Posted on 10 July 2019

Time Resolved in Situ Chip Temperature Measurements during Inverter Operation

Free Bodo's Power Magazines!




The IR camera is the best choice for temperature measurements at lower voltages

The device temperature is one of the most critical parameters in the dimensioning of an inverter. However, experimental studies focusing on the virtual junction temperature Tvj in running inverters are scarcely found. The feasibility of four different practical methods will be compared. Special focus is set on routines with high time resolution that enable tracking of the time-dependent temperature during one period of the sinusoidal output current. Details of the procedures are explained and compared with simulations.

By Waleri Brekel, Thomas Dütemeyer, Gunnar Puk, Oliver Schilling, Thomas Schütze, Infineon Technologies


The dimensioning of an inverter application requires qualified knowledge of the stress imposed on the semiconductor devices. All electrical parameters are easily accessible by current and voltage probes and standard oscilloscope data recording while the chip temperature during inverter operation at the moment is seldom determined. The thermal dimensioning is normally done using typical or worst case values specified by the supplier (e.g. thermal resistances of IGBT module and cooler) in combination with simulations of the generated losses. In this work theoretical predictions for the junction temperature are compared with four different methods to measure the Tvj during inverter operation:

1) Infrared camera
2) Thermocouple
3) Infrared sensor
4) Internal IGBT gate resistor (RGINT)

Test conditions

The temperature measurements are performed on a 3-phase pulsecontrolled inverter using water cooled 6.5kV IGBTs. To measure the temperature with the IR sensor or IR camera, the surface of the IGBT must be uncovered. Therefore the dielectric, that is necessary to assure high insulation capability, has been removed from the module. To avoid a flashover, the applied DC-link voltage for the measurements has been limited to VCC=2kV. Operating conditions are: ICmax=980A; f0=20Hz; fSW=400Hz; cosϕ=0.01; Ta=30°C.

Figure 1. shows the black coated DUT whose IGBT chips are investigated by different methods. IGBT a) is measured with thermocouple, b) with IR sensor and c) with RGINT method.

Black coated 6.5kV module (DUT) with marked IGBTs measured by different methods

Temperature calculation with IPOSIM

IPOSIM is an Infineon simulation tool for power loss and thermal calculations of IGBT modules [1]. It provides switching and conduction losses of IGBTs and free-wheeling diodes operated in several circuit configurations, e.g. for three-phase inverters with sinusoidal output current. The corresponding Tvj under defined operating condition which can be set by user (e.g. VCC, IC, fSW, f0…) is calculated as well. For best comparison, thermal values Zth acquired by experiment are used for the calculation. Figure 2. illustrates the time depended power losses p(t) within one period and the corresponding IGBT junction temperature.

Power losses and Tvj as calculated with IPOSIM

Infrared camera

The temperature dependent intensity of the emitted electromagnetic radiation of a body is given by Planck’s equation. This can be used to determine the surface temperature of an IGBT chip. To achieve an emissivity close to one, the surface has to be coated with a suited material. The respective wavelength maximum of the emitted radiation is in the infrared spectral region. Infrared cameras allow to get the temperature distribution of the module and measure the temperature of all IGBT chips in parallel. A picture of the temperature distribution is shown in the table 1) B.

The prerequisite for high time resolution with IR camera is its low integration time of the focal pane area. A compromise has to be made between required time resolution and the necessary intensity to measure the temperature itself. An integration time of 0.6ms is chosen to achieve a fine resolution of the 50ms period of the output current. The sample rate of the IR camera is set to 19.5Hz in contrast to f0=20Hz load current. The small difference between the two frequencies leads to sequentially sampling of the temperature over many periods under steady state conditions. In this case the time between two sampling point is 1.28ms, resulting in 39 sampling points for a 50ms period. The temperature of IGTB a) and b) is analyzed over their chip area, whereas IGBT c) is investigated in the middle of the die where RGINT is positioned. The result is presented in table 1) C.

The mean temperature of IGBT a) with ~64°C fits very well to the simulated value of 65°C. It is noticeable that Tvjav of chip a) which is located near the centre of the module is higher about 3K than the chip b) with Tvjav~61°C. This is a well known effect caused by lateral temperature spreading in the module and taken into account in the module characteristics and specification.

Table 1 part 1

Overview of all investigated methods - table 1 part 2


A common way to measure the temperature is given by the thermo electric effect. To record Tvj a thermocouple of type K is glued on the surface of a single IGBT chip position a). The applied glue is characterized by low thermal impedance. The picture of an IGBT chip with a glued thermocouple close to the center of the emitter area is shown in 2) A.

The advantage of the thermocouple is its linearity in the usual chip temperature range. To use the thermo voltage as a temperature proportional signal, it needs to be amplified. The calibrated characteristic line of the thermo voltage amplified by a transducer is shown in 2) B. As the time constant of the thermocouple is in the range of ~200ms, the 20Hz temperature ripple can not be resolved. In 2) C the measured temperature of ~65 °C matches well with the Tvjav of the simulation and IR camera.

Infrared sensor

Infrared sensors covering a defined solid angle are commercially available. The infrared sensor allows a contact free determination of the chip surface temperature. 3) A shows the mounted infrared sensor on top of a 6.5kV IGBT chip. The sensor has a ratio of 1:2 between the distance to the surface and the diameter of the measured area. The distance is chosen to limit the investigated area to the active area of the chip. The sensor generates a voltage that corresponds to a thermocouple of type K. This allows to utilize the same transducer and data acquisition as used for the thermocouple measurement. The corresponding calibration line is shown in 3) B.

The output of the sensor is fairly linear correlated to the temperature in the relevant range. Due to the time constant of the sensor (~50ms) only an averaged chip temperature can be determined. 3) C shows the surface temperature of the IGBT over two periods of the load current.

The measured mean temperature is ~61.4 °C. The observed small waviness with a frequency of 20Hz corresponds to the load current. Due to high time constant of the measurement the amplitude of the temperature swing is expected to be damped to a large extent. The high frequency ripple correlates to the switching frequency of 400Hz. As this frequency is far too high to be recorded with this method, it is generated by interferences between the thermoelectric voltage and the module switching operation.

Internal gate resistor

The investigated 6.5kV IGBT chip contains an internal gate resistor (RGINT) in the centre of the die. As this resistance has a well known temperature dependency it can be used to determine the chip temperature. A major advantage of this resistor is the absence of an additional heat capacitance and its immediate vicinity to the semiconductor junction inside the same die. The accurate measurement of the resistance in presence of transient high voltage and current in the IGBT module is a major challenge. A sophisticated circuitry is developed for data recognition and safe data transfer from the measurement point inside of the inverter to the external laboratory periphery. A sketch of the measurement system is shown in figure 3.

Temperature measuring system for the RGINT method

The evaluation shows that it is possible to achieve high temperature and time resolution at the same time. The principle of the measurement is shown in 4) A. A constant test current of I0=500mA is applied at the force connections F1 and F2 which generates a temperature dependent voltage drop across RGINT. This voltage is used as temperature indicator and is measured by sense contacts S1 and S2. The temperature dependence of the resistor is determined to be ΔRGINT /ΔT ~ 1.5mΩ/K. The voltage resolution of 500mA·1.5mΩ/K= 0.75mV/K allows to resolve temperature variations down to ~1K accuracy. Since the internal gate resistor of each IGBT chip varies within the specified tolerance, an exact calibration line is needed.

The temperature measured by RGINT is presented in 4) C. The maximum of temperature is about 67.4°C with a ripple of 5.9K and average value of 64.5°C. Thus the temperature of an IGBT in inverter operation can be well resolved. The course over time is in excellent agreement with the simulated results. The ΔTvjRipple is about 2K higher than expected by simulation. The average temperature of IGBT c) measured with IR camera matches perfectly to the RGINT value. The resistor is placed in the middle of the die at the gate pad. According to the IR measurement there is no hot spot at this position. Therefore the RGINT method provides a local temperature which is approximately the mean chip temperature of the IGBT.


To determine the junction temperature of an IGBT during inverter operation, four different measurement methods are applied and compared. Two of these methods (RGINT and IR camera) are capable to resolve the temperature ripple of the IGBT caused by the periodic alternating output current in inverter applications. Due to their high time constants, thermocouple and IR sensor achieve the mean temperature but are not able to resolve the temperature ripple itself.

The measurements are most valuable as a complimentary method to check simulation results. This can help to improve the careful design of an inverter application. Table 2. gives an overview of all measurement methods in comparison with simulation.

Overview of all measured value (Simulated temperatures)

The measurement with the IR camera is the best choice for temperature measurements at lower voltages. The RGINT method is a reliable way to achieve time resolved temperature informations where no optical access to the module is possible or measurements at high voltages are requested.



1) Th. Schütze, Th. Barucki, and U. Knorr: IPOSIM Web-based Design Support Tool for IGBT Applications, Bodos Power Systems 12.09.



VN:F [1.9.17_1161]
Rating: 0.0/6 (0 votes cast)

This post was written by:

- who has written 791 posts on PowerGuru - Power Electronics Information Portal.

Contact the author

Leave a Response

You must be logged in to post a comment.