Design of Optimal Feedback Filters with Guaranteed Closed-Loop Stability for Oversampled Noise-Shaping Subband Quantizers
49th IEEE Conference on Decision and Control December 15-17, 2010 Hilton Atlanta Hotel, Atlanta, GA, USA Design of Optimal Feedback Filters with Guaranteed Closed-Loop Stability for Oversampled Noise-Shaping Subband Quantizers SanderWahls HolgerBoche. Abstract—We consider the oversampled noise-shaping sub- T T T T ] ] ] ] band quantizer (ONSQ) with the quantization noise being q q v q,q v q,q + Quant- + + + + modeled as white noise. The problem at hand is to design an q - q izer q q - q q ,...,w ,...,w optimal feedback filter. Optimal feedback filters with regard 1 1 ,...,v ,...,v to the current standard model of the ONSQ inhabit several + q,1 q,1 w=[w + w=[w =[v =[v shortcomings: the stability of the feedback loop is not ensured - q q v v and the noise is considered independent of both the input and u ŋ u ŋ -1 -1 the feedback filter. Another previously unknown disadvantage, q z F(z) q q z F(z) q which we prove in our paper, is that optimal feedback filters in the standard model are never unique. The goal of this paper is (a) Real model: η is the qnt. error (b) Linearized model: η is white noise to show how these shortcomings can be overcome. The stability issue is addressed by showing that every feedback filter can be Figure 1: Noise-Shaping Quantizer “stabilized” in the sense that we can find another feedback filter which stabilizes the feedback loop and achieves a performance w v E (z) p 1 q,1 p R (z) equal (or arbitrarily close) to the original filter.
[Show full text]