SiC-based MOSFETs offer benefits in automotive, power applications
By Jeffrey Fedison, PhD, Sr Applications Engineer, STMicroelectronics
Wednesday, 08 January, 2020
As conventional silicon-based MOSFETs mature, they are now reaching their theoretical limits of performance. Wide-bandgap semiconductor devices represent an interesting alternative for improving performance due to their electrical, thermal and mechanical properties.
Since the first commercially available silicon-based power MOSFETs were introduced almost 40 years ago, they (along with their cousins, IGBTs) have been the primary power-handling control component in switching supplies for powering circuits, driving motors and countless other applications.
However, their roles have also grown to where they have become victims of this success. As these semiconductor switches have improved in overall performance, especially in areas such as reduction of on-resistance and switching loss, their applications have expanded across many dimensions. As a result, more and more has been expected of these Si-based MOSFETs and IGBTs, with continually increasing demands for even better performance.
At some point, of course, the law of diminishing returns takes over, despite the efforts of leading researchers and vendors to further improve these MOSFETs/IGBTs. In the past few years, the incremental advances have been modest despite hard work and investment. This is not unusual: eventually, a technology and products reach a point where the gains are small compared to efforts, thus setting the stage for a new, disruptive approach and new devices.
For MOSFET devices, this disruptive cycle is a result of developing and mastering a new base material. MOSFETs based on silicon carbide (SiC) semiconductors rather than silicon alone have demonstrated significantly better performance than using silicon alone. Note that this is not just based on R&D samples or prototype demonstration, as these SiC-based MOSFETs are already becoming accepted in commercial use.
One of the major and rapidly growing applications which benefit from these devices, and in turn is driving their development and production, is electric/hybrid vehicles (EV/HEV). These battery-laden automobiles are much more than a large battery pack connected to electric traction motors (for HEVs, with a small gasoline engine for charging) even if that is how most consumers think of them. Instead, they require many electronic modules for operation, management and specialty functions.
Among the many power-switching converter systems in the EV/HEV are:
- traction inverters for the wheel motors (200 kW/up to 20 kHz);
- AC-input onboard charger (20 kW/50 kHz to 200 kHz);
- optional fast-charging function (50 kW/50 kHz to 200 kHz);
- power for auxiliary functions: driver’s console, battery management and control, air conditioning, infotainment, GPS, connectivity (on the order of 4 kW/50 kHz to 200 kHz).
Why focus on efficiency? Obviously, one priority for an EV/HEV is driving range. Even a small improvement in inverter performance corresponds to a meaningful increase in that basic figure of merit as seen by the customer.
But it is much more than just that. Increased efficiency is desirable due to multiple factors:
- Lower operating temperatures for enhanced reliability.
- Reduced thermal load, meaning less heat to be dissipated via heat sinks, radiators, cooling fluids and other techniques.
- Less charging time and basic electricity use.
- More flexibility in overall packaging due to the demands and constraints which inherently come with hotter systems.
- Greater ease in meeting regulatory mandates.
SiC addresses the challenge
Fortunately, SiC offers a path to the higher efficiency needed as well as delivering related benefits. How does SiC differ in structure and performance compared to mainstream silicon-only MOSFETs? In brief, SiC uses an SiC n+ substrate, topped by an SiC n-doped epitaxial layer (also called the drift layer; see Figure 1). Critical on-resistance parameter RDS(ON) is largely determined by the resistance RDrift of the drift layer, and the resistance of the channel between the source/body contact and the drift layer.
For a given RDrift value, at a junction temperature of 25°C, a SiC transistor has a practical die area many times smaller than that of silicon super-junction alone, which translates to far higher performance with a comparable-sized die. Another perspective is a comparison of SiC to Si using the well-known figure of merit (FOM) RDS(ON) × die area (lower is better). At a blocking voltage of 1200 V, a SiC MOSFET from ST has a FOM value which is about one-tenth that of the best available high-voltage silicon MOSFET (900 V superjunction).
If we compare the SiC MOSFET against the silicon IGBT mostly used in traction inverters, the main benefits include:
- reduced switching loss, along with lower conduction loss at low and medium power levels;
- no PN junction voltage drop as there is for the IGBT;
- the SiC device has a robust and fast intrinsic body diode, eliminating the need for an external diode; this intrinsic diode has almost negligible recovery charge;
- the ability to operate at higher temperatures (200°C), which reduces cooling requirements and heatsinking demands while enhancing reliability;
- the ability to operate at four times the frequency of an IGBT at the same efficiency, resulting in weight, size and cost reduction due to smaller passives and fewer external components.
What about the drive?
Experienced engineers know that the power device alone is only one of the many components critical to a complete system. Achieving a reliable, efficient and cost-effective design also requires the appropriate drivers for the MOSFET — one which is tailored to the unique current slewing, voltage levels and timing constraints of the MOSFET and its load. Since the Si-based MOSFET technology is mature, there are many standard drivers available from multiple sources to make the driver/MOSFET pair work, and work well.
Therefore, one legitimate concern about SiC MOSFETs is how easy or hard they are to drive, and more importantly, whether the drivers are available. The good news is that driving an SiC-based MOSFET is almost as easy as driving the Si-based MOSFET: it just needs a gate-source voltage of 20 V, with current drive on the order of about 2 A (max) for an 80 mΩ device. As a result, simple and standard gate drivers can be used in many cases. ST and others have developed additional gate drivers that are optimised for SiC MOSFETs, such as the ST TD350.
This advanced gate driver includes an innovative, active Miller-clamp function that eliminates the need for negative gate drive in most applications, and allows the use of a simple bootstrap supply for the high-side driver. It also incorporates a two-level turn-off feature with adjustable level and delay, which protects against excessive overvoltage at turn-off in case of overcurrent or short-circuit conditions. The same delay which is set in the two-level turn-off feature can be applied at turn-on to prevent pulse-width distortion.
Reality, not just speculation
Process advances sometimes fail to make the jump from lab R&D to production reality and volume use. This is not at all the case with SiC-based MOSFETs. They are in full production and already used in HEV/EV designs with tangible benefits related to efficiency, performance and operating conditions that translate to tangible benefits at the circuit and system levels.
A comparison of an 80 kW traction-motor inverter power module used in HEV/EV applications shows how 650 V SiC MOSFETs outperformed silicon IGBTs across many key operating parameters. The three-phase inverter design used a bipolar PWM topology with synchronous rectification mode for the reverse current. Both devices were sized to yield a junction temperature of about 80% of their absolute maximum rating. The design used four paralleled 650 V/200 A IGBTs with the relevant freewheeling silicon diodes of the same rating versus seven paralleled 650 V/100 A SiC MOSFETs without any external diodes (just using the intrinsic body diode). Peak-power rating was 480 Arms (10 seconds) and a normal load working at 230 Arms. Other operating conditions were:
- DC-link voltage: 400 Vdc
- Switching frequency: 16 kHz
- Vgs of +20 V/-5 V for SiC, Vge of ±15 V for IGBT
- Cooling fluid temperature: 85°C
- RthJ-C(IGBT-die) = 0.4°C/W; RthJ-C(SiC-die) = 1.25°C/W
- Tj ≤ 80% × Tjmax°C at any condition
Typical power losses at the peak-power rating are shown in Table 1.
SiC MOSFETS offer lower conduction loss compared to IGBTs because of the benefits of paralleling: when MOSFETs are used in parallel, the resultant RDS(ON) is divided by the number of MOSFETs, thus driving the conduction losses down towards zero; in contrast, when IGBTs are used in parallel, the resultant VCE(SAT) does not decrease linearly, and the minimum on-voltage drop is limited to about 0.8 to 1 V.
SiC-based MOSFET solutions show much lower losses across the entire load range. These MOSFETs also show a drop in conduction loss from 125 W to 55 W at 100% load level, due to the lower on-voltage drop (see Figure 2a and 2b).
SiC-based devices offer up to 3% improvement in efficiency at low load and a better aggregate efficiency (>1%) across the entire load range. While 1% may not seem like much, at these levels, it represents considerable power, associated dissipation and heating. As engineers know, higher temperatures are the enemy of sustained performance and reliability. In addition, this translates into extended range for EVs, which is a big value proposition, appreciated by both car makers and consumers. In the SiC versus IGBT comparison at 16 kHz, the former is the clear winner from low to full load (see Figure 3), with the cooling fluid at 85°C for both; the data shows that the IGBT cooling system must be made more efficient due to its higher losses.
The SiC devices are also cooler across nearly the entire operating frequency range (see Figure 4) and run cooler than IGBTs even when operating as low as 8 kHz, and the Si-based IGBT is already out of its specification range at 46 kHz.
If the SiC MOSFET solution is dimensioned so that the junction temperature is kept below a set maximum level (typically, 80% of Tjmax of 200°C) at the peak power-pulse condition where its conduction losses are greater than that of the IGBT, then the SiC MOSFET offers distinct benefits including:
- smaller semiconductor area, for a more compact solution;
- much lower losses at low-to-medium loads;
- much longer battery autonomy and therefore range;
- lower losses at full load, for a smaller cooling solution;
- lower difference between junction temperature Tj and cooling-fluid temperature Tfluid across the entire load range, for enhanced reliability.
These attributes yield tangible value to the user, with an efficiency improvement of at least 1% (75% lower loss); a smaller, lighter cooling system on the inverter side (roughly 80% reduction); and a smaller, lighter power module (50% reduction).
Then there is cost
No discussion of any technology advance and its benefits is complete without looking at the cost factors. At present, the cost of the SiC-based MOSFET is four to five times that of a silicon IGBT. This differential in basic component cost is often outweighed by total system savings in BOM, cooling and energy costs. In the two- to five-year outlook, this price ratio should decrease to 3x or even 2.5x as the industry moves to larger-diameter wafers — ST has already begun this transition — and with improvements in RDSON × Area figure of merit, along with higher volumes. Over the longer-term span of 5–10 years, costs will continue to be driven down due to progress along the same factors.
SiC-based power switches offer both the promise and reality of performance benefits with few design trade-offs in application and installation. Given the intense activity related to HEV/EV development and their many associated power modules, as well as other higher-power motor-centric applications, they can play a large role in successful designs where even small improvements yield large system-level benefits.
Fugaku has taken out the top spot on the TOP500 list — a ranking of the world's fastest...
Costing aside, choosing the right industrial joystick when such requirement is called for can be...
Australian researchers have successfully tested and recorded what is claimed to be the...