interference alignment configurations for distributed crosstalk cancellation in g.fast

In this paper, we evaluate various system configurations for crosstalk cancellation using interference alignment (ia) in uncoordinated and partially coordinated downstream g. fast. in the first configuration, which can be implemented in a distributed fashion, signal-level coordination between users is unnecessary and ia is applied in the frequency domain. in the second configuration, partial signal-level coordination is available; however, it is entirely consumed by vectoring and ia is applied in the frequency domain. in the third configuration, partial signal-level coordination is available, but vectoring is not applied. instead, ia is applied jointly in both the space and frequency domains. considerably higher sum bit rates are achieved using the third configuration; however, the computational complexity is noticeably higher than the other two configurations. to tackle this problem, we propose a joint partial vectoring design based on the third configuration which outperforms the diagonalizing pre-compensator (a.k.a. zero-forcing precoder) considerably. then we employ it in the second configuration to obtain a high-performance crosstalk cancellation scheme with much lower complexity. our simulation results show that depending on the level of coordination, ia increases the achievable rates of g. fast loops significantly compared to available solutions with comparable complexity. in this paper, we evaluate various system configurations for crosstalk cancellation using interference alignment (ia) in uncoordinated and partially coordinated downstream g. fast. in the first configuration, which can be implemented in a distributed fashion, signal-level coordination between users is unnecessary and ia is applied in the frequency domain. in the second configuration, partial signal-level coordination is available; however, it is entirely consumed by vectoring and ia is applied in the frequency domain. in the third configuration, partial signal-level coordination is available, but vectoring is not applied. instead, ia is applied jointly in both the space and frequency domains. considerably higher sum bit rates are achieved using the third configuration; however, the computational complexity is noticeably higher than the other two configurations. to tackle this problem, we propose a joint partial vectoring design based on the third configuration which outperforms the diagonalizing pre-compensator (a.k.a. zero-forcing precoder) considerably. then we employ it in the second configuration to obtain a high-performance crosstalk cancellation scheme with much lower complexity. our simulation results show that depending on the level of coordination, ia increases the achievable rates of g. fast loops significantly compared to available solutions with comparable complexity.