15
$\begingroup$

What type of computers are used (generally) in robotic ( i.e. space probes and rovers) vs human craft (such as the ISS, soyuz)?

Subquestion: What kind of specs are important on computers to operate in space?

I ask because I was curious as to the differences between say the laptops that they sometimes use on the ISS and space probes and also what kind of electronics principles do and don't work (or just are and are not efficient) in space.

$\endgroup$
15
$\begingroup$

The laptops on the ISS can only be used for non-time critical uses. Those laptops are used for some mission critical tasks, but only if the task can handle the time needed to replace a toasted laptop with another. Those laptops are commercial off-the-shelf, and they do not fare well against cosmic rays. Most cosmic ray hits result in a "single event upset", where bits in memory get flipped. Reboot the computer and it's usually fine. Usually. Some cosmic ray hits do more damage and the computer truly is toast.

Most flight computers used to control safety or mission critical aspects of a spacecraft or a robot are extremely expensive, radiation-hardened computers whose ground counterparts were state of the practice (not state of the art) in some previous millennium.

The one exception is SpaceX (see http://aviationweek.com/blog/dragons-radiation-tolerant-design), which achieved the required redundancy by using lots of computers. The tradeoff is a bit more expensive software to keep all those computers in sync and find the ones that have gone off the ranch versus much less expensive and much, much, much more capable computing hardware. SpaceX chose the off-the-shelf flight computer route. Flying with computers that were state of the practice fifteen to twenty years ago puts severe limitations on what can be done onboard.

$\endgroup$
  • $\begingroup$ Ran Ginosar's "Survey of Processors for Space" (2012, PDF) provides a decent overview of how radiation issues are handled and some of the tradeoffs involved. $\endgroup$ – Paul A. Clayton Jul 4 '14 at 15:29
12
$\begingroup$

NASA Built-in Hardware

The dedicated mission computers are tested extensively for vibration resistance, resistance to cosmic and solar radiation (both particulate and EM), magnetic fields, and are also prioritized for low energy consumption and high reliability. Generally, this results in processors a generation or three older than current desktops, and often in larger architecture versions and slower speeds.

Further, the motherboards and enclosures are often custom made designs, like the Shuttle's upgraded AP-101B, using a 400 kHz single core main processor, and 24 dedicated IO processors, in a radiation resistant case with a vibration resistant motherboard, and a very small amount of memory - a few hundred kilobytes.

Many satellites use off-the shelf processors; for years, the Zilog Z-80, Intel 8080, and Motorolla 68000 were staple processors, because they were both inexpensive (having outlived the patents), and had already passed the NASA testing requirements. Newer processors have since replaced them in common use, but missions into the late 1990's made use of NASA-specific production 8080 chips. (A top-end Z80 is 20 MHz, 8-bit clean, single core, and $12. This has resulted in a couple of students looking to use it on microsat projects - it's cheap and space rated.)

Off the Shelf Hardware

NASA has, since about 1992, allowed the use of off the shelf IBM laptops for some mission critical applications. They've found that the devices are unreliable in orbit, due to magnetic field fluctuations. Still, they are reliable enough to be used in general purpose non-flight-essential uses.

Hand calculators, including HP-41's, have been used by several astronauts over the years; these devices have larger circuits, slower processors, less memory, and have always been used only for non-mission-critical applications. (The HP-41 also was used to simulate the AP-101. It was also the backup plan in case of failure of the AP-101 cluster.)

References

http://www.nasa.gov/mission_pages/shuttle/flyout/flyfeature_shuttlecomputers.html
http://spaceflight.nasa.gov/shuttle/reference/shutref/orbiter/avionics/dps/gpc.html
http://hpinspace.wordpress.com/category/hp-41/
http://www.hq.nasa.gov/pao/History/computers/Ch4-6.html

$\endgroup$
  • $\begingroup$ How would magnetic field fluctuations cause unreliability? $\endgroup$ – Hobbes Jul 5 '14 at 17:05
  • 1
    $\begingroup$ @Hobbes two methods: 1, by allowing more particulate radiation in, and 2, rapid changes directly induce currents. #1 is the main issue. #2 is almost a non-issue, but note that I've seen certain low power processor devices become unreliable due to power lines. (A friend's smartphone was built on a 1.2v processor, and wasn't well shielded - passing under a pair of high power lines would cause a reset.) $\endgroup$ – aramis Jul 5 '14 at 17:12
  • 2
    $\begingroup$ Correct me if I'm wrong, but I thought the off-the-shelf Z80, 8080, and 68000 were all NMOS/CMOS/HMOS, vulnerable to cosmic ray interactions, and therefore unsuitable for use in space. However, I believe a number of "classic" CPU designs (perhaps all of the above) had derivative implementations which were radiation-hard, but these versions would not have been available as "off-the-shelf" items alongside the originals (only produced as special/custom orders, or available "off the shelf" only much later), and when available, would have been much more expensive. $\endgroup$ – Anthony X Jul 5 '14 at 18:17
  • 2
    $\begingroup$ @AnthonyX ALL semiconductors are subject to cosmic radiation interactions. CMOS isn't as much an issue as voltage (5v is better than the current high end 1.5v; an induced cascade can more easily overload 1.5v machines), and size of wiring (which correlates to amps drawn; further more amps and wider gates is both more power and more resistant to random single particle interaction failures). The newer versions of the Z80 are more prone to failure due to smaller junctions and lower amperage and voltage. $\endgroup$ – aramis Jul 7 '14 at 22:43
  • $\begingroup$ It's a bit hyperbolic to say the HP calculator simulated the AP-101. It simply had some of the same equations programmed into it. And if the "cluster" had failed, the shuttle would have been lost since it was a fly-by-wire design. No calculator would be of any benefit in that scenario. $\endgroup$ – Organic Marble Dec 13 '16 at 13:18
2
$\begingroup$

For mission-critical computers (flight control, etc.) ESA uses the LEON processor, based on Sun's SPARC architecture. They used other custom designs before LEON.

$\endgroup$
  • 1
    $\begingroup$ It might be worth mentioning that LEON uses radiation-hardening by design (special circuit techniques et al.) rather than a radiation hardened manufacturing process, distinguishing it from something like the RAD750. Using hardened by design allows using less expensive and more advanced ordinary manufacturing processes at the cost of some area (which is typically more than compensated by the more advanced process) and significant design effort. $\endgroup$ – Paul A. Clayton Jul 7 '14 at 17:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.