The time and frequency resolution of a covariance-driven wavelet analysis technique for identifying the modal parameters of linear systems from random vibration response measurements was discussed. The technique starts with the calculation of structural covariance responses from random vibration response of the system. The covariance was then converted into time- frequency domain using a continuous Morlet wavelet transform. The phase angles and the magnitudes of the covariance wavelet coefficients along the ridges were used to estimate structural modal parameters. To set an enough frequency resolution for modal separation, the central frequency of the mother wavelet function was proposed to be adjustable. To deal with the end effect of wavelet analysis, the negative covariance response was estimated as end padding. The numerical study on a single-pylon double-cable-plane cable-supported bridge verifies that the time-frequency resolution of the wavelet analysis can be fixed by setting a reasonable central frequency of the analyzing mother wavelets and the proposed negative covariance estimation technique is of good capability to eliminate the end effect of the wavelet analysis.