GaN HEMTs Promising Future for Next-Gen Power Semiconductor

Article By : Maurizio Di Paolo Emilio

GaN HEMTs feature low operating resistance and a high breakdown voltage, which makes them promising as a next-generation power semiconductor.

Silicon power MOSFETs have not kept pace with the evolutionary changes in the power electronics industry, where factors such as efficiency, power density, and smaller form factors are the main demands of the community. Silicon MOSFETs have reached their theoretical limits for power electronics, and with board space at a premium, power system designers need alternatives. Gallium nitride (GaN) is a high-electron-mobility transistor (HEMT) semiconductor that is adding real value in emerging applications. GaN transistors are significantly faster and smaller than silicon MOSFETs, enabling efficiency gains that have opened the door to applications not possible with silicon technology. Efficient Power Conversion’s (EPC’s) eGaN FETs are supplied in low-inductance, low-resistance, small, and low-cost land grid array (LGA) and ball grid array (BGA) packages. The new EPC FETs offer designers best-in-class performance compared with silicon MOSFETs in both hard- and soft-switching applications.


GaN switching devices are available in two types: enhancement mode (e-GaN) and cascoded depletion mode (d-GaN). An e-GaN transistor works as a normal MOSFET, even with a reduced gate-to-source voltage. It offers a simpler package and low resistance, with a bidirectional channel and without a body diode. The d-GaN transistor is normally switched on and needs a negative voltage. You can overcome this problem by connecting the HEMT transistor in series with a low-voltage silicon MOSFET, as shown in Figure 1. In contrast, the e-GaN transistor is normally off and is turned on with a positive voltage applied to the gate. Unlike d-GaN, e-GaN devices do not need a negative startup bias; with a zero bias on the gate, the device is turned off and does not conduct current.
Figure 1: e-GaN (left) and d-GaN configurations (Image: EPC)
The threshold of an e-GaN FET is lower than that of a silicon MOSFET, yielding a very low gate-to-drain capacitance (CGD). The low-capacitance structure permits switching hundreds of volts in nanoseconds at megahertz frequencies. GaN FETs’ large gate-to-source capacitance (CGS) relative to CGD gives the devices good dV/dt immunity. The dV/dt sensitivity of power-switching devices is caused by the various parasitic capacitance and gate-drive circuit impedance levels. The gate-charge (Qg) parameter, on the other hand, indicates the ability of the device to change states quickly, reaching a higher dV/dt with minimal switching losses. The gate charge of an e-GaN device is 10× higher than for an equivalent MOSFET, while d-GaN devices have about a 2× to 5× higher Qg than MOSFETs. To determine the dV/dt sensitivity of a power switch, you can use a figure of merit called the Miller charge ratio (QGD/QGS1). A Miller charge ratio of <1 will guarantee the theoretical dV/dt immunity. The gate-drive circuit layout is a critical factor in improving dV/dt immunity. The d-GaN transistor has the gate structure of a low-voltage silicon MOSFET. Therefore, existing commercial MOSFET gate drivers can easily operate d-GaN switches. The downside is that the addition of the silicon MOSFET ignition resistance raises the overall ignition resistance. The increase can be significant for low voltages (<200 V). For higher values (600 V), the additional resistance may be only about 5% of total on-resistance. A d-GaN transistor also has increased packaging complexity. Parasitic inductance and capacitance between the MOSFET and the GaN HEMT may cause delay during switching transients. The reverse-conduction characteristics of a switching device are important. In a MOSFET, the voltage drop of a body diode is low and its reverse recovery is very slow, resulting in high switching losses. GaN devices do not have a reverse body diode; they are able to conduct in the reverse direction because of their physical nature. In the case of reverse conduction, it will be necessary to have dead time. A d-GaN device in cascade has reverse recovery by means of the low-voltage silicon device. In hard-switching converters, the output charge is dissipated in the FET at each power-on transition. This loss is proportional to the output charge (QOSS), bus voltage, and switching frequency. GaN FETs have a significantly lower QOSS than silicon FETs, reducing the output charge loss per cycle and thus allowing higher frequencies (Figure 2).
Figure 2: Parasitic capacitance and current flow in a GaN device (Image: EPC)

Automotive and consumer solutions

Emerging computing applications demand more power in much smaller form factors. In addition to the expanding needs of the server market, some of the most challenging applications are multi-user gaming systems, autonomous cars, and artificial intelligence. Automotive systems are increasingly moving toward 48-V devices, driven by the increase in electronically controlled energy-hungry functions and the emergence of autonomous vehicles that create additional demands on the energy distribution system from systems such as LiDAR, radar, cameras, and ultrasonic sensors. These energy-hungry processors represent an additional burden for traditional 12-V power distribution buses in the automotive sector. For 48-V bus systems, GaN technology increases efficiency, reduces system size, and reduces system cost. A 250-kHz GaN solution with double the frequency allows a 35% reduction in size, lowering inductor DC resistance (DCR) losses as well as cutting system cost by approximately 20% compared with the MOSFET solution. GaN’s exceptionally high electron mobility and low temperature coefficient allow low QG and zero reverse-recovery charge (QRR)). The final result is a device that can manage tasks with a very high switching frequency for which a low on time is beneficial, as well as those in which on-state losses dominate. “High-power EPC eGaN FETs’ lower resistance in the on state, lower capacitance, higher current, and excellent thermal performance characterize these power converters with over 98% efficiency,” said Alex Lidow, CEO and co-founder at EPC. “This family of eGaN FETs halves the drain-source on-resistance to enable high-current and high-power-density applications.” The latest generation of eGaN FETs also halves the hard-switching figure of merit compared with the previous generation for improved switching performance in high-frequency power conversion applications. An increase in performance to 30 V enables GaN to be used to build high-power DC/DC converters, point-of-load (PoL) converters, and synchronous rectifiers for isolated power supplies, PCs, and servers. The smallest, most cost-effective, and highest-efficiency non-isolated 48-V to 12-V converter, suitable for high-performance computing and telecommunication applications, can be achieved by employing eGaN FETs such as the EPC2045, according to EPC. The EPC2045 has an operating temperature of –40°C to 150°C with thermal resistance of 1.4°C/W. The drain-source on-resistance (RDS(on)) is 5.6 mΩ typical (Figures 3 and 4).
Figure 3: Representation of a 48-V mild hybrid system
Figure 4: Efficiency versus current (Image: EPC)
In the consumer market, portable solutions are becoming increasingly energy-hungry. Efficiency and thermal management are critical in small platforms with minimal cooling solutions. The need for fast and efficient chargers has led the consumer market toward GaN solutions (Figure 5).
Figure 6: EPC9144 development board (Image: EPC)


The eGaN FETs and integrated circuits are the logical choices to use when turning on a laser in a LiDAR system because FETs can be activated to create high-current pulses with extremely short pulse widths. “The short pulse width leads to higher resolution, and the higher pulse current allows LiDAR systems to see further,” said Lidow. “These two features, together with their extremely small size, make GaN ideal for LiDAR.” EPC provides various development boards for its eGAN FETs. The EPC9144 is mainly designed to drive high-current laser diodes with high-current pulses at a total pulse width of 1.2 ns and current up to 28 A (Figure 6). The board is designed around the 15-V EPC2216 eGaN FET, which is automotive-qualified to AEC-Q101. The EPC9126 and EPC9126HC development boards are primarily intended to drive laser diodes with high-current pulses and total pulse widths as low as 5 ns (10% of peak). They are designed with the 100-V EPC2212 and EPC2001C enhancement-mode eGaN FETs, capable of 75-A and 150-A current pulses, respectively.
Figure 6: EPC9144 development board (Image: EPC)
Cepton Technologies’ Helius LiDAR solution, based on the EPC technology, delivers advanced object detection, tracking, and classification capabilities to enable applications for smart cities, transport infrastructure, security, and more. It combines industry-leading 3D LiDAR sensing powered by Cepton’s patented Micro Motion Technology (MMT), edge computing for minimum data burden and maximum ease of integration, and built-in advanced perception software for real-time analytics. “LiDAR has become a very significant market,” said Lidow. “It is probably most recognized as the solution for autonomous cars. However, a faster-growing market is for short-range LiDAR, which is being used for things like robots that only need to see a few feet, drones for collision avoidance, and driver-alertness systems. “Short-range LiDAR systems do not require as much current as long-range LiDAR systems. But seeing [over a] short distance means you need an even faster pulse, because if you’re measuring something that is 1 meter away, [for example], the return signal will come back in just nanoseconds. We’ve demonstrated systems for short-range LiDAR that have pulse widths less than 1.2 ns.”

Wireless power

“Wireless energy is ready to be incorporated into our daily lives,” Lidow said. Transmitters can be placed in furniture, walls, and floors to power or charge electronic and electrical devices efficiently and economically over large areas and across multiple devices.
Figure 7: EPC2037 enhancement-mode power transistor (Image: EPC)
Wireless energy transfer has been studied for more than 100 years; in fact, the concept dates back to the invention of the Tesla coil. A key factor in making viable wireless energy transmission systems viable is efficiency: To define such a system effectively, a large portion of the energy transmitted by the generator must reach the receiving device. Magnetic resonance technology is the linchpin to ubiquitous implementation, enabling transmission over large areas, spatial freedom for positioning reception devices, and the ability to power multiple devices simultaneously. EPC offers a full range of transmitter and receiver reference designs from single device charging to multiple devices powered simultaneously across a large surface area. GaN enables high efficiency for both the low-frequency (Qi) and high-frequency (AirFuel) standards, supporting a lower-cost, single-transmit amplifier solution that can wirelessly charge devices regardless of the standard used in the receiving device. Wireless charging systems that rely on the Qi standard operate by inductive coupling at frequencies in the 100- to 300-kHz range.

Audio applications

The lower power dissipation of Class D audio systems produces less heat, saves space and costs for printed circuit boards, and extends battery life in portable systems. Now that GaN HEMT devices with much better physical properties have become a reality, a leap in Class D amplifier performance is on the doorstep. The low resistance and low capacitance of eGaN FETs offer low transient intermodulation distortion. Fast switching capability and zero reverse-recovery charge enable higher output linearity and low crossover distortion for lower total harmonic distortion.
Figure 8: Isolated power-conditioning unit, the Cesium PCU-1C28, designed with EPC devices (Image: EPC)
“The first Class D amplifiers were designed for cars, because they wanted to have more speakers and more power in cars,” said Lidow. “Class A amplifiers were just too big to produce more than about 25 W and still fit in the dashboard. Class D was first introduced in the 1980s and enabled cars with 16 speakers and 250 W of power. Its sound quality, however, was never as good that of a Class A amplifier. That’s because MOSFETs can’t switch fast enough, and therefore, the relatively low switching frequency means relatively poor-quality reproduction. And with GaN devices, of course, you can go to much higher frequencies.”

Space applications

Enhanced-mode GaN is widely used in device development for space applications. Commercial GaN power devices offer significantly higher performance than traditional radiation-hardened devices based on silicon technology. This allows the implementation of innovative architectures with applications on satellites, data transmission, drones, robotics, and spacecraft. Smaller than equivalent MOSFETs, eGaN FETs provide radiation tolerance, fast switching speed, and improved efficiency, leading to smaller and lighter power supplies (smaller magnets and reduced heat sink size or even elimination of heat sinks in many cases). Faster transient response can also reduce capacitor size. Using these FETs, power supply designers have the choice of increasing the frequency to allow smaller magnets, increasing efficiency, or designing a satisfactory balance of both.

Leave a comment