As processors gain performance and draw ever-increasing power, it becomes more difficult to cool high-speed systems. Putting standards in place helps companies create interoperable, effective cooling techniques, which was a large part of the conversation heard in the embedded technology world at the Embedded Tech Trends conference in January. After hearing about what board designers are doing in order to manage thermal issues, PD&D talked to Ken Grob of Elma Electronic this week about how cooling techniques and standards are changing.
Cooling has to be considered at a different stage of design than it did before, he said. In the past, when power was lower, thermal dissipation was less of an issue and not an integral factor driving the system design.
“Now it has to be considered early in the design process to check for feasibility,” he said in an interview on Feb. 10. “It becomes a governing factor. We need to consider it much earlier than we did, and we need to model it with tools like FlowTherm or thermal analysis software in a different way.”
Embedded technology standards allow for a variety of cooling methods that have been developed to try to mitigate the problem. Air flow-through cooling passes air over the surface of the board. Liquid draws fluid through a heat exchanger. Heat can also be directed; heat pipes serve this purpose in conduction cooled systems.
Sometimes, new materials like diamond carbon films are used to create a low-impedance path between the chip and its thermal path to the outside world, such as fins or a heat sink.
“This is a really big challenge,” Grob said.
In order to measure the hot spots on a card, Elma uses thermal modeling to simulate the thermal behavior of a mated board in a system. In addition, thermal cameras and board-mounted sensors are used to determine hot spots.
Once hot spots are identified, the next step is to make contact with the area and move heat away from it with a lower impedance thermal path, a heat pipe or thermal plate for a conduction cooled card, where heat sinks are used for a convection cooled card. Convection or air cooled designs also consider techniques for directing air across the heat sink surface to improve cooling efficiency.
Which type of cooling is best depends on the end use case. Larger machines could be best served by liquid cooling, when available. Liquid cooling could also be used through a heat-exchanging jacket that wraps around the chips, a technique that has been used for a long time. Boxes that need to be sealed are especially challenging, especially if fans can’t be factored into the design. On aircraft, forced air can be directed through heat exchangers to remove heat from the chassis.
Cooling schemes and dissipated power can become the performance gate for system designers. In OpenVPX, for instance, more power is fed in than can be removed. However, eventually the need for more power and wide temperature ranges will limit system performance when it comes to cooling. For now, people are finding more and more ways to work around those problems and keep up with the increased power capability driving the world of computing.
Filed Under: Rapid prototyping