Abstract
Power cycling lifetime in a semiconductor module accounts for the progressive fatigue of the device due to repetitive
thermo - mechanical stress. Mathematical models for the lifetime express the number of cycles that the device can withstand
before failing under predefined repetitive conditions. These models reveal an exponential dependency of the lifetime
with the amplitude of the swings of the chips junction temperature. Thus, an accurate estimation of the lifetime requires
a precise knowledge of these temperature swings.
During the design phase, the junction temperature time course over time is derived from the device losses with models
of the thermal propagation inside the device package. The standard approach is based on discrete thermal models synthesized
as a Cauer or Foster network equivalent. More advanced modelling techniques rely on Finite Element Methods
(FEM). The paper compares three common thermal modelling approaches regarding their influence on lifetime prediction.
In particular, it is demonstrated that these thermal models can present significant differences in predicting the cross
couplings terms between the chips in the same module. Copyright VDE Verlag GmbH
thermo - mechanical stress. Mathematical models for the lifetime express the number of cycles that the device can withstand
before failing under predefined repetitive conditions. These models reveal an exponential dependency of the lifetime
with the amplitude of the swings of the chips junction temperature. Thus, an accurate estimation of the lifetime requires
a precise knowledge of these temperature swings.
During the design phase, the junction temperature time course over time is derived from the device losses with models
of the thermal propagation inside the device package. The standard approach is based on discrete thermal models synthesized
as a Cauer or Foster network equivalent. More advanced modelling techniques rely on Finite Element Methods
(FEM). The paper compares three common thermal modelling approaches regarding their influence on lifetime prediction.
In particular, it is demonstrated that these thermal models can present significant differences in predicting the cross
couplings terms between the chips in the same module. Copyright VDE Verlag GmbH