Study on power cycling test with different control strategies
Introduction
As an important reliability assessment method, power cycling test is the cornerstone in the lifetime estimation procedure. Power cycling test results are usually extrapolated to the real application condition to calculate expected lifetime. Therefore, as power cycling test results, not only the number of cycles to failure Nf but also test condition and test method should be documented. Taking the test of IGBTs as example, test conditions of the power cycling test normally refer to gate-emitter voltage VGE, junction temperature swing ∆Tj, mean junction temperature Tjm, load current IL, load pulse duration ton and load pulse off duration toff.
In 2010, Schuler and Scheuermann have first published results about the impact of control strategies on power cycling lifetime [1, 2]. It was shown that by changing the control strategy, power cycling results of devices of the same type under the same start test conditions could even vary by a factor of three. In 2014, Sarkany et al. have confirmed the impact of control strategy by using IGBT chip itself as temperature sensor [3, 4]. From both research groups, there was only one parameter used for one control strategy. It is believed that adjusting different parameters in order to compensate different degradation mechanism of the device will lead to different results. For standard IGBT power modules, chip topside Al bond wire connection and soft solder layers are the main weak points in power cycling test.
Section snippets
Design of experiments
In the performed power cycling tests, heat sink temperature Ts was measured 2 mm beneath the surface of the adapter plate with thermocouples in order to determinate the thermal resistance Rthjs (junction to heat sink). The IGBT chip temperature was measured by using the forward voltage drop of the collector side pn-junction at a measurement current of 100 mA. Due to the required time for the recombination of excess charge carriers after conducting a high load current, a measurement delay of
Test results
The different control strategies were realized by a PID controller written in LabVIEW. The chosen parameter was adjusted every 20 cycles according to measurement data from the last 20 cycles. For each control strategy, at least two devices were tested in order to reduce the impact of fluctuation of test results on the effect to be investigated.
Post-test analysis
In this work, all devices failed with an increase of Rthjs by 20% due to degradation of chip solder (see Fig. 12). A summary of all test results is given in Table 1. From the same group, devices failed at comparable cycles. Only the device 5_3 and 9_4 are conspicuous. As already mentioned before, since the temperature swing of device 5_3 is about 2 K lower than 11_4, the achieved lifetime of 5_3 is therefore slightly higher than 11_4. For the device 9_4, a strong fluctuation of ton can be seen
Conclusions
It is confirmed in this work that control strategy has a great impact on power cycling lifetime. By using different parameters to compensate the degradation, different results can be produced. It is found that using ton to compensate solder degradation is most effective. For bond wire failure, load current is believed to be more effective as shown in this work for the solder failure. Further investigation on the impact of control strategy in packages with bond wire failure is of great interest.
References (13)
- et al.
Power cycling results for different control strategies
Microelectron. Reliab.
(2010) - et al.
Requirements in power cycling for precise lifetime estimation
Microelectron. Reliab.
(2016) - et al.
Impact of load pulse duration on power cycling lifetime of Al wire bonds
Microelectron. Reliab.
(2013) - et al.
Impact of test control strategy on power cycling lifetime
- et al.
Thermal transient analysis of semiconductor device degradation in power cycling reliability tests with variable control strategies
- et al.
Comparison of different power cycling strategies for accelerated lifetime testing of power devices