It would be awesome if we could have made it thinner -- the wing is there as the lowest-drag shape that we can put around the other components in the car -- suspension, steering, driver, etc. Its designed to be a lifting body because of the ground effect which would otherwise result in a negative lift. The cambered wing counters the negative lift generated by the ground effect.
This is what you need our research for. We can actually work out when the best time to change to a lower FSB/memory/IO/CPU frequency is with one nice tunable like you're used to... http://ertos.nicta.com.au/research/power
We did some work recently where we showed that in a lot of cases, running the CPU at a lower performance point actually resulted in more energy usage -- scaling down the CPU frequency means everything takes longer to run, which means that you get less time to spend in low-power idle modes. There are also a lot of other complexities with frequency scaling... Particularly on a platform like the Android where there would be multiple scalable frequencies, etc.
There's a whole lot of other problems with the slower-is-better approach... But check out the paper we've just published.
As a measure of QoS, I think this is quite cool work, but the way they translate this into frequency scaling seems broken.
There are 1034 cut down Sunpower A300 cells in the array. (They're cut to take off the corners and allow us to get 5% more active area into our 12m^2 - we ended up with 11.5m^2 active area). They're encapsulated by Gochermann Solar Techonology in Germany (for reference: I can't speak highly enough of these things. Having built several of our arrays, including the vacuum-formed curved panels for Sunswift 2, I can assure everyone that this is easily the best experience I've ever had with solar cells. Everything just works the way its supposed to).
Oooh. We were also looking to see how much of an advantage making the car thin gives you, and then optimising the upper/lower curvature for each thickness to give the minimum drag (which is approx 0 lift).
We weren't able to use the machines in the lab's opening hours. Mostly because the simulations required large amounts of RAM. We wrote up a submission system which started the machines up each night, ran the simulation until the labs opened, and shut the simulations down again. This would happen each night until the job was done and the results were added to the pile. Each job was run on 8 machines, so we were able to do 10 in parallel. That's 10 slightly different designs which were tested in parallel. Each job would take a few nights to complete. (For those interested, we were running a 15 million element model in Fluent). In total we would have tested several hundred permutations. The main things we were looking for were the car's ride-height, the length and position of the spats/wheels, the canopy shape/length, and the shape of the nose. We needed to run multiple simulations to see how the car would handle in cross winds, etc. You can see a pretty picture of the results in the Gallery page on sunswift.com.
Linux itself has a number of issues which have been outlined in various other comments. Lack of proper real-time support, speed on ARM, reliability and security. While these probably limit how useful Linux can be, there is a solution being developed. Two companies in particular are building it: Jaluna and ERTOS at NICTA. That solution is para-virtualising linux on top of a fast, real-time, and secure operating system. Performance, I hear you say? The ERTOS solution is so fast that in some cases it out-performs native linux, and in most others performs comparably. These systems have already started to make it into mobile phone manufacturer's hands
National ICT Australia is a new research organisation set up by the government to work on "use-inspired" research. They're a nice combination between academia and industry. While CSIRO is a general science research organisation, NICTA is focussed on ICT. I work for the Embedded Real-time and Operating systems group within NICTA, and (with all modesty) we're doing some kick-arse stuff (focussing on seriously fast microkernels, virtualisation, power management, etc) with a very smart group of people. Our research is starting to make it into industry, with our software being used in Qualcomm platforms. There's too much stuff to write about here, but have a look at the website and, even with a moderate appreciation for OS technology, you'll be impressed. Other groups within NICTA are also doing some interesting things, but I don't know so much about them.
"Which part of CARBON FIBER IS NOT CONDUCTIVE DO YOU NOT UNDERSTAND?!"
If only this were true!! I've some not-insignificant portion of my life thus far trying to trace down shorts between solar cells and carbon fibre in the nearest thing to F1 that exists outside of F1 -- solar car racing. Unfortunately, a live-chassis, when dealing with >100VDC, ain't so healthy for the car, its electronics, or the driver.
Oh yeah. Did I mention the shocks the car used to give me? Or the multiple fires its started, or...
(Sorry. Carbon fibre being conductive has caused me that much grief...!)
Does this smell of the Copland debacle prior to Mac OS X's introduction? Apple worked for years on their own OS (Copland). They gave up and bought NeXT and Jobs.
Actually, I think ASC is longer. WSC is tougher though. WSC is 3010km whereas ASC is closer to 4000.
WSC is somewhat tougher though. Honda did it in 4 days (kind of - they raced overtime on the last day and copped a penalty, but still won the race). It also goes through the middle of the australian outback, past alice springs, etc... So there's no spot to stop for repairs if something goes wrong.
And that damn red dust gets in every possible crevice you can imagine! I had it coming out of EVERYTHING after the race for about a month! (Plus the arrays get dirty).
Dave.
I assume we'll get another (well deserved IMHO) post regarding the World Solar Challenge nearer the time of the race... But I thought I'd point out that the event is also on the way.
ASC is looking great this year, with some new rules allowing more liberal use of higher technology. (e.g. space grade cells, etc...). (Similar to the WSC;-).
Our car will be up against a number of these in a few months (in the race from Darwin to Adelaide - WSC)! Looking forward to seeing you all over here...
Sunswift 2 (the University of New South Wales Solar Racing Team's solar car) will be competing in the race...
Our website is "in progress" (pending approval)... But is at: www.sunswift.com. Check it out in months to come. I'm looking forward to showing you the new car. It should be fairly special.
Back to that abstract.;-).
Dave.
Re:Finally, a story with some real value
on
Geeks vs. Nerds
·
· Score: 1
Slashdot Mirror
It would be awesome if we could have made it thinner -- the wing is there as the lowest-drag shape that we can put around the other components in the car -- suspension, steering, driver, etc. Its designed to be a lifting body because of the ground effect which would otherwise result in a negative lift. The cambered wing counters the negative lift generated by the ground effect.
This is what you need our research for. We can actually work out when the best time to change to a lower FSB/memory/IO/CPU frequency is with one nice tunable like you're used to... http://ertos.nicta.com.au/research/power
This is, of course, where you start to use a virtualised OS with a proven-secure hypervisor designed for exactly this job... Like OKL4.
Dave.
We did some work recently where we showed that in a lot of cases, running the CPU at a lower performance point actually resulted in more energy usage -- scaling down the CPU frequency means everything takes longer to run, which means that you get less time to spend in low-power idle modes. There are also a lot of other complexities with frequency scaling... Particularly on a platform like the Android where there would be multiple scalable frequencies, etc.
There's a whole lot of other problems with the slower-is-better approach... But check out the paper we've just published.
As a measure of QoS, I think this is quite cool work, but the way they translate this into frequency scaling seems broken.
Koala: A Platform for OS-Level Power Management
There are 1034 cut down Sunpower A300 cells in the array. (They're cut to take off the corners and allow us to get 5% more active area into our 12m^2 - we ended up with 11.5m^2 active area). They're encapsulated by Gochermann Solar Techonology in Germany (for reference: I can't speak highly enough of these things. Having built several of our arrays, including the vacuum-formed curved panels for Sunswift 2, I can assure everyone that this is easily the best experience I've ever had with solar cells. Everything just works the way its supposed to).
Oooh. We were also looking to see how much of an advantage making the car thin gives you, and then optimising the upper/lower curvature for each thickness to give the minimum drag (which is approx 0 lift).
We weren't able to use the machines in the lab's opening hours. Mostly because the simulations required large amounts of RAM. We wrote up a submission system which started the machines up each night, ran the simulation until the labs opened, and shut the simulations down again. This would happen each night until the job was done and the results were added to the pile. Each job was run on 8 machines, so we were able to do 10 in parallel. That's 10 slightly different designs which were tested in parallel. Each job would take a few nights to complete. (For those interested, we were running a 15 million element model in Fluent). In total we would have tested several hundred permutations. The main things we were looking for were the car's ride-height, the length and position of the spats/wheels, the canopy shape/length, and the shape of the nose. We needed to run multiple simulations to see how the car would handle in cross winds, etc. You can see a pretty picture of the results in the Gallery page on sunswift.com.
Apologies. We've moved it to a much more appropriate server and updated the DNS. It should filter through to you all soon.
G'Day,
Linux itself has a number of issues which have been outlined in various other comments. Lack of proper real-time support, speed on ARM, reliability and security. While these probably limit how useful Linux can be, there is a solution being developed. Two companies in particular are building it: Jaluna and ERTOS at NICTA. That solution is para-virtualising linux on top of a fast, real-time, and secure operating system. Performance, I hear you say? The ERTOS solution is so fast that in some cases it out-performs native linux, and in most others performs comparably. These systems have already started to make it into mobile phone manufacturer's hands
Dave.
National ICT Australia is a new research organisation set up by the government to work on "use-inspired" research. They're a nice combination between academia and industry. While CSIRO is a general science research organisation, NICTA is focussed on ICT. I work for the Embedded Real-time and Operating systems group within NICTA, and (with all modesty) we're doing some kick-arse stuff (focussing on seriously fast microkernels, virtualisation, power management, etc) with a very smart group of people. Our research is starting to make it into industry, with our software being used in Qualcomm platforms. There's too much stuff to write about here, but have a look at the website and, even with a moderate appreciation for OS technology, you'll be impressed. Other groups within NICTA are also doing some interesting things, but I don't know so much about them.
Except that the same microkernel also runs on x86, alpha, Itanium, blackfin, mips, powerpc (lots of variants), etc.
Apart from the dozen embedded OSes that would be in the average home. (car, phone, TV, DVD player, etc, etc, etc).
Cheers,
Dave.
Does this smell of the Copland debacle prior to Mac OS X's introduction? Apple worked for years on their own OS (Copland). They gave up and bought NeXT and Jobs.
If you drove for two days across Australia, you'd be less than half the way there unless you were driving _ALL_ day.
WSC is somewhat tougher though. Honda did it in 4 days (kind of - they raced overtime on the last day and copped a penalty, but still won the race). It also goes through the middle of the australian outback, past alice springs, etc... So there's no spot to stop for repairs if something goes wrong. And that damn red dust gets in every possible crevice you can imagine! I had it coming out of EVERYTHING after the race for about a month! (Plus the arrays get dirty). Dave.
Can't seem to find it. :-|.
Dave.
I'm fairly sure that Honda's qualifying speed is not the record... I remember being told of one team (perhaps Biel!?) that had been 160km/h...
The records as set in the Guiness Book, are fairly long in the tooth...
Dave.
ASC is looking great this year, with some new rules allowing more liberal use of higher technology. (e.g. space grade cells, etc...). (Similar to the WSC ;-).
Our car will be up against a number of these in a few months (in the race from Darwin to Adelaide - WSC)! Looking forward to seeing you all over here...
Sunswift 2 (the University of New South Wales Solar Racing Team's solar car) will be competing in the race...
Our website is "in progress" (pending approval)... But is at: www.sunswift.com. Check it out in months to come. I'm looking forward to showing you the new car. It should be fairly special. Back to that abstract. ;-).
Dave.
So jobs was a Nerd, and Woz was a Geek???