A Critical Look At The SiC, High-voltage MOSFET
Merits of the high-voltage SiC MOSFET include low switching losses and a high operating temperature. But is this transistor sufficiently rugged and reliable for widespread deployment? BY MUNAF RAHIMO FROM ABB SWITZERLAND LTD SEMICONDUCTORS
For decades, silicon devices have dominated the power electronics market. They have maintained pole position through tremendous advances in the starting material quality, in the techniques used for process fabrication, and in the architecture of the device.
However, silicon is not the perfect power electronics material: its inherent material properties are notably inferior to those of wide-bandgap materials. This has spurred the research and development of SiC devices, which deliver superior performance. Advantages include a wider range of voltage ratings, lower losses and higher operating temperatures.
These attributes are welcomed by designers of electrical systems, because they enable increased operating efficiencies and higher current densities, frequencies and temperatures. Savings that result from all these attributes are not only measured in terms of energy, but extend to the cost, weight and size of the unit, thanks to the opportunity to downsize the semiconductor content, the cooling system and the passive components/magnetics.
The first SiC device to hit the market was the SiC-based Schottky barrier diode (SBD). Launched in the early 2000s, it initially sported ratings below 1200 V. It can be purchased as a discrete component, or as part of a hybrid module, where it is paired with silicon IGBTs. Many low power inverters have been built with these devices, including those that serve the solar PV market.
More recently, the SiC power device portfolio has expanded through the launch of SiC MOSFETs with ratings below 1700 V (see Figure 1). These transistors are available as discretes, and also as full SiC MOSFET modules.
Many of the leading producers of SiC MOSFETs are now trying to increase the voltage rating of the devices to 3.3 kV and beyond (see Figure 2 for an example of the capability of these devices). This move has the potential to provide similar improvements to those in the lower power range. However, these MOSFETs will be targeting different applications with megawatt powers, such as: grid systems, both HVDC and FACTS; railway traction converters; medium-voltage industrial drives; and high-power renewable energy conversion and storage.
Figure 2. Rohm is one of the pioneers of the high-voltage SiC MOSFET. Its performance is compared to a IGBT produced by ABB.
The majority of the high-power applications in the megawatt range are based on a voltage source converter (VSC) topology, and operate at switching frequencies below 2 kHz. Today, at the heart of these systems are high-voltage silicon switches, such as IGBTs and IGCTs, and diodes with voltage ratings up to 6.5 kV (see Figure 3). The capability of all these devices continues to increase, with evolutionary steps on silicon device/package technologies providing continual improvements in system performance, reliability and cost.
Figure 3. SiC devices are targeting various applications in the high-power semiconductor market that are currently served by a portfolio of silicon products.
Traditionally, depending on the requirements of the particular application and optimum performance-to-cost ratio, VSC-based applications adopt either a two-level (see Figure 4) or three-level topology (note that the level refers to the number of output voltages). With this technology, the route to higher voltages is the series connection of many devices. HVDC systems adopt this approach.
Figure 4. A typical voltage-source-converter H-Bridge with four IGBTs and diodes.
An important trend in recent years, which has been seen in many grid system and industrial drive applications, is to turn to multi-level topologies. This allows operation at much lower frequencies, such as below 300 Hz, and results in very low losses and far higher power levels.
One consequence of having a wide range of high-power applications is that it leads to differences in specifications, including those for power handling, efficiency and cost. Design engineers must carefully select from many available device and package solutions to satisfy the requirements of a particular application. However, there is some common ground, as all high-power applications share a design-in practice that is best described as a meticulous and relatively slow process. The focus is on the overall performance and reliability, rather than just considering basic power ratings and associated loss calculations.
At the global power technology giant ABB, we have extensive experience in evaluating the requirements for power electronics for mainstream, megawatt applications. In the remainder of this article we will detail the key characteristics for power electronics in general, before considering and evaluating the opportunities for the SiC MOSFET in the high power range.
The primary considerations for power electronics for megawatt applications may be divided into three categories: the power density handling capability; the capability for controllable, soft switching; and ruggedness, fault-handling and reliability. Devices that handle high powers should ideally exhibit low conduction and switching losses, a low thermal resistance, and the ability to operate at high temperatures while not compromising reliability.
When it comes to the second category, controllable and soft switching, there needs to be a soft-controllable turn-off that is accomplished with low overshoot voltages and minimal EMI levels, and a control of the turn-on that addresses transients while minimizing losses. The third category – ruggedness, fault-handling and reliability – involves considerations of the safe operating area associated with the turn-off current, fault-handling issues that include the short-circuit capability of IGBTs and surge current limitations of diodes, and reliability requirements associated with: current and voltage sharing, for devices arranged in parallel and series; stable conduction; and stable blocking performance.
While considering all of these issues, design engineers must also evaluate the attributes of the package. There is a shift towards more compact packages, which combine a higher packaging density with low parasitic elements and the opportunity to optimize the electrical layout. Additional trends are the development of more powerful packages, judged in terms of current, voltage and temperature handling capabilities, and reliability to temperature and power cycling.
If a power semiconductor is to be designed into a high-power application, it must fulfil every one of the requirements detailed above. Given this condition, it is our view that SiC MOSFETs must overcome a number of challenges, which depend on the specific application, before they offer strong competition to the well-established, and still improving, silicon devices.
SiC MOSFETs: Pros and cons
While there is still work to do, there is no question that unipolar SiC devices, and in particular the SiC MOSFET and SBD, are the most promising options for voltage ratings of up to 10 kV. Measurements on devices show that for voltages of up to 10 kV, drastic reductions in switching losses are realized in conjunction with low conduction losses, even at high temperatures. From a practical perspective, this is encouraging, because the voltage-controlled gate drive needed for the SiC MOSFET is similar to that for the widely adopted IGBT.
Lagging behind, in terms of performance, are high-voltage SiC bipolar devices, such as p-i-n diodes, IGBTs and thyristors. They are all impaired by high on-state losses, which stem from the built-in voltage that is inherent in the p-n junction: it exceeds 2 V. Making matters worse, SiC bipolar devices are plagued by forward-voltage degradation, which can be traced to crystal staking faults that materialise during device conduction.
Given this state of affairs, it is clear that there is a compelling case to focus on just SiC MOSFETs and SBDs. Here, there is increasing interest in running the SiC MOSFET in diode-mode operation. That’s because of the ease in implementing a gate drive for the MOSFET conduction during freewheeling mode, the low reverse recovery losses, and the resulting lower costs associated with extra SiC diodes. We believe that for high-power applications, this approach offers a perfect opportunity. To understand why, consider our following analysis.
Handling the power
As power devices improve, they trim the overall losses. This enables an increase in power density, and the opportunity to boost system efficiency. For both silicon IGBTs and SiC MOSFETs, realising this requires the optimisation of the channel density, resistance, MOS cell pitch and device thickness. The key difference between the two transistors is that: with the silicon IGBT, there is also the need to fine-tune the excess carrier enhancement in the base region; while with the unipolar SiC MOSFET, efforts must be directed at lowering the on-resistance via judicious selection of the doping and the thickness of the base region.
For higher voltage devices, the loss component is more prominent, strengthening the case for SiC. Assuming a similar conduction loss at typical nominal currents, compared to the silicon diode, the SiC MOSFET delivers a total reduction of more than 75 percent in switch-mode switching losses, while providing negligible diode recovery losses. The low overall losses of the SiC MOSFET make it an attractive candidate for many mainstream applications that utilize different topologies operating below 2 kHz.
Another advantage of the SiC MOSFET over both the silicon IGBT and diode is the absence of a built-in voltage. This equips the SiC MOSFET with the potential to cut losses with each technological improvement – an accomplishment that is beyond the reach of silicon devices.
Armed with this attribute, the SiC MOSFET is an attractive option for modern multi-level topologies that require low conduction losses. Furthermore, the SiC MOSFET low conduction losses at low currents are actually very attractive in many applications, including FACTS/STATCOM and urban traction, where a large percentage of losses are dissipated in sub-load or idle conditions.
A well-known strength of wide bandgap materials is their superior thermal properties to silicon. However, this advantage has minimal impact, because the thermal resistance is largely dominated by the device area, packaging and cooling. Making more of a difference is the low leakage current at high temperatures, because this allows SiC devices to operate beyond 200°C.
One of the limitations of the SiC MOSFET is the packaging technologies employed, including the encapsulation materials and joining techniques. Reliability is held back, just as it is for packaged silicon devices. Consequently, the majority of system improvements that come from simplified cooling stem from the lower losses of SiC devices. Note, however, that there is also a weakness afflicting the SiC MOSFET at elevated temperatures, associated with its strong positive temperature coefficient. Due to this, conduction losses increase at higher temperatures, and have to be addressed by increasing the device area – an approach that adds to chip costs.
Unipolar devices are a concern with regard to their switching characteristics. This is blamed on oscillatory behaviour during switching transients, originating from the absence of excess carriers. Bipolar devices, in contrast, are known for softer characteristics. However, while there is truth in these views – and it is important to develop low inductive packages and circuits – the SiC MOSFET benefits from adjustments in the gate drive set-up, with the greatest gains coming from optimisation of the turn-off gate resistor.
Although a higher gate value increases the energy consumed in the off-state, losses are still far lower than for those with an IGBT. So, system performance in high-power applications is enhanced significantly, due to the substantial reduction in losses.
Additional virtues of the SiC MOSFET are that it responds well during turn-on, and when doing so, it produces far lower energy losses than the silicon IGBT. These benefits come from the low reverse-recovery peak current, and the ability to ramp down the voltage very fast during switching.
Ruggedness, reliability and fault-handling
For high-power applications, ruggedness, reliability and fault-handling all play a major role during the design-in phase of power devices. For SiC MOSFETs, these requirements could be the stumbling block to their introduction in the high-voltage range.
One of the strengths of the SiC MOSFET is its substantial safe-operating-area during turn-off. Due to the unipolar nature of this device, it does not suffer from a dynamic avalanche limitation that impairs safe-operating area capability – note that this does hamper the silicon IGBT. What’s more, the SiC MOSFET has a higher junction built-in voltage, so, compared to its silicon cousin, it has extra protection against parasitic n-p-n transistor failure modes during switching. Due to these characteristics, much higher currents are needed to forward bias the n-source, due to the higher built-in voltage of the wide bandgap material. In turn, there is a potential improvement in safe-operating-area capability.
When it comes to fault handling, one of the key performance requirements for high-voltage applications, SiC MOSFETs are yet to reach the required levels, so there is a need to either improve device capabilities, or alternatively, employ new solutions at the system/gate drive level.
Encouragingly, low-voltage SiC MOSFETs have already some capability of withstanding short circuit conditions. But there is a need to employ new gate drive conditions to limit the pulse durations, because these devices have high short-circuit levels – and this leads to higher thermal stress, due to the high MOS cell densities that are used to ensure a low on-resistance. The downside is that improvements to short-circuit capability will come at the expense of increases in conduction losses, which are considerable in high-voltage devices.
With respect to diode surge current capability, there is a need to accommodate a high current, which is up to ten times the nominal current for different pulse durations. When the SiC MOSFET is utilized in diode mode as discussed earlier, the integrated p-i-n diode is better suited to providing the required surge current capability than a separate unipolar diode, due to the lower conduction losses at such high currents.
Concerns over reliability must consider how the SiC MOSFET is deployed within the circuit. If it is used in a parallel configuration involving many chips in a high current module, the device’s strong positive temperature coefficient and its low temperature dependency on the switching losses should ensure good current sharing. If it is used in series, there should also be no major issues, thanks to low leakage current levels and little dependency on increasing temperatures. The greatest challenges are to realise low levels of voltage changes with time, and minimal oscillations. This is accomplished by reducing the stray inductances and optimizing the gate drive parameters, as they can influence the voltage sharing of series-connected devices during switching events.
Another key characteristic of semiconductor devices is their stability over time. This is evaluated with standard qualification tests. SiC MOSFETs suffer from higher electric fields at the MOS cell, so to combat this, developers of these devices have optimized processes and material preparation techniques to ensure a low shift in the gate parameters, such as the threshold voltage. An additional concern is a shift in the on-state forward voltage. As the SiC MOSFET p-i-n diode will be functional during reverse recovery switching and in surge events, the device will be flooded with excess carriers (bipolar mode), which could produce stacking faults that lead to shifts in the forward voltage drop.
In addition, there are major concerns associated with the stability of the blocking behaviour of the SiC MOSFET. To ensure stability, there is a need to increase the protection of the device through superior junction termination/passivation designs, and the introduction of better packages, which provide extra protection against harsh environmental conditions, such as high humidity and temperature variations.
To increase cost competiveness with silicon, SiC manufacturers have capitalized on the wide bandgap properties and shrunk the junction termination region, so that they can maximize the usable active area. The downside is that the electric fields outside the semiconductor bulk material, such as those in the passivation layers and package filling materials, are higher than those in silicon. What’s needed are optimum chip designs and packaging materials for SiC device protection.
A further consideration is to prevent failures that are induced by cosmic rays. This is possible through optimisation of the thickness and doping levels of the base region. Getting this just right is essential, as it will impact the conduction losses associated with the on-resistance.
In short, while there is no doubt that the high-voltage SiC MOSFET has tremendous promise to deliver substantial efficiency savings, alongside reductions in the weight, size and cost of power systems, today doubts remain over the reliability, ruggedness and fault-handling capability of these wide bandgap transistors.