Symbol vs Block Level Precoding in Multi-beam Satellite Systems

Precoding techniques for multi-beam satellite systems have received a considerable attention in recent years as a tool to mitigate the interference among the beams, and hence increasing the throughput. Our goal is to compare two main categories of precoding schemes, namely, the conventional linear block level precoding and the symbol level precoding. Focusing on power minimization problem with signal to interference plus noise ratio (SINR) constraints, symbol level precoding (SLP) has significant gains with respect to the zero forcing (ZF). However, the lower transmit power is achieved with a price: A higher computational complexity. Therefore, several sub-optimal SLP techniques have been proposed in the literature to overcome the complexity. While ZF has the lowest complexity among the techniques chosen in this paper, it is not an optimal linear block level precoder as far as power minimization is concerned. Therefore, in order to have a more complete picture, one needs also to consider optimal block level precoders. Our results indicate that in order to have a fair comparison, one needs to consider two different scenarios, namely, low and high SINR threshold regimes. While for low SINRs the optimal linear block level precoding scheme may provide a good solution with reasonable complexity, for high SINR threshold, the SLP techniques become more attractive. Our results also indicate that the performance of SLP highly depends on the chosen constellation space, and therefore a final conclusion can be achieved only after appropriately optimizing the constellation set.