Productivity and code quality improvement of mixed-signal test software by applying software engineering methods

Typical nowadays mixed-signal ICs are approaching 1000 or even more parametric tests. These tests are usually coded in a procedural or a semi-object oriented language. The huge code base of the programs is a significant challenge for maintaining code quality which inherently translates into outgoing quality. The paper presents software metrics of typical mixed-signal power management and audio devices with regard to the number of tests conducted. It is shown that classical ways to handle test programs are error prone and tend to systematically repeat known mistakes. The adoption of selected software engineering methods can avoid such mistakes and improves the productivity of the mixed-signal test generation. Results of a pilot project show significant productivity improvement. Open-source based software is employed to provide the necessary tool support. They establish a potential roadmap to get away from proprietary tester specific tool sets