Contents I. Introduction
One of the main reasons for the malfunctioning and failure of power electronics is the increase in temperature [1]. Hence the knowledge of the operating temperature of the device is important. One of the applications of this information is thermal management where the device temperature is used to get the thermal resistance or thermal impedance [2] and provide feedback to the cooling system. Based on the temperature feedback information, the device can be cooled by different means. Generally, this is done by switching on a cooling fan or by starting the flow of coolant liquid once the critical temperature threshold is crossed and keep the cooling process on until the device cools down to the safe operating temperature. Another area of usage is in the field of research and development of semiconductors to make assumptions on the reliability and lifetime of the device. To achieve this, the devices are aged under thermal stress in an artificially controlled environment. This leads to thermo-mechanical stress which is due to the difference in the thermal expansion coefficient between the different layers in the component package. The degradation of the device under test (DUT) is continuously monitored under the accelerated ageing tests until the device fails. This data can be further processed by a various prognostic algorithm such as Gaussian process regression algorithm to estimate the lifetime of the device [3]. There are many more applications where the information on the temperature of the power device can be utilized. These cases show why the temperature monitoring of a device is very important. Depending on the application and usage, various kinds of temperature sensing methods can be implemented. This can be either the measurement of junction temperature of the device or the case temperature itself. This can be broadly divided into the following categories: physical contact, optical and electrical methods [4]. Physical contact method is one of the simplest ways of measuring the device temperature where the sensor includes thermocouples, scanning nanoprobes, thermographic phosphorus, etc. The measurements here rely on the transfer of heat from the DUT to the temperature sensor contacting the device. The advantage of this method is that it has a good spatial resolution. However, the disadvantage is that the surface of the device has to be accessible so that the contacting sensor can be mounted. Another disadvantage in this mode of measurements is that the dynamic response of these sensors is comparatively low [4]. There are also modules in which, to monitor the junction temperature, negative temperature coefficient (NTC) or positive temperature coefficient (PTC) resistors are placed inside the package. These thermistors have to be placed very close to the device for accurate measurements but several limitations have to be considered such as EMI, voltage isolation distance between two chips [5]. Usage of thermographic cameras is another method to measure the device temperature optically. It brings a new dimension of measurement when compared to the physically contacting ones i.e. no physical contact is necessary. The infrared (IR) cameras have a good spatial resolution and it is possible to have a two-dimensional temperature map of the semiconductors. But the main disadvantage is that the surface of the device must be exposed to the camera to capture the image and in return the temperature. Also, the setup becomes bulkier, complex to use and expensive for quick measurements.
