Characterization of frequency-dependent network impedance is vital for harmonic emission evaluation, stability studies as well as for the optimum design of power-line communication technologies. This work presents a simulation-based performance comparison of three techniques that are used to obtain the network harmonic impedance. Two broadband methods, the passive pseudo-random binary sequence and the active pseudo-random impulse sequence, and the single-frequency injection are investigated. The methods are first applied to a passive network and then extended to a network with nonlinear equipment such as power electronic converters. The methods' performance is evaluated in terms of mean impedance and sub-cycle impedance. While the results of all three methods are comparable for the passive network, it differs in the presence of nonlinear devices, especially for the sub-cycle impedance, with the single-frequency injection method demonstrating superior results. The passive-pseudo-random binary sequence method is inferior in the characterization of both average and sub-cycle impedance due to the absence of perturbations around the zero-crossings of the recorded line voltage. These results point to the need to carefully evaluate the perturbation techniques for accurate measurement of network impedance.