<html><body><div style="color:#000; background-color:#fff; font-family:arial, helvetica, sans-serif;font-size:12pt">On Saturday, November 24, 2012 1:18 PM spike <spike66@att.net> wrote:<br><div><div>>>... On Behalf Of Dan<br>>>...I have no problem with looking at past failed efforts and see this a<br>>> problem with some research... it might be easier just to do all kinds of<br>>> trial and error tests that lead to solutions that might never be found if<br>>> one worked from known theories or known working applications alone.<br>><br>> Dan this is one of those special cases where a lot of trial and error would<br>> produce very predictable results. We have a microprocessor, we have in<br>> enclosed in something. Whenever a sufficiently high energy particle comes<br>> in, it hits a solid surface which is held together by chemical bonds. If<br>> the kinetic energy of the particle is
sufficiently high, it does not matter<br>> what those chemical bonds are or how strong: the electromagnetic attraction<br>> that forms bonds is negligible compared to the energy of the incoming<br>> particle. So the particle whacks loose a shower of secondary particles, all<br>> of which have lower energy, but they are charged, since that secondary<br>> shower contains electrons and protons, along with plenty of other oddball<br>> ons which we never see outside of some super high energy event, such as in a<br>> particle accelerator.<br>> <br>> So we get this shower of secondary particles that hit the microprocessor,<br>> and we know by Maxwell's equations an incoming charged particle creates an<br>> EM field which in some rare cases can flip a bit, which does unpredictable<br>> things to the process or memory. Sufficiently geezerly types here may<br>> recall playing Core War back in the 70s. The
space environment plays a<br>> pumped up version of Core War with our processes, but rather than just flip<br>> a bit, in some cases it might actually fuck up the processor hardware<br>> itself. So it is possible that there is a smallest practical integration or<br>> smallest practical interconnect that can be used for space applications in<br>> which the mission depends on the processor working perfectly, if we keep<br>> with our present software paradigm.<br><br>I understand the physics. This wasn't my point. My point was rather that, if the price goes down low enough, all sorts of different solutions can be tried _in situ_ and solutions might crop up that one wouldn't expect from first principles or from ground tests.<br> <br>> To make matters worse, usually whenever there is a holy grail to be found,<br>> there are many seekers, and they are not eager to tell the others where they<br>> have already
searched. The failed attempts are valuable trade secrets. I<br>> can imagine the lads up at Berkeley must be amused at the parade of geeks<br>> that come thru with the same old tired ideas that the accelerator lads<br>> already know will fail. But they are not at liberty to divulge that<br>> information, and wouldn't in any case, for that geek parade pays the bills<br>> up there on the hill and keeps the lights on up at the cyclotron.<br><br>That's another related problem and partly falls under the general issue of intellectual property.<br> <br>> I am of the notion that the answer is to go to massssssive parallelism,<br>> massive. We need not three but about thirty processors, where a majority<br>> vote means something. Then we can go ahead and let them shrink to these<br>> super-tiny integrations we see on the market today, knowing that some of<br>> them will get zapped by cosmic
rays. In the meantime, the good old i486 is<br>> still used in some flight applications.<br><br>But that is the solution that's in place. (Of course, when you write "massssssive parallelism" I wasn't expecting you to talk about "thirty processors." That's not massive in most applications today, is it? I mean, back in the 1990s, massively parallel meant systems with an order or two higher in number of processors.) Even before the link Stephen sent in, I'd already read (and years ago at that) that the STS had, IIRC, five computers on board that crossed-checked each other and voted when there was a disagreement. The same is going on on the Falcon system.<br><br>I do see a problem with this if you get to the point where many components have degraded to the point where overall whole system reliability, even with voting, is affected. But that shouldn't need to be pointed out.<br><br>Anyhow, the whole thing is a known problem. My initial response here
was not about what the physics was or not understanding the problem, but my surprise that this was a news item. Next, I expect a news story on the Sun not orbiting Earth. :)<br><br>Regards,<br><br>Dan<br> </div> </div> </div></body></html>