With Conduction Cooling, Signal and Image Processing Choices No Longer Gasping for Air

One trend in deployed signal processing is to move the processors closer to the sensor, enabling real-time data processing that's much faster and more efficient than downlinking raw data from aircraft or uplinking from vehicles to remote processing systems. Putting more signal processing near the sensors also allows fusing that data from multiple sensors, performing hyperspectral imaging and enhancing target cuing. Such functionality requires scalable multiprocessor systems with very high-bandwidth communications between the processors.

The problem with this approach is that the environment near the sensor is rarely friendly to COTS signal processors. And even with modifications to increase their ruggedness, most aircooled COTS solutions may survive the vibration, humidity, temperatures, dust and other modest environmental factors, but they still need air to cool high-power DSP chips and boards.

On the other hand, multiprocessing signal processing systems can require from tens to hundreds of GFLOPS in a system, and in any system offering more than a few GFLOPS, cooling the processors becomes a major concern. In these deployed multiprocessing applications, supplying air close to sensor platforms has been an obstacle nearly impossible to overcome cost effectively.

The answer, of course, rests with conduction cooling. But while COTS conduction-cooled DSP modules have existed for several years, what's been missing have been options to incorporate large multiprocessing DSP systems interconnected by RACEway and other switched fabrics. That's changed recently, because DSP industry heavyweight Mercury Computer Systems announced plans to build conduction-cooled modules and even offered to license its technology and intellectual property (IP) selectively to the user community.

View Entire Paper | Previous Page | White Papers Search

If you found this page useful, bookmark and share it on: