View Full Version : Radiation Resistant Computers

2005-Nov-21, 07:08 PM
SUMMARY: Modern spaceflight is dependent on reliable computers to handle navigation, life support, and other functions. The problem is that radiation in space, such as cosmic rays can cause computer chips to calculate incorrectly. NASA is working a solution that would run multiple redundant computers to do the same calculation several times over and then vote on which is the correct result. If a cosmic ray caused one processor to make a mistake, the other processors would still be correct, and the error would be prevented.

View full article (http://www.universetoday.com/am/publish/eaftc_resistant_computers.html)
What do you think about this story? post your comments below.

Dark Jaguar
2005-Nov-23, 09:19 PM
Now that is a clever solution. I'm reminded of the solution to early issues with laser disks leading to the standard file system on CD-ROMs. Firstly the data was encoded in such a way that if a bit was incorrectly read, a lot of the time it would result in an "invalid" byte, due to the formatting standards, leading the drive to read from the disk again. Another fix was redundancy. This cost some storage space overhead (I believe that a byte is 10 bits in that system, though the extra bits mean that it is effectively 8 bits for programmer's purposes), but the idea was simply that if a lot of the bits were repeated, then if they couldn't be read then the data may still be there. These two things combined to make CD-ROMs reliable enough for the average consumer to buy them. RAM also has the parity solution for determining data storage errors (bits are always paired, if the total bits in a section come up odd, the computer knows there is an error, at least that's how I understand it). These methods work well on Earth anyway. As for space, I guess redundant processors are the only way to go. Random electrical signals in completely unpredictable parts of a computer could certainly produce some major errors (most of them fatal I'd imagine), so I'm really surprised computers in space operate as well as they do. I suppose the cosmic rays hitting an electrical component are rare enough that we can at least count on some of that stuff working some of the time.

2005-Nov-24, 05:15 AM
tell-me-three-times has been around a long, long time - just about as long as computer themselves. I remember hearing about it when visiting my dad at one of the physics plants he was working at (working for his PhD)..

The concept of 'how important is this decision' is pretty keen, though.

DJ - the article mentioned that NASA uses hardened chips - the transistors require more energy to flip than is normally derived from the passage of a cosmic ray. The problem is they are slow, hot, and heavy - makes much more sense to sacrifice a little weight for a redundant system (heck, I use redundant (2x) systems at work every day)

2005-Nov-24, 03:20 PM

Radiation is a problem in space.
An astronaut explained that one of the problems of living in space is that if you close your eyes to catch some sleep, you have to get used to those flashes of light which are caused by the flow of radiation which is hitting the retina.
Near Jupiter, the radiation is so strong that its dose, just applied for a second, would be deadly for a human being.

The best protection against radiation may be to apply effective shielding. Careful design, redundancy of systems and checking the correct operation of every system by comparing and weighting the results are effective, too.
Processors are not intelligent. Itīs the designers and the programmers which are intelligent. One should not mix up speed with brain.

Reducing data by processing them by powerful technology "in the field" means that you only get certain results, but not the data from which the results were derived. If your idea about what to be found "in the field" and what to process in the field is different from that what actually exists "in the field", you system may give answers which satisfy you, but which are absolutely imperfect, because you may only see what you were looking for, but everything else may evade your perception. If your concept is limited to the detection of elephants, you may find elephants, but you may overlook the existence of ants.

I find that most consumer chips produced nowadays are in production and application for a very limited period of time. Every chip generation is usually replaced by the next chip generation after some months. At such short life cycles it is not possible to detect all faults and bugs and understand or even repair them. The industry cannot afford to have an interest in its products to last, as products that last and thus satisfy the customer will lower the demand for new products.

In space, the equipment must last, and canīt be replaced easily. Consumer processors, no matter how powerful and cheap, may be inadequate for the application in unforgiving and rough environments.



Dark Jaguar
2005-Nov-25, 06:55 PM
Ah, so not only has this technique been around a while, it has a cute nick name too? Good to know. I should have guessed as much...

2005-Nov-26, 05:55 PM
Do you mean Environmentally Adaptive Fault-Tolerant Computing (EAFTC) ?
Certainly trips off the tongue.

Or do you mean Tell-me-three-times perhaps?
Yes, that is kinda cute.

of course one day, we all might end up as programs running on such a computer, so I shouldn't knock them...