SiC transistors: Futuristic potential or practical solution?

For a decade, the industry has witnessed the healthy progression of silicon carbide discrete devices gaining adoption into energy-critical power conversion applications. Due to unparalleled parameter advantages over silicon incumbents, SiC JBS diodes are now widespread in higher-power computing PFC circuits and solar inverters, almost always paired with best-in-class silicon MOSFETs to achieve highest powertrain efficiency. This has driven the industry to consume >150 million units of SiC JBS discrete diode in 2012 alone. However, designers have had little progressive choice in transistor technology because, although SiC JBS products are commercially entrenched, SiC transistor devices such as JFETs, MOSFETs, and BJTs have not yet achieved any real stabilization in offering, manufacturability, or long-term reliability and hence have not achieved mainstream acceptance. Theory, case studies and suppliers web sites all present compelling reasons for SiC transistor adoption but still questions still linger : what is the best SiC transistor solution, when will the industry see truely viable products that are genuine “drop in replacements” and will they live up to the expectation?

Problem
Due to well-known wide bandgap material properties, fundamentally all SiC transistors outperform their silicon rivals in all key parameters that pertain to conduction and switching behavior; so realistically there is no bad choice of SiC transistor. However, like the proverbial shoe – not one size fits all, and there are three competing SiC transistor technologies entering the market today, each with pros and cons. Thankfully, the design community is familiar with these technologies in their silicon forms, so no huge leaps of faith are required. Designers are asking what are the expectations of each SiC transistor technology, what problems do they solve, and — last but not least — what are the challenges?

First and foremost, no supplier has yet developed and released the “WonderSiC” transistor that has zero on-state resistance, has infinitesimal switching loss, and is simpler to drive than silicon. In addition, whatever SiC die is used, it will be performance limited from its true possibility by the package vehicle housing it. SiC transistors in traditional through-hole packages such as TO220 and TO247 are absolutely restricted from their best ability due to the high package inductance limiting the achievable device switching speed and poor package thermal conductivity limiting the die power dissipation. Any progressive breakthrough for SiC transistors needs to be enabled by parallel breakthroughs in packaging and, on a system level, progressive nonstandard-form-factor power dense modules.

Study
Taking the implied remit that every power circuit manufactured today is minimizing footprint while increasing power and efficiency, switch selection is paramount to deliver system needs. Table 1 simply compares and contrasts the three main SiC transistor technologies, in “like for like” standard discrete packages in perspective to important device parameter and design considerations.

Table 1: Comparison of the three main SiC transistor technologies

In essence, any SiC switch offers many progressive advantages over its silicon equivalent, but when one looks into the second- and third-order properties, differences are apparant. All three technologies offer low conduction and switching loss, can theoretically operate up to several thousand kilovolts in breakdown, and are immensely rugged in avalanche capability over that of conventional silicon switches. However, firstly, the SiC MOSFET will require a larger die area per R_DS(on) than the JFET or BJT.

Gate driving any SiC transistor is a complete topic in itself, but in summary we can make these observations:

• The MOSFET is more conventional in approach, but still requires a large gate voltage swing (–5 to +25 V).

• The JFET is still a voltage-driven device, but either normally on or normally off; it also requires the use of very well controlled unique voltage levels that are perhaps classed as awkward to most designers.

• The BJT, despite having excellent current gain, is still a current-driven device and hence the least efficient to drive.

Body diode performance — or even the availability to reverse conduction, favors the SiC MOSFET because it does contain an inherent body diode that has excellent reverse-recovery characteristics, but also struggles with a high V_F . The BJT is unable to freewheel, and the JFET can operate in the third quadrant, but that cannot be classed as true body diode capability.

Operation at junction temperature ≥175°C is a great advantage of any SiC switch, but the MOSFET gate oxide can exhibit challenges at high temperatures and so does long-term bipolar degradation in the BJT. The unipolar JFET doesn’t suffer from either issue, and this stance can be reflected in the device's long-term reliability and is the strongest candidate and consequently seen in many high-temperature, high-reliability applications.

So in essence there is no clear winner among the three leading SiC switch technologies when comparing each on a general form and function basis. However, testing and observing actual device switching waveforms in real circuit conditions shows a very different picture.

Investigation
Analysis was performed grouping similarly performing SiC switch variants: 1,200 V_DS , 80-mΩ R_DS(on) . The BJT was not included in the exercise as, at the time of evaluation, the BJT was not commercially available. Figures 1 and 2 show device turn-on and device turn-off, respectively, with a dc-bus of 600 V and a 12-A load current.

It can be seen during device turn-on that the SiC MOSFET No. 1 (green trace) exhibits the least turn-on switching loss, shows some current overshoot and rests to steady state within 0.2 μs. The JFET (red trace) has identical Di/Dt to the MOSFET No. 1, but shows a slower Dv/Dt — hence more switching loss. MOSFET No. 2 unfortunately is by far the slowest device and exhibits considerably more turn-on loss than any other DUT.

During device turn off, again MOSFET No 1 offers by far the lowest loss, followed by MOSFET No. 2 and then finally the JFET. 
 

Fig 1: Device turn-on of a dc-bus of 600 V and 12-A load current.

  Fig. 2: Device turn-off of a dc-bus of 600 V and 12-A load current.

From the simple yet significant testing, it can be seen that a series of 80-mΩ SiC switches will naturally exhibit similar conduction losses (I² R_DS(on) ), but during switching transitions we see significant important differences that relates to device loss and hence overall performance. It can be noted there is a fourth trace shown on both graphs. This trace (blue) shows a single packaged 80 mΩ SiC normally on JFET in cascode with a 30-V 4-mΩ Si MOSFET. It can be seen that the cascode turn on matches a virtually identical trajectory to the best-performing SiC MOSFET, and the same situation is charted during turn-off. So a JFET cascode and MOSFET No. 1 are clear winners in the switching-loss contest. However, it is worth noting that the cascode solution is driven with a simple 0 to 10 V_GS , the cascode can happily operate to 175°C T_J and finally the cascode offers genuine excellent body diode performance of 0.7 V_F .

Any SiC switch will eclipse its silicon counterpart in performance, but all SiC switch offerings are not made equal and the user has to accommodate certain nuisances in implementation and occasionally restrictions in use. There are, though, novel answers to these problems and a single packaged SiC JFET cascode can provide a practical solution.