Abstract: A variation in the cancellation coefficents of an echo cancelling modem indicate a change in the impedance of a line. In an echo canceller with T spaced samples (i.e. producing a cancellation signal at the data rate) the resolution is too poor to provide satisfactory identification of faults. A set of coefficients is stored and the sampling phase adjusted to produce a second set. This may be repeated producing eight sets in the preferred embodiment. A higher resolution picture of the line impedance is built up allowing comparisons with an earlier or optimum set of values. The test facility may be included in an operating modem providing line test facilities when not transmitting data.