Step Stress Accelerated Life Testing methods for lifetime predictions of SAW and BAW filters

摘要

We present unique step stress test and data analysis methods for evaluating power handling of acoustic devices. Our test system characterizes SAW and BAW filters at highly stressed conditions of rf power and ambient temperature T_(amb), stepping and holding power until parts fail while maintaining constant inputs of T_(amb), frequency, and source and load impedances. The system monitors each part at each power step recording large-signal output delivered power, input reflected power, and harmonics, while simultaneously measuring small-signal s-parameters over frequency, capturing high-speed changes in output power, and taking either high-speed optical or infrared images. Small and large signal data and images are used to determine the failure power step and the failure time interval for each part. Analyzing the stress and failure data, we determine parameters to a failure time model. The model is composed of a lognormal distribution for the failure times and a nonlinear relationship linking the failure times to the thermal and power stresses. In the underlying model, an Arrhenius term includes a self-heating temperature dependent on both power and T_(amb). To estimate the model coefficients, we use the method of maximum likelihood ML. In the ML method, the likelihood is the joint probability function of the failure times and the nonlinear relationship. The model coefficients are a constant, a thermal activation energy E_a, a power acceleration coefficient n, and a lognormal shape factor. Across technologies, manufacturing, design and layout, operating bands, and the various input conditions, we have determined model parameter estimates for E_a ranging from 0.6eV to 2.3eV and n ranging from -2.9 to -17.8. From the lifetime model and extracted parameter estimates, we make lifetime predictions of the SAW and BAW filters under reasonable powers and temperatures. We use various techniques to determine confidence bounds on the model parameters and the predicted lifetimes. Our test system and analysis methods can be used in device research and product development and can be easily extended to other acoustic and non-acoustic devices.