John Keller, Editor in Chief
For years now we’ve lived with the assumption that computer processing power-and hence systems capability-doubles about every couple of years. This is the essence of Moore’s Law, which has been an electronics maxim for the past four decades.
Few principles in the electronics business have been so constant, dependable, and predictable as Moore’s Law, but a sea change in its guidance may be in the offing.
The problem is heat. Fast electronics and tightly integrated packaging that are typical in embedded systems in military and aerospace applications generate substantial amounts of excess heat, and the pace of improvements in integrated circuitry is outstripping our ability to remove the unwanted heat. Tightly packaged electronics, in fact, are becoming more commonplace almost daily.
Gordon Moore, then-director of the Fairchild Semiconductor Research and Development Laboratories, first set forth his theory in 1965. He wrote that integrated circuit transistor numbers double every two years. That theory holds true today, and has accurately predicted the evolution of microprocessors from Intel and many other industry leading lights.
Somehow, circuit designers always have been able to shrink geometries, double the number of transistors on chips, and step up processing power at a rate that validates Moore’s Law, despite some naysayers who along the line have insisted that such advancements simply are not possible. Thus far the circuit designers always have proven the skeptics wrong.
Systems designers have come to rely on Moore’s Law in their planning for incremental upgrades. Designers take it for granted that technology will always be available for them to double computational performance on a reasonably predictable schedule.
Yet after such a phenomenal 40-year run, Moore’s Law soon may no longer apply-not because of physical limitations to stuffing ever-more transistors on a chip, but because of the industry’s lagging ability to remove heat from chips at the same pace of new processor development.
More processing power means increasing amounts of heat that designers have to remove, and today’s conventional technologies to cool chips and boards simply are not progressing as fast as processor technologies.
The upshot may be, in the not-too-distant future, a technological smashup in advanced embedded computers for military and aerospace applications that threatens to bring to a screeching halt the rapid upgrades that we’ve all become accustomed to seeing.
Far more so than I can ever remember, electronics designers from the chip to the platform level are spending far more time and investing far more passion on topics like innovative cooling techniques and improving system efficiencies.
Some are even starting to discuss actually slowing the performance of systems components such as the latest microprocessors and field-programmable gate arrays (FPGAs) just so they can stay within their power- and heat-management budgets.
There are many viable options out there, depending on the application. Sometimes today’s cooling techniques are just fine. Heat sinks, fans, cold walls, heat-conducting board stiffeners, and other traditional cooling techniques for rugged embedded systems often are adequate for even the most advanced processors.
For other applications, simply slowing down processor speeds or otherwise compromising system performance also might work. Yet for some requirements, simply putting the brakes on system performance may not be an option.
For those embedded applications that cannot rein-in performance, systems designers are paying increasing attention to new kinds of cooling techniques, such as spray cooling, flow-through liquid cooling, and thermoelectric devices that pump heat away from electronics hot spots.
Some of these approaches are not new. In fact, original plans for what has become the U.S. Air Force Lockheed Martin F/A-22 Raptor strike fighter called for flow-through cooling for the aircraft’s avionics. This approach calls for physically piping cooling fluid onto printed circuit boards to remove heat quickly.
Other kinds of liquid cooling are also under consideration. Spray cooling from companies like ISR Inc. in Liberty Lake, Wash., bathes hot electronic components in a fine spray of inert fluid, which recirculates through the system to keep high-performance systems cool.
Other options include thermoelectric device technology, a thin-film superlattice material made from a semiconductor alloy that provides heat-pumping devices for hot electronics.
How successful these new cooling techniques prove to be-in terms of cost, size, weight, and efficiency-undoubtedly will be telling factors in how electronics development unfolds over the next several years.
Already some engineers are starting to quip about some of the dilemmas these new cooling approaches may create.
One of my favorites involves the future “all-electric” aircraft, inwhich designers seek to substitute electric motors and actuators for the hydraulic systems that move flaps, rudders, and other control surfaces.
One of the motivations for the all-electric approach is to reduce subsystem size and weight by getting rid of heavy hydraulic systems, which also are notoriously leaky and difficult to maintain.
Ironically, however, the all-electric aircraft may require the same types of fluids and piping as those this design approach is intended to replace. Instead of using pipes and circulating fluid to move control surfaces, the all-electric aircraft just may require pipes and circulating fluid to cool electric and electronic systems.
I doubt that future design challenges ultimately will be quite that extreme, but it’s clear that thermal experts have a lot of work ahead of them to come to grips with future generations of high-performance electronics.