Co-design of analog to digital interface and digital signal processing architecture

Conventional signal processing optimizes the operation of A/D conversion and digital signal processing as two separate tasks with two separate objectives and performance measure such as signal reconstruction and signal detection. Oversampling linear feedback coders can be viewed as simple precise encoders for analog signals. While oversampling technique has received wide attention, its full potential is not realized because of the lack of an optimum decoder. The objectives are design of new decoders and decimation filters to increase performance and/or feasibility of digital implementation of several signal processing tasks. A new decoding method is proposed in this dissertation which uses the inherent signal flow structure of an encoder to achieve an optimal decoding. The proposed decoding method can achieve the following advantages in comparison with the other decoding methods: (A) high resolution, much lower mean square error; (B) consistent output compared to the sinc filter or other linear decoding methods; (C) much shorter latency compare to FIR filters; (D) input noise reduction. A methodology for design and performance evaluation of a noise-shaping encoder/decoder system is introduced which is able to take into account distortion of input signal passing through a nonlinear encoder. Improving signal processing realizations of signal acquisition system and data communication system blocks such as analog to digital interface and embeded filtering, match filtering and symbol detection are the ultimate goals of our research. Having an A/D interface with a large dynamic range lets us transfer many signal processing tasks such as channel selection and interference rejection to the digital side in a digital receiver. This will reemphasize the importance of the efficient decimation filter design from overall performance point of view.