Spectrum scarcity and the inefficient use of the electromagnetic spectrum motivated the development of Cognitive Radio (CR), which aims to extend the spectral efficiency, with opportunistic access to the available frequency bands. Energy Detection (ED) is the most adopted spectrum sensing technique for cognitive radio applications due to its simplicity. However, fading effects are usually simplified or discarded when evaluating the energy detector performance in spectrum sensing. This paper presents the performance evaluation of cognitive spectrum sensing, based on energy detector, for fading channels. The objective is to analyze how the use of various fading models affect the spectral detection in cognitive networks. Rayleigh, Rice, Nakagami-m and Lognormal fading channels are considered, and the results favor the use of ED when the channel is subject to Lognormal fading.