Diagnosing faulty cable network segments from modem power readings

This paper describes our investigation of the possible use of modem status data to discover faulty behavior in a cable network. Early analysis of the data revealed that the modem power signal had a daily variation which was congruent with the diurnal temperature variation. Based on this observation, we hypothesized early on that variations from this daily cycle could potentially indicate an abnormal (faulty) behavior. We developed specific metrics to quantify the magnitude of the aberration on a per modem basis. We have used this metric to develop a segment interest metric applicable to the segment level. We conclude that the proposed segment power interest metric derived solely from status information obtained from the installed modem base, was proven to be implied by a corresponding WSR interest metric. This is an important outcome, as it establishes the segment power interest metric to be a strong candidate to be used as predictor of the "health" of the corresponding segment.