Real - Time Distributed Implementation of Interference Alignment with Analog Feedback

Interference alignment (IA) is known to achieve the maximum capacity for the interference channel. IA is a precoding technique to make interfering signals be aligned at receivers. The resulting sum rate is linearly scaled with the number of users. The performance of IA, however, depends on the practical issues such as the performance of synchronization, channel estimation and feedback. In this paper, a prototype is implemented for the IA system with three users. There have been IA prototypes in recent years, but the previous prototypes have not considered the distribution of the nodes in IA network. The nodes are physically distributed in our prototype not sharing their time and frequency references with any other, thus working independently, which enables the experimental study of IA under the most practical setup. For the distributed system, the over-the-air schemes for time and frequency synchronization and analog feedback are studied and implemented. According to the measurement from our prototype, it is shown that IA achieves the sum rate from the previous analysis on imperfect channel information. In addition, some other measurements are performed considering the accuracy of IA solution, synchronization of nodes, and CSI feedback. For the accuracy of IA solution, the performance of IA versus the number of iterations for an iterative IA method is measured. For synchronization accuracy, the performance with different residual frequency offset is measured. Finally, for the CSI feedback quality, analog feedback and scaler quantization-based limited feedback is first compared, and then the performance with Doppler frequency offset is also measured.