The doctoral thesis deals with development and design of novel architectures of digital filters for low-power integrated circuits. The main goal was to achieve optimum parameters of digital filters with respect to the chip area, power consumption and operating frequency. The target group of the proposed architectures are application specific integrated circuits designed for signal processing from sensors using delta-sigma modulators. Three novel architectures of digital filters optimized for low-power integrated circuits are presented in the thesis. The thesis provides analysis and comparison of parameters of the new filter architectures with the parameters of architectures generated by Matlab tool. A software tool has been designed and developed for the practical application of the proposed architectures of digital filters. The developed software tool allows generating hardware description of the filters with respect to defined parameters.

This work presents novel continuous-time delta-sigma modulator architectures with low-power consumption and improved signal transfer functions which are suitable for wideband A/D conversion in wireless applications, e.g., 3G and 4G receivers. The research has explored two routes for improving the overall performance of continuous-time delta-sigma modulator. The first part of this work proposes the use of the power efficient Successive-Approximations (SAR) architecture, instead of the conventional Flash ADC, as the internal quantizer of the delta-sigma modulator. The SAR intrinsic latency has been addressed by means of a faster clock for the quantizer as well as full-period delay compensation. The use of SAR quantizer allows for increasing the resolution while reducing the total power consumption and complexity. A higher resolution quantizer, made feasible by the SAR, would allow implementing more aggressive noise shaping to facilitate wideband delta-sigma A/D conversion at lower over-sampling-rates. As proof of concept, a first-order CT delta-sigma modulator with a 5-bit SAR quantizer is designed and implemented in a 130 nm CMOS process which achieves 62 dB dynamic range over 1.92 MHz signal bandwidth meeting the requirements of the WCDMA standard. The prototype modulator draws 3.1 mW from a single 1.2 V supply and occupies 0.36 mm2 of die area. The second part of this research addresses the issue of out-of-band peaking in the signal-transfer-function (STF) of the widely used feedforward structure. The STF peaking is harmful to the performance of the modulator as it allows an interferer to saturate the quantizer and result in severe harmonic distortion and instability. As a remedy to this problem a general low-pass and peaking-free STF design methodology has been proposed which allows for implementing an all-pole filter in the input signal path for any given NTF. Based on the proposed method, the STF peaking of any feedforward modulator can be eliminated using extra feed-in paths to all the integrator inputs. A major drawback of the conventional feedforward topology having low-pass STF is the large sensitivity of the STF to the coefficients. In particular, component mismatch, due to random errors in the relative values of individual resistors or capacitors, can significantly degrade the anti-aliasing of the CT modulator and give rise to the unwanted STF peaking. To solve this problem two new architectures, namely dual-feedback and dual-feed-in are proposed which allow us to synthesize a low-pass STF with a smaller number of coefficients than the feedforward structure. The dual-feedback structure which shows significantly lower sensitivity to coefficient mismatch is extensively analyzed and simulated. Also for proof of concept a third-order modulator is implemented in a 130 nm CMOS process which achieves 76 dB dynamic-range over 5 MHz signal bandwidth meeting, for example, the requirements of a DVB-H receiver standard. In addition the modulator shows 77 dB anti-aliasing and less than 0.1 dB worst-case STF peaking. The measured power consumption of the modulator is 6 mW from a single 1.2 V and the die area is 0.56 mm2.

No description available.

Dans le cadre de la radio logicielle, un transmetteur numérique, basé sur la modulation ΔΣ, est proposé. Son architecture est construite autour de deux modulateurs ΔΣ passe-bas suréchantillonnés du 3ème ordre qui fournissent un signal multiplexé sur 1 bit à haute cadence, qui code directement le signal RF dans le domaine numérique. La séquence de sortie peut ensuite être appliquée à l'entrée d'un amplificateur de puissance commuté ayant une bonne efficacité.<br />Le standard UMTS a été choisi comme exemple d'application et un générateur de signaux RF 1 bit à 7,8Géch/s a été réalisé dans une technologie 90nm CMOS. Une arithmétique redondante comprenant des signaux complémentaires, une quantification de sortie non exacte et une évaluation anticipée de la sortie ont été implémentées pour parvenir à la cadence désirée. Une logique dynamique différentielle sur 3 phases d'horloge, générées par une DLL, a été utilisée au niveau circuit.<br />Le circuit intégré du transmetteur prototype démontre une fonctionnalité complète jusqu'à une fréquence d'horloge de 4GHz, permettant ainsi d'atteindre une bande passante de 50MHz autour d'une fréquence porteuse de 1GHz. Si la bande image est utilisée, la fréquence d'émission peut être déplacée jusqu'à 3GHz. Avec une fréquence d'horloge de 2,6GHz et un canal WCDMA de 5MHz modulé autour d'une fréquence porteuse à 650MHz, 53,6dB d'ACLR sont obtenus pour une puissance de canal en sortie de -3,9dBm. Pour la bande image (1,95GHz), l'ACPR est de 44,3dB pour une puissance maximale du canal en sortie de -15,8dBm, ce qui rentre dans les spécifications UMTS. L'aire active du circuit est de 0,15mm² et sa consommation de 69mW sous 1V à cette fréquence.

La chaîne de réception des téléphones mobiles de dernière génération utilisent au moins deux étages de transposition en fréquence avant d'effectuer la démodulation en quadrature. La transposition en fréquence augmente la complexité du système et engendre de nombreux problèmes tels que la limitation de l'échelle dynamique et l'introduction de bruit issu de l'oscillateur local. Il est alors nécessaire d'envisager une numérisation du signal le plus près possible de l'antenne. Cette dernière permet la conversion directe d'un signal analogique en un signal numérique à des fréquences intermédiaires. Elle simplifie ainsi la conception globale du système et limite les problèmes liés aux mélangeurs. Pour cela, des architectures moins conventionnelles doivent être développées, comme la conversion analogique-numérique utilisant la modulation Sigma-Delta à temps continu. La modélisation comportementale de ce convertisseur analogique-numérique, ainsi que la conception des principaux blocs ont donc été l'objet de cette thèse. L'application d'une méthodologie de conception avancée, permettant la simulation mixte des blocs fonctionnels à différents niveaux d'abstraction, a permis de valider aussi bien la conception des circuits que le système global de conversion. En utilisant une architecture à multiples boucles de retour avec un quantificateur multi-bit, le convertisseur Sigma-Delta passe bande à temps continu atteint un rapport signal sur bruit (SNR) d'environ 76 dB dans une large bande de 20MHz. / Wireless front-end receivers of last generation mobile devices operate at least two frequency translations before I/Q demodulation. Frequency translation increases the system complexity, introducing several problems associated with the mixers (dynamic range limitation, noise injection from the local oscillator, etc.). Herein, the position of the analog-to-digital interface in the receiver chain can play an important role. Moving the analog-to-digital converter (ADC) as near as possible to the antenna, permits to simplify the overall system design and to alleviate requirements associated with analog functions (filters, mixers). These currently requirements have led to a great effort in designing improved architectures as Continuous-Time Delta-Sigma ADCs. The behavioural modeling this converter, although the circuit design of the main blocks has been the subject of this thesis. The use of an advanced design methodology, allowing the mixed simulation at different levels of abstraction, allows to validate both the circuit design and the overall system conversion. Using a multi-feedback architecture associated with a multi-bit quantizer, the continuous-time Bandpass Delta-Sigma converter achieves a SNR of about 76 dB in a wide band of 20MHz.

Ce travail de recherche porte sur la conversion analogique numérique delta sigma à temps continu passe-bas, et plus particulièrement sur les difficultés de réalisation de ce type de convertisseur. L'objectif global de ces recherches était la mise en place d'une méthode de conception adaptée. Dans un premier temps, le travail s'est focalisé sur le calcul d'architecture et l'obtention de coefficients adaptés à une spécification donnée. L'impact des imperfections d'horloge sur les performances de ces convertisseurs a ensuite été étudié et une méthode analytique d'estimation des dégradations introduites par l'intermédiaire de l'horloge a été proposée. Ces deux étapes clefs lors de la réalisation d'un delta sigma à temps continu ont été intégrées à un flot de conception complet allant des spécifications à l'implémentation sur silicium. Finalement, ce flot de conception a été utilisé pour réaliser un modulateur delta sigma à temps continu en technologie CMOS065 pour une application WLAN.

Advances in electronic systems have lead to the demand for high resolution, high bandwidth Analog-to-Digital Converters (ADCs). Oversampled ADCs are well- known for high accuracy applications since they benefit from noise shaping and they usually do not need highly accurate components. However, as a consequence of oversampling, they have limited signal bandwidth. The signal bandwidth (BW) of oversampled ADCs can be increased either by increasing the sampling rate or reducing the oversampling ratio (OSR). Reducing OSR is a more promising method for increasing the BW, since the sampling speed is usually limited by the technology. The advantageous properties (e.g. low in-band quantization, relaxed accuracy requirements of components) of oversampled ADCs are usually diminished at lower OSRs and preserving these properties requires complicated and power hungry architectures. In this thesis, different combinations of delta-sigma and pipelined ADCs are explored and new techniques for designing oversampled ADCs are proposed. A Hybrid Delta-Sigma/Pipelined (HDSP) ADC is presented. This ADC uses a pipelined ADC as the quantizer of a single-loop delta-sigma modulator and benefits from the aggressive quantization of the pipelined quantizer at low OSRs. A Noise-Shaped Pipelined ADC is proposed which exploits a delta-sigma modulator as the sub-ADC of a pipeline stage to reduce the sensitivity to the analog imperfection. Three prototype ADCs were fabricated in 0.18μm CMOS technology to verify the effectiveness of the proposed techniques. The performance of these architectures is among the best reported for high bandwidth oversampled ADCs. / Graduation date: 2011

The quest for multi-standard and software-defined radio (SDR) receivers calls for high flexibility in the receiver building-blocks so that to accommodate several wireless services using a single receiver chain in mobile handsets. A potential approach to achieve flexibility in the receiver is to move the analog-to-digital converter (ADC) closer to the antenna so that to exploit the enormous advances in digital signal processing, in terms of technology scaling, speed, and programmability. In this context, continuous-time (CT) delta-sigma (ΔƩ) ADCs show up as an attractive option. CT ΔƩ ADCs have gained significant attention in wideband receivers, owing to their amenability to operate at a higher-speed with lower power consumption compared to discrete-time (DT) implementations, inherent anti-aliasing, and robustness to sampling errors in the loop quantizer. However, as the ADC moves closer to the antenna, several blockers and interferers are present at the ADC input. Thus, it is important to investigate the sensitivities of CT ΔƩ ADCs to out-of-band (OOB) blockers and find the design considerations and solutions needed to maintain the performance of CT ΔƩ modulators in presence of OOB blockers. Also, CT ΔƩ modulators suffer from a critical limitation due to their high sensitivity to the clock-jitter in the feedback digital-to-analog converter (DAC) sampling-clock. In this context, the research work presented in this thesis is divided into two main parts. First, the effects of OOB blockers on the performance of CT ΔƩ modulators are investigated and analyzed through a detailed study. A potential solution is proposed to alleviate the effect of noise folding caused by intermodulation between OOB blockers and shaped quantization noise at the modulator input stage through current-mode integration. Second, a novel DAC solution that achieves tolerance to pulse-width jitter by spectrally shaping the jitter induced errors is presented. This jitter-tolerant DAC doesn’t add extra requirements on the slew-rate or the gain-bandwidth product of the loop filter amplifiers. The proposed DAC was implemented in a 90nm CMOS prototype chip and provided a measured attenuation for in-band jitter induced noise by 26.7dB and in-band DAC noise by 5dB, compared to conventional current-steering DAC, and consumes 719µwatts from 1.3V supply.

Continuous-time (CT) delta-sigma (ΔΣ) analog-to-digital converters (ADC) have emerged as the popular choice to achieve high resolution and large bandwidth due to their low cost, power efficiency, inherent anti-alias filtering and digital post processing capabilities. This work presents a detailed system-level design methodology for a low-power CT ΔΣ ADC. Design considerations and trade-offs at the system-level are presented. A novel technique to reduce the sensitivity of the proposed ADC to clock jitter-induced feedback charge variations by employing a hybrid digital-to-analog converter (DAC) based on switched-capacitor circuits is also presented. The proposed technique provides a clock jitter tolerance of up to 5ps (rms). The system is implemented using a 5th order active-RC loop filter, 9-level quantizer and DAC, achieving 74dB SNDR over 20MHz signal bandwidth, at 400MHz sampling frequency in a 1.2V, 90 nm CMOS technology. A novel technique to improve the linearity of the feedback digital-to-analog converters (DAC) in a target 11-bits resolution, 100MHz bandwidth, 2GHz sampling frequency CT ΔΣ ADC is also presented in this work. DAC linearity is improved by combining dynamic element matching and automatic background calibration to achieve up to 18dB improvement in the SNR. Transistor-level circuit implementation of the proposed technique was done in a 1.8V, 0.18μm BiCMOS process.

No description available.