Archive for September, 2006

Symposium on Direct Fluid Cooling

Monday, September 4th, 2006

I attended the Symposium on Direct Fluid Cooling, co-sponsored by the Silicon Valley Section of ASME, at HP Labs. After 24 years in the valley, where the mantra has always been “get a larger fan,” it was gratifying to see that multi-phase flow and heat transfer has become a vital part of high tech. I greatly appreciated the opportunity to see how my first love (technically speaking) is being carried forward, and I was impressed by the overall quality of the speakers and their work. I especially appreciated the work of Van Carey and Phil Kuekes, both of which were new to me.

In fact I have to confess to a degree of wistfulness as I heard the presentations from Purdue, Wisconsin, Stanford and Maryland. The scale of the heat transfer phenomena involved in cooling semiconductor packages is so much better suited to university research than the large-scale phenomena I studied. With the dramatic cost and performance improvements in instrumentation, I am very optimistic that industry sponsored university research will lead to very significant advances in boiling heat transfer.

I was especially intrigued by the concept of Thermal Ink Jet technology to reliably provide demand-based spot cooling without any moving parts. Throughout the symposium, I pondered the benefits and development hurdles of using ink-jets for cooling.

  • Drazen Fabris related the relatively low efficiency of the TIJ as a pump, and it seemed to me that a substantial amount of energy was going into the dilatation and contraction of the vapor space above the reservoir. In thinking about how to implement a fast acting check valve, without any moving parts, I thought that a second heater might be used to form a bubble in the feed tube between the reservoir and the jets, helping to reduce the coupling to the free surface above the liquid.
  • Another way to improve volumetric efficiency might be to use the TIJ jet to entrain a larger volume of liquid in a jet pump configuration, potentially doubling or tripling efficiency.
  • Also, a major difference between printing and cooling applications is the degree of subcooling of the fluid. In printer applications, the liquid subcooling is greater than 10%, whereas in a cooler with an integral condenser, the coolant will be close to saturation. This may well influence bubble formation, growth and collapse characteristics.

Finally, I came away with concern that there were varying interpretations and definitions for Critical Heat Flux. It’s natural that various definitions will apply depending on the cooling method being investigated, but we will need to distinguish between burn-out and dry-out conditions. And manufacturers will need to specify some margin of safety, such as a Critical Heat Flux Ratio, to assure that systems to do not approach CHF conditions. To be meaningful to system designers, some standard method of determining CHF needs to be developed.