Neural Network-Based Compensation of Time Base Jitter in High-Speed Equivalent Sampling Oscilloscopes

The sampling clock of the equivalent sampling oscilloscope is typically not provided directly; rather, the clock signal of the same frequency and phase must be recovered from the data signal. In the process of sampling and reconstructing high-frequency signals, multiple modules are passed through, all of which may introduce the time base jitter during signal processing. Therefore, it is necessary to reduce its impact by technical means and accurately estimating and compensating for this error is a challenging task. Recently, Bilal Moussa implemented time-based correction by employing a sampling technique through two reference channels that exhibited one phase difference of about 90 degrees, thereby reducing jitter to a level of around 260 fs. Concurrently, Jiangmiao Zhu enhanced this orthogonal distance regression algorithm through the adoption of an innovative weight distribution approach, which was subsequently utilized for the estimation of time jitter in sampling oscilloscopes. However, the time base jitter can be even smaller. To rectify the time base discrepancies in wideband sampling oscilloscope readings, an optimized back propagation (BP) neural network, enhanced by the sparrow search algorithm, is utilized for the analysis and compensation of oscilloscope's sampled signals. After simulation verification, the neural network showed a reduction in jitter to 221 fs.