Solar cell doubles power output of terrestrial applications

The US government's National Renewable Energy  Laboratory (NREL) and Spectrolab, Sylmar, CA, have achieved a conversion efficiency of 32.3% for a solar cell that could double the power output of terrestrial applications in operation today. The cell is Spectrolab's triple-junction gallium-indium-phosphide on gallium arsenide on germanium (GaInP2/GaAs/Ge) concentrator solar cell. Spectrolab took the basic cell design concept, used for high-power space satellites over the last five years, and made it cost-effective for terrestrial applications by combining with a concentrator system. These concentrators replace semiconductor area with less expensive optics, and by doubling the power generating efficiency of the cell, the size of the collection system can be reduced by 50%, thereby lowering overall cost of the infrastructure. Concentrator systems can afford the slightly higher cost of multi-junction cells, yet still be manufactured at lower dollar-per-watt cost compared to other flat-plate modules. With higher efficiency multi-junction cells in the concentrator modules, only about one-half of the real estate is required to generate the same power output, compared to crystalline silicon or thin-film flat-plate modules. A nearly 40% conversion efficiency has been predicted for four-junction space cells -- optimizing for terrestrial use may surpass that mark. Circle 400.

New hypercomputers: "Blue Pacific" monitors nation's nuclear stockpile, "HAL" employs FPGAs

Lawrence Livermore National Laboratory has dedicated the ASCI Blue Pacific supercomputer, built by IBM, Armonk, NY, while previewing an even more powerful 10-teraOPS system, ASCI White. Blue Pacific is part of DOE's Accelerated Strategic Computer Initiative or ASCI Program, and will be used for maintaining the safety, reliability, and performance of the nation's nuclear stockpile. This massive supercomputer, using an IBM RS/6000 SP with IBM PowerPC 604e processors, can run nearly four trillion floating operations per second, while applying 5,856 processors in parallel to a single problem. Nearly five miles of cable and connecting hardware are used in an area covering 8,000 square feet. Meanwhile, Star Bridge Systems, Inc., Draper, UT, is expecting to deliver their first hyper-computer, based on an architecture employing Field Programmable Gate Arrays, in February 2000. (FPGAs were last mentioned here as applied to the brain of a robot cat, DFX, Sept. 99.) An example of their devices, nicknamed HAL, would be the HAL-300GrW1, which runs 12.8 teraOPS on just 280 FPGAs. The unit draws just 1600 watts of 110 VAC power, and occupies 3.78 cubic feet of real estate. SBS is also introducing their VIVA software, which can be applied to programming FPGAs from any manufacturer. Circle 401.