How Gallium Nitride (GaN) Enables Smaller, More Efficient Power Supplies

Hello and welcome to CUI Inc's training module on how GaN enables smaller, more efficient power supplies. My name is Ron Stull and I'm the Power Product Marketing Engineer at CUI. We're a power supply company and we specialize in customizable wall mount and desktop adapters. Today, I'm going to go over what GaN is and how it benefits power supplies and show you how CUI is taking advantage of GaN in our latest desktop adapters.

So what is GaN? GaN, or Gallium Nitride, is a metallic alloy of gallium and nitrogen. These elements have three and five valence electrons respectively and form a mixed ionic covalent bond. The resulting GaN molecule is a semiconductor material, meaning that it has a conductivity between that of a conductor and an insulator, and because it is composed of more than one element, it's classified as a compound semiconductor. This is compared to the more abundant silicon semiconductor, which is composed of a single element. It is additionally classified due to the number of valence electrons, as a group 3-5 semiconductor and within this group, belongs to the nitride subset. Group three five materials are known as direct band gap semiconductors and are able to emit light, whereas indirect semiconductors, such as silicon, cannot. This makes semi-conductors in this category potentially suitable for LED applications. Other semiconductors in the 3-5 group include Gallium Phosphide, which is used in LEDs, Gallium Arsenide, which is used in RF devices, and NDM Phosphide, whose applications include photovoltaics.

There's one more category of materials that GaN belongs to, and that is the class of wide bandgap semiconductors, and it's this characteristic associated with this group that is most valuable to power supply manufacturers compared to conventional semiconductors, such as silicon, which have band gaps around one to one and a half electron volts. Wide band gap semiconductors have band gaps greater than two electron volts. Other wide bandgap materials include Silicon Carbide and Carbon as shown in the table to the right. The band gap is the energy separation between valence and conduction bands within the material and a wider band gap makes something more insulative and capable of withstanding higher voltages. Elevated temperatures will also act to close the gap by applying thermal energy to the valence band and so having a wider gap means that a material is able to withstand higher temperatures as well. A third benefit of wideband gap semiconductors is higher electron mobility, which means that electrons are able to move through the material faster for the same applied electric field, which translates into higher frequency capabilities than that of conventional semiconductors.

These many characteristics of GaN as a compound direct wide bandgap semiconductor material make it beneficial to many applications. As we said, the fact that it is a direct band gap material means it can emit light, which has led to its use as the laser diode in blu-ray players. The high frequency and voltage and temperature ruggedness have made it useful in military applications as well. For example, Lockheed martin has been using GaN for the last decade and is using it to upgrade US Military radars to help improve performance and reliability. And, also in the last decade or so, the development of high electron mobility transistors (HEMTs) has seen GaN enter the world of power electronics, where GaN HEMTs replace the conventional silicon MOSFET. The wide bandgap enables higher switching frequencies, voltage ratings, and temperatures, all of which are a benefit to power supply designs.

So where will you find GaN in a power supply? Power adapters and other switching power supplies convert power using semiconductor switches, most commonly in the form of silicon MOSFETs. These switches must conduct large amounts of current at high frequencies, and because of this, are one of the main contributors to power loss and heat generation inside a power supply. Because of this, manufacturers are always looking for better switches to make their power supplies smaller, lighter, and more efficient. Traditionally, improvements to efficiency would come incrementally through technological improvements to MOSFETs. MOSFETs have been around for a long time, originally replacing bipolar junction transistors, and have stuck around by continually improving. However, there is a physical limit to these improvements, and MOSFET manufacturers are quickly approaching it. GaN HEMTs offer a big improvement in performance over silicon and key areas that enable improvements to efficiency and form factor over conventional silicon MOSFETs, and their relative newness means that they have a lot of room to improve, providing power manufacturers with a solution to improve their devices well into the future.

So how does GaN improve efficiency? Efficiency is a key feature of a power supply and one that GaN significantly improves by reducing the losses of the power switch. Losses in the power switch are generally separated into two categories: switching and conduction. One difficulty in improving the efficiency of the switch is that these two loss mechanisms are inversely related, that is, improving one will negatively affect the other. This makes comparing devices difficult and so a figurative merit is used to compare the relative performance of these devices. The figure of merit is the product of gate charge and on resistance of the FET and a lower value is considered better. GaN, as can be seen on the right, offers a significant improvement in the figure of merit over silicon MOSFETs, meaning that it should be more efficient in a power supply application. So now let's take a look at the individual loss categories and see how GaN improves them.

Conduction losses are the simplest of the two to understand. They occur when the switch is fully on. In this state, the switch resembles a resistor, which is called rds, and the losses are equal to the resistance times the rms current squared. From this, we can see then that the conduction losses are proportional to the RDS value and we therefore desire it to be low. The curve to the right shows the theoretical limit of RDS per unit of die size with respect to the breakdown voltage. This represents the lowest value of RDS that can be achieved for a given breakdown. Actual device characteristics will lie somewhere to the left of these curves. MOSFETs have been decreasing RDS steadily for decades, but they're fast approaching this theoretical limit, meaning that future improvements will be small and tougher to achieve, but for GaN, the naturally higher breakdown voltage means that for a given breakdown, the RDS again is much lower and because of this its theoretical limit is much further to the right. This means that while GaN already shows a better performance than silicon, they can expect to see further improvements for some time to come. Switching losses compared to conduction are a little more complicated and have several parts. These losses occur as the switch transitions between on and off states and are proportional to the switching speed and switching frequency. Switches do not change state instantaneously and the rising and falling of voltages and currents during this time produce switching losses, which can limit switching frequency and efficiency. The higher the frequency, the higher the average losses will be over time. The components of these losses are the transition time, reverse recovery, and the dead time.

Transition time losses occur because as the switch turns on or off, there is a period of time when the current is conducting but the voltage is still high. For example, when transitioning from off to on, the switch voltage starts high, around 400 volts usually, and while it's high, the current ramps up to its on on state value. It is not until after the current has ramped up that the switch voltage will fall down to its on state. The on to off transition is similar and the voltage increases to the off state voltage before current stops conducting. This results in short periods of very high losses and we therefore want to keep this period of time as short as possible to limit the average value. The high electron mobility of GaN helps increase the switching speed, which decreases the amount of time that the current and voltage are high at the same time. This both decreases switching losses and also enables higher switching frequencies.

In addition to the transition losses, there are also dead time and reverse recovery losses. In half bridge configurations, a period of time when both switches are off is needed in order to avoid a short circuit condition. During this time, current flows through an intrinsic diode of the MOSFET, known as the body diode. This diode is much lossier than the resistance of the channel and so to limit these losses, we want the dead time to be as short as possible. Here GaN helps us in two ways. First, there is no body diode. During the dead time, the current flows through the more efficient RDS channel, which reduces the losses compared to a MOSFET, which has a body diode, and again because of the high electron mobility, the GaN device can switch faster and operate with a shorter dead time, further improving the switching losses.

The last aspect of switching losses is the reverse recovery of the body diode. After the body diode has conducted, it must be turned off and there are losses associated with this. In order for the diode to turn off and be able to block current again, it must reverse briefly to reset itself internally. During this time, the current and voltage go negative and incur losses. Because GaN lacks the body diode, there's no body diode to turn off and therefore we don't suffer the same reverse recovery losses that we've seen in the MOSFET.

We've now seen that GaN helps to improve all aspects of losses inside the power switch and this not only improves efficiency, but can allow for significant increases in switching frequency that previously were hindered by extreme switching losses. Being able to increase the switching frequency allows the form factor of power adapters to be decreased. Increasing the switching frequency means that less energy needs to be stored to sustain a switching cycle, and therefore the size of energy storage elements, such as transformers and capacitors and inductors, can be decreased. These components are the largest in a power supply and shrinking them can lead to significant form factor reductions. The frequency is able to be increased both because of the high electron mobility and because of the reduction in switching losses. GaN is also thermally more conductive as a lateral device relative to its vertical MOSFET cousins and has a higher maximum temperature as a result. Both of these characteristics make it much easier to manage heat compared to MOSFETs. Operating a MOSFET at these high frequencies would lead to unsustainable switching losses and temperature rises, requiring large amounts of thermal management, such as heat sinks, and airflow which would negate the benefit of the increased frequency. You can see on the right the package size is also considerably smaller, which is a result of the higher voltage withstand capabilities.

So with all these benefits to GaN, why is it taking so long to see GaN power supplies? From the development side, there are a few complications. One is the fact that GaN is a depletion mode device, which means that it's on by default. Power supplies require enhancement mode devices, which are off by default. There can also be some difficulties in driving a GaN device compared to that of a MOSFET, which have kept some designers from implementing them. Manufacturers have dealt with this in different ways, but a common one is to insert a low voltage MOSFET in a cast code configuration, as shown on the right. This makes it so that the user can drive the GaN device in the same way they've driven MOSFETs in the past, making it easier, and converting to an enhancement mode device at the same time. Besides developmental issues, the cost has been a factor. Silicon is an abundant material and MOSFETs are a mature technology. Gallium, on the other hand, is not as abundant and GaN doesn't have the scale or infrastructure yet to bring the cost down to that of a MOSFET. But, as the technology matures and becomes more widely adopted, the cost will continue to fall and the number of solutions continue to grow.

Finally, to show the realized impact that GaN has had on adapters, take a look at CUI's latest GaN-based adapter: the 200G Series. These adapters switch up to six times faster than conventional silicon-based adapters and this has led to an increase in power density of 216%. This is primarily the result of reducing transformer size, which is made possible by the increased switching frequency. The smaller magnetic components also led to a decrease in weight of 25% and while a conventional silicon-based adapter may sacrifice efficiency to try and fit into a smaller form factor, with GaN, this tiny form factor is realized while still achieving and exceeding the most stringent efficiency standard - COC tier 2. These GaN adapters peak at over 96% efficiency.