The Modulated Wideband Converter (MWC) is a blind sub-Nyquist sampling system used especially to monitor wideband spectrum. This system can be realized by existing analog components. From a theoretical point of view, all analog components are assumed ideal. However, this hypothesis is false in practice. Indeed, some imperfections are introduced by analog components such as nonlinearities of mixers, phase/attenuation/selectivity of low-pass filters and desynchronization between modulating waveforms. Consequently, it is necessary to correctly estimate the sensing matrix to ensure correct spectrum reconstruction performance. Conventional calibration methods are based on the measurements of different single tones. However, these approaches need to record several measurements with single-tone inputs. To avoid this problem, this paper presents a new hardware calibration method of MWC by using a single measurement of a white noise signal for radio frequency spectrum monitoring. A preprocessing method is performed to resynchronize input, output and modulating waveforms signals with each other. Our hardware calibration method is applied to our prototype of Compressed Sensing (CS) scheme in order to estimate the corrected sensing matrix. To prove the efficiency of our hardware calibration, reconstruction performances were evaluated with respect to the probability of correct reconstruction and false alarm criteria. These results are compared with those obtained with the theoretical MWC scheme, with our prototype without calibration and with reference calibration.