Experience: Redesigning Radio Test Specifications via Field Performance Correlations

Products manufactured in Telecommunications companies need to go through a large number of test cases, often running into thousands. These test cases have test specifications and test limits decided by designers from their understanding of the components and experience. Test limits used are usually binary thresholds so that if a manufactured unit is on the correct side of the test limit then the corresponding test is said to be passed. The impact of variation in test performance on field performance has not been extensively studied. To study this, we consider the field performance of a large number of Remote Radio Units (RRU) and try to correlate the field performance with the test performance. We describe here a system which can integrate data from production tests and field performance and run large number of statistical experiments automatically. These statistical experiments allow us to narrow down on a very small number of production tests which are clearly correlated to system performance. The process makes it feasible to bring down the number of production test cases which need further analysis and identify possible changes in thresholds. In particular, we look at downlink performance degradation and linearization faults as one set of field performance results and channel quality indicators as another. Initial results on our system show orders of magnitude difference in the number of test cases directly impacting field performance results to the actual number of tests performed. We believe that incorporation of such systems and checks would lead to better telecommunications equipment performance across the industry and therefore better end user experience.