Bayesian Optimization for High-Speed Channel Equalization

Equalization methods are used to recover the signal attenuated by channel loss. Different optimization algorithms are applied to find the best tap coefficients for each equalization that will improve eye opening and reduce bit error rate. The goal function for most equalization optimization algorithms is to reduce the difference between input and output signals, which is a linear optimization problem and can be solved relatively easily. This indirectly will increase eye height and improve eye diagram. Directly optimizing eye height is a non-linear problem and cannot be solved with analytical method. We are proposing FFE and DFE combined equalization optimization algorithm that optimizes directly eye height using Bayesian Optimization (BO). The proposed algorithm can be generalized for multi-level signals.