Extraction of Junction Temperature of SiC MOSFET Module Based on Turn-On dI_DS/dt

Abstract

In this paper, a method of extracting the junction temperature based on the turn-on current switching rate (dI_DS/dt) of silicon carbide (SiC) metal-oxide semiconductor field effect transistors (MOSFETs) is proposed. The temperature dependence of dI_DS/dt is analyzed theoretically, and experimentally to show that dI_DS/dt increases with the rising junction temperature. In addition, other factors affecting dI_DS/dt are also discussed by using the fundamental device physics equations and experiments. The result shows that the increase of the DC-link voltage V_DC, the external gate resistance R_G-ext, and the decrease of the driving voltage V_GG can increase the temperature sensitivity of the dI_DS/dt. A PCB (printed circuit board) Rogowski coil measuring circuit based on the fact that the SiC MOSFET chip temperature and dI_DS/dt is estimated in a linear way is designed to obtain the junction temperature. The experimental results demonstrate that the proposed junction temperature extracting is effective.

1. Introduction

In recent years, power semiconductor devices such as the insulated gate bipolar transistor (IGBT) and the metal-oxide semiconductor field effect transistors (MOSFETs) have been widely applied in many industrial fields especially fields with high reliability requirements in new energy, wind power generation, and automotive and aerospace industries [1]. Therefore, the reliability issues have become the focus of increasing concern. In particular, the novel wide band gap (WBG) devices, which are high electron mobility transistors such as silicon carbide (SiC) power MOSFET or gallium nitride (GaN), can significantly improve power density and conversion efficiency. However, due to the lack of related technology and the instability of the gate oxide layer, the WBG devices require significant condition monitoring to ensure its reliability [2,3,4]. Studies have shown that 31% of failures in power electronic converters can be attributed to the failure of power devices while 60% of device failures are due to thermal stress. In fact, due to the different thermal expansion coefficients of several materials in power semiconductor modules, the accumulation of thermal stress will eventually lead to aging failure. In an actual operating condition, the higher the junction temperature is, the smaller the safety margin of the operation is. Additionally, the greater the junction temperature fluctuations are, the shorter the life of the thermal cycle is. Therefore, the measurement or estimation of junction temperature is essential for its condition monitoring [5,6]. Obtaining this temperature in real time during the operation of the converter can effectively improve the status monitoring system of the SiC MOSFET. In addition, using the junction temperature-based control algorithm can enhance the lifetime of the module [7].

The current methods for obtaining the power semiconductor junction temperature are mainly through a direct measurement method and an indirect measurement method. Direct measurement methods are based on optical and physical contact methods that use infrared cameras (IR) and thermocouples to measure the junction temperature, but they need to destroy the module package structure [8,9]. This method uses an internal implanted thermistor sensor to obtain temperature, but it is limited by the accuracy and bandwidth of the sensor and the dynamic response is slow (a few milliseconds are required). This means the methods based on optical and physical contact are limited in practical industrial applications. Indirect measurement methods mainly include the RC thermal impedance model [10,11] and the thermo-sensitive electrical parameter (TSEP). The RC thermal resistance network model mainly includes the Foster and Cauer models, but they do not consider changes in the thermal impedance caused by aging, which results in the inability to obtain accurate junction temperatures. Based on the TSEP, the temperature estimation of the standard package module can be performed non-invasively without modifying the module itself and only the electrical parameters of the device need to be obtained. Many studies have found that the TSEP method is the most promising solution for monitoring the junction temperature of the module [12]. The high-precision and fast-response temperature measurement of the power device can be performed, and the junction temperature can be measured online or offline using TSEP. This is a very extensive research field. Reference [13] uses the on-resistance R_on to obtain the junction temperature of the IGBT. However, due to the low on-resistance of the SiC MOSFET (several tens of mΩ), a very precise measurement circuit is required and the switching frequency of the SiC MOSFET is high. It is difficult to measure the junction temperature accurately. In Reference [14], the IGBT junction temperature was measured using the Miller plateau voltage (V_GP). However, due to the high switching speed of the SiC MOSFET, the turn-on transient is only few hundred nanoseconds, which means the length of the Miller platform is very short and is difficult to measure. Reference [15] provides a method for measuring the junction temperature using the threshold voltage (V_TH) of the IGBT. Physically, V_TH is the minimum gate bias inducing an inversion layer of free electrons underneath the gate oxide. It creates a conductive channel between the drain and source in MOSFET or the collector and emitter in IGBT. Since this voltage is not well defined, a quasi-threshold voltage (gate voltage of drain current reaching 10% of rated current) can only be obtained by detecting the value of the drain current. Reference [16] uses the turn-off delay of IGBT to measure the junction temperature, but SiC MOSFET is a unipolar switching device and does not include any extracting minority carriers during the turned off transient. Therefore, the junction temperature cannot be measured using the temperature sensitivity of the minority carrier during the turn off transient. Reference [17] proposed a method to obtain the junction temperature based on the peak value of the gate current I_{G peak} during the SiC MOSFET turn-on transient. Based on the positive temperature coefficient of the internal gate resistance (R_G-in), SiC MOSFET can generate different gate currents at different temperatures. The junction temperature is extracted by obtaining the peak value of the gate current.

Reference [18] includes an analysis of the relationship between dI_DS/dt and the junction temperature during the SiC MOSFET turned on and off transient. However, it lacks the experimental research on the large current module and does not design a measurement circuit to define the junction temperature. In summary, compared to silicon semiconductors, there are few studies on the junction temperature of silicon carbide semiconductors. Due to the wide bandgap characteristics of SiC MOSFET, some methods of extracting junction temperature can be used on silicon IGBT but cannot be applied to SiC MOSFET. In this paper, based on the switching characteristics of SiC MOSFET, the junction temperature is obtained by the turn-on current switching rate (dI_DS/dt). Firstly, it is analyzed that the temperature sensitivity of dI_DS/dt is mainly related to the negative temperature coefficient of the threshold voltage (V_TH). Second, an experimental platform was established to demonstrate the good linear relationship between dI_DS/dt and the junction temperature. The effects of gate drive resistance R_G, the supply voltage V_DC, load current I_DS, and the drive voltage V_GG on dI_DS/dt have been experimentally verified. Lastly, a small and economical Rogowski coil measuring circuit for obtaining the dI_DS/dt is proposed. The circuit can be embedded in the SiC MOSFET driver. The feasibility of the circuit to obtain the junction temperature is verified by experiments. In practical engineering applications, considering the switching loss and frequency, an intelligent driver method is proposed to measure the junction temperature of the SiC MOSFET without increasing the switching loss.

2. Theoretical Analysis of Turn-On dI_DS/dt Temperature Dependence

Figure 1a shows the cell structure of the planar SiC MOSFET. It can be seen that the cell of the SiC MOSFET is mainly composed of the gate, source, drain, oxide layer, N-drift region, N-base region, and Junction Field-Effect Transistor (JEFT) region. The figure also shows the route of the electron carrier flow, the position of inversion layer channel, and the distribution of parasitic capacitance.

Figure 1b shows the typical SiC MOSFET turn-on transient waveforms under an inductive load. In order to facilitate the theoretical analysis, the parasitic inductance of the switching loop is neglected [19]. This switching process can be divided into four phases to analyze further. Before t₀, SiC MOSFET is blocked, the driving voltage V_GG = U_off, the drain-source voltage V_DS = V_DC (DC-link voltage), and the gate current I_G is zero.

In Phase 1 (t ∈ [t₀,t₁]), the turn-on pulse trigger at t₀, the driving voltage V_GG = U_on, and the gate current I_G first performs a step to its maximum value and then starts to decay. At the same time, the gate voltage V_GS begins to increase. Due to the forward bias of V_GS, holes in the P base region are first squeezed out to form a depletion layer. With the increase of V_GS, the electrons (minority carriers) in the P base region begin to gather under the gate oxide layer to form an inversion layer channel when the energy band is bent to the surface potential, which is equal to the body potential. V_GS rises to the threshold voltage V_TH when the inversion layer channel is formed at t₁. t_TH is the turn-on delay time. When initiating the turn-on until V_GS reaches V_TH, it can be calculated by using Equation (1) where R_G is the gate total resistance and C_ISS is the input capacitance. The expression for V_TH can be obtained from reference [18] where q is the charge constant, V_FB is the flat-band voltage (related to the oxide layer and semiconductor interface charge and gate material), ξ_sic is the dielectric constant of the semiconductor, and N_A is the doping concentration. C_OX is the oxide layer capacitance, T is the absolute temperature (thermodynamic temperature), the Fermi potential Ψ_B is the potential difference between the mid-gap and the Fermi level far from the surface, and К is the Boltzmann constant.

$$t_{TH}=t_1-t_0=R_G{C}_{ISS}\ln (\frac{V_{GG}}{V_{GG}-V_{TH}})$$

(1)

$$V_{TH}=V_{FB}+2{\psi }_B+\frac{\sqrt{4{\xi }_{sic}KT·N_A\ln (N_A/n_i)}}{C_{OX}}$$

(2)

$${\psi }_B=\frac{KT}{q}\ln (\frac{N_A}{n_i})$$

(3)

Using Equation (3), it can be seen that Ψ_B is related to the doping concentration and temperature of the semiconductor and n_i is the concentration of the carrier. If the start time of turn-on is 0, the changing rate of the gate voltage can be calculated by using the equation below.

$$\frac{d{V}_{GS}}{dt}=\frac{V_{GG}}{R_G\cdot C_{ISS}}{e}^{-\frac{t}{R_G\cdot C_{ISS}}}$$

(4)

Phase 2 (t ∈ [t₁,t₂)) is the inversion layer channel established at time t₁. The electron carriers start to enter in the JFET region and the drain-source current I_DS starts to increase. When I_DS reaches the rated current, the freewheeling diode starts to reverse recovery and the I_DS reaches its maximum value at time t₂. Due to the influence of the parasitic inductance between the drain and source, the drain-source voltage V_DS drops by a small step, but since V_DS >> (V_GG−V_TH), SiC MOSFET is in the saturation region, which means the expression of the drain current I_DS can be calculated by using the following equation:

$$I_{DS}=\frac{\beta }{2}{(V_{GS}(t)-V_{TH})}^2$$

(5)

$$\beta =\frac{W_{CH}{\mu }_{CH}·C_{OX}}{L}$$

(6)

In the above equations, β is the gain coefficient, W_CH is the channel width of the inversion layer, μ_CH is the effective mobility of the channel electrons, C_OX is the capacitance of the oxide layer, and L is the channel length. The current switching rate dI_DS/dt can be calculated together with Equation (4) and the derivative of Equation (5), which is shown below.

$$\frac{d{I}_{DS}}{dt}=\beta (V_{GS}-V_{TH})\frac{V_{GG}}{R_G\cdot C_{ISS}}{e}^{-\frac{t}{R_G\cdot C_{ISS}}}$$

(7)

In Phase 3 (t ∈ [t₂,t₃)), when the drain-source current I_DS reaches the maximum value at time t₂, the gate voltage V_GS rises to the Miller plateau voltage V_GP, V_GS remains basically unchanged, and I_DS starts to fall to the rated current. After the reverse recovery of the freewheeling diode is complete, the gate current I_G charges to the Miller capacitance reversely. At the same time, the drain-source voltage V_DS begins to drop and drops to a very low on-state voltage at t₃. The Miller plateau stage ends.

In Phase 4 (t ∈ [t₃,t₄)), the gate capacitance C_GS is charged by the gate current I_G after t₃. V_GS starts to increase exponentially and rises to the driving voltage V_GG at t₄. In addition, I_G decreases to zero. At the end of this phase, the SiC MOSFET is fully turned on.

From the SiC MOSFET turn-on process, it can be seen that temperature-related factors include the threshold voltage V_TH and the drain current change rate dI_DS/dt. When Equations (2) and (3) are combined, the temperature is derived to obtain Equation (8). From Equation (2), the threshold voltage V_TH is a function of temperature. As the temperature is raised, the band bending (2Ψ_B) required to induce an inversion layer decreases due to the rapid increase in the intrinsic carrier concentration (n_i) in Equation (3). This is partially offset by the temperature pre-factor (KT) and the increased ionization of acceptors N_A when the temperature is raised. However, as the temperature rises, the Fermi level moves closer to the mid-gap. Therefore, the band bending (2Ψ_B) needed to reach the threshold decreases. This reduces the threshold voltage, which is shown in Equation (2) [20]. Therefore, the threshold voltage decreases with an increasing temperature and shows a negative temperature coefficient.

$$\frac{d{V}_{TH}}{dT}=\frac{d{\psi }_B}{dT}(2+\frac{2}{C_{OX}}\sqrt{\frac{{\xi }_{sic}q{N}_A}{{\psi }_B}})$$

(8)

The derivation of the temperature is obtained by using Equation (7).

$$\frac{d^2{I}_{DS}}{dt\cdot dT}=\frac{V_{GG}}{R_G\cdot C_{ISS}}{e}^{-\frac{t}{R_G\cdot C_{ISS}}}(\frac{d\beta }{dT}(V_{GS}-V_{TH})-\beta \frac{d{V}_{TH}}{dT})$$

(9)

From Equation (9), it can be concluded that the temperature coefficient of dI_DS/dt is related to the gate threshold voltage V_TH and the temperature coefficient of the gain coefficient β. Based on Equation (6), the gain coefficient β is proportional to the channel width W_CH, the effective mobility of the channel μ_CH, and the oxide capacitance C_OX, but is inversely proportional to channel length L. W_CH, μ_CH, and C_OX are related to the manufacturer’s process and are constant under the same module, which means β only relates to μ_CH. In the silicon semiconductor, μ_CH is about half of the mobility of the drift region while, in 4H-SiC, μ_CH is only 5–10% of the mobility of the drift region. It is mainly affected by the surface roughness scattering, Coulomb scattering, and phonon scattering of the semiconductor and oxide interfaces [21]. The main effect of μ_CH is Coulomb scattering at room temperature. As the temperature increases, the surface roughness scattering and Coulomb scattering decrease and a large number of phonon scattering increases, which results in a decrease of the carrier mobility. Therefore, μ_CH decreases with an increasing temperature. From Equation (6), β displays a negative temperature coefficient, but due to the wide bandgap characteristics of SiC MOSFET, μ_CH is very low and the temperature sensitivity of β is also low. Therefore, the negative temperature coefficient of V_TH dominates the negative temperature coefficient of β [21]. The temperature coefficient of dI_DS/dt ((dI_DS)²⁄(dt·dT)) is mainly affected by the threshold voltage V_TH.

Reference [22] proposed that the input capacitor C_ISS is also affected by the temperature because C_ISS includes Miller capacitance and gate capacitance C_GS where the gate capacitance C_GS does not change with temperature. The Miller capacitance is composed of the oxide layer capacitance C_OX and depletion layer capacitance C_dep. In general, C_OX does not change with the temperature. The expression of C_dep is shown in the equation below.

$$C_{dep}=A\sqrt{\frac{{\xi }_{sic}q{N}_A{N}_D}{2V_{DS}(N_A+N_D)}}$$

(10)

In Equation (10), A is the effective area of the capacitor area, ξ_sic is the dielectric constant, q is the unit charge, N_D, N_A is the doping concentration of the donor and acceptor, and V_DS is the drain voltage of SiC MOSFET. When the temperature rises, the doping concentration increases, and the Miller capacitance increases with a rise in temperature. It shows a positive temperature coefficient. However, since the rise of the drain-source current I_DS mainly occurs in Phase 2 (t₁, t₂), the drain voltage V_DS is large during this period, which means the depletion capacitance C_dep is small. The Miller capacitance is mainly the oxide capacitance C_OX [20]. Therefore, the temperature effect of the input capacitor C_ISS can be ignored.

In summary, due to the negative temperature coefficient of the threshold voltage (a negative sign before dV_TH/dT in Equation (9)), the turn-on dI_DS/dt shows a positive temperature coefficient and the SiC MOSFET turns on faster at a higher temperature.

3. Experimental Verification of dI_DS/dt Temperature Dependence

3.1. Relationship between dI_DS/dt and Junction Temperature

In order to demonstrate dI_DS/dt temperature dependence, a 1.2 kV, 180 A SiC MOSFET module (BSM180D12P2C101, ROHM Semiconductor, Kyoto, Japan) [23] was used to build a double pulse experimental platform, which is shown in Figure 2. The module has two SiC MOSFETs switches in series including one as a test device and the other as a freewheeling diode (D₁). The module has a parasitic inductance L_S between source S and auxiliary source S’ and it can be seen that L_S is 13 nH from the manual. In the experiment, the DC-link voltage was measured using a high-voltage differential probe (N2891A, Agilent Technologies, Santa Clara, CA, USA) and the drain current was measured using a Rogowski coil (CWTMini1B, PEM, Nottingham, UK). The oscilloscope used for capturing the data was Agilent MSO8064A with a bandwidth of 600 MHz. The SiC MOSFET module is placed on the adjustable heating plate and coated with thermal grease on the substrate. The substrate of the module is closely attached to the heating plate and each temperature is adjusted. The heating process takes a long time in order for the junction temperature (T_j) of the chip can be considered equal to the temperature of the heating plate. Each experiment is performed from a low temperature to a high temperature, which ensures that there is no heat wasted inside the chip. In addition, it improves the measurement accuracy. The experimental circuit is shown in Figure 3.

The DC-link voltage V_DC = 350 V, the load inductance L_LOAD = 200 μH, the external resistor R_G = 60 Ω, and the driving voltage V_GG = 18 V. The SiC MOSFET was tested at 60 °C, 75 °C, 100 °C, 125 °C, and 150 °C for double-pulse measurements. The drain current I_DS waveforms (measured by the Rogowski coil) of several temperatures are shown in Figure 4. It can be seen that the dI_DS/dt of 150 °C is greater than 60 °C. Therefore, the dI_DS/dt increases when the temperature is raised due to the negative temperature sensitivity of the threshold voltage V_TH. In addition, at higher temperatures, the inversion layer forms earlier and the drain-source current I_DS takes the shortest time to reach the rated current. At the same time, the intrinsic carrier concentration increases with a rising temperature, more carriers flow through the channel, and dI_DS/dt becomes larger. Therefore, dI_DS/dt shows a positive temperature coefficient, which is consistent with the above theoretical analysis during the turn on transient. In the experiment, the voltage of the parasitic inductance L_S between the auxiliary source S’ and the source S is measured, which is shown in Figure 5. It can be seen that the peak value of V_S’S increases with an increasing temperature and shows a better linear relationship. The dI_DS/dt can be obtained by Equation (11). The voltage V_S’S is 1.6 V at 75 °C and 1.9 V at 150 °C. Therefore, it has a difference of 300 millivolts. The resolution of V_S’S is 4 mV/°C and the junction temperature can be obtained in real time by using Equation (12) where K₀ is the temperature resolution. This is related to the manufacturer’s production process and product type. $V_{S^{\prime }S}{|}_{T_0}$ is the V_S’S value at temperature T₀.

$$\frac{d{I}_{DS}}{dt}=\frac{V_{s^{\prime }s}}{L_S}$$

(11)

$$T=K_0\left(V_{S^{\prime }S}-V_{S^{\prime }S}{|}_{T_0}\right)+T_0$$

(12)

3.2. Influence of Other Factors on dI_DS/dt

If dI_DS/dt is used as the parameter of TSEP for SiC MOSFET condition monitoring, its relationship with other parameters must be calibrated. The temperature sensitivity of dI_DS/dt needs to be removed from these factors in order to be used for the junction temperature detection. From Equations (7) and (9), it is clearly shown that the factors affecting the dI_DS/dt are gate resistance R_G, the DC-link voltage V_DC, the load current, the gate voltage V_GG, and the junction temperature T_j. The relationship between them is shown in Figure 6. The current change rate dI_DS/dt increases when the junction temperature T_j, the drive voltage V_GG, and the DC-link voltage V_DC increase. However, dI_DS/dt decreases with increasing gate resistance R_G. The load current cannot be determined by using the formula. The following experiments are used to verify their relationship.

3.2.1. Effect of Load Current and DC-link Voltage on dI_DS/dt

The effect of the load current on the temperature sensitivity of dI_DS/dt was analyzed in the double-pulse experiment, and the load current can be adjusted by changing the pulse duration of the turn-on voltage V_GG. The experiments were performed at the DC-link voltage V_DC = 350 V, the driving resistance R_G-ext = 60 Ω, the junction temperature T_j = 150 °C, and a range of load current from 50 A to 160 A. The experimental waveform is shown in Figure 7. It can be seen that dI_DS/dt did not change substantially under the different current at the same temperatures. Figure 8 shows that, at the same current, the higher the temperature is, the greater the dI_DS/dt is and the better linearity is. Therefore, dI_DS/dt is not disturbed by the fluctuation of the load current and it is beneficial to use it as TSEP for the junction temperature extraction of the SiC MOSFET under practical conditions.

In order to verify the effect of the DC-link voltage V_DC on the SiC MOSFET, the temperature of SiC MOSFET is kept constant at 150 °C and the drain-source current I_DS = 80 A at a steady state. Double-pulse experiments were performed at different DC-link voltages (100 V, 120 V, 180 V, 200 V, 250 V, and 380 V), respectively. The turn-on I_DS and V_DS waveforms are shown in Figure 9. It can be seen that, at the same temperature and rated current, there is only a slight increase in dI_DS/dt of the DC-link voltage with a difference of 280 V. The voltage V_S’S induced by parasitic inductance L_S is measured and shown in Figure 10. It can be seen that the peak value of V_S’S is also a slight increase with the increase of V_DC, which means that dI_DS/dt increases slightly with V_DC. As the DC-link voltage increases, more charges are discharged in the N- drift region increasing the depletion layer width. This will also decrease the depletion layer capacitance C_dep. Based on Equation (10), it can also be seen that C_dep decreases with increasing DC-link voltage. Therefore, the Miller capacitance is reduced by increasing V_DC, which, in turn, reduces the input capacitance C_ISS and speeds up the turn-on transient of the SiC MOSFET. However, C_dep is small in this stage (in Phase 2, the drain voltage V_DS is large) and dI_DS/dt shows a slight increase. However, the DC-link voltage V_DC is often constant in voltage source converter applications, which means the influence of V_DC on dI_DS/dt is less important. The dI_DS/dt temperature coefficient is usually calibrated to accommodate different DC-link voltages in a new operating environment. In addition, the slight fluctuation of the DC-link voltage has little effect on the dI_DS/dt in the actual operating conditions, which also facilitates the use of the dI_DS/dt as a TESP parameter for junction temperature extraction.

3.2.2. The Effect of Drive Resistance on dI_DS/dt

As known in silicon MOSFETs and IGBTs, the external drive resistance (R_G-ext) also affects the turn-on dI_DS/dt of SiC MOSFET. In order to observe the effect of the R_G-ext on dI_DS/dt, the heating plates were adjusted to 50 °C, 75 °C, 100 °C, 125 °C, and 150°C, respectively. Each adjustment waits for one hour for the junction temperature of the chip to reach the set temperature. The rated current remains at 150 A and the waveforms of I_DS observed from different R_G-ext are shown in Figure 11. As observed from Figure 11 and Figure 12, the dI_DS/dt reduces as the resistance increases and the switching speed becomes slower. Since the increase of R_G-ext reduces the gate current I_G, the gate voltage V_GS charges slowly to the gate capacitance C_GS and V_GS reaches the threshold voltage for a longer time (introduced from Equation (1)). In addition, the time that V_GS reaches the Miller platform voltage also becomes longer (in Phase 2), which means dI_DS/dt will decrease when R_G-ext increases (introduced in Equation (7)). However, a small R_G-ext increases dI_DS/dt, but the gate parasitic inductance cannot be suppressed in a way that reduces the temperature sensitivity of dI_DS/dt by decreasing the gate voltage that reaches the chip [24]. It can be also seen from Figure 11 and Figure 12 that temperature sensitivity of dI_DS/dt shows better linearity under a large R_G-ext and there is less oscillation in the waveform of I_DS at a large drive resistance, which creates an easy way to obtain dI_DS/dt. In addition, R_G-in is the equivalent of the distribution resistance of the gate polysilicon and metal connection in MOSFET. Reference [25] confirmed that R_G-in shows a positive temperature coefficient. If R_G-ext is large, the temperature effect of R_G-in is suppressed and it can be defaulted as R_G (R_G = R_G-ext + R_G-in). This does not change with increasing temperature, which means dI_DS/dt has better temperature sensitivity under larger R_G-ext. Equation (9) also suggests that the increase of R_G decreases the coefficient in front of dβ/dT, which increases the temperature sensitivity of dI_DS/dt.

In summary, the experimental and theoretical analysis are consistent and dI_DS/dt shows better temperature sensitivity under the larger R_G. Based on the experiment of this module, it is found that R_G-ext shows a better temperature sensitivity when it is around 60 Ω (relative to R_G-in, which is already very large). Since R_G-ext is so large, dI_DS/dt is too small, which makes the measurement difficult.

3.2.3. Driving Voltage V_GG Effect on dI_DS/dt

Combined with the theoretical analysis of the second part, it can be seen from Equation (7) that the reduction of the driving voltage V_GG will reduce the dI_DS/dt. However, from Equation (9), it can be deduced that the decrease of V_GG reduces the coefficient of dβ/dT, which increases the positive temperature coefficient of dI_DS/dt. In order to verify the influence of V_GG on the temperature coefficient of dI_DS/dt, the DC-link voltage V_DC is kept at 350 V, the driving resistance is R_G-ext = 60 Ω, and I_DS at a steady state is 100 A. The double-pulse experiment was performed at different driving voltages of 12 V, 15 V, 18 V, and 20 V and different temperature points of 50 °C, 75 °C, 100 °C, 125 °C, and 150 °C, respectively. Figure 13 shows the waveforms of I_DS with different drive voltages of 12 V and 20 V, respectively. Figure 14 shows the dI_DS/dt values for different temperatures under the four driving voltages. It can be seen from Figure 13 that, as the driving voltage decreases, dI_DS/dt becomes smaller. However, there is a greater range of variation with the junction temperature when the driving voltage V_GG = 12 V. Using the dI_DS/dt measured at 50 °C when V_GG is 12 V as a reference, the value of dI_DS/dt at 150 °C increases by 1.2 times. Similarly, using the dI_DS/dt measured at 50 °C when V_GG = 20 V as a reference, the value of dI_DS/dt at 150 °C increases by 1.12 times. At the same time, the reduction of V_GG reduces the switching speed and the oscillation of I_DS that facilitates the measurement of the dI_DS/dt. Therefore, dI_DS/dt shows better temperature sensitivity at low drive voltages. However, it can be seen from Figure 14 that the dI_DS/dt is too low when the driving voltage is reduced, which makes it difficult to measure. In addition, if V_GG is too low, the module will not be completely turned on and it will work in the active area. Therefore, it is best to select a driving voltage of 12 V based on this module.

4. Design Measuring Junction Temperature Circuit and Experimental Verification

The dI_DS/dt as a real-time junction temperature extraction for SiC MOSFETs can be achieved using the following method. For the SiC MOSFET module with a source auxiliary terminal S, it is possible to measure the peak value of the induced voltage V_S’S generated by the parasitic inductance L_S between the source assist S’ and the source S during the SiC MOSFET turn-on transient. The value of L_S can be found in the device manual. According to the formula V_S’S = L(dI_DS/dt), the value of dI_DS/dt can be obtained. Figure 15 shows the peak value measurement circuit of V_S’S, but the above circuit is only suitable for SiC MOSFET modules with an auxiliary source. In order to target all modules, a measurement circuit based on the printed circuit board (PCB) Rogowski coil [26] was proposed. The measurement circuit is shown in Figure 16a. It consists of two high-precision operational amplifiers (LM7171), a diode, a storage capacitor of 4.7 nF, and a discharge resistor R₁. The first LM7171 is mainly used to acquire the parameters of the dI_DS/dt sensed by the PCB Rogowski coil and is used to isolate the high voltage. The second LM7171 mainly amplifies the resolution according to the actual needs. It facilitates the acquisition of the one-to-one correspondence between dI_DS/dt and the junction temperature. From Figure 17b, it can be seen that the circuit occupies a small space and the cost is low so that it can be embedded in the drive module. The PCB Rogowski coil can be directly mounted on the terminal of the SiC MOSFET module, which offers a new idea for engineering applications.

In order to verify the feasibility of the circuit, the driving voltage is 12 V under the voltage and current level of 350 V and 100 A and the double pulse experiments are carried out at the working junction temperature of SiC MOSFET at 50 °C, 75 °C, 100 °C, 125 °C, and 150 °C, respectively. The magnification of the measurement circuit is 1 (This value can be changed by adjusting the ratio of R₂ to R₃) and the output waveform is shown in Figure 17. It can be seen that each temperature corresponding to the measured voltage shows better linearity. The measured voltage is 0.4 V at 50 °C, and 0.8 V at 150 °C. In addition, the resolution is 4 mV/°C. The resolution can be increased by increasing the amplification of the operational amplifier. The Junction temperature of SiC MOSFET can be obtained by using Equation (13) where T₁ is the calculation temperature, K₁ is the resolution (250 °C/mV), and V₁ is the measuring voltage by this measuring circuit.

$$T_1=K_1(V_1-0.4)+50$$

(13)

To further test the feasibility of extracting the junction temperature of this circuit, the heating plate is adjusted to different temperatures. The measured voltage is obtained by the measuring circuit of Figure 16 and the measured temperature can be calculated by using Equation (13). The result is shown in

Table 1. The measured voltage by the measuring circuit at different temperatures and the temperature calculated by using Equation (13).

Junction Temperature	60 °C	70 °C	110 °C	130 °C	140 °C
Measured Voltage	0.43 V	0.5 V	0.655 V	0.73 V	0.755 V
Calculation Temperature	57.5 °C	75 °C	114 °C	133 °C	139 °C
Deviation	−2.5 °C	5 °C	4 °C	3 °C	−1 °C

. It can be observed that, compared with the actual junction temperature, the maximum deviation of the calculated temperature does not exceed 5 °C. Therefore, this circuit for extracting the junction temperature of SiC MOSFET is feasible. In an actual operation, the junction temperature and the measurement voltage corresponding to the calculated temperature are converted into the digital signals by the A/D module (Analog to Digital Converter) and are stored in the Field－Programmable Gate Array (FPGA) of the driving module. The junction temperature is obtained by looking up the table.

As mentioned above, dI_DS/dt has better temperature sensitivity under a small drive voltage and a large drive resistance, but it also brings about an increase in the switching loss. Since the real work does not need to extract the junction temperature at every moment, an intelligent driver [27] can be used to set a small time period to make the driver open SiC MOSFET under the large resistance and small drive voltage for the junction temperature extraction. After the data is acquired, the original small drive resistance and large gate voltage are restored. Therefore, the accurate junction temperature can be extracted while satisfying fast switching frequency and small switching loss. In summary, it is feasible to extract the junction temperature from SiC MOSFET based on dI_DS/dt.

5. Conclusions

In this paper, the temperature dependence of the turn-on dI_DS/dt of SiC MOSFET is discussed using device mathematical models and experiments. It is shown that dI_DS/dt increases with a rising temperature as a result of the negative temperature sensitivity of the threshold voltage V_TH. Afterward, other factors affecting dI_DS/dt are analyzed. It was found that rising DC-link voltage V_DC values and external gate resistance R_G-ext as well as the decrease of the driving voltage V_GG can increase the temperature sensitivity of dI_DS/dt. However, the load current has no effect on dI_DS/dt. Therefore, there is a good temperature sensitivity of dI_DS/dt under large drive resistance and small drive voltage. Lastly, a small and low-cost PCB-based Rogowski coil measurement circuit was designed. The dI_DS/dt obtained by this circuit exhibits a near linear dependency on temperature with a resolution of 4 mV/°C (the resolution can be increased by adjusting the amplification factor). When the actual application needs junction temperature extraction, the intelligent driver sends a detection signal to turn on the SiC MOSFET under a small drive voltage and a large drive resistance. The value of dI_DS/dt is obtained by measuring the circuit based on the PCB Rogowski coil. Afterward, the operating junction temperature of SiC MOSFET is obtained by the junction temperature database.

Future work will focus on designing intelligent drivers that can be embedded into the junction temperature detection (based on the junction temperature extraction of turn-on dI_DS/dt) to use for SiC MOSFET health monitoring (the dI_DS/dt temperature sensitivity of aging devices is different). This further tracks the aging of devices and provides an estimation of remaining life.