ERTG: A test generator for error-rate testing
Shideh Shahidi, Sandeep Gupta · 2007
A new concept called intelligible testing has been recently proposed to improve yields for a class of error tolerant systems, including audio, speech, graphics, video, and digital communications. Error rate, defined as the percentage of clock cycles for which the value at a circuit's outputs deviates from the corresponding error-free value, has been identified as a key measure of error severity. In error-rate testing, every fault that causes error rate greater than a threshold specified by the application is called an unacceptable fault; all other faults are called acceptable. The objective of error-rate testing is to detect every unacceptable fault while detecting none, or a minimum number, of the acceptable faults. In this paper we present a test generator for combinational circuits with the objective of detecting every unacceptable fault while minimizing the number of acceptable faults detected. Our test generator embodies new algorithms and cost functions that capture the properties of acceptable and unacceptable faults. Via experiments on several benchmarks we show that our approach can provide coverage of all unacceptable faults (high test quality) while significantly reducing the number of acceptable faults detected (high acceptance rate). In this manner, our approach helps improve yield via error-rate testing.