Conference Paper

A new cycle test system emulating inductive switching waveforms

KAI-Kompetenzzentrum Automobil- und Industrieelektronik GmbH, Villach
DOI: 10.1109/EPE.2007.4417742 Conference: Power Electronics and Applications, 2007 European Conference on
Source: IEEE Xplore

ABSTRACT Cycle life testing of smart power switches requires significant hardware effort to provide the required ohmic-inductive load patterns. A new reliability test system for research purposes is therefore introduced that generates arbitrary current waveforms to emulate inductive switching behavior. This allows flexible cycle stress testing of integrated power switches under arbitrary application conditions. The current drivers of the proposed "ARCTIS" test system are protected from thermal overload in case of failure of a stressed device using a combination of case temperature sensing and a thermal equivalent circuit. Therefore the power MOSFETs in the output stage may be utilized to the limits of their dynamical safe operating area. All devices under test are continuously monitored for short circuit and open load failures. The respective waveforms and failure events are digitally recorded by the PXI-based control system to obtain a statistical basis for the evaluation of cycle life time.

0 Bookmarks
 · 
42 Views
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper Bayesian networks are used to predict complex semiconductor lifetime data. The data of interest is a mixture of two log-normal distributed heteroscedastic components where data is right censored.\newline\indent To understand the complex behavior of data corresponding to each mixture component, interactions between geometric designs, material properties and physical parameters of the semiconductor device under test are modeled by a Bayesian network. For the network's structure and parameter learning the statistical toolboxes \textit{BNT} and \textit{bayesf Version 2.0} for MATLAB have been extended. Due to censored observations MCMC simulations are necessary to determine the posterior density distribution and evaluate the network's structure. For the model selection and evaluation goodness of fit criteria such as marginal likelihoods, Bayes factors, predictive density distributions and sums of squared errors are used.\newline\indent The results indicate that the application of Bayesian networks to semiconductor data provides useful information about the behavior of devices as well as a reliable alternative to currently applied methods.
    BAYSM2013; 06/2013
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The problem of contemporary semiconductor reliability testing is twofold: on one hand demands on the device lifetime increase steadily implying longer testing times and on the other hand resources are limited (devices, testing time, ...). Therefore it seems unavoidable to apply advanced statistical methods to gain a reliable lifetime model. To increase the model quality significantly, we propose a combination of optimal Design of Experiments (DoE) and Bayesian statistical modeling. Optimal DoE ensures that the data for the model contain as much information as possible, whereas Bayesian modeling provides the possibility to include available prior information. With this approach resources can be saved because lifetime testing can be reduced to a necessary minimum.
    SCo 2013, Milano; 09/2013
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Based on physical failure inspection, known physical relationships, correlation analysis and posterior predictive distributions, a valid model for semiconductor lifetime is developed. It is shown that the data follow a mixture of two normal distributions and that the mixture weights depend on the destruction level. The mean lifetime of each component is modeled with an extended Coffin-Manson model. As priors, a combination of normal and hierarchical inverse gamma distributions is used. Since the given data show censoring, this is not a conjugate setting and the posterior distributions are simulated with MCMC methods. Model verifi�cation, based on the posterior predictive distribution, con�firms good prediction quality for interpolations and shows potential for improving extrapolations.
    SMTDA 2012: Stochastic Modeling Techniques and Data Analysis; 06/2012