Forget the Integrated Circuit: Power Conversion Is Where the Action Is
by Derek Lidow, CEO International Rectifier Corporation

The Solid State Revolution began 50 years ago with the invention of the transistor and selenium rectifiers, the first commercial solid state devices. The genealogy of the adoption of solid state devices goes something like this: selenium diodes begot germanium diodes, which in turn begot silicon diodes, which then resulted in commercial transistors and thyristors, and then begot ICs which begot memory and microprocessors, which spawned scores of new opportunities and changed the direction of our world.

Fifty years ago, solid state was power and electronics meant muscle. For the past twenty- five years the electronic revolution has been driven by the tumbling cost of silicon brain power. Although the technological paths of brain power diverged from those focused on providing muscle power, today the paths are once again crossing.

The semiconductor business has now come full circle. Power, which was the original semiconductor technology, now, 50 years later, is once again critically important to the perpetuation of the world's electronic revolution. Indeed, without a closer alliance between brain and brawn, the electronics revolution will lose momentum and could even come to a halt.

The world loves the pretty graphics and fast access to all types of information. But how we handle and manage power holds the key to further improvements in the world. Over the past few decades, we have all changed the way we think because of how cheap it is for machines to calculate for us. But to change the way we live requires physical changes to our world, and this requires that we manipulate power as easily as we manipulate numbers.

In order for the electronics revolution to further improve our world and to improve our standard of living, while enabling us to clean up our environment, will require a major leap forward in how we process and deliver power.

The criticality of power derives from three independent forces:

• First of all, as more sophisticated ICs are introduced on the market, they require more sophisticated forms of power and more power coverall. In order to pack more information or intelligence into a piece of silicon now requires: lower voltages and higher currents, more voltage levels, tighter regulation, and all delivered within nanoseconds. Today, come ICs draw more than 10 amps and go from sleep state to full on in tens of nanoseconds. If ICs are going to further improve, they'll require even lower voltages, even higher currents, delivered faster and with tighter tolerances. This is tough stuff.

• Next, the demand for electricity is growing faster than our planet's ability to safely and cleanly produce it. The world's leaders are currently discussing this problem. At the same time, the world wants more electronics, and wants to make more of its electrical infrastructure, yet we don't want to produce more energy and don't want to live near any new power plants. In order to reduce greenhouse gasses and to make it economically feasible to use more electricity that will necessarily become much more expensive, we must look to power electronics to make all electronic devices much more efficient. Otherwise, our electronic revolution will come to a halt.

• Finally, size and thermal constraints in many products are now dictated by the power conversion function. Products like the lap top computer run so hot that they can no longer be used on a lap. The size of batteries and their chargers make products bulky and limit consumer acceptance. If the electronics revolution will continue to entice the consumer, then we need to find a way to shrink and make more efficient the power circuits within these products.

The key to realizing this second wave of the electronic revolution will come from making improvements in the Power Conversion Process. This process is analogous to refining oil. As oil comes from the ground, we cannot use it; we must refine it into gasoline to power automobiles and jet fuel to fly airplanes. Similarly, as power comes from the power grid (or battery) we must convert it into the exact form necessary for the particular load.

Power Conversion is a five-step process. First we receive the electricity from the source, then we measure and control it, next we use switches to cut it into smaller more manageable pulses, and then we configure it into the correct form and finally deliver it to the particular load.

Our power conversion challenges are big ones. Over the next ten or so years, I believe we will see major changes in how we convey electricity to loads. Let me make some predictions on where power conversion technology will take the electronics revolution in the next decade or so:

• First, Moore's Law will be permitted to continue undiminished as voltage ratings will shrink to well under one volt to facilitate faster speeds and ever-shrinking semiconductor geometry. Ten years from now we'll be talking about half-volt buses.

• Improvements over the next ten years in load management will permit portable instruments to run three times longer with the same size and types of batteries as today.

• Next, I believe we will be able to cut power supply losses and their bulk by two thirds. We will convert AC line voltages to almost any DC voltage at an average 95 percent efficiency.

• In the next ten years, we will replace over five billion light bulbs with energy- saving electronically ballasted lights. Lighting uses over one quarter of all the power generated in the world, and converting just five billion of the 20 billion bulbs now in use would save the world around 200GW of power. We're talking about not needing scores of power plants, just by changing light bulbs.

• We will also see in the next ten years over one billion motors converted to speed control. Speed controls will fall in price to the point that most new motors will include them. Motors use over one third of all electricity we generate and motor controls make motors much more energy efficient, while improving their performance. One billion motors with speed controls will save the world another 100GW or so of power.

• Within the next decade, electro-mechanical and hydraulic devices will become obsolete. Solid state relays will replace them as cost-effective and reliable alternatives in virtually all applications. Today, solid state relays have lower contact resistance than copper wound relays and hydraulics are even being replaced in cost critical power steering applications. We've made silicon as efficient a conductor as aluminum, so electro mechanical devices have no brains for competition in the future.

• Ten years from now, electric and hybrid automobiles will be common, but still not yet the most popular form of transportation. Nonetheless, even a modest adoption of vehicles with electrical propulsion will cut smog and make our streets quieter. It will also test our capacity to produce huge quantities of sophisticated power semiconductors that even a modest quantity of these electric vehicles will require.

• On the other hand, because of power electronics, over the next ten years, conventional cars will get 20 percent better mileage and will be safer and more fun to drive. Power semiconductors can improve mileage because every watt saved is the equivalent of reducing the car's weight by one pound. In addition, power electronics will replace heavy and less reliable wire harnesses and electro mechanical components. Our cars will also have electronic load leveling and electronic power steering that will make the ride much more enjoyable.

To see these changes in our world, over the next ten years improvements to the Power Conversion Process must come in several areas:

• First, power packaging, both discrete and multi-chip, will improve as we design packages that take much less space, dissipate more heat and cost less. Power packaging has not kept pace with the improvements made in silicon. Today, the package contributes as much to resistance as the leading-edge silicon. This must now change by necessity. As another example, we waste on average over three quarters of the area in power packages. In the future as can improve the die size to footprint ratio to at least 50 percent, thereby saving considerable space and cost.

• Power integrated circuits, including high-voltage devices that operate off-line, will permit the more sophisticated power conversion architectures. We'll need to deliver lower voltages at higher currents in smaller spaces in shorter times. Power ICs will also dramatically reduce component count and shrink the size of power conversion products.

• Wafer processing of power devices will continue to improve basic device efficiency which we require to handle the lower voltages of tomorrow's ICs. On-resistance will continue to fall at the same past as in the past. We've seen a 50 fold improvement in on-resistance in the past 15 years. This is equal to a 30 percent improvement per year. We can count on this continuing and we can count on the cost per amp to fall in lock step.

• Power conversion materials will change over the next decade. Certainly silicon will improve in purity and defect density, but you can also expect silicon carbide to find its way into mainstream applications. Silicon carbide's high bandgap can translate into improved efficiency and dynamic performance. We must learn how to control this more exotic material in order to keep the revolution rolling.

• Finally, we will design power conversion systems differently. All these improvements will give rise to standard architectures. High-performance and highly integrated chip sets will be available at low cost, where today hundreds of variations of discrete designs exist. Standard designs will exist to provide the power required for each type of load. This will greatly reduce power conversion design costs and time to market, not to mention the reduction of warranty costs by eliminating untested new designs.

This all leads me to suggesting a structural change in how we design electronics. We need to view power systems as integral parts of the entire design, not just as stand-alone sub-systems. Power system design experts must become full members of your product design teams. These design experts will in the future be made available from power semiconductor suppliers as part of the services they provide. The power semiconductor industry can and will change to meet the needs of future generations of electronics customers, but we will generations of new, exciting, and much needed products to market faster, at lower cost and with fewer problems if we begin working as an integrated team today.

By uniting the two sides of the semiconductor design problem - brawn and brains - our electronics revolution will once again find itself on solid ground.