NASA’s Mars Data Mission
When all the hugs and rounds of cheering at NASA’s Jet Propulsion Laboratory in Pasadena finally died down late last Sunday night, the scientists, engineers and navigators who had just improbably placed a car-sized planetary rover onto the surface of Mars dutifully returned to their terminals to commence years of surface exploration.
The details of the entry, descent and landing, almost preposterous on the surface, make it no surprise that many years of planning, calculations and building led up to the daring touchdown of Curiosity. The 14 minute transmission delay between Mars and Earth meant the landing would meet its fate before the signal arrived back at Earth. It inspired a viral movie JPL dubbed “7 Minutes of Terror.”
But to any follower of computing technology and especially to a computing or information professional, the data now flowing back and forth between Mars and Earth is a telling study in IT engineering minimalism using technology as simple as could be practical to do many things very reliably.
Ann Devereaux was deputy lead for entry, descent and landing (EDL) stage of Curiosity’s mission. She also built the Electra-Lite radios that are on the Mars Science Laboratory (MSL) and Mars Reconnaissance Orbiter (MRO), a project she started in 2004. Her job up through EDL was to make sure that computers would be doing what they were supposed to and that all the data interfaces on the rover and orbiters would be working.
Image courtesy FlordaToday and Tom Walters
With EDL accomplished she’s turning to tactical duties with the same mission equipment that took Curiosity to Mars, a slim stack of working gear that all but the earliest Internet users would find quaint. The central processor aboard the MRO, an IBM PowerPC 750 hardened to deflect radiation, clocks at 133Mhz and comes from a generation of chips introduced in the 1990s. The code footprint is 20 MB with 128 MB of total execution space. The entire file storage for telemetry and science data storage is 4.5 GB.
Put another way, the computing that powers NASA’s most sophisticated space exploration mission ever is pretty much the dead opposite of the technology you’d use to make a Hollywood movie about space exploration.
“When people saw the data volumes we process and the memory space they’d inevitably ask, ‘You're taking this to Mars?’” Devereaux says. “Why isn’t the latest gear? My memory stick has more memory than your whole computer.”
It Has to Last
As Devereaux explains, NASA’s big design constraint is finding parts that are “space qualified,” a term that figuratively translates to “bulletproof.” Everything on board including the radios has some computing inside just as an iPhone does, but there is no getting around the use of older technologies for a few reasons.
For starters, she began building mission components eight years ago, which predetermined the lineage. And, newer electronic circuit designs have some drawbacks in space. Before microcircuits were commonplace, components ran at higher voltages and were less sensitive to radiation bombardment and extremes of temperature. For military or NASA use, manufacturers spend a lot of time and money to be approved as space qualified; it takes a long history and a big database of performance reliability.
“You need to have super controlled processes to know parts are still going to work in 10 years,” says Devereaux. “If you are Dell or Apple, and 10 percent of a chip run fails, you just throw it on the floor. And if one out of 10,000 iPads goes bad because of a memory chip or a video driver, you just replace it.” NASA, she confirms, has no service centers on the Red Planet.
Like the CPU and memory, the bandwidth of data coming from the rover is modest and arrives from two transmitters. One points directly to Earth and operates at about half the speed of a home telephone modem. The other handles UHF data relays between Curiosity, the orbiters and JPL in Pasadena, and is Devereaux’s mission focus.
The rover’s UHF transmitter sends data to either the Mars Reconnaissance Orbiter or Odyssey Orbiter, an older space craft with an older radio. All the transmissions have to take place in 15 minute windows three or four times a day depending on geometry “when the orbiter goes screaming overhead,” in Devereaux’s words.
The orbiters act like cell towers with more power and storage than the rover and a longer look at Earth’s horizon. “They suck up whatever data Curiosity can give them every day and when have a view they can dump all the data back to JPL at once,” says Devereaux. “It makes for a very chunky sort of day where you plan everything around when you uplink orders to Curiosity to do stuff during the day and when you downlink and assess all the work you did that day and how it went.”
Even the 15-minute windows are unreliable with signal strength rising as the orbiter moves overhead and falling as it regains the horizon, says Devereaux. “We get a lot of junky links until the orbiter is straight above, so on the newer Reconnaissance Orbiter we use adaptive data rates where the orbiter monitors the signal strength and can tell the rover to speed up or slow down its transmission.” The aim is to get the absolute most out of 15 minutes, and NASA is starting very slow in order to optimize and capture the most important information cleanly.
Data comes in three streams: telemetry to report temperature, power and other onboard conditions; computer process reporting (of the type you'd see in a DOS or UNIX computer booting); and the actual science data, a pipeline that will grow only slowly.
It is slow enough that NASA measures bandwidth in bits rather than bytes (8 bits = 1 byte). Curiosity can transmit up to two megabits per second to Mars Reconnaissance Orbiter; the older Odyssey radio can only handle 256 kilobits, one-eighth as much. Right now the downlinks reported at press conferences are on the order of 10s of megabits of data a couple of times per day.
“We are operating the vehicle very cautiously and we want to see the health data first,” says the scientist. “There was a lot of data transmission to reconstruct from entry, descent and landing and that had the highest priority by far.”
Transmission rates are partly a function of power, and Curiosity’s nuclear battery puts out a steady 100 watts, just a fraction of what’s needed to run a blow dryer. So, solar panels charge other batteries, but every day’s planning has to carefully account for activity.
“If we are going to run the [robotic] arm, god forbid, we have to calculate the power it takes to heat the mechanisms that move it, to use an instrument the arm is providing samples for plot all this out,” Devereaux says. “When you get to a certain battery state you have to stop and recharge and recalculate.”
All the variables that made the successful beginning of the Curiosity mission so remarkable are based on sound engineering, science and mathematics. It could not have been done with wholesale stacking and layering of infrastructure as happens in mainstream IT and data management.
“Compared to most IT development you have to program in very clever ways to work on Mars. The best result is that you end up with just the key crisp points of data.” Devereaux compares her task to an email that might read, “Hi, hope you are well, I wanted to tell you X,Y and Z and remind you of a couple of things we are working on…”
”All we get to say is, ‘Hi’ and we’re done,” Devereaux laughs. Her analogy is that anyone can generate large amounts of numbers and load them to a server farm, “but it takes a clever person to do Sudoku where everything is tiny and has to fit in boxes.”
NASA’s mission is also to inspire the public to future scientific exploration and it prioritized activities to be able to deliver that first grainy shot of Curiosity’s wheels on the Martian surface almost immediately. It was a bit of fortunate timing that also lifted the whole project team.
“It was funny because we’d tested entry, descent and landing hundreds of time in the test bed and dozens of times in dress rehearsals where the computer says everything is working,” says the scientist. “We knew it was real but it was a weird experience to see the data you’ve looked at dozens of times and then you see that picture and it’s really sitting on Mars and it hits home.”
Devereaux, third from left, celebrates at JPL.
Now, she says, an impatient consumer audience is expecting streaming video, but like everything else, only in due time. “That video stream would be kind of a killer app that people want to do, but the storage and bandwidth will dictate when it’s possible. Plus, I’ve got to tell you, the surface of Mars doesn’t change that much so the rover drivers might go crazy for it but you might not see much new.”
Also, compressed video can’t give scientists the accuracy they want. It might be fine “for watching ‘Battleship’ on NetFlix,” but scientists want lossless compression because they want to see every shadow and detail that uses the sun to estimate position and reveal textures.
The audience engagement part she gets personally. “I will tell you when we were doing EDL and there was this pause and blank screens until the UHF radio kicked in once we’d gotten rid of the parachute that was covering the antennas. Someone called out, ‘We’re getting Odyssey back,' which meant UHF data through Odyssey to the ground, all the screens started populating and I jumped up and screamed oh yeah, Electra! This is what we waited eight years for from the time we built this radio, the first time we’d used it and seen it work. The whole team felt this way.”