Sharing Experience on the Reliability of IGBT Modules

With the rise of emerging applications such as wind power generation, smart grid construction, electric vehicles, and high-voltage inverters, the use of high-power IGBT modules is becoming increasingly prevalent. Correspondingly, the importance of IGBT reliability in high-power power supply design has grown significantly. System reliability has become one of the most crucial design metrics, and the reliability issues of high-power switching devices are of paramount importance.

The failure mechanisms of IGBT modules can be broadly categorized into two classes, comprising nine aspects:

First class, four issues due to insufficient parameter margins:

1. Coupling current interference caused by excessive rate of change of transformer junction capacitance concerning voltage variations. The consequence of this problem is logic errors, interference with control circuits, and circuit failure.
2. Insufficient working frequency (minimum pulse width) of the driver circuit relative to the IGBT switching frequency (duty cycle range), or insufficient average output power of the auxiliary power supply, resulting in unstable outputs. The consequence is fluctuating driver states, increasing the probability of the system’s worst-case scenario.
3. Mismatch between the rise/fall time of the driver circuit output voltage and the IGBT switching speed or insufficient peak power of the auxiliary power supply, preventing the driver circuit from reaching full swing drive. The consequence is a decrease in product batch consistency and an increased probability of the system’s worst-case scenario.
4. Insufficient rated output power density of the driver chip, leading to accelerated device aging. The consequence is an increase in delay time, resulting in a relatively insufficient dead time.

Second class, five issues related to application technologies:

1. Issues related to device selection, including reliability problems of energy storage capacitors, equivalent DC resistance of capacitors, aging and reliability issues of photosensitive devices, and environmental dust and mechanical strength of optical interfaces.
2. Issues related to the reliability of output logic, including some suggested measures for memory logic errors and recommendations for the installation position of the driver board.
3. Issues related to coupling current paths, including the installation environment and position of each unit, grounding issues, coupling current guidance issues, system sensitivity bandwidth, latch, and power integrity issues.
4. Issues related to the value of the output resistance, including constraints on the upper limit and lower limit values, and the relationship between IGBT temperature and value range.
5. Issues related to IGBT installation, including failures due to uneven heat or mechanical stress, and failures due to the uniformity of thermal resistance and heat dissipation conditions.

The following provides detailed explanations for each item:

I. Coupling current interference caused by excessive rate of change of transformer junction capacitance concerning voltage variations.

When discussing the isolation effect of the driver, some friends often think of the isolation voltage specified in the parameter manual or the maximum rate of voltage change it can withstand. However, these parameters essentially indicate under what working conditions the driver will not be damaged, rather than the isolation effect of the driver.

Any driver, including those using optocouplers, must have an isolation transformer that provides power to the output stage. The transformer itself will inevitably have a coupling capacitance between the primary and secondary sides. When the IGBT switch causes a significant voltage change on both sides, the charging and discharging of this capacitance will inevitably generate a current. This will disturb the circuit on the side that shares the ground with the transformer.

The rate of change of the collector-emitter voltage of the IGBT depends on the rate of voltage change of the equivalent capacitor between the gate and emitter under the action of the drive current. When the gate-emitter voltage of the IGBT changes to the point where the gate-emitter current is equivalent to the operating current, the gate-emitter voltage will no longer change.

The current output by the driver will charge and discharge the equivalent capacitance between the gate and collector, achieving a change in the gate potential. Therefore, this potential change process itself corresponds to the constant current charging process of the capacitor under this condition, which has a step-like nature at the beginning and end. Therefore, the overall function of this interference current has the characteristics of a gate function.

To analyze the impact of this interference current on the circuit system, various analysis tools such as wavelet transform should be used from the perspective of instantaneous frequency spectrum analysis to identify the characteristics of those instantaneous frequency components that carry more energy. Full-time domain analysis based on Fourier transform should not be used.

The reason is that the results of this type of full-time domain analysis are essentially based on the results of instantaneous frequency domain analysis, further averaged over time. This will result in distortion and loss of real-time characteristics of the signal, unable to truly reflect the problem.

Regardless of the instantaneous frequency analysis method used, it will be close to the macroscopic current function characteristics. That is, the main instantaneous frequency components exist above the frequency corresponding to the gate function period and are relatively close. At the same time, due to the presence of rising and falling edges, there is a considerable portion of components in the relatively high frequency range.

This makes the main instantaneous frequency components of this interference current concentrate on two large parts: low frequency and high frequency. Among them, the frequency of the low-frequency part is roughly corresponding to the duration of the current, in the range of hundreds of nanoseconds to microseconds, roughly corresponding to the frequency range of 1 to 10 megahertz.

The high-frequency part is from the frequency determined by the coupling channel itself and should be significantly higher. Considering the low-pass capability of stray parameters in actual situations, the high-frequency components in practice should be at the level of several hundred megahertz.

The range of 1 to 10 megahertz is a sensitive frequency band. It is the overlapping zone of common grounding and multiple grounding in PCB layout. This means that the inductive component in the line system in this frequency band has reached or even surpassed the impedance component, becoming the main factor.

The distribution path of the current becomes more complex and relatively concentrated due to the pronounced characteristics of both inductive and impedance in this frequency band. Because of the characteristics of inductance and impedance, there is a condition for forming a large voltage between two interconnected points.

Although this interference current occupies the main part and has a large amount of energy, the frequency range is relatively low. The main impact is still concentrated on the potential difference formed between the signal reception and transmission ends. This will lead to a loss of the margin of the logic error detection threshold, increasing the probability of logic errors.

The high-frequency component in the range of several hundred megahertz will show clear high-frequency current characteristics. It should be higher or close to the operating frequency of most control chips.

As we know, the resonance frequency of decoupling capacitors in high-frequency digital circuits should be at the highest operating frequency of the circuit. If the frequency of the interference current is higher than the highest operating frequency of the circuit, it is likely to make the decoupling capacitor behave inductive.

The result is that in the process of supplementing charges to achieve charge balance (forming an equipotential body), it will cause a significant fluctuation in the power supply voltage (especially when the circuit grounding is poorly handled). In terms of the power level of this current, it has the nature of a current source because it comes from the IGBT switching action.

Its energy is sufficient to trigger power integrity issues, such as the most dangerous latch-up problem in CMOS devices. The magnitude of its harm can be imagined.

In summary, the charging and discharging current of the transformer junction capacitance caused by the IGBT switching process has a significant impact on the circuit system sharing the ground with it. When selecting IGBT drivers, it is necessary to consider this factor based on the actual situation of the system.

Special attention should be paid to complex control circuits. It needs to be emphasized that when comparing the differences between different drivers in this aspect, one cannot only pay attention to the numerical value of the junction capacitance. It is necessary to pay special attention to the differences in the transformer structure.

For mature driver products, different levels of drivers are expected to have different levels of isolation capability. As long as there is no situation where a small driver carries a large load, it should be fine. However, for self-made driver products, it is necessary to compare the differences in the transformer structure with similar mature products.

For example, factors such as the winding spacing, winding projection area, and winding structure. This will enable a more reliable self-assessment. It should be noted that when the concept is vague and the basic method of analyzing the problem is problematic, errors are likely to occur. In technical issues, conclusions and rules are secondary because specific conditions vary.

However, ideas and methods are important because they remain constant amidst change. Therefore, it is highly anticipated to discuss viewpoints, thoughts, and deep mechanisms of some technical issues together. I believe it will be more helpful.

II. Insufficient working frequency (minimum pulse width) of the driver circuit relative to IGBT switching frequency (duty cycle range), or insufficient average output power of the auxiliary power supply, leading to unstable outputs.

I wonder if everyone has a question. Generally speaking, the factors limiting the output frequency are response speed and dissipated power. However, in comparison, the specified upper limit of the output frequency of many drivers appears to be much smaller. Why is this? One reason is that the driver cannot return to a steady state immediately after a single output flip. If it flips the output again before the driver returns to a steady state, it may cause some reliability issues.

Another aspect is the storage charge issue of the junction-type transistor. Due to the control advantages, the driver circuit often contains bipolar transistors instead of all field-effect transistors. Bipolar transistors have a characteristic that their turn-off process depends on the total charge flowing through.

This process is the depletion process of the base storage charge. The output of the driver is not continuous; there is no output once it reaches the given potential. This essentially cuts off the channel for releasing stored charge. Therefore, many times the driver needs a long time to deplete the stored charge after a single output.

If it flips the output again before it returns to a steady state, it may cause delayed response, insufficient output amplitude, and a sudden increase in dissipated power. It needs to be emphasized that if the above mechanism is the primary limiting factor for the working frequency of a driver, then the correlation between the driver’s maximum working frequency and temperature will be relatively large. Correspondingly, attention should be paid to the actual test temperature of the highest working frequency, leaving room for discretion.

In summary, the driver’s output frequency should have a certain margin. It is best to include the duty cycle change rate in the calculation. For example, if the duty cycle may change from 33% to 66% between adjacent cycles, then the corresponding maximum working frequency should be 1.5 times the current value.

On the other hand, aluminum electrolytic capacitors externally connected to the driver should preferably use higher-quality products and should not be randomly purchased in the market. Especially recommended are some specially-made switch-mode power supply output filter capacitors. These capacitors have advantages in ESR. Furthermore, if the application’s temperature range is wide, such as power equipment used in the field, which may be used in extreme cold and heat, it is recommended to leave a more generous maximum working frequency margin according to the situation.