Parasitic-Based Active Gate Driver Improving the Turn-On Process of 1.7 kV SiC Power MOSFET

Abstract

This article discusses an active gate driver for a 1.7 kV/325 A SiC MOSFET module. The main purpose of the driver is to adjust the gate voltage in specified moments to speed up the turn-on cycle and reduce the amount of dissipated energy. Moreover, an adequate manipulation of the gate voltage is necessary as the gate current should be reduced during the rise of the drain current to avoid overshoots and oscillations. The gate voltage is switched at the right moments on the basis of the feedback signal provided from a measurement of the voltage across the parasitic source inductance of the module. This approach simplifies the circuit and provides no additional power losses in the measuring circuit. The paper contains the theoretical background and detailed description of the active gate driver design. The model of the parasitic-based active gate driver was verified using the double-pulse procedure both in Saber simulations and laboratory experiments. The active gate driver decreases the turn-on energy of a 1.7 kV/325 A SiC MOSFET by 7% comparing to a conventional gate driver (V_DS = 900 V, I_D = 270 A, R_G = 20 Ω). Furthermore, the proposed active gate driver lowered the turn-on cycle time from 478 to 390 ns without any serious oscillations in the main circuit.

1. Introduction

Nowadays, medium-voltage (MV) silicon carbide (SiC) MOSFET power modules are used in a significant amount of power applications, including electric vehicle power systems and small renewable power plants based on photovoltaics [1]. Manufacturers of MV SiC MOSFETs provide new devices with short switching times counted in hundreds of nanoseconds which is related to low switching energies and power losses. On the other hand, high values of the drain current and drain-source voltages of those devices are the reason for significant di/dt and dv/dt values observed in power circuits. Parasitic elements under large di/dt and dv/dt ratios cause oscillations and overshoots [2,3,4]; thus, fast switching MV SiC MOSFETs are the source of EMI problems [5,6,7]. There are many ways to suppress these negative effects, for example, by decreasing the switching speed through an increase in the gate resistance, but the price is higher power loss. Another possibility is active gate driving, discussed in this paper. A conventional gate driver (CGD) with fixed gate voltage and resistances has no dynamic control of the gate current of the device. However, some drivers can manipulate the gate current in open loop by the change in the gate resistance in the specified moments of the switching cycle [8]. A number of open-loop gate drivers are dependent on the fixed value of the feedback signal which makes them impractical in most power converter applications. Nonetheless, the most preferable way is to use an active gate driver (AGD) which adjusts the gate resistance or the gate voltage in a closed loop. A number of AGD topologies have been presented in the literature [9,10,11,12,13,14,15]. They differ from short circuit protection [12] by means of feedback signals and, especially, considering the main purpose of those circuits, improvement in the switching performance of the transistors [15]. Thus, AGDs can be used in many applications to solve the commonly found issues in power electronics. Some of the mentioned problems are short circuit protection, v_DS and i_D overshoot or oscillations caused by a short switching period due to parasitics. In [9], active control of the gate-source voltage was realized via voltage injection to the gate: at the turn-on cycle, the AGD pulls down the gate potential to the ground to slow down the di/dt ratio. The aim of this research was to reduce the switching speed of a 1.2 kV/33 A SiC MOSFET in order to suppress voltage and current overshoot at turn-off and turn-on, respectively. Both simulation and experimental results showed that the authors reduced those negative effects significantly. On the other hand, switching losses were increased, since di/dt and dv/dt ratios were reduced as well. Furthermore, the AGD proposed by the authors was designed for a low-power SiC MOSFET. The switching trajectory improvement of an AGD for a 1.2 kV/40 A SiC MOSFET was discussed in [10], where the feedback loop of the driver was based on the resistive measurement derived from the drain-source voltage v_DS and drain current i_D. Then, the gate resistance was changed to manipulate the gate current and, in consequence, the AGD influenced the di/dt and dv/dt ratios. This type of measurement would be hard to implement in a MV SiC MOSFET since an auxiliary circuit is necessary. Furthermore, resistive measurements of the drain current and drain-source voltage generate additional power losses, which is undesirable in power electronic applications. The same type of the feedback signal was applied in an AGD where the purpose was to minimize the effects of parasitic inductances in a converter layout with a 1.2 kV/40 A discrete SiC MOSFET [11]. The authors used the same method as in [10] to modulate the gate current value in specified moments of the switching cycle. The researched AGD was experimentally validated by the authors with a double-pulse test and SiC MOSFET-based inverter with motor load. The AGD provided promising results in case of the turn-on and turn-off cycles. Unfortunately, the authors examined this solution only for low currents so it is not known if this approach could be used in high-power applications. The AGD can be also used as a short circuit protection of the device through the employment of a Rogowski coil to measure the current and use it as a feedback signal [12]. In this particular case, the AGD was designed for a 1.7 kV/325 A MV SiC MOSFET. Unfortunately, this solution generates additional expenses to the summary AGD cost as it needs an auxiliary circuit to be used in case of a feedback loop. In addition, the AGD can be used for voltage balancing in series connection of a 1.7 kV half-bridge SiC MOSFET module, which is the main topic in the paper [13]. As the closed-loop test results showed, the authors accomplished balancing of the drain-source voltage values in series connection in comparison to natural voltage sharing. In this case, the feedback signal of the AGD was acquired with a switch voltage sensor based on an RC branch connected to the drain of the SiC MOSFET. The main weakness of this solution was that the quantity of switching energy losses increased by less than 7.5%. Furthermore, the AGD can be used for SiC MOSFET oscillations damping [14] as well. In [14], the authors solved problems with overshoot and ringing in the body diode under inductive load. The subject of the experimental studies in this case was a 1.2 kV/36 A discrete SiC MOSFET which was tested with only a 400 V DC bus. The results indicated that the AGD had an impact on the oscillations and successfully suppressed them. However, in case of MV voltages, power modules are usually characterized by higher parasitic inductances and capacitances, and it is much more complicated to use closed-loop solutions due to disturbances in the feedback signals. Moreover, there are studies with an AGD for a 1.7 kV SiC MOSFET where the feedback loop is based on the parasitic inductances of the source and drain [15], and the multi-level gate voltage method is used for turn-on and turn-off cycle improvement. The application of the AGD reduced the turn-on energy by 30% for the 1.7 kV SiC MOSFET and showed a positive impact on EMI performance. To avoid high propagation delays, the main control unit of the AGD in this paper was based on a CPLD; alas, it made the whole device more complex. A multi-level gate voltage method for a 3.3 kV/450 A SiC MOSFET in open loop was described in [16]. This method was verified in a double-pulse test (DPT) and in a half-bridge inverter. The turn-on energy decreased by half through the manipulation of the gate driver supply. As a result, the PWM frequency of the inverter can be doubled. However, the gate driver works in open loop which limits the system capabilities for variable drain currents.

According to the literature review, it can be seen that active gate drivers are discussed mostly for low-voltage and -current transistors, while those for 1.7 kV SiC MOSFET power modules are rare [12,13,15]. Moreover, a lot of the articles presented discrete 1.2 kV SiC MOSFETs [9,10,11,14]. Some of the AGDs proposed in the papers are for MV; however, they have limited drain current capabilities, thus excluding them from most high-power applications. Therefore, the subject of this paper is to show simulation results and experimental validation of an AGD for MV and a high-current SiC MOSFET based on a power module. The closed-loop AGD for a 1.7 kV/325 A SiC MOSFET proposed in this paper uses voltage measurement across the parasitic source inductance to simplify the measuring circuit. Unlike most of the presented AGDs in the literature, this kind of feedback loop does not generate any additional energy losses to the system. Additionally, the AGD improves the turn-on cycle of the device by manipulation of the gate voltage to decrease the dissipated energy by increasing the dv_DS/dt ratio and suppressing the drain current oscillations. The proposed solution can be used in mostly all power SiC MOSFETs, affecting instant losses depletion without strengthening EMI. Modulation of the gate current is acquired by application of the multi-level gate voltage method. This approach allows having a constant value of the gate resistance, which optimizes the layout of the gate driver for the MV SiC MOSFET. The model of the proposed AGD was designed without any complex circuits such as CPLD as in [15] and is based on low-cost, high-speed, off-the-shelf integrated circuits. The main assumption was to improve the turn-on process with the proposed method since turn-on is usually the cause of much higher power losses in comparison to turn-off in the case of MV SiC MOSFETs in most hard switching applications [17]. The proposed AGD operates independently from the maximum drain current value in a range of operational conditions, which will be also discussed later in the paper. Thus, it can be used in highly efficient MV high-power electronic applications. The paper is organized as follows. The Section 2 of this paper provides a theoretical introduction of the methods used in the developed system. Then, the concept of the driver is verified in simulations in the Section 3. The experimental results and comparison of the AGD with CGD are shown in Section 4. Finally, the paper is concluded in Section 5.

2. The Active Gate Driver

Usually, the AGD input is connected to some feedback signal. A closed-loop signal can be obtained directly from the drain-source voltage, drain current or gate-source voltage [10,11]. However, all of those signals except the gate-source voltage are hard to be measured in case of MV SiC MOSFETs. For example, to measure the v_DS voltage, the measuring resistance must be connected between the drain and the source, which makes the complete circuit more complicated and influences the main circuit performance [10]. An approach that is based on the feedback signal in the form of measurement of i_D and v_DS requires a complex subcircuit to be included in the AGD. In the proposed AGD, the closed loop is provided with the V_Ls voltage (1). In order to explain this more clearly, a SiC MOSFET connected in the double-pulse test, as presented in Figure 1, is considered. The SiC MOSFET active gate driver is connected to the gate of the transistor through R_G and its reference ground is a Kelvin source (S_Kelvin). S_Kelvin in SiC MOSFETs is terminal, placed as close as possible to the actual chip inside the module and used for driving purposes. If the device was driven in reference to the power source (S), the source parasitic inductance L_S would have undesired effects on the gate current i_G. The existence of the parasitic inductance L_S between S_Kelvin and S is naturally caused by internal connections of the SiC MOSFET. Dependence of the temperature for the value of L_S is very limited, and it decreases slightly with rising temperature [18]; hence, temperature fluctuations can be omitted when considering this method. From (1), it can be considered that the value of V_Ls is proportional to the value of L_S and di_D/dt. Thus, the value of i_D and its dynamics can be estimated from V_Ls. In an ideal case of the turn-on cycle, before i_D starts to rise, the V_Ls value would be always 0 V; at the moment when i_D begins to rise, it should have a constant value proportional to the slew rate of i_D. After the turn-on cycle, the value of V_Ls should be 0 V. A feedback loop based on V_Ls makes the AGD independent from the drain current values during the switching cycle because the drain current dynamic is defined by the changes in the V_Ls value from 0 V to positive and negative values. Furthermore, V_Ls peak values are not dependent on the maximal drain current values, but strictly on the di_D/dt ratio (1) which is only reliant on the fixed values, thus making it an adequate signal to sense the individual stages of the turn-on cycle. Summarizing, the depicted closed-loop method simplifies the whole structure of the AGD and does not affect the power circuit of the SiC MOSFET.

$$V_{\mathrm{Ls}}=L_{\mathrm{s}}\frac{d{i}_D\left(t\right)}{dt}$$

(1)

Gate Voltage Manipulation for the Turn-On Cycle Improvment

The measurement of the V_Ls voltage can be used as a feedback loop for the method presented in Figure 2. The multi-level gate voltage (V_GG—see Figure 1) method’s purpose is to decrease power losses, drain current overshoot (I_rr) and oscillations of the drain current at the turn-on cycle of a SiC MOSFET by changing the V_GG voltage from 15 V to higher at times specified by the regulation system. For the chosen device (CAS300M17BM2 [19]), the V_GG value recommended by the manufacturers of SiC MOSFETs is 15 V; thus, such level was chosen. The second level of V_GG may be selected between 15 and 30 V—the peak gate voltage allowed for this module; in this case, 30 V was chosen to obtain the maximum increase in the switching speed without damaging the gate oxide layer. Similar to other AGD methods for SiC MOSFET switching improvement, this approach is based on gate current manipulation of the device at the right moments of the switching transient. Thus, in that case, the aim is to improve the turn-on cycle without an unnecessary increase in di_D/dt that would cause overshoot and oscillations of the drain current. What distinguishes this method from the many others is that the R_G value is constant. Usually, the gate resistor R_G is placed as near as possible to the gate terminal of the SiC MOSFET to make the i_G path as short as possible. The sequence of the proposed method appears as follows: at t₀, the driving signal (DS) is changing from the low state to the high state which begins the turn-on cycle of the SiC MOSFET. The V_GG voltage at the same time is changing from −5 (low stage V_GG level) to 30 V. The purpose of the first boost of V_GG to the value of 30 V is to reduce the time delay between DS and v_GS reaching the threshold voltage V_Th. At t₀, the gate-source voltage v_GS starts to rise from zero. After the v_GS voltage crosses the threshold voltage V_Th at t₁, the drain current i_D starts to rise. The V_Ls voltage at t₁ will start to rise from 0 V to the value proportional to di_D/dt and L_S; thus, the beginning of the rise of the current can be sensed. In order to decrease the di_D/dt ratio during the interval between t₁ and t₂, V_GG is reduced to 15 V at time t₁.

This gate current manipulation stage target is to not cause any serious oscillations of the drain current in the main power circuit. Then, the gate-source voltage is latched from t₂ to t₃ at Miller’s plateau. t₂ is the moment when the drain-source voltage v_DS is starting to decrease and when the i_D has reached its maximum value, which is sensed based on V_Ls changes from positive to negative values as the drain current overshoot occurs. The V_GG voltage is switched again to 30 V, which increases the dv_DS/dt ratio and reduces the period between t₂ and t₃ which in fact decreases the turn-on power losses in the device and its switching period. Notice that t₂ occurs when the drain current has reached its maximum value. In order to suppress further negative effects of i_D overshoot, the gate current is increased through modification of the value of V_GG to 30 V. To prevent damaging the gate of the device, V_GG is again set to 15 V at t₃. Finally, at t₄, the gate-source voltage v_GS reaches 15 V, equal to V_GG. The feedback loop of the studied device is based on sensing the two from moments in the turn-on cycle occurring at t₁ and t₂, based on the V_Ls measurements and thus responding with changes in the V_GG values appropriately, as depicted in Figure 2.

The multi-level gate voltage method impact can be theoretically explained using Equations (2)–(6) based on fundamental knowledge [20]. From Equation (2), it can be concluded that the higher the voltage V_GG and the lower the resistance R_G, the greater the gate current i_G. However, the maximum V_GG should be taken into consideration.

$$i_{\mathrm{G}}\left(t\right)=\frac{V_{\mathrm{GG}}-v_{\mathrm{GS}} \left(t\right)}{R_{\mathrm{G}}}$$

(2)

In the proposed method, R_G is fixed so the gate current is only dependent on the difference between V_GG and v_GS. The instantaneous gate-source voltage v_GS in the switching transient (3) also mainly depends on the V_GG, C_GD and C_GS values. In the proposed method, the V_GG voltage is set to 30 V from t₀ to t₁ to minimize the delay time between the DS signal and the t₁ moment where the v_GS crosses the threshold value and the drain current starts to rise.

$$v_{\mathrm{GS}}\left(t\right)=V_{\mathrm{GG}}\left(1-e^{\frac{t-t_0}{\tau }}\right),\hspace{1em}\tau =R_{\mathrm{G}}\left(C_{\mathrm{GS}}+C_{\mathrm{GD}}\right)$$

(3)

Delay time between t₀ when v_GS starts to rise and t₁ when the MOSFET starts to conduct the drain current that is mainly dependent on the V_GG voltage and constant values of the device, as shown in Equation (4). To compensate the delay between the rising edge of the DS signal at t₀ and t₁ in Δt₁₀, V_GG is set to 30 V.

$$\Delta t_{10}=t_1-t_0=-\tau Ln\left(1-\frac{V_{\mathrm{Th}}}{V_{\mathrm{GG}}}\right)$$

(4)

Another Equation (5) represents the rise time of the drain current from time t₁ to t₂, where I₀ is the drain current in the steady state after the turn-on cycle (see Figure 2), g_m is the transconductance and V_Th is the threshold voltage at the turn-on cycle. A low value of Δt₂₁ leads to a high di_D/dt ratio and, in consequence, current overshoot caused by the Schottky body diode (SBD) and the parasitic elements of the device. Thus, it is the reason for the V_GG voltage depletion in t₁ from 30 to 15 V.

$$\Delta t_{21}=t_2-t_1=\tau Ln\frac{g_{\mathrm{m}}{V}_{\mathrm{GG}}}{g_{\mathrm{m}}\left(V_{\mathrm{GG}}-V_{\mathrm{Th}}\right)-I_0}$$

(5)

The time of the drain-source voltage falling in the turn-on cycle can be presented by Equation (6), where V_DD is the drain-source voltage in the time before t₀ and r_DS(ON) is the drain-source resistance of the SiC MOSFET. The value of the Δt₃₂ time should be as low as possible in order to increase the dv_DS/dt ratio, which will finally decrease the turn-on energy of the MOSFET. Small values of Δt₃₂ do not have any undesirable effects on the switching transient. Therefore, equations show that V_GG manipulation based on precise moments in the turn-on cycle can decrease Δt₃₂, which in consequence reduces switching losses. Moreover, the proposed method does not generate drain current overshoot and provides the possibility to increase the switching frequency of the SiC MOSFET only by replacing the gate driver in any SiC MOSFET with a Kelvin source. The SiC MOSFET with an AGD using the multi-level gate voltage technique and feedback loop based on V_Ls can be switched at same frequency as in normal conditions. To provide an adequate feedback signal, the AGD response time must be shorter than the whole time of the turn-on cycle of the device; thus, the operating frequency of the power device is not constrained. Usually, the maximum switching frequency of an MV SiC MOSFET power module is limited by the thermal dependency of the device on power losses. In the case of a transistor in power modules, the switching frequency is much lower (100–500 times) than the sum frequency of the turn-on and turn-off cycles.

$$\Delta t_{32}=t_3-t_2=R_{\mathrm{G}}{C}_{\mathrm{GD}}\frac{\left(V_{\mathrm{DD}}-I_0{r}_{\mathrm{DS}\left(\mathrm{ON}\right)}\right)}{\left(V_{\mathrm{GG}}-V_{\mathrm{Th}}\right)}$$

(6)

3. Simulation Study

The proposed AGD circuit in the DPT configuration is presented in Figure 3. The chosen device is a half-bridge module (CAS300M17BM2), so the AGD is controlling only transistor T₂ (see Figure 3) during the DPT. T₂’s gate current i_G is actively controlled by the AGD with the previously described method and T₁’s gate is connected constantly to −5 V (in reference to its source). In this particular case, L_S is equal to 15 nH based on the datasheet.

Simulation verification of the proposed AGD multi-level gate voltage method based on the parasitic inductance voltage measurement was conducted in Saber RD software. The CAS300M17BM2 module simulation model was provided by the producer of the device. The first steps of the research were to observe the waveforms of V_Ls with CGD and, basing on this, conclude if the method can be used in all SiC MOSFETs. The purpose of the simulation study was to verify the assumptions about the decrease in the turn-on time and energy in the chosen method. Finally, the main aim of the simulation was to measure power losses of the CGD at the turn-on cycle and compare them with the AGD. In simulations, the SiC MOSFET half-bridge module was connected in the DPT setup (Figure 3). Transistor T₁ was turned off and T₂ was the DUT. The gate resistance R_G of both the AGD and CGD was set to a fixed value of 20 Ω. The main assumption of the similar value of the R_G in tests of the AGD and CGD was to obtain the same di_D/dt ratio at the turn-on of the device for both cases. T₁ was connected in parallel to the inductor L = 110 µH. The CGD gate voltage V_GG was fixed at a value of 15 V at the turn-on (according to datasheet recommendations). The AGD gate voltage V_GG (Figure 2) had a variable value of either 15 or 30 V. The AGD simulation was conducted using the feedback signal from V_Ls, which was connected to the Saber RD logical and electrical elements to provide the active control of the V_GG voltage. Delays of the components were not considered in that stage of the research. The DPT was conducted for four operating conditions at the turn-on cycle for the V_DD voltages of 700 or 900 V and I₀ currents of 170 or 270 A. However, the observations of the response of the AGD on the increasing I₀ current had smaller step of change to verify that this solution is independent from the variation in the drain current. The turn-on cycle using the CGD for V_DD = 700 V, I₀ =270 A is shown in Figure 4. In comparison, the turn-on cycle with the AGD for the very same conditions is shown in Figure 5, where the difference in the voltage falling time is visible. The vds/dt ratio increased without any serious oscillations of i_D, thus shortening the turn-on cycle of the device. Notice that in Figure 4 and Figure 5, the i_D rising time is marked, and in both simulated cases, the di_D/dt ratio was the same (1 kA/µs) due to identical R_G values and V_GG levels in Δt₂₁ (see Figure 2). Additionally, the simulations show that the AGD compensates the delay between the rising of the v_GS voltage and the moment when the device starts to conduct the drain current.

Table 1. The turn-on energy losses of CGD and AGD comparison from the simulation study.

VDD	I0	Eon CGD	Eon AGD
700 V	170 A	24.0 mJ	19.4 mJ
700 V	270 A	46.2 mJ	36.5 mJ
900 V	170 A	32.4 mJ	27.4 mJ
900 V	270 A	65.5 mJ	45.9 mJ

presents the results of the previously mentioned simulation. The turn-on energy mostly decreased for V_DD = 900 V, I₀ = 270 A. The turn-on energy settled at 65.5 mJ for the CGD, whereas for the AGD, it reached just 45.9 mJ, resulting in a difference of roughly 30%. The smallest percentage disparity was measured for V_DD = 900 V, I₀ = 170 A, and it was equal to 15%.

Simulations of the AGD in Saber RD software confirmed that the multi-level V_GG method is able to decrease the turn-on energy of the transistor from the CAS300M17BM2 SiC MOSFET module, and there is the possibility to use the feedback signal from the V_Ls voltage measurement, thus confirming the presumptions.

4. Experimental Verification

After the simulation study, the proposed AGD was verified by means of experiments—the simplified main electrical circuit of the proposed AGD for the MV SiC MOSFET is presented in Figure 6. After the simulation research and datasheet-based calculations, it was concluded that the V_Ls voltage will be in the operational range of the integrated electronical circuits (0/5 V) after adequate signal conditioning. The AGD feedback loop is provided by the connection of the filtering circuit of the AGD to the power source S of the DUT. All depicted voltage levels in Figure 6 are referenced to the potential of the Kelvin source S_Kelvin. During the short switching time of the chosen SiC MOSFET with high-level power turn-on conditions, V_Ls is filtered by ferrite beads (FBD) from high-frequency noises to prevent damaging the vulnerable logical gates. The filtered signal is then delivered to the operational amplifiers circuit. In this particular case, two LM-7171 high-speed operational amplifiers were used in the differential configuration. The purpose of this stage is to separate the measuring signal into two independent measuring paths and to add the offset to both of them. Two measurement paths are the reason to have independent signal loops for sensing t₁ and t₂ (see Figure 2).

The outputs of the amplifiers are connected to low-propagation delay comparators, TLV3501 (4.5 ns), with the reference voltages V_ref1 and V_ref2. Values of V_ref1 and V_ref2 were obtained based on the datasheet of the device (switching speeds and L_S = 15 nH). Note that these parameters may differ and the driver applied for different modules may need additional tuning. For the considered CAS300M17BM2, approximate maximum and minimum values of V_Ls should be in the range ±7.8 V for R_G = 20 Ω. However, this value is too high to be provided for the TLV3501 comparators (0/5 V). Operational amplifiers suppressed this voltage and added offset, meaning that their output signals have maximum values of 5 V and minimum values of 0 V. It means that in steady state, where the device is not switching, the outputs of the amplifiers are 2.5 V. Thus, V_ref1 has a value of 2.5 V + 10% of 2.5 V, which is 2.75 V, and its role is to identify when i_D starts to rise during the turn-on cycle. The 10% tolerance secures it from any inappropriate disturbances that could cause switching of the comparator due to the parasitic in the device. On other hand, the V_ref2 role is to sense moment t₂ (see Figure 2). The value of this reference voltage was set to 4 V after approximated calculations of the expected V_Ls value. t₂ is sensed based on the change in the operational amplifier output from a value above 2.5 to below 2.5 V, which corresponds to the V_Ls change from positive to negative voltages. Comparators are responsible for sensing the t₁ and t₂ time moments (see Figure 2) of the V_GG voltage modification. Thus, one of the comparators detects the drain current rising moment t₁ which corresponds to V_GG depletion from 30 to 15 V_. The role of the second comparator is to sense when i_D starts to fall, which is caused by the overshoot of current i_D at t₂ and it is responsible for the V_GG increase from 15 to 30 V. Since voltage V_Ls represents the dynamics of the drain current V_ref1 and V_ref2 can be set to persistent values, the comparator answering to sense the t₂ role is not to define the exact value of V_Ls, but rather to sense the moment when di_D/dt begins to reach a negative value (see Figure 2). Subsequently, the signal is provided to the logical gate circuit and then connected to an X4340 high-speed driver. The logical subcircuit is based on high-speed logical gates, which are responsible for detecting the rising edge of the comparators’ outputs and for protecting the V_GG switching sequence. The purpose of that is so that V_GG would not set inappropriate values at unsuitable moments of the turn-on cycle. It is shown as a block to make the diagram more simplified and clearer. X4340 handles switching the voltage from 15 to 30 V dependently on the logical signal on its input with a low propagation delay. Voltage levels 15/−5 V are supplied by isolated DC/DC converters, MGJD241505. The supply voltage of 30 V is provided from the boost DC/DC converter TP61175. The last part of the circuit is the IXDN630 current amplifier whose role is to turn on or turn off the device according to the value of the DS signal. The gate resistance R_G of the AGD was constant and equal to 20 Ω. Summary propagation delays of the used components were calculated from the data provided from the manufacturers. The response times of the AGD were under consideration during the design stage of the research, in correspondence with the turn-on time of the studied SiC MOSFET. Even with the low switching period of the chosen device components, the AGD operated properly.

A photo of the designed AGD with the detailed marks of its visible components is shown in Figure 7. The AGD was optically isolated from the logical signal DS using HFBR-2531 optical receivers. As it can be noticed, the AGD consists of two connected PCBs (Printed Circuit Boards). The role of the first two-layer PCB (blue) is to provide a power supply to all integrated circuits of the device, switch the voltage from 15 to 30 V and amplify the logical signal DS. The second subcircuit of the AGD is a four-layer PCB (black), and it is directly connected to the power source of the T₂ MOSFET. Filtering, conditioning and logical circuits were placed onto this board of the AGD. The main purpose of that was to suppress any disturbances in the V_Ls measurements.

The experimental setup with a half-bridge module (CAS300M17BM2) and a suitable inductor (L = 110 µH) connected in the DPT circuit is presented in Figure 8. Both the CGD and AGD were controlled by a microcontroller (uC) unit. Measurements were made using a Tektronix MSO56 scope with differential medium-voltage probes (P5205A) with a measurement tolerance of ±2% and a Rogowski coil (CWTMini HF3B) with a tolerance of ±2%. Delay times of the probes were calculated from the datasheets and were further included in the measurements to provide correct results. The AGD and CGD were compared in the same experimental conditions. Thus, the tolerances of the used probes can be omitted when both devices are compared. The drain current i_D, drain-source voltage v_DS and positive bus of the V_GG voltage were measured. A 2 kV DC power supply was provided by a Magna Power TSD 2000-22.5/380 with the output connected in parallel to the energy storage (C = 150 µF). Next, Figure 9 presents the photo of the experimental setup in the DPT configuration during experiments.

The first test was conducted with the CGD—Figure 10 (V_DD = 700 V, I₀ = 270 A) presents the turn-on cycle for a standard gate voltage of V_GG = 15 V. After the drain current overshoot, continuous, weakly damped oscillations can be observed, and they are continued after the end of the turn-on process where the impact on the drain-source voltage is also visible. The voltage fall time is slightly above 300 ns. Comparing it to the same turn-on cycle from Figure 11, the impact of the AGD on i_D and v_DS can be noticed. As expected, the AGD increased the speed of the v_DS voltage decrease (<200 ns) and shortened the cycle without causing any negative effects on the transient. Individual influences of the AGD on the di_D/dt and dv_DS/dt ratios and drain current oscillations are marked on the waveforms presented in the following Figure 12 (V_DD = 900 V, I₀ = 270 A). The shown voltage V_GG+ is the positive rail of the gate driver and, as it can be seen, the switching from a value of 15 to 30 V was obtained in proper moments of the cycle. The first decrease in the V_GG voltage was delayed on purpose to increase the di_D/dt ratio which occurs in the turn-on energy depletion without i_D overshoot in crucial moments. The AGD had the most visible impact on the dv_DS/dt ratio, decreasing ipso facto the falling time of v_DS. Moreover, the AGD delay between the ideal point of the switching V_GG+ voltage can be noticed in Figure 12. In addition, the response time of the AGD can be noticed, which is visible as a result of the summary delays of used electronical components and filtering subcircuits. Comparing both obtained waveforms for the AGD, the damping of the drain current oscillations is greater in the case with V_DD = 900 V, contrary to the case with V_DD = 700 V. However, in both shown AGD cases, the impact on the drain current oscillations can be seen.

In summary, the experimental measurements of the turn-on energy E_on of the AGD and CGD in the same switching conditions (V_DD, I₀ and probes’ tolerances) are shown in

Table 2. The turn-on energy losses of AGD and CGD comparisons from the experimental test.

VDD	I0	Eon CGD	Eon AGD
700 (±14) V	170 (±3.4) A	23.5 (±0.94) mJ	21.5 (±0.86) mJ
700 (±14) V	270 (±5.4) A	37.9 (±1.52) mJ	36.0 (±1.44) mJ
900 (±18) V	170 (±3.4) A	32.3 (±1.29) mJ	29.0 (±1.16) mJ
900 (±18) V	270 (±5.4) A	53.9 (±2.16) mJ	50.0 (±2.0) mJ

. The turn-on energy of the CAS300M17BM2 MV SiC MOSFET using the AGD was decreased in all of the measured conditions compared to the CGD. The turn-on energy was decreased even by 10.1%, with an average of 7% for all measured turn-on energies. Comparisons of the obtained results from experimental and simulation studies are presented in Figure 13.

5. Conclusions

In this paper, an AGD for a 1.7 kV/325A SiC MOSFET power module was proposed and validated via simulations and experiments. A mix of multi-level gate voltage and proper measurements of the voltage across the parasitic source inductance proved a positive impact on the turn-on process of the transistor. The AGD was able to deplete the turn-on energy of the power SiC MOSFET by 7% on average and shorten the turn-on cycle without interference in the main circuit. Furthermore, experimental validation showed a higher dvds/dt rate and, subsequently, measurements confirmed that the chosen method has no negative effect on the drain current oscillations and even suppresses them. Moreover, the device shown in this paper was designed based on off-the-shelf integrated circuits, which makes it a low-cost solution. Additionally, due to use of an integrated boost converter, the AGD needs only one level of the positive supply voltage (15 V) which leads it to be less complex in terms of structure. Finally, the proposed driver can also be used in other models of MV SiC MOSFETs where the Kelvin source and the power source of the device are provided. Further research in the case of turn-off cycle improvement with the shown methods is under consideration.