keynote lecture:

what lies ahead

franklin was wrong - there is at least one other thing that is as certain as death or taxes: we will continue to have problems cooling electrical and electronic systems, and will continue to need innovative solutions to those problems.
here is what i see in my crystal ball for the next ten years.
there are several cooling strategies available today (direct air-cooling, spray cooling, single-phase liquid cooling, boiling, etc.) and several “facilitators” that can be used in these systems (heat spreaders, heat sinks, thermoelectric coolers, heat pipes, etc.). the bottom line, though, is fixed. the ultimate heat sink for all of these strategies is either the ambient air or the local water supply and the way to get there is through a heat exchanger.
liquid cooling and spray cooling clearly involve heat exchangers, and their size and weight are important. even a simple finned heat sink is a heat exchanger.
i think we will see a lot of progress towards ultra-high-compactness heat exchangers in the next ten years. using metallic or non-metallic foams, or micro-scale manufacturing techniques we can greatly reduce the size of heat exchangers while keeping their thermal performance and pressure drop constant. an “order of magnitude” (factor of ten) reduction in core volume is not impossible. to make these techniques practical, we need to reduce the cost of the new manufacturing methods and develop some new design approaches. cooling system designers will have to become fluent in heat exchanger design and we need to get much smarter about the design of compact headers and low-loss ducting.
these reductions in heat exchanger size will affect the air movers we need. i think we will move toward lower flow, higher head blowers with active flow control. noise management will become more important.
direct air cooling, which seems to be the simplest cooling system, is actually the most difficult to pull off. direct air cooling faces two very difficult problems: predicting the flow paths inside the enclosure, and the predicting the heat transfer from the components. most people in the field rely on some level of cfd for predictions but the results are not always reliable. small wonder: even the best of “research-level” cfd codes still can’t handle separated and reattaching flows with good accuracy, and those situations abound in direct air cooling.
the dominance of cfd may be jeopardized by recent advances in rapid prototyping and rapid experimental methods. two new experimental techniques may combine to replace cfd as the “tool of choice” for determining the flow distribution in a direct air-cooling situation. in the near term (this year and next) this approach may be limited to the most critical “high value” situations but it will become more widely available as more facilities are assembled.
the two techniques i have in mind are magnetic resonance velocimetry and precision stereo lithography. mrv can quantitatively document the velocity field (magnitude and direction at thousands of points) within a model enclosure in just a few hours of run time. starting with a geometry file, an appropriately scaled high precision stereo lithography model can be made and the unit tested within a week. within the next ten years, i think this experimental approach will be routinely considered as an option.
given an accurate sla model it is relatively straightforward to attach thin-film heaters to the critical components and directly measure the heat transfer.
the “desk-top” heat transfer laboratory may be faster, cheaper, and more accurate than even a very good cfd program. time will tell - and the clock is ticking