Cyclostationary Stochastic Jitter Measurement Method With Uncertainty Prediction.

This paper presents a completely new approach to measure the jitter at the output of a sample-and-hold circuit. A cyclostationary stochastic process is used to model the entire noise of the measurement system. This approach relies on the calculation of a cyclostationary variance, representing a complete description of the noise of the measurement system. This provides the basis for conducting powerful analyses of how the various noise sources contribute to the cyclostationary variance, enabling the user to gain a deep insight into the noise behavior of the system. A new jitter estimator is derived that provides some new and unprecedented features. First, it is robust against the second harmonic of the input signal, and second, it allows for the elimination of cyclostationary noise contributed by the data acquisition device. A stochastic model of the method is derived and a thorough analysis of the uncertainty of the proposed estimator is presented. This results in a formula that allows for the prediction of the measurement uncertainty and another formula that allows for the calculation of the required number of samples in order to meet a given uncertainty. The method proved to be accurate in Monte Carlo simulations and in measurements as well. A jitter in the range of 22 fs out of a 10-GS/s sample-and-hold integrated circuit was successfully measured with a standard deviation of 230 as. [ABSTRACT FROM PUBLISHER]