9JUNE 2024is already widely known, although the lower values of capacitance and inductance needed are only part of the story.Typical switching frequencies of converters based on ordinary silicon power semiconductors have been in the range of a few tens of kilohertz: say, 30-80kHz. At these frequencies, polypropylene capacitors are suitable and widely used, being proven, reliable, and above all cost-effective. However, above this frequency range, parasitic effects cause excessive resistive losses and self-heating. More Materials ScienceWe have worked with most of the leading power electronics teams as they developed new prototype converters around SiC power transistors. Investigating the new demands these power switches impose on supporting circuitry enabled us to develop our KC-LINK ceramic capacitors, based on a proprietary high-voltage C0G dielectric that ensures extremely low effective series resistance (ESR) and very low thermal resistance. They can operate with minimal losses at frequencies into the low megahertz range and handle very high ripple currents with no change in capacitance versus DC voltage. The capacitance is also extremely stable over temperature. Capable of operating up to 150°C allows mounting close to fast-switching semiconductors in high power density applications. The available series offer rated voltages from 500V to 2000V to cover a wide range of applications including use with 400V and 800V EV battery systems. We also developed transient liquid phase sintering (TLPS), a non-solder interconnect technology that enables building high-capacitance MLCC leadless stacks that have a small footprint and exploit the temperature stability of the class-I C0G dielectric to address high-power applications that can reach temperatures of 150°C and more without cooling.WBG penetration in data-center server applications, on the other hand, is typically predicated on GaN technology. Typical switching frequencies have been stuck at about 300kHz for many years. This has increased with GaN's arrival, although still is only at about 900kHz. Here, we find that the performance of magnetic components is the major limiting factor. Inductors have two loss mechanisms, comprising resistive losses due to the winding as well as energy losses experienced as heating of the ferrite or metal-composite core. The ideal is to minimize the core loss without compromising the core's magnetic permeability, which is the basis of its ability to resist changes in current within the circuit and to store energy in the magnetic field.It's another challenge for materials science that our teams have taken up and are ready to announce a solution. While retaining high magnetic permeability, this new material is optimized for the lowest losses in the 1-5MHz frequency range and so allow the switching frequencies of GaN-based converters to increase. Just as in a SiC converter, raising the switching frequency allows smaller values of capacitance and inductance, ultimately resulting in greater power density. There are additional benefits to be gained from raising the power-supply switching frequency. The load-transient decoupling capacitance needed to protect critical parts such as the main processor can be greatly reduced. Historically, these have been tantalum or aluminum polymer capacitors. Reducing reliance on decoupling capacitance enables a small array of class-II MLCCs such as X5R, X6S or X7R devices to be placed directly adjacent to the processor. The next goal we are currently working towards is to embed aluminum polymer decoupling capacitors into the chip carrier inside the package, to work in conjunction with on-die silicon capacitors. This could overcome the decoupling challenges processor designers are dealing with today and allow higher converter frequencies; possibly up to 10MHz and beyond in the future. It could take about another five years of engineering effort.We also find that improving the performance in one part of a system can reach an impasse, causing designers to look more closely at other parts of the system for continued improvement. Our materials department formulated the U2J ceramic dielectric specifically to help develop the first switched-tank converters. With custom inductor geometries added into the mix to reduce magnetic core loss, these converters unleashed a drastic increase in the efficiency of the 48V-to-12V conversion in distributed power systems for data-center servers.These converters define the upper limit ­ for now - in terms of 48V-12V conversion efficiency. As that limit was reached, attention transferred to the Point-of-Load (POL) converters. Here, high-performing processors and FPGAs operate with a combination of low digital supply voltage and high clock frequencies that cause the current demand to change rapidly, reaching a high peak value. Multi-phase voltage regulators typically used to power these ICs force designers to trade transient response against ripple current. Transient response is limited as all phases need time to settle in sequence. Moreover, these multi-phase regulators mitigate against power densification as it becomes impractical to reduce the inductor width while preserving mechanical stability. Dual-winding, four-terminal inductors have enabled the development of the trans-inductor voltage regulator (TLVR) in which all phases respond simultaneously for faster transient response. Pulse Electronics, part of the Yageo Group, is a leader in TLVR inductors. WBG and Noise EmissionsThe rapid switching transions of WBG semiconductors creates an unwanted challenge for designers ­ electrical noise emissions or EMI/EMC. To address this design challenge to bring converters and inverters into compliance, KEMET's magnetics group has developed Nanocrystalline core materials for use in EMI Common Mode Chokes which offer broadband performance in a smaller package.What the Future BringsThe developments we are seeing, including advanced materials, new circuit topologies, and new demands on capacitors and inductors, are very much interrelated. Together, they are enabling continuous progress in the drive to increase energy efficiency and power density. Who knows when, or even if, we will ever reach a limit beyond which no further improvement is possible?
< Page 8 | Page 10 >