High-performance configurable Zynq-based framework with Pseudo-Ping-Pong mechanism and DMA zero-copy strategy for Raman distributed temperature sensing system

The core component of a Raman distributed temperature sensing (DTS) system is a high-performance and low-cost signal processing unit, which is utilized to extract high signal-to-noise ratio (SNR) signals from original high-frequency sampled signals with high noise levels. To address this requirement, we propose a configurable Zynq-based signal processing framework incorporating the Pseudo-Ping-Pong data processing mechanism and the DMA zero-copy strategy. The proposed Pseudo-Ping-Pong mechanism is designed for the digital filter unit in the programmable logic section, which can fully leverage the parallel processing capability of the FPGA, thereby significantly improving processing performance while drastically reducing BRAM resource consumption. The proposed DMA zero-copy strategy is implemented in the processing system section to enhance the real-time data processing performance of the Linux system. Experimental results based on the constructed Zynq-7020 experimental platform demonstrate that, for 50,000 sampling points, the proposed framework improves the signal SNR from 11.28 to 31.86 dB via mean filtering (with 10,000 accumulations). The maximum processing time is 2.2 s, which meets the industrial safety standards (e.g., fire detection profiles). The mean absolute error of temperature is 0.23°C, and the maximum error is 0.41°C. The comprehensive performance of our proposed framework outperforms existing state-of-the-art methods. The proposed hardware-software co-design framework not only achieves high signal processing performance but also provides a scalable Linux-based application platform, making it a cost-effective solution for Raman DTS systems.