| 摘要 |
DSL, crosstalk reduction. When using DSL vectoring like MIMO, diagonalization of the channel matrix is achieved through joint signal precoding at Tx and joint signal post-coding at Rx, implying perfect channel estimation. Parallel channel estimation is performed where all users send pilot signals concurrently. Orthogonal pilot sequences are therefore necessary.
However with new higher spectrum used, crosstalk increases and thus the precompensation values as well that may create PSD mask violation at Tx and thus clipping. To avoid this, Non Linear Precoding (NLP) based on modulo arithmetic (Tomlinson Harashima and also Cioffi for vectored DSL) is widely used in DSL. However NLP when applied, destroys the mutual orthogonality of pilot sequences.
Solution: A new so-called Generalized Walsh-Hadamard pilot sequence matrix is proposed, with sequence scaling that guarantee that the precompensated signal stays within modulo bounds preserving orthogonality of pilot signals. |
