Use of the coherence function for a comparison of test signals for frequency domain identification

The coherence function has proved useful for aiding in the design of a frequency domain identification experiment. For consistent and reliable estimation of system parameters, making use of frequency domain identification techniques the main requirement is that the coherence function be close to unity and have a flat characteristic over a sufficiently wide bandwidth. For efficient estimation the input test signal length and peak factor, along with the type of signal which is best suited to the experiment are also important factors. The paper provides a comparison of these and other properties of commonly used test signals such as linear and logarithmic swept sinewaves, binary m-sequences and multi-harmonic test signals