Your Cellphone May be Smarter Than Mars Science Lab
Keith’s note: NASA JPL Employee @doug_ellison recently tweeted “Based on some MIPS benchmarks, and the RAD750 cpu on MSL…my iPhone 4S is 12x more powerful than Curiosity.”
Gee … NASA is always bragging about all of the advanced gizmos they have (justifiably) – but they never talk about how outdated some of their things are – and why. Among other things, the reasons why old stuff is used have to do with the brutal nature of the space environment (and what will reliably “work”). But the reasons also have to do with NASA’s slow-motion design practices and mission delays due to self-induced cost overruns.
Then again, my toaster is much smarter than Voyagers 1 and 2 (together) and yet they will both last far longer – and JPL wizards keep updating their apps and OS as they enter interstellar space – 35 years on. That said, NASA probably flies yesterday’s technology a bit more often than it should. Gotta work on that, NASA. Maybe that $700 million in the FY 2013 budget for “technology” will give that process a kick start.
Hmm … cellphones smarter than Mars robots. Why isn’t NASA putting cellphones inside of spacecraft? Just wrap them in Lead.
We must consider that to test the hardening and survivability of new or upgraded equipment, specially things as delicate as computers, takes considerable time and money, and ideally requires operating the new hardware in space first, i.e., on somebody else’s mission, even if it is only a milk run to the ISS. Doing environmental testing only in Earth simulations adds a significant element of risk to very expensive missions.
You know, if it hasn’t already been done, it might be worth while to run a bunch of off-the-shelf cell phones, tablets, single-board computers and such in orbit for a couple of months, land them, and then do extensive testing on them. It should be relatively cheap if you can piggyback it on some other launch. We know cell phones work at the balloon ceiling. It’s a small investment that may show us how to save big money. Using a GEO or Molniya orbit would be even better.
Steve
That’s exactly what we’re doing at Surrey Space Centre with our STRaND cubesat mission — apart from landing & recovering them, that stuff is *expensive*.
Actually they brought up a few non rad harden laptops to the ISS. They were pretty fried after a few days.
Actually I do not think that is true.
The ISS has had over 70 laptops at one time being used for various science experiments and personal use. I have not heard of any of them being “fried”. If they didnt work just fine, NASA would stop sending them up there.
“if it hasn’t already been done, it might be worth while to run a bunch of off-the-shelf cell phones, tablets, single-board computers and such in orbit for a couple of months, land them, and then do extensive testing on them.”
Violates FAR. You can’t just buy what you need. You have to have a proper bid offer with detailed specifications and assessment of bidders. It’s more efficient, doncha know.
[And yet DoD can just buy a bunch of Playstations, when they were cutting edge, to cannibalise into a Beowulf supercomputer.]
DING DING DING DING DING DING
And we have a winner – “It violates FAR”. The best thing that could happen is we take the FAR and burn it. Burn to ashes and then mix the ashes in water so no one can ever read that mess again.
Compare organizations that are not enslaved by FAR to those who are.
Respectfully,
Andrew Gasser
TEA Party in Space
Andrew,
I think FAR, like ITAR, is a valid concept and (in theory) serves a necessary purpose. The problem is that the people responsible for administering FAR (and ITAR, likewise) have been allowed to become self-important little dynasties, and it ends up like having your own dog bite you. One of the most damaging aspects I see in them is also a flaw in the federal government — more laws, legislation, rules, and precedents are added year after year, but none are ever reviewed and then rescinded or modified when they are no longer relevant or reasonable. If FAR and ITAR were executed for the reasons they were created and kept current I’d say they were beneficial, keeping honest people from making honest mistakes, instead of it all being the irrelevant power game that it is now.
Steve
And communism is a wonderful idea in theory only It doesn’t work for humans at Alll!!!!
Funny how much of what we talk about at NASA watch is about how good ideas and plans don’t work because of human nature.
You can buy these things – and NASA has. You just have to justify sole sourcing them. Happens all the time – all in accordance with the FAR.
“
Why isn’t NASA putting cellphones inside of spacecraft” – Check out “phonesat” at NASA Ames. They’re launching an android as a proof of concept.
EXACTLY!
And yet people like to think NASA Ames is a waste. I would remind my friends in Washington that Ames is the best thing going for NASA.
Respectfully,
Andrew Gasser
TEA Party in Space
“Why isn’t NASA putting cellphones inside of spacecraft? Just wrap them in Lead.”
I know you were joking but a bit of lead won’t stop electronics killing radiation. Anything short of a metre of so thickness of lead merely turns a cosmic ray “bullet” into a shotgun blast of secondary particles.
The RAD750 is getting a little old (2001), smartphones with their high volumes and turnover have sparked rapid increases in performance, including multicore cpus. I expect one could design a multi cpu 750. Also, keep in mind that your cellphone is sleeping most of the time to save power and will run the battery down pretty quickly if you play Angry Birds in Space all the time.
http://www.engadget.com/201…
Perhaps Rovio should help sponsor ExoMars?
PowerPC architecture development took a hit when Apple moved over to x86. There isn’t much money to upgrade that architecture, let alone from a radiation hardening sense.
PPC is used in the Playstation 3, GameCube, Wii and XBox360. That’s roughly a hundred million units – how many did Apple ever sell? They’re also in IBM’s servers and supers. The Cell, for example, is a 970 with additional (non symmetrical) cores.
They’re just not RAD-hard.
James,
Whenever I’m not on the road, my cell phone always sits connected to a charger. A low tech solution that works in many cases, all it takes is developing the habit. Even in the car you can have it plugged in. The key is to keep it at full charge or plugged and charging whenever it’s not in use or you are stationary, like at your desk, at home, or at a workstation on the ISS. I always look for the low tech solution first and avoid adding new hardware whenever possible. This is a case of thinking inside the box.
Steve
Ok, Keith. You’re going waaay outside of your expertise here with pointless criticisms. As already noted, lead makes the problem worse, not better. You have no clue what it takes to design, fabricate, and test components to operate in these environments. It costs enormous amounts of money for parts that are made in relatively small quantities for a small market. That market can only afford to update the technology occasionally, and then only with good justification. That market includes military space, and even with all of their money, it doesn’t happen often. Since you’re all over cost control, you are contradicting yourself here by suggesting frequent technology updates at high cost for no compelling reason. If you want cost control, you need to be able to accept good enough.
Another anonymous comment by someone who seems to be claiming that they know more than I do – which is quite possible, of course. You are the one who is outside your comfort / knowledge zone. When people work inside the NASA system they start to believe that different rules apply to what they do and solutions that are obvious to others elude them.
It is very expensive to develop rad-hard parts. You can test existing parts to see if they can survive high radiation levels, but to design rad-hard parts from the beginning is expensive. Look into rad-hard Xilinx FPGAs. I’m not sure the latest costs but I believe that a commercial part that costs $100 costs about $10,000 each. Parts designed only for hi-rel rad-hard have a very small market.
Using some of the commercial parts have other issues. Environmental concerns have discourage the use of tin-lead IC leads which as I understand it can lead to tin whiskers. This can cause shorts and has been known to do so in early missions.
What problem are you trying to solve by going with faster processors? As the joke says, pessimists say the glass is half-empty; optimists say it’s half-full; and engineers say the glass is twice the size it needed to be. The processor needs to run a spacecraft in deep space, not a high-resolution GUI display on Earth. The engineers picked the processor that matched the requirement.
Didn’t the Russians just lose Phobos-Grunt because someone used unqualified electronic components in space?
As for why NASA folks tend to post anonymously, the answer should be obvious. A lot of contracting companies have policies that require any statement invoking NASA affiliation to be cleared with company HQ and/or NASA PAO. This is a lot of fuss to go through just for posting to the Internet. Failing to do so can result in being fired. It’s a lot smarter to post anonymously and avoid that mess. The site formerly made it easier to do so; why did you remove that option? Are you trying to discourage NASA participation?
BTW: the “can’t speak to the press” thing is not unique to NASA or even government contracting. A lot of large companies have similar policies.
Kibitz,
I think you’re actually the one contradicting himself. You say, “It costs enormous amounts of money for parts that are made in relatively small quantities for a small market. That market can only afford to update the technology occasionally, and then only with good justification” and use the word market twice, which suggests that you’re a little behind the times with respect to how marketing works, and environmental verification as well.
Markets update ASAP because of the very competitive nature of electronics products. There is generally very little way to get a competitive edge on price, so you improve performance and features in order to compete. Computers have an average 18 month life cycle (basically determined by Moore’s law). Smaller Items like cell phones typically introduce new models annually.
As for environmental verification, you don’t do a large production run and then start testing to see if your new design/manufacturing process has produced something that will make the grade; you do a pilot run, a small quantity for testing, a quantity that is generally independent of your market size. You don’t do a full production run until you’ve proved out your new product with extensive pilot run testing. The cost of the testing process is therefore independent of the eventual market size, and affects only the product’s final cost (by a small amount) when the testing is rolled into the overall life cycle costs.
Steve
Steve,
Kibitz is likely referring to the space radiation hardened electronics market. In the 80s that was a sufficiently large market to inerest some of the large semiconductor manufacturers. But for the past 10+ years, very few manufacturers will develop space radiation hardened EEE parts. Not enough money to make it worth their effort.
Willie
Willie,
It’s kind of difficult for me to consider this issue any further since Kibitz’s comment seems to have disappeared.
Steve
Well said. Why put in “new and untested” when “tried and true” works just fine? I love the links posted on this site, but more often than not Keith’s spiteful comments cause me to roll my eyes.
Keith’s comments would have had a lot more teeth if there was even the slightest mention of how a faster CPU might have enhanced the mission. Fact of the matter is the that it wouldn’t. And yes, I have worked directly with an MSL software engineer and talked with him in depth about the rover. Can’t remember him once saying “damn if we only had a faster processor we could have done so much more”. LOL.
John
Sigh. Another anonymous MSL/NASA employee. For once it would be nice if they used real names.
John,
Do you happen to know the clock rate of MSL? Wikipedia lists it at 200 MHz, but I didn’t think the RAD750 went that high. (I’m only asking since it looks like you might know–I agree with your comment).
You are right, but you are also wrong. There other ways to achieve reliable tech for space missions, than doing “small quantities gold plated everything” approach.
You have to change the entire design paradigm though, and i do not see any of the usual suspects taking that leap.
Have any of the shiny new players done differently? What CPUs are being used by SpaceX and Orbital? It would be awesome if they’ve advanced the state of the art.
I haven’t heard anything, and wasn’t able to find out with a quick google. Keith, do you know?
SpaceX and Orbital are flying short term LEO missions and aren’t as sensitive to high radiation levels experienced above the Van Allen belts and when flying for long durations (such as a trip to Mars and performing a multi year mission).
It’s not about technology, it’s about discovery. No sadder thing is there than something wonderful but useless, better then to wring every part of usefulness out of what we have built.
The radiation environment on ISS is pretty benign compared to deep space (after all people live on station). You’re not taking a 5 kRad/yr hit on station. And as others have pointed out, shielding doesn’t actually help all that much (and carries a mass penalty.. which science instruments will you give up to fly shielded boxes with commercial parts?)
But radiation performance isn’t why we fly older more mature parts and not cellphones. There’s lots of factors:
1) your cellphone won’t work very well at -50C or +85C.
2) use of spares from previous missions or from group buys. You might have 2 spares from a previous mission, and then you only need to buy 1 or 2 new ones for this one. At $1M/each for flight qual boards, this is saving the taxpayer money.
3) It’s hard to get a science justification for a requirement for increased performance. Most scientists would rather have the risk item for the mission be their instrument technology, not the flight computer, so they design for what currently flying “safe” stuff can do.
4) flight qualification of a new design is expensive. You could argue that maybe NASA goes about this the wrong way, but it’s a pretty complex problem. You analyze each and every part for its performance against the environment and for manufacturing variability. You test the bejeebers out of your engineering/qual model. The other procurement model is the “build 1000, test them all, and use statistics”. Cellphone and computer DOA rates are in the 1-2% range, and failures over time are in that same range (your toaster may last 10 years, but it’s a lot simpler than a computer). flight computer DOA rates are much lower and inflight failure rates are a lot lower too. How many units will you test for statistics? 10,000? run them for accelerated life? How much will that cost?
Maybe NASA could look to the automotive industry, though. They have 1 in a million sort of failure rates, over fairly harsh environmental requirements. OTOH, development costs for a new Engine Control Unit are probably much higher than for a new flight computer. (When you’re selling 10 million ECUs at $1000 each, you can afford to spend 10s of millions to design and test it)
The ever increasing money supply for NASA is clearly at an end. Someone needs to find a way to do things in a less expensive, more responsive fashion, otherwise the agency’s ability to do all the things it wants to do will totally collapse.
Moved
Ah, as in Faster, Cheaper, Better, right? What a concept!
The agency is not going to be able to do “all the things it wants to do”. We should all get over that fantasy.
In order to avoid collapse, what the agency really needs to do is manage costs more reliably, such that it can make plans that fit credibly with agency budget futures. The failures of the agency have been less that things are expensive, but that they’ve been unexpectedly expensive.
The problem with Faster, Cheaper, Better was that NASA people only worried about “cheaper”. While you cannot get equal parts of all three if you keep all three in mind you can make more progress than if you focus only on cost. Sometimes doing something “faster” can be “cheaper”. Sometimes it is the exact opposite.
Agreed Keith. The major reason that Faster, Better, Cheaper failed big time, I firmly believe, was in the overbearing managerial oversight, which saw the plan as Cheaper, Cheaper, Cheaper. A huge mistake that should have been obvious to any rookie.
Steve
I think we all agree with this. But what you said was “less expensive, more responsive”. Dan Goldin’s NASA achieved precisely that. But what they didn’t achieve was the “better” part. That’s what I was calling you on.
It goes without saying that if you do things cheaper and more responsively, that will help you meet your priorities. But that’s not the whole thing.
The thing to remember is that “better” doesn’t equal “cheaper” + “faster”. That it does is the fallacy that Goldin’s NASA was working under.
Maybe the collapse of NASA would be the best thing for American space flight in the long run. I know every little here, but all NASA does sure seems very foolish to me.
With respect,
George H. Worthington IV
NASA is the best thing on the planet for space hands down. It isn’t likely that “creative destruction” of it will yield more “creative” than “destructive”. Be careful what you wish for, you might get it.
The ideological wars haven’t had the benefit of improving anything in America, but of sickening it. Because it is not a discourse bent on improving America through a process of arguing out a best path, but a purity movement to cutting out the impure given a faulty presumption of knowing what purity is.
We need to refine NASA (other things too), and what is making it faulty is the poor way we attempt to do it ad hoc, because we do it for the wrong reasons / motivations. Is it any reason things turn out so poorly?
Still listioning Closely !!!
Steve the question is HOW ???
Some ideas below
Nice post Mr. Whitfield 🙂 What prompted that 🙂
Perhaps it’s time to quit pointing fingers and think harder about ways to make it happen.
What a novel idea lol try to fix the problem! Lol
How do we get the public Interested in Space exploration or space settlement TODAY!?
Who are the public/ kids today? They grew up with iSS. Space is boring old news. They are living threw hard times every penny counts. They are looking for results, benefit to them, pay off/back value for their dollar. They live in a world of global warming, they are much more aware of our worlds Envernment
and the effects 7 billion people are having on our little sphere.
Soooo
What if NASA demonstrated that all electric power, solar and or nuclear could be provided from Leo by using recoverable reusable heavy lift. Solving a major world problem plus creating the anchor economic tenant for the space economy.
Today’s generation is interested in going to space themselves not just watching others explore. Push horizontal suborbital flight to become orbital flight.
As to exploration NASA should nott be the tip of the spear. NASA should be the supporter of others like Elon musk. Musk that yankee doodle dandy has the goal and has captured the publics interest Just help him do it!!! Isn’t it that easy? Don’t slow him down to make public rockets look good. Use public rocket money to provide missions for his rockets.
Nasa role should be something like the way the national geographic society supported it’s explorers
They should have a behinds the seen role in exploration adventures. It much easier for someone like musk to capture the publics imagination plus it would be better for the privite explorers to take on the risk of failure. It is that very risk that gets Joe Qs interest.
Isn’t obamas flag race what’s got the interest now? Not SLS or Orion. Heard someone on the news a while back say Spacex was our favorite rocket now etc etc
NOoC,
I agree completely. If the politicians and the public at large can lose interest in exploring and/or developing space, as it appears that the majority have, then industry will respond in kind, because their motivation is to concentrate on the markets that exist and only those new markets that offer the minimum (or at least manageable) levels of risk. I think we’re at the point where, for better or for worse, NASA is the sole source of incentive in North America of any size, not just in terms of expanding space activities in the future, but in terms of simply surviving in space. If NASA disappeared from the picture, then commsats, government-sponsored weather satellites, and GPS might well be the full extent of North America’s non-military activity in space. Not only would that take us off of the list of industrial/technology leaders, it would make a dangerous political statement, that being that our only interest in space is for military purposes.
I readily admit that there is much room for improvement in NASA’s operations, in how NASA and the federal government interact, and in how it benefits (or is perceived to benefit) the common people, but I maintain that we’d be in a much worse without NASA. With NASA gone there would certainly be no more American human space flight. Those who are betting on the commercial space companies to take up the HSF baton without NASA’s help haven’t thought it through. It’s not going to happen without NASA, because NASA is the only way we’re going to get enough funding to get through the initial stages, because tax dollars from the federal government is the only funding source that’s potentially large enough to make it happen.
So, far from getting NASA out of the picture, we need to get the public supporting NASA, so as to make it a more effective funnel of tax dollars into civil space activities. If NASA and industry and the general population were in basic agreement, and if they were to collectively pressure their elected representatives, then the White House would be inclined to pay more attention and Congress would be forced to act less arbitrarily (and less in its own self interest), and we might stand a chance of getting a sensible space program after decades of poor planning by almost everyone in the chain of command.
Step one is to, once and for all, get agreement and official recognition of NASA’s responsibilities and the level of authority they in fact have with which to uphold those responsibilities (the two must match, or we’re guaranteed to fail at the outset.)
Perhaps it’s time we all quit simply pointing fingers and thought harder about ways to make it happen.
Steve
Mr. Consequence feb 20 2012
Thanks for the Slap. I know you are right about NASA. I’m just very frustrated. See pretty clearly how the system has stopped what could have been. And is still stopping what might be. I do realize that our system of government is about the best we can have and still be free.
Had this idea about energy stations.
If NASA wants to build something. Build this. Turn sls and Orion into this.
Orbiting Nuclear/solar power stations power planet earth.
Steve, iPod is plugged into the wall per your suggestion. 🙂 Works good! Lol Where’s my power coming from? Coal, nuke, oil,hydro?
Easy Important space resources for us to go after if we had affordable lift to LEO.
Solar energy plus natural heat sink perfect for generating energy. The sun is the perfect place to junk nuclear waste as well as worn out nuclear cores.
Current events
Global warming, most likely man made. Possibility of gases from ocean floor to bubble up causing earth to go Venus in just a few 1000 years maybe mass extinction sooner.
Japans earthquake makes it obvious that nukes on earth are a bad idea.
40 percent of the 7 billion people/families on earth have washing machines and the other 60 percent want them. Customer demand 🙂
Why not build power stations in Leo don’t we have too?
What would one look like? How would it work?
Wouldn’t you start with a large solar array that has the ability to beam energy down to earth? Guessing your stations would have to either be in a polar orbit to get light always from the sun or in a geo sink orbit to always be over it’s energy beam receiver.
Wouldn’t you want to use the shade from your solar panel as your heat sink for your nuclear reactor? Wouldn’t you use solar energy for emergency cooling systems? Couldn’t you design your core to head for the sun should it get into trouble?
Obviously I know little about how to build a power station in space to power earth in a very green way.
But the demand is there. And our very survival may depend on us using space to provide the green energy we need or want instead, of depleting fossil fuels on earth first, thickening the atmosphere and risking a runaway greenhouse effect.
Just a thought for commercial space and yet another reason to drop the price to Leo SOON NASA!
When you start thinking energy from space it’s pretty hard to say there is no commercial opportunities in space should someone get on the stick and drop the price to LEO.
Ok I did my job trying to save the world today. Guess I’ll go work around the cabin, maybe throw some clothes in the washing machine Lol
Doesn’t take a rocket scientistt
Hummm couldn’t you shape a solar array in a parabola so that the energy that is not absorbed by your solar cells is reflected to a central core with a heat engine that uses the shaded heat sink to max your output?
Is there uranium on other near by rocks or just earth?
I think the idea of beaming electricity to earth from space should be rethought Steve.
Of course all the above ideas are subject to safety concerns do to the crappy nature of most members of our species.
Just like human nature is leading to the unraveling of NASA it will most likely lead to the end of us.
Putting power stations in Leo to light the whole world. Now there is a good reason to build TINKERs HEAVY LIFTER with recoverable oxygen tugs and thrust frame that could double as power station structure. Hummm hydro tank could be square with curved walls that can be taken apart reconfigured to focus light. Lol hydro tank with solar cells or mirrored surface lol.
Hummm what if your 300 foot tall hydro tank had double or triple walls with solar cells that opened, unfolded and telescoped out like flower peddles, couldn’t Tinkers lifter be designed to be a complete energy station all ready to go in one lift of his 6 recoverable oxygen tugs?
So NASA you want to build SLS well change the design to something like Tinkers lifter and make the first generation be a prototype energy station that beams back electricity to earth. Show the world that green power from space is possible and affordable by making reusable heavy lift.
What are we building SLS for again Mr. Bolden? I forgot!
Y’all be smart!
Well this simple idea got a little long so I guess I should apologize.
Anyway I’ll leave the light on for ya 🙂
I doubt that a rad hard computer for the mars mission is the cost leader. More likely it’s the science equipment and the many moving parts designing them, qualifying them, and building them. The way to make it cheaper is to not do exotic things and keep it simple.
Jimlux,
General Motors tried that approach when the purchased Hughes Aircraft Company in ~1985. At that time, Hughes was an enormous company with 50000+ employees. GMs focus was on the space business. Hughes Space and Comm was producing 20+ GEO comsats a year, and GM though it could use is automotive electronics procurement muscle and make comsats for much less. They quickly found out that automotive electronics were not usable in a space environment. They were very dissapointed, but left Hughes basically alone as long as they sent in the required profit to GM headquarters every quarter.
Willie
Use three cellphone processors, arranged in a triangle so that no single cosmic ray can hit all three. Then run them in parallel. Reset any processor that disagrees with the other two. For extra reliability, add more processors. One doesn’t need lead. Treat this as a software problem, not a hardware issue.
This suggestion relies on two assumptions. One is that cosmic rays only cause temporary glitches, not permanent damage. Two is that they arrive as a trickle, not as a sudden shower that could be expected to disrupt all the processors at once.
Neil,
Nice idea. I would add a third assumption, though. I believe (although I may be wrong) that cell phone processors are proprietary and the code is hard wired. If this is the case, you’d have to back to the chip (phone) manufacturer every time you wanted a software revision for your own mission. And if that’s the case, it could end up being much more expensive than at first guess, particularly for small quantities. So you have to assume that this issue can be worked out.
This would suggest to me going to the various companies who design this kind of chip and inquire as to possibility that a user programmable chip can be made into a commercial product. If that can survive the space environment, all you need is an emulator and a programmer to do many different missions, and it might well be much cheaper and far more powerful than what NASA is doing now. Put three (or more) of these chips (majority redundancy) on a standardized CPU card and stop reinventing the wheel for each mission.
Steve
In that case – how do they update the software of the phone over the air ? Blows that comment ..
Michael,
Do they in fact update “the software” over the air, or just a set of data parameters? I’d bet on the latter. It’s not common, that I’m aware of, to update “the software” on a product with a one-year life cycle.
Executing code from rewritable memory means you take a significant hit on execution speed. Copying a set of data values to RAM on initialization instead means no significant speed hit. (Code and data are two different worlds, each with its own characteristics.) The SIM card in your phone is just data (no code) like your phone number, phone book, etc., and that, as far as I know, is what gets updated, not the code. That’s why you can swap your SIM card into a different phone.
Using the chip for something other than a cell phone, like NASA would need, requires different code, not just a data update, so they would need a chip programmer. Comment unblown?
Steve
Software in EEPROMs can be changed in the field. Although the Army does normally brings the machine back to base first. The equipment was expected to last 15 to 20 years.
The computer copies the update software from PROM to RAM and runs the update software in RAM. The new software is now read in and written to EEPROM. We used an RS232 port. If the down load works the microprocessor’s jump table is updated and the machine restarted.
Read Only Memory (ROM) came out in the 1970s. Programmable Read Only Memory (PROM) chips soon followed. It took a decade or so before Electrically Easeable Programmable Read Only Memory (EEPROM) were made nuclear hard, but the chip manufactures did it in the end. Military specification Flash chips can now be purchased off the shelf.
“This suggestion relies on two assumptions. One is that cosmic rays only cause temporary glitches, not permanent damage. “
Yep. Look up “latch up.”
If a cosmic ray trail latches the chip up with a short between the power supply and ground through the crip, you fry the chip. There are well-known semiconductor design methods to mitigate latch-up… but it’s unlikely that cell phone processors use them.
Here’s the problem: you could make a design and be ninety percent sure that it would be robust against a worst-case cosmic ray event. Are you willing to risk a five hundred million dollar mission on that ten percent chance that you misjudged? You can do a lot of testing to shave down that ten percent… but tests are expensive, and when you do, it’s would be be cheaper to just have used the rad tolerant part in the first place. And there’s the possibility that the tests show that the part fails, so you spent all the testing the money and *still* have to go back to the old legacy processor.
Now, if you do all that testing, spend all the money, and do qualify the part, then other people get the benefit: they can use the part later. But that doesn’t help *your* budget.
Geoffrey,
I absolutely agree. In addition to songle event latch up, there are other single event effects that cause catatrophic device and circuit failure. Burn out, gate rupture, functional interrupt (though this one may be recoverable, as long as you do not go into an unknown, unrecoverable state).
Single event part testing is expensive as you pointed out, and in missions where you must be able to prove that you have a >90% likelihood of mission survival are not structured for using unhardened parts and meet those reliability requirements.
Willie
Keith, it also costs an insane amount of money to develop a rad hard chip and qualify it. The volume is particularly low, so technology just moves slowly. Your cell phone might seem to work ok in ISS, but once you get beyond LEO the single event effect (upsets, transients and latchups) would make it unusable.
There is some work going on called rad hard by design that would let us use bulk CMOS processes instead of rad-hard processes for fabrication, and allow faster higher transistor count chips, but it’s not ready for a multi-billion dollar mission to depend on it yet.
As far as shielding goes, the mass needed to prevent upsets would be insane. It’s just not practical or we would already be doing it!
It only takes “an insane amount of money” because smart business people know that NASA and the military will pay those insane costs because they have bought into the notion that it is expensive.
It’s expensive to develop any chip. Starting with the costs of just the EDA tools, you’re talking about $2M (unless you want to do it all by hand which would require a much more expensive army of engineers).
Then you have the cost of engineering design and verification. The estimated cost to develop the ARM chip used in the iPod was $1B.
Space simply don’t have the economy of scale to sustain chip design starts as often.
If you have any ideas on how to make this cheaper, let’s here it, but I don’t see any hope for space flight parts to be anywhere close to the cost or speed of COTS until we hit the point where SEE will impact chips at sea level (apparently not yet the case at 22nm, probably below 16). It’d be interesting to look at the performance (including power) of a faster PPC, Sparc, or ARM core in an RTAX-4000, and seeing what we could do with rad-hard by design in bulk CMOS.
There’s a whole raft of LEON3 and LEON4 cores (in Fault Tolerant form) for RTAX. That’s if you don’t just buy the processor as an asic in the form of a UT699 or 712, etc.
Check out Aeroflex/Gaisler’s offerings.
Keith, the issues here have already been pointed out. I had a nice post that Disqus looks like it ate.
Short version, if MSL was powered by a cell phone chip, and died in a day, what would your reaction be?
Sorry “SomeGuy42” but I never said that there should be a cell phone chip in MSL. Nice try.
There are a small number of electronic parts that computers and instruments use. Sending a test mission to the Van Allen radiation belts every few years would remove having the parts certified from the critical path of missions, hopefully saving money.
Whether the mission would have to fly every 2, 5 or 10 years is something that would have to be determined, mostly on cost grounds.
The components would be items included in 2 or more missions. This will bring in microprocessors, rom memories, ram, analog to digital convertors for sensors, digital to analog convertors for transmitters, power supplies, batteries, solar cells and field programmable gate arrays (FPGA). Motor and motor control will have to be thought through. For components with lots of values like resistors and capacitors it will probably be simpler to test the entire family rather than over restrict the mission equipment designers.
Mixed feelings about this. Yes we should find ways of using commodity technology because its more than 10x the performance. But risk reduction of spaceflight means we also need to have 99.999% reliability in extremely harsh environments those commodities can never be speced for.
Let me also add that spacecraft failures have been traced to radiation fractured solder joints because of flares – it is often the “low tech” around the “high tech” that kills you here. And that when you design for the “shake and bake” testing, you can’t pick and chose as much as you’d like.
To drive this issue you need to put more into IR&D projects that keep pushing the state of the art by clever reuse of commodity components. Funding these have gotten impossible – that’s the issue.
Keith, are you suggesting that it’s possible to get an arbitrary chip (a 970 or Cell or whatever) RAD-hardened and bet the mission it (I’m not asking rhetorically) or that there ought to be some budget to freshen up the available chips more frequently – like, say, order some RAD-hard Cells?
I was under the impression that keeping on-site systems as ‘dumb’ as possible was normal (beyond NASA) for all the reasons everyones’ gone into below.
Would could a faster chip do for the mission? The internet says it already has autonomous navigation & hazard avoidance and I don’t imagine they’d compress or process anything on-site.
There’s also a lot of odd processing demands on the phones in question.
Real-time movie decompression and 3D games involve tremendous amounts of math, but they’re also designed to accommodate haphazardly written software and strange user requests.
Users will do things like leave 3 dozen inactive apps resident in memory simply because they don’t know any better and there’s a lot of pressure on the phone vendors to make it “just work” when subject to ridiculous demands.
Meanwhile, ISVs rely a lot on garbage collection to take care memory. You have languages like Obj-C that don’t permit stack-allocated objects and resolve method calls at run-time (“messages”) and basically torch CPU performance and slightly threaten reliability (considering memory constraints) but it’s OK because it lets everything be cheap. But I don’t think these allowances (and their associated specs) really fly in this case.
Where am I wrong? I’m interested in this topic but don’t know a lot.
Wouldn’t be better to look at how MSL was spec’d in terms of capabilities? Did they determine what they wanted MSL to do and followed that with looking to see what hardware was available, or did they say, “We have this chip, this XXX, now what can we do? If the first was what happened, then it doesn’t really matter what someone’s cell phone can do but rather that MSL has sufficient resources and hardiness to get the job done for the designed mission duration.
Six Sigma. CMM.
If either of those mean anything to you, there isn’t anything surprising in this story…
The EPOXI (formerly Deep Impact) mission is flying one of these RAD750s and has not experienced a reset in over 6 years of continuous operation.
Lets see ANY cellphone do that.
AND as for the remarks about “faster better cheaper” being a failure… There were a few failures, but many more successes under that ideal. Dan Goldin himself said that if we didn’t have few failures under FBC. then we weren’t taking ENOUGH RISKS!.
If your cellphone dies, you shell out a few bucks and buy a new phone, at worst you are inconvenienced for a day or two. If the processor on a billion dollar spacecraft fails, you lose a billion dollars, possibly your career, and possibly a once in a lifetime science mission. No space mission has ever been limited by processor speed. Flying a spacecraft is actually much simpler than booting up Microsoft Windows. So you don’t need to have the latest and fastest chips onboard, you use the tried and proven chips. Engineering is all about choosing the appropriate technology for the job.
Radiation tolerance can vary from one manufacturing batch to another. NASA tries to buy enough chips from one single batch to allow for destructive testing of a sample of chips along with fabrication of the flight units from the same batch. If chips from a previously flown batch are available, it is much more cost effective and much lower risk to use the proven chips.
“But the reasons also have to do with NASA’s slow-motion design practices and mission delays due to self-induced cost overruns.”
Interesting. The last time I checked, RAD750s were manufactured by BAE Systems, not NASA.
The reason that rad-hardened chips get updated so infrequently has little to do with “slow-motion design practices”, and everything to do with market demands. There’s just not a whole lot of applications that demand a rad-hardened processor. Spacecraft, both civilian and military, and nuclear reactors, and…that’s about it. Did I miss anything?
The comparison to cellphone technology is an interesting one. Cellphones push a far greater number of units. If you believe Wikipedia, over 37 million iPhones of various versions have sold since the initial release in 2007. In contrast, the number of RAD750s sold since the chip’s release in 2001 is, if I’m not mistaken, in the hundreds. That’s an enormous difference. What that means is that a cellphone chip maker can afford to design a new chip on a yearly cycle, as they can expect to sell millions of units and quickly cover the cost of the R&D investment. The rad-hard chip makers don’t have this luxury; the time for return on investment is much slower, definitely not fast enough for yearly development cycles. A design cycle can really only be justified when the existing chips can no longer support upcoming demands.
Also, rad-hard chips aren’t designed from scratch; they’re usually a modification of an existing, proven design. What this means is that by the time a rad-hard chip is ready for market, it’s already at least a generation behind conventional processors.
Honestly though, the RAD750s are still doing just fine, meeting their demands, and getting the job done. We’re running flight software, not trying to play Angry Birds in Space or stream Netflix on these chips. Just because computing power has grown exponentially doesn’t mean that the algorithms needed to fly the spacecraft have as well…last I checked, the laws of physics haven’t changed recently, and, as always, the KISS principal still applies. Limits on spacecraft performance tend to be based more on sensor noise and limitations than processing power…from what I understand, most space-based RAD750s are sitting well below 100% processor utilization.
So, the upshot of all of this is that the reason that your cellphone is better than MSL’s processor is that there’s simply no market demand to make anything better at the moment. Commerce doesn’t breed innovation if nobody’s buying.
Lots of good comments in this thread I think, although some of them are irrelevant to the topic. The one thing that bothers me, though, is that there seems to be an assumption that what has been done to date in space in terms of processors and electronics is sufficient, period. I sincerely hope that doesn’t turn out to be true.
If we can somehow manage to get back on track with space exploration, and especially if we can get started into space development, we’re going to need a lot more computing horsepower and sophistication than what has been used to date. Some of the control systems currently in use in satellites and the ISS are very old designs using components that are equally dated. If we continue to fly only Apollo/Soyuz level technology, I guess it doesn’t much matter. But if we hope to fly more sophisticated and long-lived systems in the future, with increased flexibility and the ability to be upgraded, then we’ll have to have a more affordable, faster and more reliable source for new electronics components that will operate in the various space environments.
Think about the future people; there’s little sense in continuously repeating what our parents and grandparents did.
Steve
I’m a little late on commenting on this, but I’d encourage anyone who is interested to view to video or review the slides from the presentation “Flight Software Challenges in the Next Decade” on the Flight Software Workshop website. I found this to be a particularly good presentation on where flight software (and hence processor need) is going.
http://www.flightsoftware.org/