Low-Resolution ADC Quantization for MLSE Optical Receivers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-speed and high-resolution ADCs in optical communication systems face challenges with excessive power consumption and high costs, and low-resolution ADCs introduce significant quantization distortions that degrade Bit-Error-Rate (BER) performance, particularly in datacenter interconnects.
Innovation Solution
A method for optimizing non-uniform quantization thresholds of ADCs in MLSE-based receivers, using a Quantized Noise distortion model to combine quantization and channel additive noises, calculating transition probabilities, and performing non-uniform quantization to achieve maximal statistical separation and minimize BER, with DSP circuits computing MLSE metrics and post-processing transition probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution ADCs are used, then measurement precision is improved, but power consumption and cost increase
Solution Approach 1:
The patent applies non-uniform quantization by changing the parameter distribution of quantization thresholds from uniform to non-uniform, optimized specifically for MLSE detection. This allows low-resolution ADCs (1-4 bits) to achieve effective high-resolution performance by concentrating quantization levels where the signal probability density is highest, thereby maintaining measurement precision while using fewer bits and reducing power consumption
2Use of energy by moving object
If low-resolution ADCs are used, then power consumption is reduced, but quantization distortion increases and BER performance degrades
Solution Approach 1:
The patent transforms the quantization parameter distribution from uniform to non-uniform, with thresholds optimized for MLSE detection. This parameter change ensures that quantization distortion is minimized in the regions where the signal is most likely to occur, thereby maintaining BER performance even with low-resolution ADCs
Solution Approach 2:
The patent creates a statistical copy of the continuous signal distribution through non-uniform quantization levels. By optimizing the quantization thresholds to match the signal probability density function, the discrete quantized values effectively represent the continuous signal statistics, allowing MLSE to achieve accurate detection despite coarse quantization
3Ease of manufacture
If uniform quantization is used, then implementation is simplified, but BER performance is suboptimal for low-resolution ADCs
Solution Approach 1:
The patent changes the quantization thresholds from uniformly spaced to non-uniformly spaced values optimized for MLSE detection. This parameter transformation improves BER performance by aligning quantization levels with the signal probability density function, while the optimization process itself can be pre-computed and stored, keeping implementation complexity manageable
Data Source
AI summary
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.