Quantization and Noise-Shaped Coding for High Efficiency Transmitter Architectures

Sammanfattning: Three driving parameters in the development of transmitter technologies for wireless communication today is: energy efficiency, flexibility and footprint reduction. This thesis investigates the use of 1-bit oversampled signal representation, mainly as a power efficiency enhancement technique for the transmitter, but also as a path toward reduced footprint and increased flexibility. Common transmitter architectures comprising this representation of the signal on either baseband or RF level are discussed and analyzed. Two common methods of mapping a complex baseband signal to different characteristics of a pulse-train, Pulse Width Modulation and Pulse Density Modulation are discussed. Noise-Shaped Coding, an inherent feature of Pulse Density Modulation, is illustrated and discussed. The implementation of Pulse Density Modulation and the determination of the Noise-Shaped Coding properties are explored within a class of oversampled A/D-converters commonly referred to as the ??-modulators. A few different approaches of analyzing the ??-modulator is discussed and a novel ??-modulator representation is introduced. Further on, an algorithm designed to achieve arbitrary Noise Shaped Coding by optimized selection of the loop-filter coefficients in the ??-structure is developed. This algorithm uses a Monte Carlo-based technique in combination with a differentiable approximation of the quantizer element and a damped Gauss-Newton search. Finally, a novel method for quantization noise suppression close to the RF-carrier is proposed. The method is based on the superposition of a selective amplitude variation onto the quantized signal. Simulations are performed on an equivalent baseband model comprising a generalized memory polynomial model for linearization. The method is then successfully demonstrated with measurements in a carrier bursting transmitter working on a high efficiency, 10 W Si-LDMOS power amplifier.