Recent industry developments, such as the significant growth in wide-bandgap semiconductor commercialization, have underscored the fact that GaN, for example, is available today from several companies in high volume manufacturing. Although it is not yet mature, GaN devices are also far from being a new technology. The packaging is becoming very standard, capacity is there, and the price point is well on its way.
By Alix Paultre, Editor Power Electronics News (based on a technical paper presented by Steven Oliver, Navitas Semiconductor)
One can arguably state that 1977 was the year we will mark in our history as the time power started its rise to the prominence in the industry it enjoys today. Once at the periphery of embedded design, power is now enjoying an important prominence. Where a digital board designer used to simply request from the power guy (or gal) that 12 volts come from the space (if you were lucky large enough) left for you on the board, now power an enabling function that lets the electric vehicle go further, or the cell phone charge faster, or enables the facility manager to save double digits in energy costs.
A revolution in power
The year 1977 was when the HEXFET came out, developed by a team led by Alex Lidow and Dan Kenza at International Rectifier. Silicon General and Unitrod came out with the PWMIC that year as well. There were also advances in magnetic materials, enabling things to go faster, and novel typologies, once only in universities, suddenly became mainstream industry things. The core technologies that enabled the commercialization of the switch-mode power supply really happened in the late 1970’s.
This new combination of power semiconductors, typologies, control, and magnetics gave the industry an amazing shift in efficiency, with the associated power density increase, while the cost came right down. This paved the way for the flood of development, bringing us technologies like the resonant power supply and the LLC converter, and Infineon’s groundbreaking CoolMOS technology.
Wide-bandgap semiconductors represent the second revolution in power electronics. These new power semiconductors inspire new controllers from the control companies and improved topologies like active-clamp flyback, introduced in the late ‘90’s. In addition, the latest magnetics can easily operate at one or two megahertz. Again, in this revolution efficiency goes up, frequency goes up, and cost will also come down .
There have been major steps forward in magnetic performance over the last few years, and improved materials are available. In the case of GaN, as recently as two years ago one could make the case that the magnetics were not there, so why try to go faster, but that is no longer the case.
With the semiconductors and magnetics picking up speed, the controller presents an area for development as well, if only to keep up with the capabilities of the other devices involved. In the case of Navitas’ AllGaN technology, the company takes an enhancement-mode GaN FET and pairs it with an integrated driver on the same die. Navitas can also integrate other devices like a level shifter or ECD diodes, or even in-logic under-voltage lock-out, making it closer to a true GaN power IC .
Since this all occurs on a single chip, it’s relatively to manufacture and package, bringing other system benefits as well. With the gate of the FET directly attached to the output of the gate driver, there is zero impedance between driver output and gate input. It also means that GaN is not the easiest thing to work with on a discreet basis, but if you can have a complete solution in one box, it means you can carefully control the voltage going to the gate, which means you also protect the gate.
An integrated solution also means that you can go high frequency, without worrying about gate impedance getting in the way in terms of performance, in terms of efficiency, and also stability. Another advantage is that an integrated drive increases robustness, and with a single chip solution, layout is flexible.
In high-frequency applications, a soft switching resonant topology is a good way to go. However, in real-world situations there may be start up routines, maybe burst mode, where you have hard switching. So it is important that everything is very tough and can reliably do hard switching and soft switching as well.
A discrete solution with no gate impedance or resistor present creates an erratic, unstable circuit. If you add an impedance, it becomes stable, but it also creates a very long switching cycle. So if you have directly-connected driver output and gate input, than you can go fast and stay very stable. The slew rate is a simple thing to do with just a resistor. So integration means a very controlled, high performance device.
One other thing about the integration is when we think about ESD. ESD with standard silicon MOSFETs is very well understood. ESD with PWMICs is very well known and understood. GaN devices in discrete form, not so much. We have to be careful about that, but if you make the power IC, than you can put ESD diodes on it. So this becomes another way that using that integration to make life easy, enabling integration of things like a level shifter, or shoot-through protection for a half bridge.
In this case the solution footprint is six by eight millimeters with a complete half bridge system and bootstrap charging function, with under voltage lock out and ESD protection. In terms of frequency, these devices are rated at two megahertz. One of the notes about high frequency level shifting is that there are technologies where there’s inductive coupled or capacitive coupled, which do this level shifting, but by their nature are different technologies.
In terms of ease of use, a discrete solution versus an integrated solution with driver results in an obvious reduction in components, makes life more predictable, can be faster to design, and results in higher performance. Integration is the key for control and also performance. An integrated solution gives you a very predictable building block for the future.
The next figure shows an example of a 65W reference design using an active-clamp flyback running at around 300 kilohertz in a soft-switching typology. In the case of a standard flyback or a quasi-resonant flyback, a snubber network handles the flyback voltage. By replacing the snubber network with a second switch, it becomes a half-bridge able to push the frequency up to 300 kilohertz.
Active clamp was introduced by Virginia Polytech in 1996, and now have we have control ICs that can not only achieve a high power density, but also meet the DoE level six requirements and European Union requirements as well. The design is presented as the world’s smallest power supply with full USB-PD output and has more complexity and function than a standard power supply, providing a 20V output from a worst-case condition of 90Vac input at over over 93% efficiency full load.
The nnext two figure show a design in megahertz using the same typology. This one was actually done in 2016 by the CPES group of Virginia Polytech. This one megahertz, active clamp flyback used a DSP instead of a control IC was a project to prove frequency and power density of full load, but in the past, they couldn’t make it an industry product, because it would not do light load efficiency. Now with the control ICs, we can do that.
In the case of EMI, one of the things that you can do by going high frequency, and we believe there’s probably a tipping point around 500 kilohertz, where you can go from a bobbin-based transformer or a toroid-based transformer. If you go to planar transformers around the 500 kilohertz mark, you can do a few things. One is you can get very low profile, which helps your power density. The second is you can actually get the same or potentially lower cost transformer solution, because you don’t have to go through the extra steps of making the whole wound transformer.
You can just do PCP design with equals on either side. But within this design, inside the planar transformer, there are also some EMI shielding layers, so it traps the EMI, or cancels the EMI at the source. So in this case, we have a very, very small single side EMI filter, which means it meets all of the specs, and if anyone would like more information we can show you this.
Since this was a university project using a DSP, because that’s what was available, using new controllers and the new half bridge, we can actually shrink this design down again.
So now, in the same size as is the standard 5W or 7W sugar cube power supply, we can squeeze in 25 watts. So this is using GaN, using new magnetic materials, in this case, this is Hitachi ML 91S. Using a new typology, previously academic, now industry, and with the control ICs available, we can now take advantage of all of this with that integration of the driver and the moss fet to push the boundary of performance.