Prof. Dr.-Ing. Andreas Lindemann, Head of the Chair for Power Electronics, Institute of Electric Power Systems (IESY), Otto-von-Guericke-University Magdeburg
Andreas Lindemann holds a diploma and doctoral degree in electrical engineering. He has worked on R&D in power semiconductor industry for ten years. Since 2004, he holds the Chair for Power Electronics at Otto-von-Guericke-Universität Magdeburg. His research interests include the use of novel power sem-iconductor devices in circuits and systems for various modern applications with a special focus on aspects like reliability and EMC. He is among others engaged for IEEE PELS, VDE and ECPE, co-chairing the biannual International Conference on Integrated Power Electronics Systems (CIPS) and the ECPE SiC & GaN User Forum.
Can you give us an overview of current developments in GaN power devices?
GaN devices in production have voltage ratings up to some 650 V and current ratings up to the order of magnitude of 100 A. All of them are lateral transistors, i.e., HEMTs (which means high electron mobility transistor) or GITs (which means gate injection transistor). HEMTs are purely voltage controlled with low gate capacity and can be normally-on or normally-off, while GITs require a small gate current to fully turn on, which makes a differ-ence for the driver design. Current is carried exclusively by electrons which means that the devices are purely unipolar, thus fast switching and not avalanche-rated. Even normally-off devices can conduct in reverse direction similar to a diode but remaining unipolar, i.e., they don’t show a reverse recovery current peak. Normally-on devices can either be controlled directly – i.e., be turned off by applying a negative gate voltage – or via a low voltage MOSFET in a cascode circuit, while the control of normally-off devices is basically similar to MOSFET control which is often preferred for straightforward circuit design even though the driver parameters are different.
The lateral structures can be integrated, allowing to join multiple switches and parts of the control circuit. For cost reasons they are usually manufactured on silicon substrates with an intermediate buffer layer.
Beyond this state-of-the-art research is on-going: For example, vertical devices have been demonstrated to e.g. achieve higher voltage ratings.
Why are applications – such as E-Mobility –suitable for GaN transistors?
GaN power devices have been frequently used in switched-mode power supplies up to the kW range. This seems natural because those often operate with DC link voltages up to 400 V which fits well to the devices’ rated blocking voltage. There often is no need for parallel con-nection of devices because of the achievable current ratings which facilitates circuit design. In addition, at least low power levels can ben-efit from the aforementioned integration. The low switching losses – in hard- and even more in soft-switching operation – allow for high efficiency even at high operating frequency which is beneficial to minimise the size of transformers, inductors, or filters.
Of course, the technology is not limited to these most typical applications. Higher currents can be achieved by paralleling devices, higher voltages with appropriate circuits like three-level converters. Drives – in particular high-speed drives – may also profit from the device properties, especially when efficiency is important. Application-specific requirements – like for instance overload or short circuit capability – need to be taken into account in addition.
This leads to your question regarding E-Mobility: An electric car needs a small onboard charger which basically is a switched-mode power supply. In addition, there is a corre-sponding converter between the traction battery, which often has a terminal voltage of about 400 V, and the 12 V system. Further, there are auxiliary drives – such as for electric power steering or air conditioning – as well as the traction drives, both requiring inverters. Power converters in electric vehicles should be as compact and efficient as possible. In view of these considerations, it seems natural to equip at least some of them with GaN devices.
Can you give us an insight into the research to solve problems such as undesirable dynamic RDSon and current collapse effects?
GaN power devices profit from the fact that the technology has been adopted from RF electronics where it already had reached a high level of maturity. Specific fundamental ques-tions have been scientifically clarified – let me mention as an example a doctoral thesis which has shown that the piezo-electric properties of the material and the related mechanical stress do not lead to a lifetime limitation of GaN power devices. Reliability tests have been car-ried out and reported and field experience has been gained with the considerable production volume. This means that a lot of research has already been carried out but of course and as always, we are not done:
Drift effects are observed like the dynamic RDSon you mention. It basically leads to a higher conduction voltage drop and thus higher conduction losses. The effect depends on the operating conditions, in particular the applied blocking voltage prior to turning the device on. It is temporary, i.e., decreases within the conduction time, but may accu-mulate in continuous switching operation. A deep physical understanding of such effects which are related to charge trapping in the individual semiconductor structures of the different devices we discussed is important to optimise the devices. Ongoing research and development have already permitted to reduce such effects as researchers and manufacturers reported e.g. on the occasion of the recent 10th ECPE SiC & GaN User Forum.
Another challenge concerning circuit design and application is device ruggedness. As men-tioned, the devices are unipolar and thus not avalanche rated. For this reason, the device manufacturers usually provide a considerable safety margin between the rated blocking and the breakdown voltage. Suitable circuit design is of course also required to avoid overvoltages, especially when considering the steep switching slopes. In a similar way, also the driver circuit must be carefully designed because the different gate structures partially are rather sensitive to overvoltages. A good match of potentially optimised devices with the circuit will be required to achieve the desired operational behaviour under all condi-tions which finally is also important on system level where e.g. EMC standards need to be fulfilled.
How has the development of SiC components –SiC MOSFETs and JFETs in particular – continued in recent years?
Generally speaking, SiC MOSFETs have clearly become the mainstream SiC transistors: They are normally-off and thus convenient for the design of most circuits. Many issues which had been observed in the earlier development phase of the technology have been mostly solved, concerning e.g. material quality; this increase of maturity is basically a similar devel-opment as observed earlier with Silicon. While SiC MOSFETs were initially commercialised in the voltage range from 650 V to 1200 V the ratings have increased in the meantime and reached 3.3 kV. The chip sizes remain smaller than we are used to from silicon which contrib-utes to the production yield.
Regarding the applications this means that the voltage range of commercial SiC transis-tors begins where GaN transistors’ ends. The devices thus are of interest e.g. for converters connected to the 400 V three-phase grid or DC link voltages significantly above 400 V. The high voltage ratings can be used for example in traction or industrial drives. Of course, parallel connection of chips is then required in many cases to cover the required power ratings.
This leads to the aspect of packaging: While lateral GaN devices – which we discussed before – mostly use some kind of chip-scale package the SiC devices are vertical and can in principle be packaged in a similar way like silicon devices. However – as has been learned quite early – the packages must be adapted. This on one hand concerns the joining tech-nique which must cope with the different material properties, such as higher stiffness and thermal conductivity. On the other hand, the electrical properties are different, too, and the package must for example be low-induc-tive because of the steep switching slopes. As of today, there are SiC transistors in a con-siderable variety of packages, beginning with conventional footprints like TO-247 – which obviously is hardly optimised, but cheap – and ending with advanced modules.
Ongoing research allowed to fabricate various samples beyond commercial products, for example 10 kV SiC MOSFETs. They demonstrate the technological possibilities and challenges. This also applies to the application circuits where for example the handling of the steep and high switching slopes with sufficient gal-vanic isolation and avoiding partial discharge is an issue.
How do electric vehicles, railroad and similar applications benefit from SiC?
It is known that suitable unipolar devices – i.e., MOSFETs – allow to increase the efficiency of drive inverters because of the lower switching losses and the resistive on-state characteristic in contrast to IGBTs where this has a knee voltage. When the losses are lower the thermal design allows to shrink the power section, i.e., to increase the power density. Although the requirements in electric cars are quite different from those in rail vehicles – also concerning ex-pected lifetime and cost – increased efficiency and power density are beneficial for both. The rated voltage range of today’s SiC MOSFETs allows to use such devices in high performant cars with battery voltages also above 400 V as well as in railway traction drives. This has been demonstrated some years ago already and, in the meantime, has reached volume production in car industry. In turn this helps to grow the production volume and decrease the cost we didn’t talk about too much.
The load profile of traction drives is cyclic. A proper inverter design must make sure that the power semiconductor components will achieve an appropriate lifetime under such operating conditions. For this purpose, they are qualified in reliability tests such as power cycling. SiC MOSFETs deliver meaningful and satisfactory results here which nevertheless also uncover effects which are subject to further research: For example, carriers can be trapped in a partially reversible way, leading to a drift of threshold voltage and a change of device characteristics. Again, it is important to understand such effects in order to on one hand develop suitable test methods and on the other hand appropriately consider the effects in the circuit design.