Slashdot Mirror
The Ancient Computers Powering the Space Race
An anonymous reader writes "Think that the exploration of space is a high tech business? Technology dating back to the Apollo moon landings is still used by Nasa mission control for comms and the 1980s 386 processors that keep the International Space Station aloft."
I'm not sure if it is still the case but for a LONG time 286 processors were the only ones available that had been hardened against cosmic radiation and were rated for space. When you're lobbing people into space, it matters most what works and is proven, not what is fastest or the newest technology.
Nobodies Prefect
Tidbits for Techs Technology Blog
The Ancient Computers Powering the Space Race
From general agreement on the definition of the Space Race:
The Space Race was a mid-to-late twentieth century competition between the Soviet Union (USSR) and the United States (USA) for supremacy in outer space exploration. The term refers to a specific period in human history, 1957-1975, and does not include subsequent efforts by these or other nations to explore space.
Emphasis mine. As to the 'ancient tech', it's stable and still working so what's the problem? People are bitching about rising taxes not the fact that we are stunting ourselves in exploring space. It's not 1975 anymore, people have moved on to other international penis/rocket/missile envy matches.
In related news, the house fails to agree on a meager NASA funding bill while space tourism continues to progress.
My work here is dung.
Adding that the CPU's are also custom made, along with it's embedded operating system, to withstand the operating environment.
http://www.cpushack.com/space-craft-cpu.html
It's the same in any long-life service, like space and military. For example the Aegis missile system runs on 286s and 386s while the busses run on a sedate 200 kilohertz speed. There have been recent upgrades to "new" PowerPCs or Pentiums, but only for a few select ships.
There are even some strange home users that still run on primitive CPUs from the Seventies! Like 6502, 8088, and 68000
"I disapprove of what you say, but I will defend to the death your right to say it." - historian Evelyn Beatrice Hall
It has been 4 + decades since the space program dominated electronics development.
Anyway, by the time any piece of electronics gets radiation hardened and goes through the "soak" - i.e., a few simulated years or decades worth of cycling through heat, usage, etc., plus fixing any uncovered problems, it is by definition not going to be cutting edge.
It's good that space computers are more commonplace, anyway. Viking 1 died because JPL couldn't afford to keep the people who understood the archaic assembly language for the landers in the ramped down extended mission team.
The last 20MHz RAD6000 flight board we bought was around $250k. A flight FPGA runs about $5k each. 10 times is actually quite an understatement for radiation hardened.
Virtually anything related to space has a huge development cycle. Contract bid to delivery is easily 5+ years. One of the first things you do is source your suppliers so you will never deliver anything state of the art. It'll be at least 5-10 years old. At pretty much the same time you have to also deliver most of your spares for the near or distant future. And there probably is no money in the contract for hardware upgrades. It is what it is until it's replaced.
... much of the flying hardware designs are decades old too - but this is IMO due to so much of it relying on govt funding or govt being a primary customer. It seems that there might be progress on this front, though - with the like of Musk, Bigelow and perhaps even Branson (suborbital now - but it's a good start). Guidance computers do not need to be terribly powerful - they need to be reliable. Witness what happened to the first Ariane 5 launch. It wasn't very long ago that the venerable COSMAC 1802 gave way in space platforms to more recent CPU/MCU designs. While quirky, it was well understood and inherently resistant to radiation upset.
A '68 Chevelle, properly optioned, will get better gas mileage than most cars built today (not a Prius, but most cars). Said Chevelle will get FAR better fuel economy than any car in its (midsize) class today.
Expect a 6 cyl Chevelle to get 20% better fuel economy than a 4 cyl Camry/Accord/Taurus, and a 307 Chevelle to get 10-15% better fuel economy than a 6 cyl Camry/Accord/Taurus.
So far, nobody has brought up the complexity explosion. To make modern processors really sing, requires a lot of work in the compilation phase. That means the instructions the coder writes are not straightforwardly coupled to the instructions on the hardware. If you really want to audit the software you rely on to make sure you don't "Need another seven astronauts" then code review is insufficient; you have to look at the hardware instructions too. That is another thing keeping mission critical paths on older hardware --- older hardware tends to be simpler hardware.
"Mil-spec" and "radiation hardened" are not hardly the same thing. A typical military system does not used radiation hardened parts - they're unnecessary. However, chips used in military hardware have to go through extensive proofing to ensure that there aren't sneak circuits, single point failures, etc. That costs money and takes a fair amount of time. You also need to understand that those "mil-spec" and "radiation hardened" pieces of hardware are not designed nor manufactured BY the military or the federal government - they're made by commercial entities, and it's those companies that charge "$30,000 for a hammer". It's called the ACQUISITION process for a reason.
They do. People are constantly making new rad-hardened chips, mostly for commercial satellites. The latest LEON (SPARCv8) chips go up to about 25MHz in the rad-hardened version. It's not just a matter of using a slightly older technology - space is an incredibly IC-hostile environment.
I am TheRaven on Soylent News
The use of the older chips has many reasons. The items above certainly apply. One the issue of cross talk due to pervasive radiation on newer smaller die chips makes them almost unuseable and certainly not for any critical system. If you write a program to be error proof it will spend all its time correcting errors rather than processing on a newer chip. Secondly you have to consider wether the power should be in the programming or in the chip. If a chip fails you can maybe replace it, though how many times may become an issue. If a program fails it can be reinstalled. Thus the use of a more solidly built and radiation resistant archetecture combined with a program that handles more processing decisions might be safer all around. Untill we work out a way to put a forcefield around a station we will be limited severely in the level of functional current technology we can put up there. I find the concept of space as a friendly place to be at odds with the level of knowledge that we have amassed, admittedly small though it is, i think about space as a place that hates Human life and will take any tiny chance and use it to kill, it becomes easier to acceppt the lower level of technology we use up there. Needless to say, if we can move beyond electricity based computing there might be hope.
It's not that it would knock out a track. A single cosmic ray hit will not ablate the metal layers. It's that the newer parts use much lower voltage to get lower leakage to get higher speed. Lower voltage == lower gate charge, in some cases the difference in charge states is < 100 electrons*. A single cosmic ray is capable of changing the charge state on these gates enough to make a bit undefined. That is a BadThing(tm).
-nB
* My info is specifically on flash and a couple years old.
(n-m)==100.
0-m electrons on the gate == logic 0
n+ electrons on the gate == logic 1
between m and n electrons on the gate == undefined value.
whois gawk date unzip strip find touch finger mount join nice man top fsck grep eject more yes exit umount sleep dump
This is the part I always wondered about. why haven't they at least tried to have new military spec radiation hardened chips created (faster procesors, etc)?
They have...
http://en.wikipedia.org/wiki/Category:Radiation-hardened_microprocessors
Specifically
http://en.wikipedia.org/wiki/Proton200k
About a gigaflop or a couple gigamips or giga-whatevers.
The problem is not finding an app to burn some mips, but finding the weight for the power supply and cooling ...
And the realistic market shipping quantity is probably triple digits at most.
And running a thousand times quicker, seems to mean on land based processors that it'll crash by memory leak or whatever a thousand times more often.
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
By federal law, any product of the Federal Government cannot be copyrighted (and thus, it's probably even less encumbered in that regard than FOSS). Of course, good luck getting them to disclose it.
First - you'll find Fed Gov't contributers to various OSS projects if you do a bit of digging. Having said that, it's not that simple.
While the Government might not be able to copyright works, individuals are free to patent inventions. One of the perks working at NASA is that they assist their employees with patent applications for whatever they're working on with the stipulation that the Government gets carte blanc to use the invention. But that's if you're a civil servant. NASA's strategy these days is to limit their Civil Service manpower to mostly oversight / management of programs. Meanwhile, much of the technical work is being shifted to contractors. Contractors hold all rights to whatever works they do under contract and are generally able to sell that work to other entities (law allowing). So not all Federal Government work goes in to the community pot with less and less doing so these days.
I should note that this off-loading strategy isn't absolute. There are still many Civil Servants at NASA doing technical work. NASA is less of a top-down directed organization than a collection of organizations within various groupings and sub-groupings with their own little fiefdoms and budgets that tend to work towards common goals. So while there may be a general trend, there will be plenty of small pockets of resistance that buck that trend if they have firm control over their own budget and the leeway with which to finance it (that and firing a Civil Servant is rather involved).
Not really. An 80386 doesn't run that fast--or hot. A 40MHz 386 chip is going to draw maybe 5 watts. Intel Core i7s draw closer to 100 watts (on up.)
Granted, processing power per watt is much higher in new chips, but that's not a worthwhile tradeoff if you don't need 95% of the computing power at your disposal.
Check out my world simulator thingy.
Admiral Grace Hopper, who wrote the world's first compiler and co-wrote the world's second compiler, advocated FOSS in the 1950s. Admiral Hopper encouraged programmers to collect and share common portions of programs.
Free Martian Whores!
It costs ten times as much because it comes with a sheet of paper.
Not a spectacularly amazing sheet of paper, it has to be said. But a sheet of paper that confirms that the chip is specced to handle a lot more abuse than anything available in the commodity market, a sheet of paper that says "You want to use this in applications where lives are at stake? Where if it goes wrong, someone is more-or-less guaranteed to die? No problem!".
You look at the paper for ordinary consumer chips - it normally says the exact opposite.
It's called the ACQUISITION process for a reason.
...because governments are required to go through all 285 rules of acquisition before finally obtaining the parts they need. When dealing with Ferengi, surely that must be a time consuming process.