Research on Gate Charge Degradation of Multi-Chip IGBT Modules in Power Supply for Unmanned Aerial Vehicles

Abstract

In recent years, with the burgeoning application of high voltage in various industrial sectors, the deployment of unmanned equipment, such as industrial heavy-load Unmanned Aerial Vehicles (UAVs), incorporating high-capacity Insulated-Gate Bipolar Transistors (IGBTs), has become increasingly prevalent. The demand for high-voltage IGBT modules in UAV is continuously growing; therefore, exploring methods to predict fault precursor parameters of multi-chip IGBT modules is crucial for the operational health management of unmanned equipment like UAVs. This paper analyzes the gate charge degradation in multi-chip IGBT modules after thermal cycling, which can be used to evaluate the operational state of these modules. Furthermore, to delve into the electrical response of a gate drive circuit caused by local damage within the IGBT module, an RLC model incorporating parasitic parameters of the gate drive circuit is established, and a sensitivity analysis of the peak current in the gate charge circuit is provided. Additionally, in the experimental circuit, an open sample of an IGBT module with partial bond wires lifted off is used to simulate actual faults. The analysis and experimental results indicate that the peak current of the gate charge is closely related to L and C. The significant deviation in the gate current, influenced by the partial bond wires lift-off, can provide a basis for the development of predictive methods for IGBT modules.

1. Introduction

Compared to other power devices, the IGBT is more favored in power applications due to its switching characteristics similar to those of Metal Oxide Semiconductor Field Effect Transistors (MOSFETs) and its high current and voltage capabilities akin to Bipolar Junction Transistors (BJTs). While the operational frequency of IGBT is significantly lower in comparison to that of Si MOSFET or GaN FET, its ability to conduct current under high-voltage conditions positions them as pivotal components in various safety-critical domains, including railway traction, power distribution, renewable energy applications, and UAVs [1,2,3,4,5,6,7,8].

The most power-consuming electrical components in UAVs are the power electronics driving the motors. Among these devices, the most commonly used modules are IGBT and Silicon Carbide MOSFET (SiC MOSFET). In contrast, SiC MOSFET modules are relatively more advanced, with transistor switching frequencies that exceed those of IGBT, which signifies their capacity for the precise control of high-speed motors. They can also control high-power motors of hundreds of kilowatts. However, the higher material costs, complex manufacturing processes, and the current limited scale of production contribute to their elevated price points. IGBT modules compensate for this deficiency. The technology of IGBT modules is relatively mature, and they are produced on a large scale. Their design is intended to meet the demands of high-power motors in the multi-megawatt range and to withstand more extensive temperature variations. These attributes are highly compatible with the requirements of UAVs, as shown in Figure 1. To meet the UAV’s high-voltage, high-power, and high-reliability demands for motor controllers (ESC), the design that employs IGBT modules as the actual power switches controlling the current flow to the motor coils is becoming increasingly sophisticated, as shown in Figure 2. In the industrial UAV design proposed by STMicroelectronics, IGBT devices replace the mechanical commutation of three-phase brushless DC motors with electronic switching commutation, achieving precise control over movement along multiple axes. To reduce power transmission losses, Texas Instruments utilizes IGBT devices as power transistors in UAV electronic speed controllers, extending UAV flight endurance. Embention has developed a series of motor controllers for UAV using IGBT modules, preventing motor control performance degradation and ensuring long-term normal operation of UAVs. MGM COMPRO uses the latest IGBT modules to develop high-voltage speed controllers for UAV motors, achieving peak efficiencies of 98–99%.

Juana et al. found that a highly efficient DC/DC converter is needed in the design of power electronic converters for hybrid UAV; thus, they select the IGBT as the transistor for controllable switch regulators in [9]. The DC/DC or AC/DC/AC converter technology based on IGBT devices strikes a good balance between efficiency and power density for small UAV power converters in [10]. In [11], by comparing IGBTs and MOSFETs, it is found that IGBTs have the capability to reduce voltage, making them highly suitable for UAV applications. The battery voltage of the UAV is regulated near the maximum voltage through a combination of a tuned Proportional–Integral–Derivative (PID) controller connected to the IGBT, aiming to prevent overvoltage charging operations. When the UAV engine is running, the IGBT operates in generator mode to power the UAV’s electronic systems, and when the engine stops, the IGBT operates in motor mode as a starter for the UAV engine. The analysis of key circuit parameters for suppressing turn-off voltage spikes ensures the safe operation of the IGBT and improves the stability of the UAV power system in [12]. Gong et al. have innovated the design of UAV motor structures and electronic speed controllers, using IGBTs with sinusoidal wave outputs in the speed controller design, which eliminates the need for a sinusoidal wave controller, reducing motor weight and increasing the power-to-weight ratio of the UAV. Furthermore, the use of IGBT modules can effectively enhance controller efficiency, achieving an integrated design of the motor and controller, significantly reducing the weight and volume of the UAV electric propulsion system. For the issue of partial discharge (PD) in high-voltage IGBTs under harsh conditions such as those faced by aviation UAVs, an effective PD testing method and anti-series connection configuration were proposed, and the insulation reliability of commercial high-voltage IGBT under low-pressure and high-temperature conditions was studied in [5]. In [13], for commercial aviation flight control power modules, task modeling analysis concluded that the use of IGBTs can significantly reduce the overall system power consumption of more electric aircraft. In [14], a 4 MW high power density generator was developed for future hybrid electric aircraft, but this was limited by the IGBT modules available on the market. In [15], two regenerative energy absorption methods based on diode rectifiers for generators and active rectifier-based starter/generator (SG) systems were proposed, using IGBTs as the power electronic converters to improve the efficiency of more electric aircraft and reduce heat generation.

Over the years, in order to analyze certain characteristics of IGBT modules in a targeted manner, many scholars have conducted research by modeling IGBT modules. The paper [16] selects the PSI25/06 IGBT module, manufactured by Powersem, as the subject of investigation, and developed a compact thermal model capable of calculating the internal temperatures of all components within the module, including self-heating phenomena and mutual thermal coupling among components. The paper also introduced an indirect electrical measurement method, utilizing cooling curves to compute the waveform of transient thermal impedance. Ultimately, it confirmed that the proposed compact thermal model and measurement technique could provide a detailed description of the thermal phenomena occurring within the IGBT module. In [17], to investigate the temperature dependence of the IGBT module, finite element simulation was employed to obtain the thermal flow curves along the central axis of the module. Based on these findings, a Cauer-type thermal network model was proposed to simulate the IGBT module. Real-time temperature data were collected using an infrared camera and precise fine-wire thermocouples, verifying the accuracy and robustness of the proposed model in conjunction temperature estimation. The paper [18] constructed a multi-physics coupled model of the IGBT module, incorporating electrical, thermal, and mechanical fields. Utilizing this model, the distribution of physical properties under various conditions was calculated, and the bonding wire failure mechanism was analyzed. The model simulated the impact of bonding wire cracks and detachment faults on the module’s characteristics. The results indicated that heel cracks in the bonding wires can significantly affect the module, accelerating its failure rate.

The reliability requirements for IGBTs in safety-critical applications previously mentioned are significantly more stringent than in standard industrial applications. For instance, IGBT modules used in railway traction applications are expected to have a lifespan of at least 30 years [19]. Consequently, research on the failure mechanisms and reliability degradation of IGBT modules is becoming the focus of research, and more and more research results have been reported in recent years. The main failure mechanisms of IGBT modules used in high-power applications are reviewed, and the main failure modes are provided in [20]. The bond wire fatigue and solder failure were identified as dominant failure [20,21,22,23]. Precursor parameters were highlighted to enable the development of a prognostic approach for IGBT modules [24]. A prognostic and warning system for IGBT modules was developed by tracing the saturation collector–emitter voltage of IGBT modules in [25]. An online ringing characterization was used for the prognostic and health management of IGBT modules in [26]. A real-time compact electronic thermal model was presented for the health monitoring of IGBT modules in [27]. Last, the condition monitoring and fault diagnostics of IGBT modules were reviewed, respectively, in [28,29]. These studies elucidate that the reliability evaluation of IGBT modules is very important, and developing prognostic methods that can anticipate failures and predict the residual life are desired.

Unfortunately, the failure mechanism of IGBT modules is very complex, and the methods mentioned above are useful but somewhat insufficient. For example, the saturation collector–emitter voltage of IGBT modules will be changed not only with reliability degradation but also with something else like junction temperature, etc., i.e., making it difficult to trace. A real-time compact electronic thermal model also has some limitations: for example, the power consumption of IGBT devices cannot be measured directly, etc. The overwhelming majority of physical wear-out mechanisms give little external indication of impending failure, so the essential step behind reliability evaluation is precursor parameters identification and the continued control of those parameters. In general, the majority of existing studies focus primarily on the overall thermal management, electrical parameter variations, or mechanical stress of IGBT modules. In contrast, the present investigation is dedicated to the analysis of gate charge degradation, which is an aspect that has received relatively less attention in the study of IGBT module reliability. Compared to traditional methods that rely on parameters such as the saturated collector–emitter voltage, this paper introduces an online monitoring approach that does not depend on isolated measurements, which will be more convenient and practical for real-world applications. A major advantage of monitoring the gate charge current rather than the gate voltage is that isolation is not required, as seen in [28].

This paper investigates the failure of IGBTs during prolonged high-temperature operation, where internal bond wires lift off due to mechanical expansion, leading to module failure. Through theoretical analysis, it was identified that a significant alteration in the maximum gate charge current occurs when partial bond wires lift-off takes place, suggesting that monitoring the gate charging current is a viable method to ascertain IGBT module malfunctions. The research is structured as follows:

(1)

We elucidate the structural composition of the IGBT, highlighting the potential for bond wire lift-off due to the substantial disparity in thermal expansion coefficients of various materials within the module, which can lead to lift-off after long-term exposure to high temperatures.

(2)

We analyze the role of bond wires within the circuit and deduce that the lift-off of these wires results in changes to the maximum charge current. However, due to technical constraints, these changes are challenging to detect.

(3)

Focusing on the monitoring of peak current changes, we construct an equivalent model of the IGBT module and derive a simplified gate charge circuit equation based on this model. The equation theoretically demonstrates that the peak current is affected by bond wires lift-off.

(4)

To validate the theoretical findings, we simulate bond wires lift-off by manually severing the wires in an unpackaged IGBT module. We record the charge current waveforms post-severing and observe changes in the waveforms that corroborate the theoretical analysis. We also discuss how other conditions do not affect the turn-on waveform of the IGBT module, validating the feasibility of monitoring current deviations.

2. Module Failure Mechanism and Gate Charge Exploration

The IGBT is a hybrid device, which can be seen as the combination of MOSFET and BJT. Its unit cell structure is very similar to that of a Vertical Diffused MOSFET (VDMOSFET), except there is an additional ${\mathrm{P}}^{+}$ layer that injects holes into the N⁻ region to reduce on-resistance, as shown in Figure 3.

Due to the power density of the IGBT silicon chip being restricted by physical reasons, it is necessary to parallel multiple chips to enhance their current-carrying capacity. Consequently, medium- and high-power IGBT modules are constructed with several paralleled chips, encased in a plastic cover, and interconnected through aluminum wire bonds between the main emitter electrode and the IGBT chips [30], as shown in Figure 4.

This architecture is preferred in commercial packages, consisting of several different materials, resulting in considerable complexity, as shown in Figure 5. The constituent elements from top to bottom include the bond wire, the silicon chip, the direct copper bonded ceramic substrate (DCB), and the base plate.

One major drawback of this multiple layered structure is a mismatching of the coefficients of thermal expansion (CTE). As listed in

Table 1. CTE.

Material	CET (ppm/°C)
Al	22
Si	3
Solder	Compliant
Cu	Not relevant
$\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3$ or AlN	7 or 4
Cu	Compliant
Solder	Compliant
Cu or AlSiC	17 or 8

, the CTE values of different materials used in the IGBT module vary greatly, as noted in [19,31]. Based on the parameters of Table 1, there are two weak points inside the IGBT module, one is the interface of Al bond wire and Si chip and the other is the interface of the DCB and base plate. High temperatures or repetitive temperature cycles will induce large thermal mechanical strain on these weak points and eventually lead to the destruction of IGBT modules as described in [32,33], and one physical phenomenon resulting from this kind of thermal degradation is partial bond wires lift-off.

Bond wires act as stray inductors in gate charge circuit, as shown in Figure 6. Partial bond wires lift-off will change the total stray inductance value. When this alteration reaches a certain amount, the total gate-emitter capacitance value will be also changed. and consequently, the maximum charge current will also be changed accordingly. So, monitoring this phenomenon via a gate charge current can provide a viable diagnostic method for anticipating failures in IGBT modules. In comparison with JESD-24-2, measuring the gate charge current can be performed online. Unfortunately, due to the bandwidths of the current sensors, the gate charge current can hardly be detected directly. An alternate practical method is measuring the voltage across the external resistor that exists in the primary side of the isolated DC/DC converter of the gate driver board.

3. Analysis and Experimental Results

In order to detect the peak current of a realistic gate driver circuit in principle, an equivalent electrical model of an IGBT module is introduced. This work can be divided into two parts: one is building the electrical model of the IGBT chip [34,35], and the other is extracting the stray inductances of interconnections inside the IGBT module [36]. Then, the integrated electrical circuit of the IGBT module used in the following experiment is achieved in this paper, as shown in Figure 7, where each element represents a different region of the IGBT.

For the PNP transistor, $C_{CER}$ is the collector–emitter redistribution capacitance, $C_{EBJ}$ and $C_{EBD}$ are the emitter–base junction and diffusion capacitance, and $R_B$ is the conductivity-modulated base resistance. For the VDMOSFET, $C_{GDJ}$ is the gate-drain overlap depletion capacitance, $C_{DSJ}$ is the drain-source overlap depletion capacitance, $C_{OXS}$ is the gate oxide capacitance of the source overlap and $C_M$ is the source metallization capacitance; these are the main components of the gate-source capacitance $C_{GS}$. The last one $C_{OXD}$ is the gate oxide capacitance of the drain overlap, which is the dominant component of gate-drain capacitance $C_{GD}$. The switching characteristics of the IGBT modules are affected by the following parameters, as shown in Formulas (1) to (9) [35].

$$\begin{array}{c}\hfill C_{\mathrm{CER}}={\displaystyle \frac{Q{\epsilon }_{si}}{3qW{N}_B}},\end{array}$$

(1)

$$\begin{array}{cc}\hfill R_B=\left\{\begin{array}{c}\hfill W/\phantom{W\left(q{\mu }_{nc}A{N}_B\right)}\left(q{\mu }_{nc}A{N}_B\right)Q<0\hfill \\ \hfill W/\phantom{W\left(q{\mu }_{eff}A{n}_{eff}\right)Q\ge 0}\left(q{\mu }_{eff}A{n}_{eff}\right)Q\ge 0\hfill \end{array}\right.& \hfill ,\hfill \end{array}$$

(2)

$$\begin{array}{c}\hfill C_{\mathrm{DSJ}}={\displaystyle \frac{\left(A-A_{GD}\right){\epsilon }_{Si}}{\sqrt{2{\epsilon }_{Si}(V_{CE}-V_{GE\left(th\right)})/q/N_B}}},\end{array}$$

(3)

$$\begin{array}{c}\hfill C_{\mathrm{GDJ}}={\displaystyle \frac{A_{GD}{\epsilon }_{si}}{\sqrt{2{\epsilon }_{Si}(V_{CE}-V_{GE\left(th\right)})/q/N_B}}},\end{array}$$

(4)

$$\begin{array}{c}\hfill {\mu }_{nc}=1/\phantom{1\left(1/\phantom{1{\mu }_n+1/\phantom{1{\mu }_c}{\mu }_c}{\mu }_n+1/\phantom{1{\mu }_c}{\mu }_c\right)}\left(1/\phantom{1{\mu }_n+1/\phantom{1{\mu }_c}{\mu }_c}{\mu }_n+1/\phantom{1{\mu }_c}{\mu }_c\right),\end{array}$$

(5)

$$\begin{array}{c}\hfill {\mu }_{pc}=1/\phantom{1\left(1/\phantom{1{\mu }_p+1/\phantom{1{\mu }_c}{\mu }_c}{\mu }_p+1/\phantom{1{\mu }_c}{\mu }_c\right)}\left(1/\phantom{1{\mu }_p+1/\phantom{1{\mu }_c}{\mu }_c}{\mu }_p+1/\phantom{1{\mu }_c}{\mu }_c\right),\end{array}$$

(6)

$$\begin{array}{c}\hfill {\mu }_{eff}={\mu }_{nc}+{\mu }_{pc}Q/\phantom{{\mu }_{pc}Q(Q+qAW{N}_B)}(Q+qAW{N}_B),\end{array}$$

(7)

$$\begin{array}{c}\hfill P_0=Q/\phantom{Q\left(qALtanh{\displaystyle \frac{W}{2L}}\right)}\left(qALtanh{\displaystyle \frac{W}{2L}}\right),\end{array}$$

(8)

$$\begin{array}{c}\hfill n_{eff}={\displaystyle \frac{\frac{W}{2L}\sqrt{N_B^2+P_0^{2}csc{\mathrm{h}}^2\left(\frac{W}{L}\right)}}{arctan\mathrm{h}\left[\frac{tanh\left(\frac{W}{2L}\right)\sqrt{N_B^2+P_0^{2}csc{\mathrm{h}}^2\left(\frac{W}{L}\right)}}{N_B+P_{0}csc\mathrm{h}\left(\frac{W}{L}\right)tanh\left(\frac{W}{2L}\right)}\right]}},\end{array}$$

(9)

where Q is the excess carrier of the total base region, ${\epsilon }_{si}$ is the dielectric constant of silicon, q is the charge of the electron, W is the width of the quasi-neutral base region, $N_B$ is the doping concentration of the base region, ${\mu }_n$ is the mobility of the electron, ${\mu }_p$ is the mobility of the hole, ${\mu }_c$ is the scattering mobility of the carrier, A is the effective area of the chip, $A_{GD}$ is the overlap area between the MOFSET part of the IGBT chip and L is the bipolar diffusion length. $V_{CE}$ is the collector–emitter voltage of the IGBT device, and $V_{GE\left(th\right)}$ is the gate threshold voltage of the IGBT device.

It can be seen from Figure 7 that there is a special parasitic parameter in the gate circuit; i.e., the gate-collector capacitance $C_{GC}$, consists of the series combination of the gate-drain overlap oxide capacitance $C_{OXD}$ and the nonlinear gate-drain overlap depletion capacitance $C_{GDJ}$, which will vary with the voltage $V_{CE}$.

$$\begin{array}{c}\hfill C_{GC}=\left\{\begin{array}{c}\hfill C_{OXD}{V}_{CE}<V_{GE}-V_{GE\left(th\right)}\hfill \\ \hfill {\displaystyle \frac{C_{OXD}{C}_{GDJ}}{C_{OXD}+C_{GDJ}}}{V}_{CE}⩾V_{GE}-V_{GE\left(th\right)}\hfill \end{array}\right.,\end{array}$$

(10)

$$\begin{array}{c}\hfill C_{GDJ}={\displaystyle \frac{A_{GD}{\epsilon }_{Si}}{\sqrt{2{\epsilon }_{Si}(V_{CE}-V_{GE\left(th\right)})/q/N_B}}},\end{array}$$

(11)

where $A_{GD}$ is the gate-drain overlap area, ${\epsilon }_{Si}$ is the dielectric constant of silicon, $V_{GE\left(th\right)}$ is the gate threshold voltage and $N_B$ is the base doping concentration. The significant change in $C_{GC}$ can be as large as a factor of 10 to 100. It makes the turn-on transient of multi-chip IGBT modules very complex.

So, based on the characteristics of collector–emitter voltage $V_{CE}$ and gate-emitter voltage $V_{GE}$, the gate charge can be divided into three intervals. The first interval corresponds to the charge of the input capacitance, during which $V_{CE}$ remains constant, and $V_{GE}$ rises from zero until it reaches the threshold value. In the second interval, as the collector current $I_C$ gradually increases, $V_{CE}$ continues to decrease to its saturated value, while $V_{GE}$ is maintained near the threshold level. During the third interval, the collector-gate capacitance increases rapidly, stabilizing $V_{CE}$ near the saturated value. At this time, the input capacitance also remains stable, and the additional charge causes $V_{GE}$ to continue to increase until it reaches the fully conductive level. The variation characteristics of these two voltage components across the three intervals are illustrated in Figure 8 [38].

In Figure 9, we have empirically validated this characteristic through waveform measurements. Under conditions where $V_{CE}$ is set at 200 V and 420 V, respectively, the turn-on waveform of $V_{GE}$ exhibits negligible differences, indicating that the magnitude of $V_{CE}$ has no influence on the turn-on characteristics of the IGBT module. On the other hand, prior to the moment $t_1$ at which $V_{CE}$ begins to decrease, $V_{GE}$ undergoes no significant changes. Subsequent to this moment, the variation in $V_{GE}$ becomes markedly evident. Figure 10 presents the waveform measurements of the gate current $i_g$ at various $V_{CE}$ levels. It can be observed that before the decrease in $V_{CE}$, $i_g$ shows no substantial changes. However, following the decline in $V_{CE}$, $i_g$ experiences pronounced fluctuations, which is consistent with the phenomenon observed in Figure 9, demonstrating that the reduction in $V_{CE}$ exerts a noticeable impact on the gate circuitry.

That is to say, the input capacitance varies with $V_{CE}$, and only the beginning of the gate charge process is the most relevant stage, because the charge current reaches its maximum value here and is independent of $V_{CE}$. Therefore, the theoretical analysis is carried out in this interval with the IGBT module modeled by input capacitance, and it is just focused on the stray inductance of bond wires as well as the gate-emitter capacitance $C_{GE}$, while all other parameters remain constant. In addition, only non-resonant gate drivers are considered, which are widely used in most of the piratical applications.

Finally, a simplified gate charge circuit is deduced from the special the interval of $t_0$ to $t_1$ as shown in Figure 9 and Figure 10, which is a very known RLC circuit and can be expressed by a second-order differential equation, i.e., the gate charge current $i\left(t\right)$ is governed by

$$\begin{array}{c}\hfill LC{\displaystyle \frac{d^{2}i\left(t\right)}{d^{2}t}}+RC{\displaystyle \frac{di\left(t\right)}{dt}}+i\left(t\right)=0,\end{array}$$

(12)

where R and L are the sum of the gate resistor and stray inductance, respectively, and C is the sum of the input capacitance of IGBT chips inside the module. And the expression of the charge current, which is the solution of Equation (12) based on initializations, is

$$\begin{array}{c}\hfill i\left(t\right)={\displaystyle \frac{V_1}{L(p_1-p_2)}}(e^{p_{1}t}-e^{p_{2}t}),\end{array}$$

(13)

where $p_1=-{\displaystyle \frac{R}{2L}}+\sqrt{{\left({\displaystyle \frac{R}{2L}}\right)}^2-{\displaystyle \frac{1}{LC}}}$ and $p_2=-{\displaystyle \frac{R}{2L}}-\sqrt{{\left({\displaystyle \frac{R}{2L}}\right)}^2-{\displaystyle \frac{1}{LC}}}$.

From (13), the time $t_m$ at which the charge current reaches its maximum as well as peak value can be expressed as

$$\begin{array}{c}\hfill \begin{array}{c}\hfill t_m={\displaystyle \frac{1}{p_1-p_2}}ln{\displaystyle \frac{p_2}{p_1}}\hfill \\ \hfill i_{\mathrm{m}}={\displaystyle \frac{V_1}{L(p_1-p_2)}}(e^{p_1{t}_m}-e^{p_2{t}_m})\hfill \end{array},\end{array}$$

(14)

Equation (12) theoretically shows that peak current of the gate charge circuit has a close relationship with L and C, which is affected by partial bond wires lift-off. To evaluate this impact, parameter sensitivity analysis based on the IGBT module used below is provided as shown in Figure 11. Figure 11 illustrates the variations in the peak magnitude of the gate charge current and the time required to reach the peak value before and after chip failure. The progression from left to right corresponds to conditions ranging from non-bond wire lift-off, to the lift-off of one bond wire, two bond wires, and up to six bond wires lift-off. From Figure 11, it can be seen that the peak gate charge current decreases from 1.11 to 0.991 A, and the time to peak value decreases from 115 to 97 ns when on chip failure. These are because the stray inductance contributed by the bond wires is not a significant value; partial bond wires lift-off can hardly influence the total stray inductance involved in the gate charge circuit, whereas when reaching a certain amount (in this case, six), one of the paralleled chips (in this case two) will be out of running, the total gate-emitter capacitance increases rapidly, meaning the peak charge current deviates greatly; i.e., the charge current deviation can be identify as an early warning signature and used as a precursor to enable the development of prognostic approaches for IGBT modules.

To verify the theoretical analysis mentioned above, a simple experimental chopper circuit is conducted on an unencapsulated IGBT module (2MBI150U4H-170, FUJI ELECTRIC HOLDINGS Co., Ltd., Tokyo, Japan), and bond wires are cut off one by one manually to simulate the bonding wires lift-off found in broken IGBT modules. A power source (Agilent 6813B, Agilent Technologies, Santa Clara, CA, USA), a gate-drive, and an oscilloscope (TDS5104B, Tektronix, Beaverton, OR, USA) are used as shown in Figure 12 and Figure 13.

The results at the same temperature are shown in Figure 14, and relevant characteristics are summarized in

Table 2. Characteristics of gate charge current.

Name	tm (nS)	im (A)
Non bond wire lift-off	144	1.164
Three bond wires lift-off	148	1.161
One chip failure	116	1.044

. Figure 14 presents the waveforms and characteristics of the charge current before and after partial bond wire lift-off. It can be observed that the results when only one bond wire is lifted off are nearly identical to those when three bonding wires are delaminated. At this juncture, the peak gate charge current achievable is approximately 1.164 A, and the time required to reach this peak is approximately 144 ns. These values are essentially consistent with the theoretical analysis derived from Figure 11 for the scenario with three bond wires lift-off, suggesting that the gate charge current has not been significantly affected when three bond wires are lifted off. When six bond wires lift off, i.e., one chip out of running, we can see that the maximum gate charge current decreases from 1.164 to 1.044 A, and the time to maximum value decreases from 144 nS to 116 nS. It is also important to note that the peak charge current deviation contributed by stray inductances of bond wires is not a significant value. As long as six bond wires are cut off, i.e., one of two paralleled IGBT chips is out of running, meaning the total gate-emitter capacitance decreases to half value, clear attenuation is observed, indicating that the experimental result is in accordance with theoretical analysis, as shown in Figure 11. In addition, in order to more intuitively analyze the impact of different $V_{CE}$ values, the peak gate charge current versus different $V_{CE}$ values is shown in Figure 15, and it can be seen that the $V_{CE}$ does not impact the peak charge until $V_{GE}$ reaches the threshold value. This also supports the theoretical analysis. To more vividly illustrate the changes in $V_{CE}$, $V_{GE}$, and $i_g$ before and after the occurrence of defects, we have provided the overall waveform changes for $V_{CE}$ = 200 V in Figure 16a,b and for $V_{CE}$ = 420 V in Figure 17a,b.

In addition, validation experiments concerning junction temperature were conducted. By utilizing a thermostatic chamber, the turn-on waveforms of the collector–emitter voltage and gate voltage at junction temperatures of 90 °C and 120 °C were measured, as depicted in Figure 18. The results indicate that within the time interval from $t_0$ to $t_1$, $V_{CE}$ and the junction temperature have no influence on the turn-on characteristics of the IGBT module. It should be considered that in actual operation of the IGBT module, the junction temperature variation due to self-heating is dynamic, and considering only junction temperatures of 90 °C and 120 °C is insufficient to reflect the real-world scenario. Therefore, in

Table 3. Changes of collector–emitter saturation voltage with defects.

Junction Temperature (°C)	40	60	80	100	120
State:before defect	2.11	2.18	2.25	2.31	2.36
State:partial breakage	2.13	2.22	2.29	2.36	2.41
State:chip failure	2.20	2.27	2.38	2.42	2.51

, by measuring the collector–emitter voltage drop before and after bond wire lift-off at every 20 °C change in junction temperature, it can further confirm that the turn-on characteristics of the IGBT module are independent of the junction temperature variations caused by self-heating.

4. Conclusions

In this study, we delved into the degradation of gate charge in multi-chip IGBT modules, which is a critical aspect of power modules utilized in UAV. The study underscored the significance of high-voltage IGBT modules in UAV and the imperative need for predictive methods to ensure the operational health management of such unmanned equipment.

We established an RLC model that incorporates parasitic parameters of the gate drive circuit and conducted a sensitivity analysis, revealing the peak current in the gate charge circuit’s close relationship with L and C. The experimental results, simulating real-world faults using a sample of an IGBT module with partial bond wires lifted off, corroborate the theoretical findings. These results indicated that deviations in the gate current, significantly influenced by the partial bond wires lift-off, could serve as a precursor for the development of predictive methods for IGBT modules.

The study’s results provided a diagnostic method for anticipating failures in IGBT modules by monitoring changes in the gate charge current. The method advanced the current understanding of IGBT module reliability and degradation mechanisms, especially in the context of UAV applications where high power density and efficiency are critical.

Although the study’s results supported the theoretical analysis, the accuracy is not highly refined, being limited to the level of a single chip. Future work will consider enhancing signal processing techniques to discern subtle deviations in the peak gate charge current, improving the sensitivity and reliability of the predictive method. This will lead to the development of practical fault monitoring systems that can be applied in critical areas such as railway traction and Unmanned Aerial Vehicles. Additionally, the relationship between gate charge current variations and service life will be investigated. By simulating and accelerating the degradation process of IGBT modules through thermal cycling or electrical stress testing, data on the gate charge current and other relevant electrical parameters will be collected. The trend of gate charge current degradation over time will be analyzed to establish a mathematical model that describes the relationship between gate charge current and the degradation of IGBT modules. Utilizing this model, continuous monitoring of the gate charge current, in conjunction with environmental parameters, will enable the prediction of the expected service life of IGBT modules under current operating conditions.

In summary, the identification of gate charge current deviation as an early warning for IGBT module failure represented a significant advance forward in the prognostics of power electronic prediction. The study not only contributed to the knowledge on IGBT module reliability but also paved the way for more sophisticated health management systems in UAV and other high-reliability applications.