Transients in reconfigurable digital signal processing systems

To solve measurement and control problems, the processing of input data is performed typically by model-based digital signal processing (DSP) systems, which contain a representation of our knowledge about the nature and the actual circumstances of the problem at hand. If the nature or the actual circumstances change, the corresponding model should also be changed. Similarly, if the amount of knowledge about the problem increases due to measurements, an improved model can be suggested which provides better performance. As a consequence, the real-time adaptation or reconfiguration of the DSP system to be applied can hardly be avoided. The transients caused by these adaptations/reconfigurations is investigated. It is shown that for feedback systems, i.e., for infinite impulse response (IIR) filters, these transients are strongly structure-dependent and that the socalled orthogonal filter structures also provide good performance in this respect.