One of the most important reliability effects observed in Metal-Oxide-Semiconductor Field-Effect Transistors (MOSFETs) is the threshold voltage shift when the device is stressed at high gate voltages at elevated temperatures, called the Bias Temperature Instability (BTI). Furthermore, two degradation mechanisms, the Positive (PBTI) and Negative Bias Temperature Instability (NBTI), are observed when the devices are stressed at positive or negative gate voltages, respectively.
In order to study BTI, a new method, Time Dependent Defect Spectroscopy (TDDS), has been recently introduced. For TDDS several stress and recovery traces are recorded on the same device, which allows for an extraction of the capture and emission times of an ensemble of defects. Very fast data acquisition equipment is necessary to obtain measurement data for this method. Evaluation and visualization of measurement results is done by detecting change points in TDDS measurement data. Out of several change point detection methods, three are selected.
The first algorithm uses the Discrete Wavelet Transform (DWT) and the Redundant Discrete Wavelet Transform (RDWT). After applying the DWT/RDWT to the measurement data, deionising techniques, such as hard and soft thresholding methods, are applied before re-transformation into the time domain. Finally the change points are extracted from the smoothed version of the measurement signal. The second algorithm is based on statistical data evaluation using histograms. Data is binned into histograms, Gaussian distributions are fitted to the histograms' peaks and a smoothed version of the signal is obtained by applying a maximum likelihood criterion.
The third, and most promising, method is a combination of cumulative sums and bootstrap analysis. After setting a detection sensitivity parameter, the steps of the measurement data are extracted. The so obtained information is visualized as a function of step height over emission time, also called spectral map. In the context of BTI modeling, spectral maps also reflect the occurrence of Random Telegraph Noise (RTN). Furthermore, the temperature and bias dependent defect behavior is investigated. A higher temperature leads to decreased emission times while the step amplitudes show only small variations.