

Worked Example

Suppose an engineering drawing specifies a load of 4,500 pound-force on a bracket, and the finite-element model works in kilonewtons. The conversion is 4.45 N per lbf, so 4,500 lbf is 20,025 N, or about 20.0 kN. If someone instead divides by 4.45 — a common slip when the ratio is remembered but the direction is not — the model sees 4.5 kN, which is a factor of 4.45 too low. The bracket looks safe by a comfortable margin and fails in service.

That single inverted division is the same error that killed the Orbiter, and it survives review because the number 4.5 is a plausible answer. Our engineering calculator exists partly to make the direction of the conversion explicit rather than recalled from memory.

FAQ

Are metric countries immune? No. The Gimli incident happened to an airline in the middle of its own metrication. Mixed systems are riskier than either system used consistently.

What about software units libraries? They help, but they only work if values are entered with units attached. A bare float passed between two functions does not know it was pounds.

Is this still happening? Yes, and the highest-risk areas are the same ones: aviation fuel, spacecraft integration, medical dosing (milligrams versus micrograms), and international construction.

The Takeaway

Every one of these failures was cheap to prevent and expensive to discover. The cost of a conversion check is seconds; the cost of missing one ranges from a rework to a $125 million spacecraft. When a number crosses a boundary — between teams, between unit systems, or between a human and a machine — treat it as a hazard, not as clerical detail.