This thesis treats several issues in the analysis, measurements and design of high-performance switching regulator systems. In Part 1 the high-frequency capabilities of two switching regulator modeling techniques, state-space averaging and discrete modeling, are compared, using the subharmonic instability in current-programmed regulators as a test. As a result of this comparison, a new small-signal, linear, time-invariant modeling technique, called sampled-data modeling, is developed. This new method retains the continuous form of state-space averaging but also possesses the accuracy of discrete modeling, and thus it combines the best features of the two techniques.
In Part II the newly developed sampled-data modeling technique is applied to the question of the interpretation of loop gain measurements in high-performance switching regulators. As a result of the inaccuracy of state-space averaging near one-half the switching frequency, conventional loop gain predictions are found to be inadequate in this regime. Sampled-data analysis is employed to develop new theoretical predictions which closely match various kinds of loop gain measurements in high-performance systems.
Finally, in Part III attention is turned to a potential sensitivity problem in high-performance switching regulators. For systems whose bandwidths lie close to one-half the switching frequency, even small changes in the operating environment of the power converter may result in dramatic degradation of the system's dynamic characteristics. Formulas for the estimation of this sensitivity are developed, and adaptive control, in which feedback gains are functions of certain important circuit quantities, is proposed as a means of curing the problem. Several different adaptive control schemes are proposed, and the implementation and testing of two of these strategies demonstrate their superiority over a conventional, non-adaptive design.
