It’s sometimes funny how things improve or worsen depending on your perspective. The perfect switch existed from the day electricity was harnessed – at least 18th century experimenters such as Volta thought so, when they fabricated an electrical isolator out of brass, wood and porcelain. It had virtually no resistance when closed and no leakage when open. It could withstand as high a voltage as you like depending on how big you made it. It was a problem solved.
Semiconductors were a step back from the ideal
The first electronic switches, vacuum tubes, were large, lossy and fragile, and early transistors were a further step backwards, with high resistance and low breakdown voltage except, of course, you could flip them a lot faster than anything mechanical. They were small, but consequently could only handle tiny currents. In the 75 years since Shockley and his team made their discoveries, engineers have worked to get back to Volta’s ideal solution, but with pressure at the same time to switch ever faster at MHz rates, keep the small size and increase current rating.
The application that drove the development of transistors to higher power levels was of course the switched-mode power supply, which allowed the possibility of DC-DC power conversion without motor-generator sets and at high efficiency. The SMPS idea was patented in 1959 and bipolar junction transistors were used in the first commercial application in 1970, in the Tektronix 7000-series oscilloscopes. BJTs were successful in the application, but at higher powers were difficult to drive efficiently and switching losses were unacceptable at anything other than a few tens of kHz. Fast and easy to drive MOSFETs were patented as early as 1960, but early versions had significant on-resistance, producing high power loss at high currents due to the ‘square’ term in I2R. A breakthrough however, was the invention of the IGBT, with the ideal combination of an easy MOSFET gate drive and the on-state characteristics of a BJT and, to this day, it remains a practical solution for very high-power converters. ‘Practical’ is not ‘ideal’ though – to avoid unacceptable dynamic losses in the highest power applications, an IGBT’s switching frequency must be kept below about 10KHz, mandating the use of large, heavy and expensive magnetic components. In the meanwhile, MOSFETs switching at up to around 500kHz have improved to the latest state-of-the-art ‘super-junction’ types which now dominate the low and medium power ranges of DC-DC and AC-DC conversion.
In a bid to close the gap between IGBT and silicon MOSFET application areas, wide band-gap semiconductors have been explored, in silicon carbide and gallium nitride. These promise lower switching and conduction losses, stemming from better electron mobility values and higher dielectric withstand rating of the materials, enabling smaller devices with lower capacitances and conduction channel lengths. Manufacturing switches with the new materials has thrown up many difficulties though, from mis-matched thermal expansion coefficients using practical substrates with GaN HEMT cells to ‘lattice defects’ and ‘basal plane dislocations’ in SiC MOSFETs, all degrading performance and reducing reliability. Refinements in fabrication continue to improve performance however and the devices, particularly SiC MOSFETs, are now mainstream and are encroaching into the traditional IGBT high-power applications.
Wide band-gap devices are a step back in some ways
In some ways however, steps back have been taken; SiC MOSFETs and GaN HEMT cells are not as easy to drive as silicon MOSFETs, the required gate voltage levels are critical for optimum performance and reliability, and for SiC, the threshold exhibits wide variation and hysteresis. Reliability of the SiC MOSFET gate oxide has also been questioned and Gan HEMT cells have no avalanche rating, forcing heavy voltage derating. Another retrograde step is the performance of the devices when conducting in reverse by ‘commutation’, the automatic reversal of current caused by an inductive load – SiC MOSFETs have a body diode dropping around 4V under forward bias and with appreciable reverse recovery loss when subsequently reverse biased. When GaN devices commutate, they conduct through the channel with no reverse recovery issue, but voltage drop is again quite high and varies depending on gate drive.
Looking back to go forwards
A step in the right direction is to ‘look back’ to an old technique – ‘cascoding’ a silicon MOSFET and a SiC JFET, dubbed a ‘SiC FET’ by the manufacturer and technology champion UnitedSiC. It has better figures of merit for overall losses than SiC MOSFETs or GaN HEMT cells, the gate drive is non-critical with a stable threshold, the body diode is fast with low recovery loss and only around 1.5V drop. Additionally, the devices have robust avalanche and short circuit ratings, independent of gate drive. Devices are available in 650V, 750V, 1200V and 1700V classes, with on-resistances down to 7 milliohms in a variety of packages, with the majority of parts AEC-Q101 qualified, to allay any possible reliability concerns.
What’s more, the challenges created by very fast and high speed switching with these devices can be solved by a simple RC snubber to manage turn-off overshoots and ringing and get the best performance out of these SiC FETs.
Is the perfect switch in view yet? Designers will always want improvements but we are surely very close with SiC FETs.
Sign up for our quarterly newsletter and receive important technical information on all new products, app notes, white papers, and blogs.