Space 4 November 2026 4 MIN READ
Why Space Computers Are So Old
Radiation flips bits in orbit and wears chips out. Two ways to cope: expensive hardened processors, or fast ones with redundancy and error correction.
THE SHORT ANSWER
Radiation in space flips bits and wears out chips, so spacecraft computers are built to expect errors. Perseverance uses a hardened processor costing over $200,000 per chip, while Orion uses commercial processors with redundancy and error correction. Voyager 2 once had a single bit fixed from Earth.
On this page
Two ways a particle can break a chip
Space is full of fast-moving particles, and electronics are not built for them. NASA Ames researchers describe two routes to failure. A single proton or heavy ion can strike a chip and cause what engineers call a single event effect, a serious problem for flight software that can end a mission. Or the chip can simply wear out as radiation accumulates: the cumulative dose.
Two kinds of single event matter here. A single event upset is a transient flipped bit: a 0 becomes a 1, and the error disappears when the memory is rewritten. A single event latchup is potentially permanent, a stuck bit. Even in low Earth orbit these disruptions are common.
A flipped bit, fixed from space
In 2010, NASA’s Jet Propulsion Laboratory traced a change in the pattern of Voyager 2’s science data to a single memory bit that had gone from 0 to 1 in the spacecraft’s flight data computer. Engineers reset it with commands sent on 19 May 2010. A signal takes nearly 13 hours to reach the spacecraft. Science data resumed on 23 May.
JPL’s page identifies the flipped bit but does not name the cause. The Planetary Society’s explanation covers the general mechanism: a cosmic ray can pass through a spacecraft’s protection and change a bit.
Strategy one: harden the chip
The classic answer is a radiation-hardened processor. Perseverance on Mars runs on a BAE Systems RAD750, a licensed design re-engineered with added logic to detect and correct memory corruption. New Scientist reported in 2021 that it has only 10.4 million transistors and that each chip costs more than $200,000. Earth-grade chips, it says, would crash regularly on Mars or fail permanently. BAE says the RAD750 flies on about a hundred satellites in Earth orbit with no failures, a claim that comes from the vendor.
The US Air Force Research Laboratory funded the technology behind the RAD750, the radiation-tolerant Xilinx FPGA and the Honeywell flight computer used on Perseverance, and says NASA uses radiation-hardened electronics because they are reliable, affordable and commercially available. A RAD750 computer flown on NASA’s Lunar Reconnaissance Orbiter carries 36 megabytes of radiation-hardened memory and 4 megabytes of non-volatile memory.
The NASA Ames paper explains the drawback of hardware-only protection: such flight computers are slow, expensive and out of date, and they draw more power and add launch mass. That is the motivation for dealing with errors in software.
Strategy two: accept errors and correct them
NASA’s Orion spacecraft, which flew the Artemis II mission, takes the opposite approach. According to NASA, its avionics are mostly software and commercial processor technology that has been ruggedised to withstand extreme radiation. The flight software is designed to keep working when radiation or a lost communication link affects it. A backup flight computer runs independently developed software and takes over if all primary computers fail. Because the backup software is written separately, a single software bug cannot take out both.
TechRadar, summarising an article in Communications of the ACM, reports a design with two vehicle management computers, each with two flight control modules, and a pair of checking processors in each module, eight in all. A faulty computer is meant to fail silently instead of sending a wrong answer, and the memory is triple-redundant, correcting single-bit errors on every read, the standard countermeasure to bit flips. NASA’s own page confirms the redundancy and the backup computer but not the count of eight; that detail rests on the secondary report.
Why it matters
Both approaches are in use, and neither is the one right way. Hardened chips are slow and costly but predictable. Redundancy lets a spacecraft use faster modern parts, at the price of more complexity.
Our take: in space, an error is an expected event. The question is not whether a bit will flip, but whether the machine notices, corrects it, and carries on.
Do you agree?
Take it to Chevlume: write your view, add your sources and let others answer. The closed alpha is by invitation.
Keep reading
The Voyager Golden Record
What is on the disc launched into interstellar space in 1977, how an alien is meant to play it, how long it may last and the arguments around it.
SPACE / DISCOVERYMars Climate Orbiter: Pounds and Newtons
One unit mix-up put a Mars orbiter about 170 km too low. What the NASA board found about the people and process failures behind it.
SPACE / DISCOVERYAriane 5 Flight 501: Thirty-Seven Seconds
How reused software and one unprotected number destroyed a new rocket in 1996, and what the official inquiry recommended.
Sources
- Expecting the Unexpected - Radiation Hardened Software (Mehlitz & Penix) · NASA Ames Research Center via NTRS · Accessed
- Engineers Diagnosing Voyager 2 Data System -- Update · NASA JPL · Accessed
- Update on Voyager 2 status · The Planetary Society · Accessed
- The Perseverance rover runs on processors used in iMacs in the 1990s · New Scientist · Accessed
- AFRL Technology traveling to Mars · Air Force Research Laboratory · Accessed
- RAD750 SpaceWire-enabled flight computer for Lunar Reconnaissance Orbiter · BAE Systems / NASA GSFC (SpaceWire Conference 2007) · Accessed
- Avionics and Software · NASA · Accessed
- NASA built a spacecraft computer that can lose three systems mid-flight and still ride through · TechRadar (summarizing CACM) · Accessed
Published
Discuss on X