Recently the rapid size reduction of electronic devices have taken place due to the high pace of development in the electronics industry. However, the power consumption of devices has not changed, so solving the thermal issues has become increasingly important. Secondly, the number of components in the electronic devices are greatly increased, therefore the power densities have also substantially increased. For this reason, monitoring the thermal behavior of devices got an important role, cooling should be planned more carefully than ever before. However, the thermal measurement of electronic devices there is more difficult because the design for thermal testability is still not a top priority.
I became familiar with a measurement method during my internship, which can measure the temperature rise of a semiconductor in time using the device’s electric parameters only. From the completed measurement we can calculate a so-called structure function, which describes the thermal system by a thermal resistance-heat capacity model.
Unfortunately, at the beginning the thermal transient of curve in each case an electric effect can be observed, which appears when switching between two power levels. This electric effect covers the initial phase of thermal transient, and without knowing it the exact temperature of the device can not be determined. The current industrial methods make a simple correction by cutting the first stage of the thermal measurement, but it is inaccurate, and the correction requires manual step, therefore it is uncertain.
My thesis aims to develop a methodology to accurately reproduce the thermal transient curve. I approached this problem using the combination of thermal transient simulation, as well and fitting the simulated results to the measured curve as we know that the second half of the measured curve certainly reflects the reality. Thus, we can assume – if we can simulate the measure with appropriate boundary conditions – that the response function result has to be same as the measured. When we use simulation there is no electrical transient of course, so we can examine what happened with the device in this interval which we have not seen. If the resulting simulation curve and the measured curve are connected in the right place, you can obtaine a combine response function, which has its initial stage equal to the beginning of the simulation curve, while his other half is equal to the measured curve.
Knowing the device’s geometry and material properties a simulation model can be prepared easily.
This principle has been confirmed experimentally in my thesis.