A Cyclostationarity-Based Method for Time Interval Error Extraction Without Clock Recovery in SerDes Links

Precise jitter identification and extraction are critical for jitter decomposition, fault diagnosis, and bit error rate (BER) estimation in high-speed serial communications. Conventional methods depend on decision threshold and clock recovery, which discard significant signal information, as only threshold crossing points are considered. In addition, the accuracy of jitter extraction is sensitive to errors introduced by histogram-based statistical computations and interpolation processes. This article presents a novel jitter extraction method based on a signal transmission model that fully exploits the cyclostationarity characteristics of the signal. By utilizing all signal points rather than just threshold crossings, this method eliminates the presetting modules and avoids statistical computations or interpolations, resulting in a simpler implementation and lower complexity. A theoretical model is developed to demonstrate that the phase of the cyclic autocorrelation function at the symbol rate frequency intrinsically encodes the jitter information for each symbol. The time interval error (TIE) jitter can be extracted by computing the phase via the fast Fourier transform (FFT), followed by least-squares correction to eliminate cumulative errors. Simulations under various jitter conditions validate the accuracy and robustness of the proposed method, showing comparable performance to conventional approaches while achieving superior accuracy in sinusoidal jitter scenarios. The effectiveness is further demonstrated through oscilloscope experiments, with relative errors below 1%, confirming its accuracy and practical applicability.