Best Paper Award at EDAPS 2016

EDAPS 2016 Best Paper Award (large view)

EDAPS 2016 Best Paper Award

(15-12-2016) Researchers from IDLab-imec, together with Amphenol-FCI, have received a Best Paper Award at the IEEE Electrical Design of Advanced Packaging & Systems (EDAPS) Symposium 2016 in Honolulu, Hawaii.

This paper introduces a new way of modeling variability in electronic devices, by applying machine learning techniques in electrical engineering.

Sources of variability in the responses of electronic devices are often introduced by its manufacturing process. Those sources can be unknown, their distribution can be unknown, or they may be too numerous to efficiently model their impact on the variability of the device response. Moreover, to build models, the knowledge of many responses (samples) is required. As manufacturing and measuring devices can be costly, and simulation is often very time-consuming, the modeling often becomes intractable.

The method proposed in the paper avoids these issues by building a model based on a small set of responses obtained by measurement or simulation for random instances of the device. The model built this way can be used to generate physically consistent responses, the distribution of which closely matches the distribution of responses that would be obtained from a large number of measurements or simulations. As such, the generated responses can be used to investigate the variability of the device in a much cheaper and less time-consuming way.

Read more articles about: