A State-Space Framework for Parallelizing Digital Signal Processing in Coherent Optical Receivers.

Area of Science:

• Optical Communications
• Digital Signal Processing
• Control Systems Engineering

Background:

• Ultra-high sampling rates in coherent optical systems bottleneck real-time processing.
• Existing methods for parallelizing digital signal processing (DSP) are complex and ad hoc.

Purpose of the Study:

• To propose a unified state-space framework for systematically parallelizing DSP algorithms.
• To analyze parallelization-induced latency and its impact on system stability.
• To derive a design guideline for balancing throughput and bandwidth in parallel architectures.

Main Methods:

• Transforming algorithm transfer functions into state-space representations.
• Deriving parallel architectures using matrix operations.
• Introducing the parallel equivalent delay (PED) metric and the Throughput-Bandwidth Product (TBP).

Main Results:

• A parallel Costas carrier recovery loop was designed and simulated, validating the framework.
• The TBP limit was confirmed, demonstrating a trade-off between throughput and stable loop bandwidth.
• FPGA implementation achieved 15.625 Gsps throughput with 50 parallelization, low logic utilization, and confirmed theoretical limits.

Conclusions:

• The proposed state-space framework effectively parallelizes DSP algorithms for coherent optical systems.
• The TBP provides a crucial design guideline for stable high-throughput parallel architectures.
• The methodology offers significant advantages over conventional methods, especially in low-SNR conditions.