Slashdot Mirror
Computers in Space Examined
Wil Harris writes "There's an article about the computers used in space missions over at bit-tech this morning. It covers the processor types and speeds, why space stations are less powerful than the laptops that astronauts take up with them and why tape storage is still de rigeur. An interesting and concise couple o' pages."
The fewer components you have, the less likely you are to encounter a failure.
I remember reading something about most space missions are pre-determined and very straight forward, there's no need for difficult maneuver like one has to execute in a X-Wing.
Having said that, there are still plenty of complicated, unexpected problems in space, but these problems have to be analysed and decision made by people on earth.
I guess it's all circumstantial, I can't even operate my 2001 Toyota electric window if the engine's dead, but my 1989 Toyota has no such problem. So if I crashed into a river, I hope I was driving the '89, but if I'm crashing into another car, I want my '01.
Rock that crushes, Paper & Scissors that don't matter.
Nuff said (but there's something to be said for the butlerian jihad, and Cmdr Adama filling his battlestar with rotary phones and manual typewriters!)
Don't blame Durga. I voted for Centauri.
Just like Cassini, the Hubble also has on-board solid state recorder (installed during one of the servicing missions), which replaced an old tape recorder. This has been really a nice addition as we can store more data into the solid state device while collecting data bits and dump them when the downlink becomes available. It really helps increase the efficiency of the satellite (and that's a big thing for science mission).
[Note that I've simplified the scheme alot here.]
Though several sections of the device have been damaged by radiation, or something, I hear. So even these things aren't too resilent to the harsh space environment, yet. Something you future engineers should think about as a project.
Anybody remember RCA's CDP1802, the weird little CMOS RISC-ish 8-bitter used in Voyager, Viking, and Galileo? These things have been running for decades, despite the radiation they've been subjected to.
Now that's engineering!
Unfortunately, radiation hardening a processor involves altering the fabrication process (some processes - e.g. SOI - are more resistant to bit flips than others), inserting guard rings, adding self-checking logic, and a bunch of other changes. Doing all of this stuff takes time (and money) so space-ready processors typically lag COTS processors by a generation or two. Example: the current "hot new" rad-hard processor is the RAD750, which is a rad-hard version of the venerable PowerPC 750.
Having said all of that, some small, risk-tolerant missions do use standard COTS processors (PowerPCs and StrongARMs are popular, as are industrial embedded processors like the Hitachi SuperH line). But you won't tend to find them in most NASA projects.
Some 9 years ago I worked on some chip design for Hughes and ESA.
Back then, we used 1.2um on 4" (or 6" in the new fab) wafers - and everything was built on a sapphire substrate instead of a silicon substrate to make them radiation hard (when they went through the van allen belt).
It was dull, as every single chip had about 12 inches of paperwork from QA. Every *instance* of every chip had its own paperwork, I mean. It was also dull because they wanted tried and tested tech, not any of this new fangled sub-micron stuff.
That was then. Can anyone let me know how much things have changed?
My experience with space rated equipment isn't all that extensive or current (I was involved in failure analysis of an AP-101 memory card that had an intermittent failure from the STS-2 and had some interactions with the engineers at IBM's old FSD division, which designed the AP-101s and wrote the flight software) but the article misses one very big point that is the really fascinating aspect (to me) of spacecraft computing hardware and I would have to challenge a number of facts in it.
1. The shuttle launch algorithms and orbital maintenance procedures are a lot more complex than the article makes them out to be. There are several hundred parameters that are continually checked, recorded and processed from tens to hundreds of times per second to make sure the flight path is correct and all systems are operating correctly. Along with monitoring the flight path, the computers were/are largely responsible for the data displayed on the astronaut/pilot's CRT displays in the cockpit.
2. It is my understanding, that in the early shuttle missions at least, there were multiple code loads during flight. The original AP-101s had a maximum of 256K words of 32 bit memory, which was enough for a separate launch, orbit and landing image, each which had to be loaded into the AP-101s before the next phase of flight. There have been issues with loading software or receiving and loading new software from the ground.
3. The original AP-101s were designed for the F-15 and could be considered "state of the art" for the early 1970s in terms of processing power and memory size. They are capable of about five MIPs and had a full megabyte of battery backed memory. They were chosen because they had been qualified for the high G-Loads and temperature extremes of the fighters. While the systems used on the shuttle were of the same design as used on the F-15 (and later the B-1B), they were inspected to much higher standards and all failures had to be resolved down to the point of having a test in place to prevent the failure from escaping the manufacturing/test processes as well root cause action plans at the component supplier.
The memory card failure that I was involved with was caused by a solder ball inside a metal RAM chip package. During the shuttle's ascent, vibration caused the solder ball to break free and intermittently touch the surface of the chip inside the package. The problem was extremely difficult to reproduce and was found by placing a microphone on the chip package and tapping the chip with the eraser end of a pencil. Chips with this solder ball defect "rang" differently than ones without this problem. After the ball was discovered and proven (by cutting open the chip package), every chip used in a shuttle AP-101 was tap tested by IBM to ensure no other solder balls were hidden inside the packages.
4. I don't know where that picture of the "Part of the AP-101S" came from as there is no way that is flight qualified hardware for an F-15, let alone a shuttle orbiter. Wire reworks are simply not allowed in high-G, high vibration environments and it looks like the surface mount components are hand soldered into place. I think this is prototype hardware that somebody pawned off on the author.
5. I don't understand where the idea that space systems having to be low power came from. The AP-101s were real power hogs (all their logic is bipolar) and were in fact glycol cooled. A significant fraction of the orbiter power generation is devoted to the compter systems (as well as the spacecraft cooling capabilies).
What is always interesting is looking at how the software for manned spacecraft is developed. A big joke is the Mars Observer and the mix up between English and Metric units, but think about how often you've heard about a software failure on board the shuttle - or any manned spacecraft for that matter. In Apollo, there were none and the software for the CM and LM computers was wire wrapped on a bed of nails instead of being burned into
Mimetics Inc. Twitter
I work for NASA on the manned programs and my experience is that hard drives are a headache on long term space missions.
u z_iss_010427.html
The laptops onboard Space Station are primarily IBM laptops (many of which will soon be running Linux - yeah!). While the drives are easy to replace, they fail fairly often (compared to other space hardware) and new ones need to be launched. The software on the drives also becomes corrupted frequently (maybe once every few weeks), requiring the crew to waste time recopying the software from CD. While these COTS laptops and hard drives were cheap up front (almost zero development cost, custom stuff would have been tens of millions of dollars) we are paying for it now because we waste a lot of operational time fixing them.
The Honeywell Command and Control computers (the primary flight computers onboard, which are triple redundant and manages core systems in the US segment) used to have a 300 megabyte hard drive to store flight software.
In 2001 during a shuttle mission, hard drive problems caused ALL THREE of those computers to crash simultaneously in a massive cascading failure. While it never got a lot of press, recovering from that took several days and an effort reminiscent of Apollo 13. You can read a contemporary article on it here: http://www.space.com/missionlaunches/launches/soy
When we got the things back and did a post-mortem, it turned out that the hard drive had a design flaw where the arm was dragging across the disk during power down and scratching it, which eventually led to failure.
They were replaced with solid state units shortly thereafter (which were already in the development pipeline). No moving parts, and much less problematic.
Worst...sig...ever!