NIST collaborates on study of adaptive characterization of jitter noise in sampled high-speed signals.

Researchers at NIST have developed an adaptive functional data analysis method to characterize random timing errors in high-speed sampled signals measured by sampling oscilloscopes. On average, jitter noise blurs a measured signal and reduces its power in the spectral domain. However, given an estimate of the standard deviation of the jitter noise, one can correct the measured power spectrum. This correction dramatically reduces prediction error and is essential for characterizing the impulse response function of high-speed sources and detectors in optical fiber communications systems at high frequencies (up to 50 GHz).

To estimate the standard deviation of jitter noise, many (hundreds to thousands) jittered signals are measured. Researchers first align the signals based on estimates of time translation errors, determined from a cross-correlation analysis of all possible pairs of signals. Based on the time-varying sample variance of the aligned signals and estimated derivatives of the noise-free signal at each time sample, they estimate the standard deviation of the jitter noise. Since there is no analytic model for the noise-free signal, the derivative is estimated using a regression spline model for the signal. Due, in part, to the non-linear nature of the estimation procedure, the initial estimate of the standard deviation of the jitter noise is biased. A computationally intensive Monte Carlo resampling scheme called the parametric bootstrap is used to estimate this bias. Based on the bootstrap estimate of bias, the jitter standard deviation estimate is adaptively corrected.

The results of this study will appear in an upcoming edition of the IEEE Transactions on Instrumentation and Measurement.

CONTACT: Kevin Coakley, [phone]; [e-mail].