To make good development decisions, system designers must know the strengths and trade-offs of competing power technologies.
In 2016, each person on the planet purchased an average of 111 semiconductor chips or integrated circuits, according to World Semiconductor Trade Statistics, and as the number of chips climbs, so does the need for greater power, space, and cost efficiency. In response, power device manufacturers have continued to evolve silicon-based technologies and develop new, wide bandgap materials such as gallium-nitride (GaN) and silicon-carbide (SiC). The result is that power systems designers are now served by technology options that did not exist just a few years ago.
Increased diversity among power solutions is natural; given that the application space for electronics and semiconductors continues to become more diverse. To make good development decisions, system designers must know the strengths and trade-offs of these competing power technologies. In addition, gaining insight into what lies ahead in power technology evolution, can help system and design engineers plan for the future.
Striving for multiple efficiencies
The need for greater power efficiency seems axiomatic, but it has not always been so. In the days when there seemed to be energy to burn, linear power supplies did exactly that by wasting unused power in the form of heat. Today, with energy costs rising and the climate at risk, priorities have changed. Now power efficiency—extracting a higher percentage of power output from input—has grown increasingly important for system manufacturers and users.
Important as power efficiency is, there are other metrics that also matter when it comes to choosing power supply solutions. For many systems, space is at a premium, so components must be smaller and heat minimized to reduce cooling requirements. Power density, measured as watts per cubic centimeter (W/cm3) or watts per cubic inch (W/in3), is an important metric for the efficient use of space in all systems, but especially in those that are highly populated, such as data centers and telecom switches.
In vehicles and portable equipment, space efficiency is coupled with the requirement for light weight. Power weight density, measured as kilowatts per kilogram (kW/kg), is another form of efficiency used to make design trade-offs for these systems.
Every system design must also meet a cost budget. This requirement is an overriding form of resource efficiency, since it may force compromises on power efficiency, space, and weight if the best-rated options are not easily affordable. Closely related to cost are quality and reliability.
The goal of power semiconductor development is to drive these efficiencies as high as possible. Or to put it another way, to drive power loss, space, weight, cost, and failure rates toward zero. Each power technology optimizes these factors differently, so the power semiconductors that offer the best design tradeoffs for one application area may not be optimal for another. This drives several types of power technologies and the need to select the appropriate technology based on application requirements.
Power improvement trends
Power semiconductors typically do not follow Moore’s Law, which has been a gold standard for digital CMOS and memory scaling. For one thing, digital circuitry voltages have fallen from 5V to less than 1V, boosting switching speeds and driving smaller lithography. On the contrary, power supplies have to continue processing increasing levels of power, which keeps the input voltages high.
Equally important is that the design of power and analog components is more multi-dimensional than digital chip design, where the effort is to repeat transistors with the same characteristics billions of times. New developments in power transistors require that the electronic “ecosystem” be completely refitted, including control circuitry, packaging, thermal characteristics, protection from voltage transients, various forms of signal interference, and magnetics. All of these environmental issues have to be resolved, sometimes involving considerable development in several areas, in order for power transistors to realize a step ahead in performance.
Even so, historic changes in power ICs have been impressive. Power losses have been cut in half every five years for at least two decades, and power supply modules used in space-critical applications, such as telecom, have doubled in density every ten years since the 1970s.
Figure 1 New power technology requires a new ecosystem
Power switching issues and technology
One important contributor to the rise in power efficiency has been the steady transition to switched mode power supplies (SMPS), which control voltage in the time domain through high-frequency pulse-width modulated (PWM) switching. SMPSs can double power efficiency over linear power supplies, but they bring new issues in the form of a higher bill of materials and greater design complexity. These factors become increasingly significant as technology pushes towards higher power ratings and higher switching frequencies.
Silicon (Si) remains the least expensive power semiconductor material because the industry has an established manufacturing base and expertise. Device innovations like the gate-controlled thyristor, insulated gate bipolar transistor (IGBT) and superjunction MOSFET have helped increase the wattage ratings for power devices. However, Si devices cannot process large amounts of power while simultaneously switching very fast. As a result, Si-based power systems continue to be limited in power conversion efficiency and tend to be bulky and heavy.
Today, the landscape for power systems is changing through the increased use of SiC and GaN power devices. Not only does the wider bandgap of these materials enable devices capable of supporting higher voltages, but both GaN- and SiC-based devices can support high switching frequencies, permitting the use of smaller passives. GaN and SiC power devices enable power systems that are not only more efficient than Si-based systems but also help make them smaller and lighter.
Although the two new materials outrun Si, they are not identical in performance. GaN operates at higher frequencies, while SiC handles greater input voltages and drives higher-power outputs (though appropriate power stage designs can enable GaN to achieve significantly high-power output). Generally, GaN seems to be the preferred choice for power supplies up to 600- to 700V range in applications such as telecom and servers. In areas such as electric and hybrid vehicles and solar inverters, where higher voltage is required, SiC begins to take over.
Figure 2 below shows the position of these different power technologies relative to power levels and frequency. The areas shown, while approximately correct, should not be taken as the final word because there is continual development of all these technologies to increase their range of support. Note that, though there is a great deal of overlap among these competing technologies, no single power transistor type is a price-performance winner everywhere.
Figure 2 Power technology scaling
[Continue reading on EDN US: Complete ecosystem solutions]
Sameer Pendharkar is a TI Fellow and High Voltage and FET technology roadmap manager inside Analog Technology Development at Texas Instruments.