6.4.1 Semiconductor Chips

The semiconductor chips form the core of a power electronic module. Fig. 6.6 shows the interior of a power electronic module with semiconductor chips mounted on a DBC substrate and interconnected with wire bonds. Semiconductor chips determine the voltage, current, and power ratings of the power module. Power modules with different internal interconnections of semiconductor chips are available to realize different power converter topologies such as half bridge and full bridge. To achieve high-current capability, multiple semiconductor chips can be paralleled inside the module. The performance of power modules in terms of on-resistance, switching speed, and power loss is critically dependent on the semiconductor chip characteristics.

6.4.2 Die Attach

Die attach is the process of attaching a die/chip to a substrate or package. For example, in the power module of Fig. 6.5, the IGBT chip and the diode chip are attached to the top copper layer of the DBC substrate. Note that in case of vertical chip devices, the device power terminals are on opposite sides of the chip, and therefore, the chip current flows through the die attach joint. Die attach is usually accomplished by using one of the following processes: (1) eutectic, (2) solder, (3) adhesive, (4) glass or silver glass:

1) A eutectic bond is formed by melting a preform consisting of a mixture or alloy of two or more dissimilar metals in the joint between the die and substrate. The preform has a melting point that is lower than the melting point of its base materials. For example, considering a typical preform composed of gold and silicon, the melting point of gold is 1640°C, and the melting point of silicon is 1414°C. However, when the materials are combined in a preform, the melting point drops to 363°C [1].

2) Solder is an attaching method commonly used in the semiconductor industry. Solder attach provides good mechanical strength, high thermal conductivity, and good electric conductivity. Lead-based solder has been popular for a long time with advantages of good performance, proved technology, and low cost. However, environmental concerns about the toxic properties of lead have recently arisen. As a result, lead-free solder is gradually replacing lead-based solder in many applications.

3) An adhesive bond is formed by attaching the die to the substrate using some type of adhesive material. It can be electrically insulating or conductive depending on the adhesive material used. An advantage of adhesive bonding is that it is performed at room temperature and does not require high temperatures.

4) A glass bond is formed by attaching the die to the substrate using glass in paste form. The glass paste may also contain silver particles that enhance thermal and electric conductivity (this is called silver-glass die bonding). The glass bonding process is similar to adhesive bonding. The differences are the material used and the need for heat. Glass bonds are heated to 350–450°C that melts the glass into a low viscosity liquid. The glass hardens as it cools, thereby making the bond [1].

6.4.3 Wire Bonds

The electric connections between the top connections of semiconductor chips and module terminals are realized by wire bonding (shown in Figs. 6.5 and 6.6). Conventional wire bonding materials include aluminum (Al), copper (Cu), and gold (Au). Aluminum is the most common material for power modules because of its low cost and convenient process, especially to form thick wires for high-current capability. Gold has the lowest resistivity and good high-frequency performance and is usually used for integrated circuit and high-frequency devices. Due to limited improvement of resistance in power module and high cost, gold is not commonly used in power modules. The electric properties of copper are intermediate between those of Al and Au, whereas the cost of aluminum is significantly lower than copper. Consequently, Al wire bonds are most common with copper used in special cases.

To satisfy high-current requirements, heavy wire bonding (Fig. 6.7) has been developed. Traditional wire size is smaller than 10 mils (1 mil=1/1000 in.) in diameter, whereas heavy wire bonding is significantly thicker and can save space on the bonding pad. An alternative solution is ribbon bonding (Fig. 6.8), which has the added advantage of better high-frequency performance [2]. Ribbon bonding is a form of wedge bonding using flat ribbon to replace the round wire. Compared with round wire, ribbon wire can carry higher current at high frequency, where skin effect is the limiting factor, and has smaller effective inductance and less signal cross talk [3].

6.4.4 DBC Substrate

The DBC substrate is an important interface between the electric components in the power module and the external cooling system [4]. It consists of a ceramic insulator layer with copper bonded to one or both sides by a high-temperature melting and diffusion process. DBC substrates provide electric conduction on the top copper layer, electric isolation from top to bottom thanks to the ceramic insulator layer, and outstanding thermal conduction from top to bottom. The top copper layer is etched to form the required electric interconnections and is electrically connected to the backside electrodes of semiconductor chips using one of the die attach techniques described above, most commonly by soldering. The excellent electric isolation is provided by the ceramic material. The two most common choices are alumina (Al₂O₃) and aluminum nitride (AlN). Al₂O₃ is inexpensive and widely used; however, it has the disadvantages of having relatively low thermal conductivity (24–28 W/m K) and of being brittle. Compared with Al₂O₃, AlN is more expensive, but it has better mechanical and thermal performance (thermal conductivity larger than 150 W/m K) and therefore has become popular in high-temperature applications [5]. The bottom copper layer is usually not patterned and is soldered to the baseplate to achieve heat spreading from top to bottom. One important thermal characteristic of DBC substrates is a low coefficient of thermal expansion (CTE), which is close to that of silicon. This ensures good thermal cycling performance and higher reliability.

6.4.5 Baseplate

One of the main functions of the baseplate is to provide mechanical support for the electric components and the DBC substrate. At the same time, it must propagate the heat generated by the semiconductor chips to the external heat sink effectively. A high degree of surface smoothness is also required to guarantee no voids between the baseplate and the DBC substrate, because voids can cause hot spots that will eventually lead to cracking and poor reliability [4]. In order to reduce the thermal resistance of the interface between the baseplate and the heatsink, a thin layer of thermal grease (also called thermal paste or heatsink compound) is applied between the baseplate and the heat sink. The purpose of the thermal grease is to fill up voids due to interface surface imperfections, because air in these voids acts as an excellent thermal insulator. However, thermal grease is not a good thermal conductor, so a thick layer of thermal grease may actually increase the thermal resistance between the baseplate and the heat sink.

CTE mismatch is a key consideration in power module design and needs to be addressed in each connection and attachment between two different materials inside the power module. To improve power module mechanical reliability, adjacent materials should have closely matched CTEs, so that, when the module heats up and cools down during operation, the adjacent materials expand and contract in a similar way, avoiding the development of mechanical stress, which can cause joint failure due to mechanical fatigue. This is particularly important for large-area connections, which may develop larger mechanical stress during thermal cycling. Therefore, the baseplate material should be thermally compatible with the DBC substrate material, and mechanical, chemical, and other factors also need to be considered.

The housing and silicone gel are used to insulate the electric components on the top of the module and prevent contamination. In modern IPMs, thermal sensors are attached to the DBC close to the chips, to measure the real-time junction temperature, which is very useful for device condition monitoring.

6.5.1 A Survey of Power Electronic Module Topologies

A brief survey of power electronic module topologies is now given:

Diode rectifier or thyristor rectifier: Four (or more) diodes or thyristors are connected in a bridge circuit configuration that converts an AC input (typically either single phase or three phase) into a DC output. A full-bridge and a three-phase bridge diode rectifier are shown in Fig. 6.9. For an example of a three-phase module, see Ref. [6].

• Half-bridge, full-bridge, or three-phase (six-pack) power module: Generally, these power modules, shown in Fig. 6.10, are implemented with either power MOSFETs or IGBTs (insulated-gate bipolar transistors). Typically, IGBTs can achieve much higher current and voltage ratings, allowing them to operate at higher power levels with reduced conduction losses, whereas MOSFETs can operate at much higher switching frequencies, making them suitable for application where fast switching and high power density are required. A half-bridge configuration can be used to implement a DC-DC power-bidirectional converter or a single-phase inverter. A full-bridge configuration can be used to implement a full-bridge converter or a three-level single-phase inverter. A six-pack power module can be used to implement a three-phase inverter or a three-phase-controlled rectifier. Note that since all switches are controllable, all these converter topologies are power bidirectional, that is, power can flow in both directions. For an example of a six-pack power module, see Ref. [7].

• Converter-inverter-brake (CIB) module: The module, shown in Fig. 6.11, contains a single-phase or three-phase uncontrolled or half-controlled bridge rectifier (C) plus an inverter (I) and a brake chopper (B). This module contains all power semiconductor devices required to implement a three-phase variable motor drive fed by three-phase power. The three-phase diode rectifier can be used to create a DC bus voltage that supplies the three-phase inverter connected to the three-phase motor. The chopper circuit allows to pulse-width modulate a braking resistor to dissipate energy during regenerative braking. See Ref. [8] for more information about regeneration in motor drives.

• Neutral-point-clamped (NPC) module: Each leg has four power switches connected in series. The applied voltage on the power switch is one-half that of a two-level converter. The NPC topology can allow operation at voltages larger than the voltage rating of a single power device. In addition, NPC converter can achieve lower output current ripples and improve the total harmonic distortion (THD) of the output.

• Power factor correction (PFC) module: PFC module combines a rectifying function and a PFC function in order to reduce the harmonics of the input current and improve power factor.

6.5.2 Power Semiconductor Devices Used in Power Electronic Modules

The most common power semiconductor devices used in power electronic modules are diodes, MOSFETs, IGBTs, and thyristors. Power bipolar junction transistors (BJTs) are less common and are briefly mentioned for completeness.

1. Power diode:
Power diodes provide uncontrolled rectification of power and are used in applications such as AC-DC converters and variable-frequency drives. Typically, the power diode can be either a P-i-N diode or a Schottky diode.
The P-i-N diode is a bipolar device having a so-called P-i-N structure, that is, it is a p-n junction with a wide, undoped intrinsic semiconductor layer (i-layer) in the middle to sustain a reverse blocking voltage. Frequently, the midlayer, rather than being intrinsic, has low N-doping and is therefore called N- (N minus) layer. Fig. 6.12 shows the basic structure of a P-i-N diode. In the forward-biased conduction, the diode has a forward conduction drop, which causes conduction loss, and the diode can be modeled as a junction voltage source, equal to the threshold voltage, and a series equivalent resistance that represents the increasing slope in the diode I-V characteristics. In the reverse-biased condition, a small reverse leakage current flows, which increases gradually with the applied reverse voltage. If the applied reverse voltage exceeds a certain voltage, called breakdown voltage, the reverse current becomes very large due to avalanche breakdown, and the diode may be destroyed due to large power dissipation in the device. The advantage of a P-i-N diode is low on-state voltage drop in high-current conduction, due to the minority carrier injection in epitaxial drift region resulting in conductivity modulation. However, the minority carrier injection during conduction causes a large reverse recovery current during diode turnoff, which is not desirable, because it causes additional turn-on loss in the active switch.
The Schottky diode is a very attractive unipolar device, formed by an electrically nonlinear contact between a metal and a semiconductor bulk region shown in Fig. 6.13. It has low on-state voltage drop, almost zero reverse recovery charge, and extremely high switching speed, but its reverse leakage current is large especially at high temperature due to its lower built-in potential barrier. Additionally, the breakdown voltage of Schottky diode is limited by the increase of drift region resistance without any conductivity modulation effect.

2. Power MOSFET
The power metal-oxide-semiconductor field-effect transistor (MOSFET) is a unipolar, majority carrier, and voltage-controlled device. It is one of the most commonly used power devices due to its easy gate driving, fast switching speed, and low loss dissipation. Most power MOSFETs feature a vertical structure (a structure of alternating p-type and n-type doping) shown in Fig. 6.14. In most power MOSFETs, the N+ source and P-well junction are shorted to prevent the parasitic bipolar transistor from turning on accidentally. When no gate voltage is applied, the power MOSFET is capable of supporting a drain-source blocking voltage through the reverse-biased P-well and N-epi junction. In high-voltage devices, most of the blocking voltage is supported by the lightly doped N-epi layer.
Generally speaking, the power MOSFET can operate in two different quadrants of its I-V characteristics: the first quadrant operation and the third quadrant operation. As the gate voltage (V_G) increases above a threshold voltage (V_th), the MOSFET starts to conduct current. When the gate voltage V_G is much larger than the threshold voltage V_th, the output I-V curve appears linear, and the MOSFET channel is fully turned on. Under low gate overdrive, the drain current reaches a saturation value when the MOSFET channel is pinched off.
The MOSFET's switching behavior is highly dependent on the parasitic capacitances gate-to-source C_gs, gate-to-drain C_gd, and drain-to-source C_ds. Gate-to-drain capacitance C_gd and drain-to-source C_ds are nonlinear and a function of bias voltages.

3. Power IGBT
The insulated-gate bipolar transistor (IGBT) is a minority carrier device with high input impedance and large current-carrying capability. The IGBT combines the advantages of both power MOSFET and BJT, and it has MOS input characteristics and bipolar output characteristics. The basic structure of a power IGBT is shown in Fig. 6.15.
The IGBT is widely used in many applications of power electronics, such as motor drives, uninterruptible power supplies (UPS), and power modules, because of the following advantages:

• A very low on-state voltage drop due to conductivity modulation and very small conduction loss

• A simple gate driver circuit design due to the MOS gate structure

• A very large current-carrying capability and smaller chip size for a given conduction current

• Good forward and reverse voltage blocking capability

However, the IGBT has some drawbacks, compared with power MOSFET:

• Slow switching speed, current tailing due to carrier combination

• Latch-up issue due to the parasitic thyristor structure

The IGBT is viewed as a serial connection of a MOSFET and a P-i-N diode, and sometimes it is modeled as a wide base PNP transistor driven by a MOSFET in Darlington configuration. In the output I-V characteristics of a power IGBT, the conduction current starts flowing when the gate voltage exceeds the threshold voltage, and collector-source voltage is approximately above 0.7 V.

4. Power Thyristor
A thyristor is a power device with four alternating layers of p-type and n-type doping and can be turned on by the gate terminal but cannot be turned off by the gate [9]. It has the largest power-handling capability.

5. Power BJT
The power bipolar transistors (BJT) are conceptually the same as the small-signal BJTs. The main difference is that the active area of power BJT device is much larger resulting in a much higher current handling capability. Power BJTs also have a thick and lightly doped collector region. Such collector regions result in a large blocking voltage.

6.5.3 SiC Power Semiconductor Devices in Power Modules

In recent times, power semiconductor devices using silicon carbide (SiC), a wide bandgap material, rather than silicon have become commercially available. The SiC material has superior material properties for power electronic applications. SiC devices are starting to be used in power modules. In particular, SiC-merged barrier Schottky diodes are used to replace high-voltage P-i-N diodes. The advantage is that, unlike P-i-N diodes, SiC Schottky diodes are majority carrier devices and do not exhibit reverse recovery at turnoff. SiC MOSFETs are also being introduced in modules. Presently, the higher cost of SiC devices limits their adoption.

For a variety of system-level reasons, modern power electronic applications require compact and high-power density power converters. In order to achieve this required miniaturization, higher integration of components and higher switching frequencies, which reduce passive filtering element size, are required. The increase in switching frequency leads to increased power losses and the physical integration makes it harder to remove the loss heat generated. This creates a significant design challenge that requires careful thermal management. Additionally, in power modules, the power dissipation is concentrated inside the small power semiconductor chips, making heat removal particularly difficult. Power module datasheets specify a maximum allowable junction temperature, that is, temperature at the semiconductor chips inside the module. Exceeding this temperature will lead to module failure. Moreover, lower operating junction temperatures lead to markedly improved module reliability. The maximum junction temperature for current commercially available silicon-based power modules is around 150°C.

Another thermal concern affecting module reliability is thermal cycling, that is, the alternate heating and cooling of the module as a result of varying operating conditions. Excessive thermal cycling leads to aging and early failure of the module. In order to avoid overheating and to improve reliability, a common recommendation is that junction temperature should not exceed 125°C. Therefore, thermal management of power module is necessary to satisfy the thermal requirements described above.

6.6.2 Equivalent Thermal Network of Power Module

Fig. 6.16 shows a cross section of a power electronic module. Power dissipation occurs at the silicon chip due to conduction and switching losses. The heat is conducted through several material layers inside the module until it reaches the baseplate. Fig. 6.16 shows that the heat flows from top to bottom through the top solder layer, the top copper layer, the ceramic substrate, the bottom copper layer, the bottom solder layer, and the baseplate. From the baseplate, the heat exits the module, crosses the thermal grease layer, and reaches the heat sink. In case of air cooling, the heat then transfers by convection to the air surrounding the heat sink. Usually, the heat flow is modeled as being unidimensional with the lateral heat spreading modeled as a spreading angle that increases the heat flow cross section. It should be kept in mind that this approximation introduces a certain error [10]. Each material layer has a certain thermal resistance and thermal capacitance. The temperature gradient across each layer drives the conductive heat transfer. Consequently, the temperature gradually decreases from junction temperature at the silicon chip to ambient temperature T_AMB.

The junction temperature in the power module can be estimated by the equivalent Cauer RC thermal network shown in Fig. 6.17. In this equivalent model, voltage represents temperature, and current represents heat flow. An RC cell represents a specific material layer [11]. The thermal resistance R_i of the ith layer is given by

$R_i=\frac{d_i}{k_i{A}_i}$

where d_i is the ith layer thickness; A_i is the effective cross-sectional area, including a spreading angle for conductors; and k_i is the thermal conductivity of the ith layer.

Thermal capacitance C_i of the ith layer is given by the equation

$C_i={\left(c\times \rho \times V\right)}_i$

where c is the constant pressure-specific heat capacity, ρ is the mass density, and V is the effective volume of the ith layer.

According to the equivalent thermal network, the junction temperature T_J is numerically equal to the top left node voltage (the top node of the input current source), while the case temperature T_C represents the reference node voltage of the network. The power module datasheets provide a numerical value for the junction-to-case thermal resistance R_th(JC). The term “case” indicates the bottom of the baseplate, where the heat sink is attached. In the equivalent model of Fig. 6.17, R_th(JC) is equal to the sum of all thermal resistances R₁ through R₇. The relationship between T_J and T_C is given by the following equation [12]:

$T_{\mathrm{J}}=T_{\mathrm{C}}+P_{\mathrm{diss}}\cdot R_{\mathrm{t}\mathrm{h}\left(\mathrm{J}\mathrm{C}\right)}$

This thermal resistance directly provides the temperature difference between junction and case. To calculate the absolute temperature, it is necessary to calculate the case temperature T_C. This temperature can be calculated based on the selected heat sink, air flow, and ambient temperature T_AMB.

As previously mentioned, power modules frequently have a thermal sensor mounted in close proximity of the chips. This gives experimental access to junction temperature in real time [13]. Note that this method can only measure the junction temperature in steady state or for slowly varying thermal transients, because the measured value is accurate only at thermal equilibrium.

According to the thermal network, the junction temperature can be controlled in an acceptable range by applying external cooling to the case of the module.

6.7.1 Reliability Tests [17]

A power module is exposed to active temperature cycles due to its operation and to passive temperature cycles due to variation of the surrounding environment. Thermal cycling and power cycling tests can be performed to represent these conditions and evaluate reliability:

6.8.1 Bypass Capacitor Considerations

Besides component selection, as is frequently the case in power electronics, proper physical layout is critical. As shown in Fig. 6.25, for the case of a half-bridge converter, the switching loop should be minimized in physical size in order to reduce overshoot and ringing during switching. The objective is to reduce the distributed parasitic inductance of the switching loop, shown as lumped inductance L_SWITCH in the figure.

As shown in Fig. 6.4, the switching loop is partially inside the power module, but it is completed by the external capacitor C_BYPASS. This capacitor should absorb the square-wave current at the switching frequency resulting from switching converter operation. It is critical to use a high-frequency high-current-ripple capacitor. Ceramic or metal film capacitors are appropriate. Polypropylene capacitors are a popular choice. It is critical to mount the bypass capacitor physically close to the module. A laminated bus bar is recommended to reduce parasitic inductance by having closely spaced flat conductors conducting opposite polarity currents.

6.8.2 Gate Driver Design Considerations

The gate driver design is similar to what is commonly done for discrete power semiconductor devices. Based on the total gate charge Q_G from the datasheets and on the desired gate voltage rise and fall time, a gate driver with adequate current capability is selected. The choice of gate resistor is a trade-off: a smaller gate resistor gives faster switching speed with lower switching losses and with increased gate noise immunity, whereas a larger gate resistor gives slower switching speed with reduced switching noise and reduced diode reverse recovery. It should be kept in mind that power modules frequently have internal gate resistances in case of multiple paralleled dies as shown in Fig. 6.26. Resistances R_g are internal to the module and are needed to ensure dynamic current balance. If a single gate resistor was used, there could be an LC oscillation caused by the input capacitances of the two IGBTs resonating with the parasitic inductance of the gate connection. In the case of Fig. 6.26, this resonant path has a 2R_g series damping resistance. The module datasheets usually provide a recommended external gate resistor range of values.

Regarding layout, it is important to minimize the gate loops shown in Fig. 6.25. A large gate loop causes a large parasitic inductance L_GATE that resonates with the IGBT input capacitance causing gate voltage ringing, which may cause spurious IGBT switching. The power module gate contacts should be directly connected to the gate driver printed circuit board if possible.

6.8.3 Gate Kelvin Contacts

Power modules typically have gate Kelvin contacts. It is essential that they be used for the gate drive as shown in Fig. 6.4 in order to have an effective dynamic gate voltage control. Fig. 6.27 shows the case of no emitter Kelvin contact. As a result, there is a parasitic emitter inductance L_E that is part of both the switching loop and the gate loop (refer to Fig. 6.25). This inductance conducts the full collector current. During turn-on, the gate drive voltage increases, and the IGBT gradually turns on. The increasing collector current induces a positive voltage v_LE on the parasitic emitter inductance that subtracts from the gate driver voltage and slows down the gate-emitter voltage rise. This negative feedback effect is undesirable because it slows down the switching transition and causes the gate driver to have less dynamic control of the actual gate-emitter voltage. A similar negative feedback effect occurs at turnoff, slowing down this transition as well. This problem can be solved by providing a separate emitter Kelvin contact as shown in Fig. 6.28. The power contact emitter inductance (in red) is not part of the gate loop, and the negative feedback effect is eliminated.

Fig. 6.29 shows the inside of a six-pack IGBT power module. On the bottom left side, one can see the power emitter pin and the Kelvin emitter pin. Notice that dedicated wire bonds go from the IGBT chip to the Kelvin emitter pin. These wire bonds conduct only the gate driver current and not the collector current and represent the E2 connection of Fig. 6.28. The figure also shows the IGBT and diode dies, the IGBT gate and emitter pads, the wire bonds, the top copper layer, the ceramic substrate, and the thermistor.

6.8.4 Other Design Considerations

In order to realize a switching converter, other design aspects must be considered, such as protection, sensing, and control implementation. The gate driver may include protections, such as short circuit protection. Current sensing may be implemented for protection and for current mode programming control. Voltage sensing is typically required for control. Either an analog or a digital controller can be used. Snubbers and clamps may be required to reduce switching waveform overshoot and ringing. EMI filters may be required.