Scattering the Code

A Repeating Whisper

Eli: November 14th, 2023. At NASA's Jet Propulsion Laboratory in Pasadena, the Voyager team expects Voyager 1's usual report, its engineering readings and its science. What comes down instead, as Sky at Night magazine later put it, is unintelligible. A repeating pattern of ones and zeros, from more than fifteen billion miles away.

Mara: At that distance, a radio signal can take twenty-three hours, one way.

Eli: And that's the listener's question. Who is still talking to the Voyager probes almost fifty years after launch, and how anyone fixes a spacecraft in interstellar space. Today's angle is the remote salvage of a 1970s interstellar scout across a two-day radio echo.

Mara: This story runs through a computer from the seventies, a team leaning on paper records, and a repair sent across nearly two days of radio round trip. How long the twins keep talking stays open.

Eli: The person in the middle is Suzanne Dodd, Voyager's project manager at JPL. In the summer of 2022 she told Insider, "Nobody thought it would last as long as it has." And then, "And here we are."

Mara: Telemetry is the spacecraft's running report on its own health.

Eli: In November 2023, that status went dark. The spacecraft was still talking, and nobody could read a word. Mara, it starts with who was listening.

The Interstellar Switchboard

Mara: They work at JPL, which Caltech manages for NASA. JPL designed and built both Voyagers, and the flight team is still based there. But Pasadena can't hear fifteen billion miles on its own. For that the team leans on NASA's Deep Space Network, which JPL also runs, with antenna complexes in California, Spain and Australia. And what they catch is faint. Voyager's routine cruise telemetry comes in at about a hundred and sixty bits a second. The smaller dishes, thirty-four meters across, can pick that up. Sending commands, or pulling down anything heavier, takes the seventy-meter giants.

Eli: So listening is the easy half.

Mara: And talking back is slow. With most of a day each way, nobody troubleshoots Voyager live.

Eli: Which puts the trouble aboard, in whatever machine was building that stream of ones and zeros.

Silicon of 1977

Mara: That machine is one of several aboard. Each probe carries three computer systems, every one built as a pair. Together they hold less than seventy kilobytes of memory, which The Guardian likened to one low-resolution image.

Eli: And they still record onto digital tape.

Mara: The one building the stream is the Flight Data Subsystem, two machines, one backing up the other. Dodd says Voyager "was designed with nearly everything redundant."

Eli: On Voyager 1, that backup was already gone. Its second Flight Data Subsystem had a memory failure on October 6th, 1981. More than forty years with no spare. Dodd has been blunt about it. As she told Ars Technica, "everything onboard is running on single-string." So when the stream turned to gibberish in 2023, there was no second machine to flip over to. Whatever had gone wrong had to be found in the only computer left.

Mara: And it was built to act alone. The delay made live ground control impossible.

Eli: Dodd has described what that buys. When things don't go normally, she said, "which is always a possibility, especially with an aging spacecraft," then "the onboard fault protection that's there can handle the situation."

Mara: Handy, unless the fault sits in the computer itself.

Eli: There is a precedent for rescues. In 2002, a part on Voyager 1 that turns readings into digital numbers was failing. The team consulted Voyager veterans, then cautiously switched on a backup sun sensor and star tracker untested since 1980. It worked.

Mara: In 2002 there was a spare to wake. This time there wasn't. Custom seventies machines like these have no modern stand-in to test on, which likely left the team leaning on the builders' paper.

The Memory Poke

Eli: And the paper was scattered. Voyager was planned for a four-to-five-year mission. Dodd has said that as the original engineers retired, "there wasn't a big push to have a project document library. People would take their boxes home to their garage." So answers lived with people. Linda Spilker, the project scientist, recalled leaning on those who remembered working on pieces of the hardware, and searching old memos to work out what to do.

Mara: Spilker goes back to the start. Voyager was her first job, in 1977.

Eli: She chose it over another mission after asking where Voyager was going. Now the team runs from veterans like her to people just starting out. Graduate students and young professionals are still making discoveries from its data. It's become a mission handed down through generations.

Mara: What nobody knew in early 2024 was which part had broken. But a lost signal had been won back before. On July 21st, 2023, planned commands left Voyager 2's antenna pointing two degrees away from Earth. Deep Space Network dishes still caught a faint carrier signal. Then controllers used the seventy-meter dish in Canberra to send a high-power command, and turned the antenna back toward Earth.

Eli: Both rescues leaned on hardware that still worked.

Mara: And no way to reach it but radio commands.

Eli: So in early 2024 the team went looking. First they tried the obvious, Spilker said, turning instruments off and on. That didn't fix it. Some engineers put up a big diagram of the computer on the wall, every path its data had to take, and as Spilker told it, "they just stuck sticky notes all over as they were figuring it out." Then, in early March, they sent what NASA called a "poke" command, prompting the Flight Data Subsystem to send back its memory.

Mara: Then nearly two days, out and back, before any answer.

Eli: What came back was a new signal. An engineer with the Deep Space Network decoded it and found a readout of the entire memory. Months after that repeating pattern began, the team finally had something to read. Whether it pointed to broken code or a broken part decided everything, Mara.

Scattering the Code

Mara: So the team had the memory, all of it.

Eli: Memory belonging to the Flight Data Subsystem, the computer that gathers Voyager's science and engineering readings and packages them for the trip home.

Mara: So it's the packer. Break it, and everything leaves scrambled.

Eli: And each of its machines holds about eight thousand words of memory. Not much room.

Mara: Which is why the readout mattered. Read alongside the wall map of the machine's paths, that dump of everything in memory narrowed the trouble to one chip. Spilker said it led them to the piece of hardware that had failed. One damaged chip, a broken patch of the computer's memory, spoiling about three percent of it. Why it failed is still open. An energetic particle from space might have hit it. Or it may simply have worn out after forty-six years.

Eli: One chip, behind all those months of gibberish.

Mara: And a chip can't be swapped from Earth, so the code had to move. The catch was size. No single free spot in memory could hold all of it. "The size of the memory was the biggest challenge in this anomaly," Spilker said. So they cut the code into pieces, tucked each piece into a different free corner of the memory, and updated every reference so the program could still find its own parts.

Eli: No backup to switch to, so the computer got rearranged instead.

Mara: The broken chip is still up there. The code just steps around it now.

Eli: On April 20th, 2024, the first result came down. The flight team was gathered in a conference room at JPL, with laptops, notebooks, doughnuts, and lucky peanuts on hand, a good-luck ritual that Spilker traces back to the Ranger missions of the 1960s. For the first time in five months, Voyager 1 sent health and status data the team could read. They cheered and clapped.

Mara: Caught by the Deep Space Network.

Eli: Science took longer. On May 19th they sent the command to restore it. Two instruments answered right away, and the others needed more work. By June 13th, all four were back. Even then, housekeeping remained: getting the timekeeping in Voyager 1's three computers back in sync.

Mara: The cruise telemetry, that steady stream of health and readings, was back.

Eli: Readable again. But fixing the computer made no new power.

Mara: Each probe runs on generators fueled by plutonium-238. Its decay gives off heat, and the heat becomes electricity. As the plutonium dwindles, output falls, by about four watts a year, NASA said in 2026. Meanwhile, Dodd has said, the transmitter alone takes "about 200 watts, approximately." NASA officials call the margins "razor thin."

Eli: And short power trips the fault protection, routines that shut off nonessentials on their own.

Mara: So the team chooses what goes dark. Each Voyager launched with ten instruments. By March 2025, each was down to three. One reading, held by Dodd, Spilker and program scientist Patrick Koehn, puts science first. NASA has said engineers avoided switching instruments off as long as possible because the data is unique, and Dodd has talked of hunting a few more watts to keep the science coming.

Eli: Mara, the other reading comes from Dodd too, and mission manager Kareem Badaruddin: protect the spacecraft first. In 2025 Dodd said, "If we don't turn off an instrument on each Voyager now, they would probably have only a few more months of power before we would need to declare end of mission." So instruments go off one at a time.

Mara: Either way, the code fix bought time, and the plutonium decides how much.

One Light-Day Out

Eli: The plutonium kept deciding. In February 2026, during a roll maneuver, Voyager 1's power dipped. On April 17th, engineers sent commands shutting down its Low-energy Charged Particles experiment, keeping its motor powered at half a watt. Badaruddin said, "While shutting down a science instrument is not anybody's preference, it is the best option available."

Mara: Four watts gone every year, and now even half a watt gets weighed.

Eli: So the team built a bigger move, called the Big Bang: swap running parts for lower-power stand-ins and push the next shutdown further out. Voyager 2 went first, in mid-2026, since it's closer and had a little more power to spare. NASA said the shifts should keep its instruments running "for at least another year." Voyager 1 is next in line.

Mara: And Voyager 1 keeps moving. It's projected to reach one light-day from Earth in mid-November 2026, about sixteen billion miles out. The Voyagers are the only working probes in interstellar space, and the longest-operating, most distant spacecraft ever. They take in situ measurements, sampling the space right where they fly instead of watching it from afar. Voyager 1's remaining instruments include ones that listen to plasma waves, picking up vibrations in the charged gas around it, and measure magnetic fields. Dodd has said what matters most is the record of how that medium changes as they travel, "that time history."

Eli: And Koehn calls each extra bit "valuable bonus science for heliophysics."

Mara: Catching it gets harder. NASA said in 2024 that Voyager 1's plasma wave data already needed five dishes arrayed together, combined into one bigger ear. As it moves farther, NASA said, it will need six.

Eli: And power sets its own clock. NASA has projected for years that all science instruments would be off by the early 2030s. The stated goal is to keep at least one running on each probe well into that decade.

Mara: So, for the listener: the ones still talking to the Voyagers are a compact team at JPL, Dodd, Spilker, Badaruddin, Koehn and the people around them, working through the Deep Space Network's dishes. And a spacecraft in interstellar space gets fixed with software, written on Earth and sent out to route around what's broken.

Eli: That stream of ones and zeros in November 2023 was one failing chip, read by a team that still knew the machine around it.

Mara: Three things to watch. Whether engineers ever settle if that chip took a particle hit or simply wore out. Whether Voyager 1's charged-particle instrument comes back once the Big Bang reaches it. And when the generators fall below what the transmitter needs, expected sometime between 2032 and 2036.

Eli: Listeners can look into the Flight Data Subsystem.

Mara: Follow science with us as aging probes push deeper into space.

Eli: Thanks for spending this time with us. I'm Eli.

Mara: And I'm Mara. Until next time, this has been Angle.