Radiation Hardened Processors and CPUs, (moved from Uranus orbiter discussion) Options

[Guest_Geographer_*]

Forgive my ignorance of computer hardware but why are such antiquated computers the only things available for space missions?

[nprev]

As I understand it, it's a consequence of both long-lead times in development needed (you gotta freeze the configuration at CDR, which is still a long way upstream from launch, and any computer is obsolete in 18 months anyhow), and the arduous process needed to achieve spaceflight qualification for a given design.

As oDoug noted earlier, this qualification process is so stringent & time-consuming that 486s are the workhorses of manned spaceflight now, and that's for an operational environment that features on-site help for problems. UMSF needs, if anything, greater reliability.

[brellis]

In my work, switching software midstream in any project is verboten. UMSF "projects" can take decades to get from planning & design to completion. Once the bugs are worked out, it's better to ride with something that works than to take a chance with compatibility issues, crashes, etc.

[edstrick]

ALSO: very not-trivial point..

The total aerospace industry's use for rad-hardened processers like a 486 or pentium is so microscopically small in proportion to corporate and consumer markets that it's a drop in the bucket. Aerospace can't AFFORD the cost of bullding rad-hard versions of modern microprocessors.

[hendric]

Yet another point, I think, is that current CPU and memory technology is getting to the point where a radiation event is more likely to cause a catastrophic failure on the device. Current processors have gate oxides on the order of a few atoms thick. Imagine what a cosmic ray, or worse, a rad shower caused by a ray hitting something right in front of the CPU, could do. It's likely that space-rated semiconductor technology will forever be stuck in the 90's.

[edstrick]

It was a somewhat big geeky-space-tech news item maybe 3 years ago when the announced completion (AND LICENCING FOR PRODUCTION) of a rad-hard Pentium 1.

[Big_Gazza]

As long as the spacecrafts OS is tightly written, even a poor lowly Pentium 1 will have excellent performance. We are too used to horribly inefficient bloatware sucking the mips out of our desktops to really appreciate the awesome power of modern CPUs. I used to program 8085s and Zilog Z80s back in the 80s, and the modern chips are Star Trek technology by comparison.

[tasp]

Good point.

IIRC, Paul Monroe Hydraulics constructed a large centrifuge drive back in the early eighties that was controlled with a modified HP pocket calculator. Efficient software and a CPU running at 44 kHz (!!) was all it took to operate tons of complex electrical/hydraulic equipment. How many things occur simultaneously on a spacecraft ?? Even during a fast flyby, in the outer solar system, camera exposure times might be several seconds, in those seconds, even a (by our standards of today) primitive spacecraft microprocessor could execute thousands of lines of machine code. 'Bloatware' is probably a greater threat to mission survivability than most space environment hazards.

{Yeah, I am still touchy about all of my PC problems}

[smile]

[nprev]

Can't last forever, though. More and more complex functions (and the need to respond to unexpected events independently) will mandate greater processing capabilities as well as fatter software for UMSF. For orbiters maybe not so much, but for outer-system landers & rovers, definitely. When you're talking multiple hours of two-way light-speed lag, it's clear that the systems need real-time adaptivity; just entering a safe mode until hearing from Earth might not be enough to assure their survival, esp. on active bodies like Io, Enceladus, Titan & Triton.

[edstrick]

Indeed, you can do a lot with efficient code. My point was that production of a rad-hard version of an considerably long-time obsolete CPU was a significant feat. RAD-HARD is... HARD!

[mcaplinger]

It was a somewhat big geeky-space-tech news item maybe 3 years ago when the announced completion (AND LICENCING FOR PRODUCTION) of a rad-hard Pentium 1.

It was more like 9 years ago that the licensing was announced, there was a production update in 2002 (see http://www.sandia.gov/media/rhp.htm and http://sandtcolloq.gsfc.nasa.gov/spring200...oll_4-30-02.pdf ) and I haven't heard anything more about it since.

I don't think the situation is as bleak as made out in this (wildly off-topic, BTW) thread. Rad-tolerant FPGAs are evolving nicely and embedded soft processors in FPGAs do as much as most of our apps need. We're using this approach in our MSL and LRO instruments and it's been working out well.

[stevesliva]

The fab that I work next to may end up doing rad-hard manufacturing with the fully-depreciated tools that were making bleeding-edge technologies 10 years ago.

A lot of the special development necessary for the rad-hard stuff piggybacks on the development for the off-the-shelf vanilla process. Once the billion dollar fabs are fully depreciated and done making chips for Nintendo, you can save them from the scrap heap by doing something like analog ICs or rad-hard ICs or solar cells...

[nprev]

I don't think the situation is as bleak as made out in this (wildly off-topic, BTW) thread.

Good point on the OT. ElkGroveDan very kindly made this discussion its own thread.

I wouldn't call the situation bleak either, just evolving; be interesting to see which way it goes, esp. within the context of the classic battle between hardware & software capabilities.

The cause & effect relationships might well be a bit inverted for UMSF. Normally, hardware advances facilitate software development, but as we've noted hardware development for spacecraft often lags behind (in terms of processing capabilities) the current software state of the art.

[mchan]

While one of the advantages of rad-hard derivatives of commercial microprocessors (versus a completely custom design) is use of commerrcial SW development tools (compilers, linkers, debuggers, emulators, etc.), those tools still must go thru qualification for space use. Current commercial SW state of the art is indeed current, i.e., it is constantly evolving with the addition of new features, targeting new HW platforms, etc. while fixing bugs from previous releases. For UMSF, stability and relative high degree of being bug-free are required more than something that offers incremental code size reduction or performance increases.

An informative description of processors for space apps is in a SpaceLinux report.

And regarding "antiquated computers" on UMSF spacecraft, never tire of noting that some of them have been up there for a long time (> 10 years) and were designed years before launch.

[stevesliva]

Incidentally:
http://www.eetimes.com/news/latest/showArt...cleID=204300138

Rad-hard FPGA collaboration in 150nm technology when the latest consumer processors are at 65nm or 45nm. Have to multiply by root-2 a couple more times to catch up to today's processes.