Design of Near-Optimal Coding Schemes for Adaptive Modulation with Practical Constraints

We consider a system with parallel flat-fading subchannels for transmission of data, similar to a multicarrier system, where the sub-channel states are known perfectly to both the transmitter and the receiver. The results presented so far in literature for this system have only considered maximizing the sum-rate with Gaussian constellations which is not realizable in practice. In this paper, we consider practical QAM constellations for transmission, the size of which can be varied across subchannels. Under this constraint, we derive the maximum sum-information-rate of the overall system and the power/rate allocation algorithm to achieve it, which has not been attempted before. A practical MIMO system can be resolved into parallel subchannels and we then extend the allocation algorithm to a MIMO case. We further constrain the system to use a single overall codebook which is more practical and optimize the proposed power/rate allocation algorithm under this constraint. The simulations with an LDPC code show that the proposed power/rate allocation method is very robust and the code performance is within 2dB of even the unconstrained Gaussian sum-rate limit for both cases.