1. Introduction
Silicon carbide (SiC) MOSFETs offer a range of advantages over silicon-based switches, including faster switching, higher efficiency, higher operating voltages, and higher temperatures, enabling smaller and lighter designs [
1,
2]. Thus, significant progress has been made in the power semiconductor industry. These advantages allow SiC MOSFETs to be applied in a range of automotive and industrial applications. However, the excessively fast switching speed of SiC MOSFETs causes high values of di/dt and dv/dt [
3]. At the same time, due to the existence of parasitic inductance and parasitic resistance in circuits, SiC MOSFETs are prone to overshoots and oscillations in current and voltage during switching, thus generating additional switching losses. These losses can even cause device damage. Therefore, reducing or even eliminating the overshoot, oscillation, and electromagnetic interference (EMI) of silicon carbide MOSFETs during switching is a priority in improving their working efficiency [
4].
Current and voltage overshoots are the biggest obstacles in SiC semiconductor applications. Stray inductance and parasitic capacitance are the main contributors to current and voltage overshoots [
5,
6]. When a SiC MOSFET works at a high frequency, the switching speed becomes too fast and the di/dt and dv/dt slopes increase. Thus, current and voltage overshoots are amplified due to stray inductance and parasitic capacitance [
7]. When the voltage and current overshoots exceed the breakdown voltage of the SiC MOSFET and the maximum recovery current of the freewheeling diode, the SiC is destroyed. However, the parasitic elements in the actual circuit cannot be eliminated [
8,
9,
10,
11]; so, the overshoot can only be eliminated by other means.
Solving the problems of overshoots and oscillations is key to improving the efficiency of SiC MOSFETs.
In the traditional approach, slowing the switching speed in the case of high gate resistance [
12] can alleviate the SiC MOSFET’s overshoots and oscillations. However, this solution means the power losses become greater with longer switching times [
6]. Adding an RC snubber circuit is also a common method [
13,
14] to eliminate overshoot and oscillation problems. Although highly efficient snubber circuits have been proposed, snubber circuits affect the overall efficiency of the system. In particular, extra devices increase the power loss of the overall circuit. Active gate driver (AGD) solutions have been developed to increase the efficiency of power devices. The main advantages of these gate drivers are reduced oscillations and overshoots. However, the extra components not only increase the volume of the SiC system but also increase the circuit’s power loss [
15,
16,
17,
18,
19,
20,
21]. In [
21], a new active gate driver was proposed that could effectively suppress overshoots and oscillation and reduce losses without compromising EMI. The main strategy of the proposed AGD was to reduce the current and voltage slope by reducing the gate driver’s voltage. However, its main disadvantages were that the process of calculating the parameters of the transformer was complicated, an additional printed circuit board was required, and the implementation of the circuit was complicated and expensive. As such, configurable digital gate drivers (DGDs) are the latest technological development in this sphere [
22]. The strengths of DGDs are their programmability, support for a wide variety of power devices, and ease of use [
23]. A key element of this technology is the ability to configure the turn-on and turn-off processes, which provides a series of steps to control the voltage level at specific times [
24,
25,
26,
27,
28]. This allows designers to digitally configure the turn-on and turn-off curve through software without requiring changes to the hardware.
In this paper, an optimized switching strategy based on gate drivers with variable voltage (OSS) was proposed to improve the switching performance of SiC MOSFETs. This switching strategy was based on and optimized for an AGD in [
21] and a DGD in [
22]. There is no doubt that the AGD performed well in [
21]. However, the gate driver circuits were too complex and too large. The ease of operation of the DGD in [
22] is very attractive. Therefore, this article aimed to combine the advantages of the voltage gate drivers of AGDs and DGDs in waveforms to further optimize the algorithm and control and reduce overshoots and oscillations in SiC MOSFETs through an OSS. Finally, we used the LTspice simulation software to verify the feasibility of our switching strategy. Compared with AGDs and DGDs, an OSS can control more variables to achieve more precise and optimal circuit control. Due to space limitations, this article focused on the impact of delay time on switching performance and lists the formula to calculate the delay time for a given calculated decrease in voltage. At the same time, the decrease in the voltage and the time of this decrease were also considered. The development potential of the OSS is significant. Compared with other switching strategies, which can only change one or two variables, the OSS can control three variables simultaneously, namely, the delay time before a decrease in voltage, the decrease in voltage itself, and the time duration of the decrease in voltage. More variables mean better control performance. However, due to various limitations, this paper only considered the influence of the delay time before a decrease in voltage on switch performance and analyzed the relationship between voltage drop and switch performance. In our next work, we will produce more simulations and experimental results to prove the advanced nature of the OSS.
This paper is organized as follows.
Section 2 and
Section 3 present the control principle, circuit model, and related formulations of the proposed switching strategy. In
Section 4, the simulation setup and simulation results are introduced and analyzed. Finally, in
Section 5, the conclusions and a discussion of this simulation are given.
3. Calculation for Voltage Drop Delay Time
An OSS can change the switching performance through unlimited control of the voltage drop delay time, voltage drop, and voltage drop time, aiming to obtain the best data for the switching performance through countless permutations and combinations. However, this is difficult to carry out. In fact, the best data can be derived by equation. The main research direction of this paper was to explore the influence of the voltage drop delay time on the switching performance and to explore the influence of different voltage drops and different voltage drop times on the switching performance. Thus, the content of this chapter focuses on deriving the best voltage drop delay time by equation.
3.1. Equivalent Slope for Current and Voltage
The slope of the current and voltage is an important factor affecting the overshoot and oscillation during the switching process of the SiC MOSFET. It can be seen from
Figure 2 that the I
OS in the turn-on stage and the V
OS in the turn-off stage were obvious and large. However, the slope of the voltage and current in the SiC MOSFET with the OSS method was not constant because of the variable gate drive voltage; so, the equivalent slope was proposed and applied in the calculation, with the aim of writing the voltage drop delay times, t
d1 and t
d2, into the equation.
In
Figure 3, when the V
GG was 20 V, the slope of the current was dI
D/dt
1; when the V
GG was 15 V, the slope of the current was dI
D/dt
2. Similarly, dV
DC /dt
1 and dV
DC /dt
2 represented the value of the voltage slope under the V
GG of −5 and 0 V. In addition, ton represented the increase in the current I
D and toff represented the increase in the voltage V
DS; t
on1 and t
off1 were the rising times under dI
D/dt
1 and dV
DC/dt
1, respectively. Moreover, dI
D/dt represented the actual current slope with OSS and dV
DC/dt was the current slope with OSS, in actuality. Additionally, dI
D/dt
eq was the equivalent slope at the turn-on stage and dV
DC/dt
eq represented the equivalent slope at the turn-off stage; t
on2 and t
off2 were the actual running times for the turn-on stage and turn-off stage. In [
21], the slope of the current was used in both the turn-on and turn-off stages because it was convenient for the subsequent calculations. In this paper, in order to fit reality, the voltage slope was used in the turn-off stage. According to the equivalent principle shown in
Figure 3, the following equation could be obtained:
and the equivalent rising time t
on2 of the current can be expressed as:
Thus, according to (5) and (6), the value of the equivalent current slope at the turn-on stage can be expressed as:
Similarly, at the turn-off stage, the equations of t
off2 and voltage slope are:
3.2. Power Losses in the Turn-On and Turn-Off Stages
The turn-on and turn-off losses of the power device could be calculated by (11) and (12), which were derived in []:
where
is the energy generated by the current passing through
and E
Irr is the energy loss of the device generated by the reverse recovery effect, which can be calculated by (14) [
29]:
where L
loop is the stray inductance in the power loop, Q
rr is the reverse recovery charge, and
is the overshoot ratio, which can be defined as:
During the turn-on transient processes, the energy losses during the current and voltage changes can be expressed as:
Therefore, according to Formulas (11) and (13)–(17), the energy loss during the turn-on stage can be expressed as:
Similarly, according to (12), the energy loss can be expressed as:
3.3. Calculation of the Voltage Drop Delay Time
In order to simplify the calculation model, it was assumed that the change of power consumption was only related to overshoot and was positively correlated. Therefore, during the turn-on stage of the SiC MOSFET, the change in power consumption is positively correlated with the change in current overshoot [
21]. Similarly, in the turn-off phase, the change of power consumption is positively correlated with the change of voltage overshoot. The formula can be expressed as:
In the turn-on phase, the most important factor affecting power consumption is the overshoot of the current; so, Equation (21) can be abbreviated as:
Because the main cause of current overshoot is the reverse recovery effect [
21], the current overshoot can be expressed as:
According to (25) and (26), (11) can be expressed as:
Similarly, in the turn-off stage, the formula can be obtained:
According to (1)–(4), (7), and (24), an equation expressed by td1 could be obtained. Similarly, according to (1)–(4), (10), and (26), an equation expressed by td2 could be obtained. Therefore, the optimal voltage drop delay times, td1 and td2, could be calculated.
4. Simulation Verification
In order to evaluate the performance of the proposed OSS, a double-pulse simulation test of the SiC MOSFET was performed using the software LTspice. The power device used in the test was CREE’s 1.2 KV/115 A SiC MOSFET module (C3M0016120D). In the simulation, the tested module was first modeled in LTspice to ensure that the simulation results would be closer to the experimental data and then the circuit was built and simulated. The detailed parameters of the tested modules in the experiment are shown in
Table 1.
The comparison of the OSS’s performance under different time delays was simulated to verify whether the calculated time delay was the best. To further compare the performance of the OSS, it was compared with hard switching (HS) and a digital gate driver (DGD) from Agileswitch.
4.1. SiC MOSFET Device
In order to facilitate future experiments based on this simulation, this simulation used the existing silicon carbide semiconductor module in the laboratory as the test equipment. The detailed parameters of the tested modules in the experiment are shown in
Table 1.
Figure 4 shows the device.
4.2. Simulation Modeling
To improve the simulation, we chose LTspice as the simulation software. As shown in
Figure 5, and according to
Figure 1, a simulation circuit was built. The main purpose of this simulation was to verify the effect on the switching performance of the OSS and compare it with other switching strategies. It was firstly particularly important to verify whether it was necessary to study the OSS in depth. A simulation can obtain more accurate results and save time. Once the excellent switching performance of the OSS was confirmed, the next step was to build and experiment with actual circuits. According to (24), (26), and data from
Table 1, we calculated an optimized delay time t
d1 of 21 ns and t
d2 of 41 ns; t
vd1 and t
vd2 used the rise time (t
r) and fall time (t
f) in the device data sheet for the simulation, respectively.
Table 2 shows the parameters of some components in the analog circuit. Among them, C
1 and C
4 represent C
ds, C
2 and C
5 represent C
dg, and C
3 and C
6 represent C
gs. In order for the oscillation in the simulation results to be closer to the actual situation, L
S, L
D, and L
G were adjusted.
Because the double-pulse simulation of the SiC MOSFET was carried out using the software LTspice, it was necessary to simulate the double-pulse signal, as shown in
Figure 6. Because of the convenience of the simulation software, the voltage waveform from the voltage source could be edited directly. When the SiC MOSFET was simulated with the OSS, the waveform could also be adjusted directly at the voltage source, which required corresponding devices in the actual experiments.
4.3. Simulation Results
In order to verify the feasibility of the proposed OSS, the simulation was run with the gate resistance 2.6 Ω, both at the turn-on stage and turn-off stages; the dc bus voltage was 500 V and the drain current was 30 A. The voltage drop delay times t
d1 of 21 ns and t
d2 of 41 ns were calculated by Equations (11) and (12). Meanwhile, as visible from
Table 1, the t
vd1 was 28 ns and the t
vd2 was 27 ns.
It can be seen in
Figure 7a that when the turn-on signal arrived, first, a high V
GG (i.e., 20 V) was applied to the gate terminal. After the delay time t
d1, which was calculated to be 21 ns, the I
DS started to rise and the V
GG dropped to 15 V. After the delay time t
vd1 (i.e., 28 ns), the high V
GG restarted to charge the SiC MOSFET. At the turn-off stage, the SiC MOSFET was first discharged by −5 V. After the delay time t
d2, which was calculated to be 41 ns, the V
DS started to rise from 0 V and the V
GG became 0 V. After the delay time t
vd2 (i.e., 27 ns), the V
DS reached the peak value and the negative V
GG was discharged again.
As mentioned in
Section 2, the voltage drop delay times, t
d1 and t
d2, are very important because they can affect the switching performance of the SiC MOSFET. In order to verify the relationship between the delay time and the performance of the switch, and to further the accuracy of the calculation results in the previous chapter, more experiments were carried out. As shown in
Figure 8, around the calculation result (i.e., t
d1 = 21 ns t
d2 = 41 ns) at an interval of 10 ns, a total of four other data points were taken before and after each transition and were simulated. Then, the simulation results were compared.
For the turn-on transition, because the calculation result of t
d1 was 21 ns, the five testing data points were 11 ns, 21 ns, 31 ns, 41 ns, and 51 ns, respectively. The reason t
d1 was not set to 0 was that this waveform was equal to the switching strategy of the digital gate driver. It can be seen from
Figure 8a that there was no obvious difference and the V
DS and I
DS waveforms generated by the five different t
d1 were not notably different. When t
d1 was 11 ns, the overshoot of the I
DS was the smallest but the overshoot produced by the V
DS was the largest.
For the turn-off transition, since the calculation result of t
d2 was 41 ns, the five test data points were 31 ns, 41 ns, 51 ns, 61 ns, and 71 ns, respectively. It can be seen from
Figure 8b that there was an obvious difference and the V
DS and I
DS waveforms produced by the five different t
d1 showed little difference. When t
d1 was 41 ns, the overshoot and oscillation of the I
DS were the smallest; the overshoot and oscillation of the V
DS were also the smallest.
As the delay time increased or decreased, the calculation results in both cases were the best. However, a trade-off between the VDS and IDS could be achieved when a td1 of about 21 ns at turn on and td2 of about 41 ns at turn off were, respectively, applied.
4.4. Comparison with Other Switching Strategies
After determining the optimal time delay of the OSS, it was necessary to compare the OSS with other switching strategies. First, we compared the OSS with the hard-switching (HS) strategy to confirm its advantages and improvements. Then, we compared it with a digital gate driver (DGD), which is widely used in teaching and research, to further verify the advantages of the OSS. A detailed experimental comparison of these different switching strategies follows.
4.4.1. Compared to the Hard-Switching Strategy
In order to demonstrate the excellent switching performance of the OSS in terms of overshoot and oscillation, more simulations were performed and the results were compared with the switching performance of hard switching. It can be seen from
Figure 9a that during the turn-on stage, the peak value of the I
DS dropped from 70.04 A to 60.23 A, a decrease of 14%, and the oscillation was also significantly reduced. However, the overshoot of the V
DS increased, which will increase the power loss; although, this will not affect the peak value of the V
DS.
Unlike in the turn-on process, the OSS successfully reduced the oscillation and peak value during the turn-off process, not only regarding the I
DS but also the V
DS. It can be seen from
Figure 9b that the overshoot of the V
DS reduced from 704.56 V to 577.52 V, a drop of 18.03%.
Combining the two processes of turn on and turn off, it can be clearly seen that the OSS effectively reduced the overshoot and oscillation of the IDS and VDS in the SiC MOSFET switch, thereby reducing the overshoot and power consumption of the entire circuit.
4.4.2. Compared to Digital Gate Driver
The digital gate drive system from Agileswitch is a product with a wide range of applications. It can control the voltage drop and voltage drop time of the gate voltage without limitation to affect the switching performance of semiconductors. The control strategy of the DGD is also very advanced. However, the existing DGD control strategy is still unable to design more complex gate drive waveforms, especially the voltage drop after a delay time, like an AGD. Therefore, it was necessary to compare the OSS with existing digital gate driver strategies.
Figure 10 shows the V
DS and I
DS comparison of the OSS and DGD at the same turn-on time. Compared with the hard-switching strategy, the overshoot of the DGD was reduced by 19%, which was better than the OSS’s 14%. Although the DGD was slightly better at reducing the I
DS overshoot and oscillation, the overshoot of the V
DS was larger, which seriously increased the power losses of the circuit. Although the two switching strategies were both effective in reducing overshoot, they increased power consumption and were low in cost performance. This is why the first generation of DGDs cannot adjust the pulse shape during turn on.
Figure 10 shows the comparison of the V
DS and I
DS for the OSS and DGD at the same turn-off time. Compared with the hard-switching strategy, the DGD had no advantage and even increased power consumption due to the longer switch-off process.
In order to better compare two different switching strategies, the closing time of the DGD was extended from 68 ns to 88 ns and was then simulated and compared. The new waveform comparison is in
Figure 11. Compared with
Figure 10b, the DGD effectively suppressed the overshoot and oscillation of the V
DS and the overshoot was reduced by 11%. Compared with the OSS, the DGD was inferior. Continuing to extend the turn-off time of the DGD will make the optimization more effective; but, this will inevitably increase power consumption and reduce the frequency of the semiconductor device.
4.5. Comparison with Different Drop Voltages
The OSS can become more complex and diverse. The previous chapter mainly considered the influence of the voltage time delay on switching performance; this section attempts to study different voltage drops.
It can be seen from
Figure 12 that, whether in the on state or the off state, the larger the voltage drop, the less overshoot. Because the increase in the voltage drop reduces the slope of the current and voltage, this proves that the formula in
Section 3 is valid. However, during the turn-off process, with the increase in the voltage drop, the oscillation of the voltage and current was obviously increased.
4.6. Comparison with Different Voltage Drop Times
In the previous simulation, in order to simplify the complexity of the simulation, the voltage drop time (tvd) used the rise time (tr) and fall time (tf) in the device datasheet. However, in the OSS, the voltage drop time is also an important parameter, similar to the delay time and voltage drop. Therefore, in this subsection, the effect of the voltage drop time on the switching performance was studied by simulation.
For the turn-on transition, because the t
r was 28 ns, five testing data points were chosen: 18 ns, 28 ns, 38 ns, 48 ns, and 58 ns. It can be seen from
Figure 13a that different voltage drop times did not have much influence on the switching performance at the turn-on stage.
For the turn-off transition, because the t
f was 27 ns, five testing data points were chosen: 17 ns, 27 ns, 37 ns, 47 ns, and 57 ns. It can be seen from
Figure 13b that in the turn-off stage, different voltage drop times had an impact on the switching performance. When the voltage drop time was 27 ns, the switching achieved the best performance. As the voltage drop time increased, the suppression effect of the OSS on voltage oscillations began to weaken. On the contrary, as the voltage drop time decreased, the voltage oscillation was suppressed; but, a larger reverse current was generated.
4.7. Power Loss Comparison
To determine whether a switching strategy is good or bad, in addition to observing whether it effectively reduces overshoot and oscillation, another important criterion is whether power consumption increases. Due to the advanced nature of LTspice, the product of current and voltage can be quickly integrated directly from the simulated circuit.
Table 3 records the power losses from simulation models in one turn-on and turn-off round.
It can be seen from
Table 3 that, compared with HS and the DGD, the power consumption of the OSS was slightly higher than that of HS. However, combined with
Figure 11, the optimization effect of the OSS remained obvious.
In the comparison of different voltage drops, it can be concluded that the smaller the voltage drop, the smaller the power consumption. However, combined with
Figure 12, it can be seen that the smaller the voltage drop, the smaller the suppression effect of the OSS on the circuit overshoot. Similarly, in the comparison of different voltage drop times, it can be concluded that the shorter the voltage drop time, the smaller the power consumption. However, combined with
Figure 13, it can be seen that the smaller the voltage drop, the smaller the effect of the OSS on circuit oscillation suppression.
Finally, in the comparison of different voltage drop delay times, it can be concluded that when the voltage drop time is the calculated value, the power consumption is close to the minimum.
5. Conclusions and Discussion
In this paper, an optimal-switching strategy was proposed for improving the switching performance of high-power SiC MOSFETs under hard-switching conditions. In addition, considering the trade-off between switching loss and switching overshoot, the voltage drop delay time was analyzed and calculated. By optimizing the voltage drop delay time of the turn-on and turn-off stages, the OSS can effectively minimize overshoot and suppress oscillation. The simulation results showed that the OSS can reduce the current overshot at the turn-on stage and minimize the voltage overshoot at the turn-off stage. In addition, compared with the hard-switching strategy, the current overshoot in the turn-on phase decreased by 14% and the voltage overshoot in the turn-off phase decreased by 18.03%. This showed that the OSS achieved a more comprehensive control strategy and better switching performance. Compared with digital gate drivers, although each has its own advantages, the OSS is more advanced and suffers less power losses.
The next step in this research is to make the actual circuit and test it. The contribution of this paper is limited due to simulation constraints. The optimization of the OSS in this paper was limited to the control voltage drop delay time; the control strategy could not be fully tested. Only the influence of a single variable of the OSS on the switching performance was tested. In future studies, multivariable control, such as delay time, voltage drop, voltage drop time, etc., could be tested in an actual circuit and jointly controlled to optimize semiconductor switching performance. However, we can confirm that the proposed OSS is promising compared to other methods.