Date of Graduation

12-2010

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Microelectronics-Photonics (PhD)

Degree Level

Graduate

Department

Graduate School

Advisor

Simon S. Ang

Committee Member

Randy Brown

Second Committee Member

Jia Di

Third Committee Member

Ricky D. Hudgens

Fourth Committee Member

Kenneth Vickers

Abstract

Production testing of Radio Frequency (RF) devices is challenging due to the complex nature of the tests that have to be performed to verify functionality. In this dissertation a methodology to replace the complex and expensive RF functional tests with defect-oriented Built-in Self Tests (BiSTs) is detailed. If a design has sufficient margin to RF specifications then RF tests can be replaced with structural tests using a new data analysis technique called quadrant analysis, which is presented. Data from the analysis of over one million production units of said System on Chip (SoC) is presented along with the results of the analysis. The BiST techniques that have been used are discussed and a Texas Instruments 65 nm RF SoC with a Bluetooth and a FM core was used as a case study. The defect models that were used to develop the BiSTs are discussed as well. The scenario in which a design does not have sufficient margin to specification is also discussed. The data analysis method required in such a case is a regression analysis and the data from such an analysis is shown. The results prove that it is possible to replace expensive RF conventional tests with structural tests and that modern RFCMOS process technology and advances in design like the Digital Radio Processor (DRPTM) technology enable this. The Defective Parts Per Million (DPPM) impact of making this replacement is 27 units and is acceptable for RFCMOS high volume products. Finally, data showing test cost reduction of about 38% that resulted from the elimination of RF conventional tests is presented.

Share

COinS