The paper describes the structure and rationale of the generalized software reliability process and a set of simulation techniques that may be applied for the purpose of software reliability modeling. These techniques establish a convenient means for studying a realistic, end-to-end software life cycle that includes intricate subprocess interdependencies, multiple defect categories, many factors of influence, and schedule and resource dependencies, subject to only a few fundamental assumptions, such as the independence of causes of failures. The goals of this research are dual: first, to generate data for truly satisfying the simplified assumptions of various existing models for the purpose of studying their comparative merits, and second, to enable these models to extend their merits to a less idealized, more realistic reliability life cycle. This simulation technique has been applied to data from a spacecraft project at the Jet Propulsion Laboratory; results indicate that the simulation technique potentially may lead to more accurate tracking and more timely prediction of software reliability than obtainable from analytic modeling techniques.>

A discussion is given on the use of software reliability models to guide engineering and management decisions on projects involving software. The focus is on experiences with using various software reliability models on actual projects, e.g. PBX project, space shuttle and software process control.>

A common intermediary connecting frequency-noise calibration or testing of an oscillator to useful applications is the spectral density of the frequency-deviating process. In attempting to turn test data into predicts of performance characteristics, one is naturally led to estimation of statistical values by sample-mean and sample-variance techniques. However, sample means and sample variances themselves are statistical quantities that do not necessarily converge (in the mean-square sense) to actual ensemble-average means and variances, except perhaps for excessively large sample sizes. This is especially true for the flicker noise component of oscillators. This article shows, for the various types of noises found in oscillators, how sample averages converge (or do not converge) to their statistical counterparts. The convergence rate is shown to be the same for all oscillators of a given spectral type.

Previous work has shown that the mean time from lock to a slipped cycle of a phase-locked loop is given by a certain double integral. Accurate numerical evaluation of this formula for the second-order loop has proved extremely vexing because the difference between exponentially large quantities is involved. This article simplifies the general formula to avert this problem, provides a useful approximation to a needed conditional expectation, and produces an asymptotic formula for the mean slip time that is moderately accurate even at low loop SNR (less than 7 dB) and small steady-state phase errors (less than 0.3 rad). The approximations extend to higher order loops, as well.