Cognitive radio has the ability to sense the environment and adapt its behavior to optimize communication features, such as quality of service in the presence of interference and noise. In physical layer to achieve this goal, different phases of sensing, channel estimation, and configuration selection are necessary. The sensing part measures the interference level, recognize the spectrum holes, and send this information to Channel estimator. In the next step, channel state information (CSI) is used for data detection and also sent to the transmitter through a limited feedback. CSI feedback consists of achievable rate, SNR value, modulation or coding schemes (MCS). Feedback link improves the system performance in the cost of complexity and delay. In this paper, we present and compare different feedback schemes for cognitive radio and study the channel capacity when an imperfect feedback link is corrupted by noise and delay.