PhD defense Evelyn Oliveira: Highly Digital Delta-Sigma Transmitters
Télécom Paris, 19 place Marguerite Perey F-91120 Palaiseau [getting there], amphi 2 and in videoconferencing
Jury
- Prof. Chadi JABBOUR, Professor, Institut Polytechnique de Paris – Télécom Paris, France (PhD Supervisor);
- Prof. Yves MATHIEU, Professor Emeritus, Institut Polytechnique de Paris – Télécom Paris, France (Co-supervisor);
- Prof. Lirida NAVINER, Professor, Institut Polytechnique de Paris – Télécom Paris, France (Examiner);
- Dr. Eric WANTIEZ, Research Engineer, Thales, France (Examiner);
- Prof. Emmanuel BOUTILLON, Professor, Université Bretagne Sud, France (Examiner);
- Prof. Antoine FRAPPÉ, Professor, JUNIA, France (Reviewer);
- Prof. Philippe BENABES, Professor, CentraleSupélec, France (Reviewer).
Abstract
This work addresses the design of highly digital delta-sigma-based transmitters for future 5G/6G wireless systems, where wide signal bandwidths, high spectral efficiency, and power-efficient digital-to-analog conversion are key requirements. In this context, delta-sigma modulation enables the use of lower-resolution Digital to Analog Converters (DAC) through oversampling and noise shaping, but it also introduces three major challenges: i) the implementation of delta-sigma modulator (DSM) architectures capable of operating at very high sampling rates, ii) the suppression of out-of-band noise, which degrades Adjacent Channel Power Ratio (ACPR) and increase power-amplifier constraints, and iii) the co-integration of the full system including the Digital Pre-Distortion (DPD) block.
To address the speed limitation, a high-throughput DSM architecture based on zero-padding (ZP), temporal unrolling, and multi-stage processing were proposed. Synthesis results in 22 nm FDSOI show that the proposed architecture can reach up to 100 GS/s, with only 0.5 dB SNDR degradation and 100 mW power consumption compared to a classical second-order DSM, demonstrating a promising trade-off between spectral performance and speed. For the second challenge, a DSM/FIR-DAC co-design methodology is proposed and evaluated within a complete transmitter chain, including realistic communication signals and system-level metrics such as ACPR, EVM, and out-of-band power. The results show that the DSM noise-shaping profile and the FIR-DAC response must be considered jointly. For a 400 MHz four-carrier aggregated 64-QAM signal, the optimized design improves the adjacent-channel ACPR by 2.86 dB and the alternate-channel ACPR by 16.19 dB compared with a quantized-coefficient FIR-DAC of comparable complexity, while maintaining an EVM of 1.02%. For the full system integration, the interaction between the DPD, DSM, FIR- DAC, and PA is investigated through a transmitter-level design-space exploration. The analysis evaluates the impact of the main architectural parameters on spectral performance, signal quality, and implementation complexity, enabling the selection of a suitable configuration for the complete transmission chain. To support the digital implementation of the DPD, a blocked fixed-point representation is proposed. The coefficient variability and dynamic range are determined based on measurements performed on 13 samples of the same PA. System-level validation shows that the resulting coefficient quantization closely preserves the floating-point performance, with maximum degradations of only 0.67 dB in ACPR and 0.02% in EVM.