Electrical & Computer Engineering     |     Carnegie Mellon

Wednesday, December 7, 12:00-1:00 p.m. HH-1112


Sounil Biswas
Carnegie Mellon University

Test Compaction for Non-Digital Devices Using Statistical Learning Methods

Successful deployment of non-digital devices greatly depends on cost-effective methods of testing them. However, in contrast to digital circuits, there are no universally accepted fault models for these non-digital components. Consequently, verifying the device specification continues to be their most accepted form of testing. However, applying all possible specification tests to a non-digital device is quite expensive. To reduce cost, we propose to use statistical learning methods for identifying and eliminating redundant tests. A test is deemed redundant if its outcome can be reliably predicted using the outcomes of other, so called kept tests. Application of the proposed method to a commercial accelerometer revealed its feasibility and accuracy. Specifically, significant test cost reduction was achieved by eliminating the MEMS shaker tests at hot and cold temperature when the proposed methodology was used to analyze the results from a set of fully tested parts for a commercial accelerometer.


Sounil Biswas received his B.Tech degree in Electrical Engineering from Indian Institute of Technology, Kanpur in May 2002 and his M.S. degree in Electrical and Computer Engineering from Carnegie Mellon University in May 2004. Currently he is pursuing his PhD degree under the guidance of Prof. R.D.(Shawn) Blanton. His research interests include digital, mixed-signal and microelectromechanical system test, statistical modeling for device performance and test cost analysis, and delay analysis and test.