Predicting the effects of task jitter in digital control systems

Abstract

ABSTRACT Many digital controllers need an operating system to perform tasks such as networking and file operations. This causes an increase in task jitter which can affect the performance of the control system. Some researchers have claimed that Root-Mean-Square (RMS) of the task jitter is a good indicator of how the control system will be affected but there has been without evi- dence to show this to be true. In this paper, Gaussian, Uniform and a Brute force distributed task jitter, all with the same jitter RMS, were tested on a second order control system. The collected step response envelopes of 10,000 simulations for each jitter type show a very similar shape suggesting that task jitter RMS is a good indicator of control system performance. The overshoot distribution is shown to be normal which means that system performance can be characterised by a much smaller number of simulation runs.The work also shows that particular patterns can have a significant impact on overshoot and settling time. A real control system subject to task jitter may randomly and occasionally show extreme behaviour which may be outside acceptable limits. Simulation is essential to find these extremes early in the design process.