Slashdot Mirror
The World's Fastest Image Processor
Roland Piquepaille writes "This image processor is not your typical digital camera. It took 6 years, 20 people, and $6 million to build the 'Regional Calorimeter Trigger' (RCT) which will be a component of the Compact Muon Solenoid (CMS) experiment, one of the detectors on the Large Hadron Collider (LHC) in Geneva, Switzerland. The RCT will fill several racks of space in order to process 4 trillion bits of information per second while analyzing a billion proton collisions per second. The camera is currently being tested at the University of Wisconsin at Madison before being shipped to Geneva in June to participate in the first experiments in 2007."
I just hope it can do math...
"all that energy is compressed into two protons, which are a million times smaller than that annoying bug[Mosquito].
Hmm, (2/(6.02*10^23grams))/(0.002grams) = 1.66112957 × 10-21 so 2 protons weigh about 1 / (1,700,000,000,000,000,000,000)th as much as those Mosquito's which means it's volume is around that much smaller as well.
How about length 15 mm vs (10^15 meters) = 1.5 × 10^ -17meters so umm nope.
While this camera was developed at the university of Wisconsin, it will be installed at a facility in Geneva, Switzerland.
We had the opportunity to deploy this in America.
The Super Conducting Supercollider project in Waxahachie, TX was a federal basic science research project that lost its funding and was dismantled in 1993. The tunnel was dug. All the technological hurdles seemed to be jumpable. But the American people were less than interested in funding stuff that wasn't directly translatable into tastier hamburgers or cooler cars. The Democrat-led congress cancelled the $2 billion budget and America resigned itself to let other countries lead in this field.
I only mention the 'democrat-led' congress because I do not believe they have earned the slurr of 'tax-and-spend-liberals'. This is one example why.
$5 / month hosted VPS on linux = awesome!
Call it a 10^4 reduction. Say you could write the data to tape. You still have to read back 10^4 times as much data to find the "interesting" stuff. It's not just I/O that's a killer, pattern rec and kinematic recon are not free. Sure you could implement the hardware solution in software as a zeroth order data filter -- and then you'll never analyze the data anyway. Rule 1) never take data you are not willing to analyze. As for that notion that maybe you'll take the data and analyze it in the future -- do you hear of people analyzing bubble chamber film from the 50's and 60's these days? Hint: no data set has ever been exhaustively mined before the itch to acquire more/newer/better data strikes. Moreover, DOE/NSF will fund new experiments but not spend a dime to support continued analysis of old data.
I did my experimental particle physics PhD on an experiment named BaBar, you know, like the elephant. Are you telling me that isn't public-friendly?
A similar experiment based in Japan is called Belle and one in upstate NY called CLEO. One of the other experiments at the LHC is called ATLAS. They all seem reasonably public-friendly names (but then I am one of the folks you are saying don't know what a public-freindly name is, so I suppose my views are irrelevant).
As to the PR, it's pretty hard to make particle physics accessible to other physicists, let alone the general public. The essence of the question that BaBar and Belle were trying to answer is "Is CP violated in strong interactions?". It generally takes several years of university physics just to understand the question. The most "successful" PR projects never even seem to get to the crux of the project.
Incidentally, the answer is "yes, maximally". Your tax dollars at work!
I'm currently working on a similar project at Fermilab. The eXtremely Fast Tracker (XFT) is a set of electronics which decides, once every 396 nanoseconds, whether or not the particle tracks that we see represent an interesting event that we want to keep, or a boring one that we don't want to bother putting on disk (well, actually tape). We are in the process of upgrading it, because the collision rate has been increasing (technically, the luminosity has been increasing), and the old XFT is not up to handling the now much higher track density. My job is writing software to test the system as it is installed.
SIGSEGV caught, terminating
wait... not that kind of sig.
So it handles 4*10^12 bits a second. And there are 40*10^6 collisions a second. So (4*10^12*bits*s^-1)/(40*10^6*collisions*s^-1)=100, 000 bits per collision. Comparitivly to a regular digital camera this is nothing but then again it is processing so much information per second. I wonder how this compares to STAR at RHIC at Brookhaven National Labs. I did research into the Rho-Muon at STAR last school year. intresting to see the advancements in computing technology.
nairb774