Low resolution adc-dsp optimization based on non-uniform quantization and mlse for data centers interconnects

Abstract

A method for optimizing non-uniform quantization thresholds of an ADC in MLSE-based receivers in an optical communication channel, according to which a Quantized Noise (QN) distortion model, in which the quantization and the channel additive noises are combined is generated. The model is applied on the channel deterministic analog states x(n) and on sequences of analog states and transition probabilities are calculated, which will be used later on to calculate the BER, from channel deterministic states and sequences of channel deterministic states into the discrete ADC quantization regions. Real value outputs of the ADC are replaced by the transition probabilities and non-uniform quantization of the ADC is performed, with thresholds that are optimized for MLSE detection, to obtain maximal statistical separation. A DSP circuit computes, the MLSE metrics and the transition probabilities of the analog states into the quantized values, for each of the channel deterministic state; and an MLSE decoder post-processes transition probabilities replacing the ADC outputs and representing analog regions, based on the derived transition probabilities.