In this paper, we consider cognitive radio network in which two cognitive radio sources communicate with two cognitive destinations via a relay node. The decode and forward (DF) relay node employs physical layer network coding (PLNC) to improve the data rate. Based on the availability of the spectrum bands at the source, relay and destination, the network employs three different diversity schemes namely source to relay diversity, relay to destination diversity and combination of earlier two diversity schemes with overall source to destination diversity schemes. The optimal allocation of channel and power with per band and sum power constraints of a node in the network is formulated as convex optimization problem to improve the end to end throughput of the cognitive radio network. Simulation results show that the resultant joint channel and power allocation are superior to the equal power allocation in terms of both end to end throughput and outage probability.