This thesis explores the clock and data recovery (CDR) for the high-speed blind-sampling ADC-based receivers. This exploration results in two new CDR architectures that reduce the receiver complexity and save the ADC power and area compared to the previous work. The two proposed CDR architectures constitute the primary contributions of this thesis. The first proposed architecture, a 2x feed-forward CDR architecture, eliminates the interpolating feedback loop, used in the previously reported CDRs, in order to reduce the CDR circuit complexity. Instead of the feedback loop, the proposed architecture uses a feed-forward topology to recover the phase and data directly from the blind digital samples of the received signal. The 2x feed-forward CDR architecture was implemented and characterized in a 5 Gb/s receiver test-chip in 65 nm CMOS. The test-chip measurements confirm that the CDR successfully recovers the data with bit error rate (BER) < 10e-12 in the presence of jitter. The second proposed architecture, a fractional-sampling-rate (FSR) CDR architecture, reduces the receiver sampling rate from the typical integer rate of 2x the baud rate to a fractional rate between 2x and 1x in order to reduce the ADC power and area. This architecture employs the feed-forward topology of the first contribution of this thesis to recover the phase and data from the fractionally-spaced digital samples of the signal. To verify the proposed FSR CDR architecture, a 1.45x receiver test-chip was implemented and characterized in 65 nm CMOS. This test-chip recovers 6.875 Gb/s data from the ADC samples taken at 10 GS/s. The measurements confirm a successful data recovery in the presence of jitter with BER < 10e-12. With sampling at 1.45x, the FSR CDR architecture reduces the ADC power and area by 27.3% compared to the 2x feed-forward CDR architecture, while the overall receiver power and area are reduced by 12.5%.

This thesis is focused on the analysis and development of Ensembles of Neural Networks. An ensemble is a system in which a set of heterogeneous Artificial Neural Networks are generated in order to outperform the Single network based classifiers. However, this proposed thesis differs from others related to ensembles of neural networks [1, 2, 3, 4, 5, 6, 7] since it is organized as follows. In this thesis, firstly, an ensemble methods comparison has been introduced in order to provide a rank-based list of the best ensemble methods existing in the bibliography. This comparison has been split into two researches which represents two chapters of the thesis. Moreover, there is another important step related to the ensembles of neural networks which is how to combine the information provided by the neural networks in the ensemble. In the bibliography, there are some alternatives to apply in order to get an accurate combination of the information provided by the heterogeneous set of networks. For this reason, a combiner comparison has also been introduced in this thesis. Furthermore, Ensembles of Neural Networks is only a kind of Multiple Classifier System based on neural networks. However, there are other alternatives to generate MCS based on neural networks which are quite different to Ensembles. The most important systems are Stacked Generalization and Mixture of Experts. These two systems will be also analysed in this thesis and new alternatives are proposed. One of the results of the comparative research developed is a deep understanding of the field of ensembles. So new ensemble methods and combiners can be designed after analyzing the results provided by the research performed. Concretely, two new ensemble methods, a new ensemble methodology called Cross-Validated Boosting and two reordering algorithms are proposed in this thesis. The best overall results are obtained by the ensemble methods proposed. Finally, all the experiments done have been carried out on a common experimental setup. The experiments have been repeated ten times on nineteen different datasets from the UCI repository in order to validate the results. Moreover, the procedure applied to set up specific parameters is quite similar in all the experiments performed. It is important to conclude by remarking that the main contributions are: 1) An experimental setup to prepare the experiments which can be applied for further comparisons. 2) A guide to select the most appropriate methods to build and combine ensembles and multiple classifiers systems. 3) New methods proposed to build ensembles and other multiple classifier systems.

Achieving a low noise level is of great interest to the forest machine industry. Traditionally this is obtained by using passive noise reduction, i.e. by using materials for sound isolation and sound absorption. Especially designs to attenuate low frequency noise tend to be bulky and impractical from an installation point of view. An alternative solution to the problem is to use active noise control (ANC). The basic principle of ANC is to generate an anti-noise signal designed to destructively interfere with the unwanted noise. In this thesis two algorithms (Feedback FxLMS and Feedforward FxLMS) are implemented and evaluated for use in the ANC-system. The ANC-system is tuned to the specific environment in the driver’s cabin of a Komatsu forest machine. The algorithms are first tested in a simulated environment and then in real-time inside a forest machine. Simulations are made both in Matlab and in C using both generated signals and recorded signals. The C code is implemented on the Analog Devices Blackfin DSP card BF526. The result showed a significantly reduction of the sound pressure level (SPL) in the driver’s cabin. The noise attenuation obtained using the Feedback FxLMS was approximately 14 dB for a tonal 100 Hz signal and 11 dB using recorded engine noise from a forest machine at 850 rpm.

There are growing industrial demands for low-voltage supply and low-power consumption circuits and systems. This is especially true for very high integration level and very large scale integrated (VLSI) mixed-signal chips and system-on-a-chip. It is mainly due to the limited power dissipation within a small area and the costs related to the packaging and thermal management. In this research work, two low-voltage, low-power integrated circuits used for data communication systems are introduced. The first one is a high performance continuous-time linear phase filter with automatic frequency tuning. The filter can be used in hard disk driver systems and wired communication systems such as 1000Base-T transceivers. A pseudo-differential operational transconductance amplifier (OTA) based on transistors operating in triode region is used to achieve a large linear signal swing with low-voltage supplies. A common-mode (CM) control circuit that combines common-mode feedback (CMFB), common-mode feedforward (CMFF), and adaptive-bias has been proposed. With a 2.3V single supply, the filter’s total harmonic distortion is less than –44dB for a 2VPP differential input, which is due to the well controlled CM behavior. The ratio of the root mean square value of the ac signal to the power supply voltage is around 31%, which is much better than previous realizations. The second integrated circuit includes two LVDS drivers used for high-speed point-to-point links. By removing the stacked switches used in the conventional structures, both LVDS drivers can operate with ultra low-voltage supplies. Although the Double Current Sources (DCS) LVDS driver draws twice minimum static current as required by the signal swing, it is quite simple and achieves very high speed operation. The Switchable Current Sources (SCS) LVDS driver, by dynamically switching the current sources, draws minimum static current and reduces the power consumption by 60% compared to the previously reported LVDS drivers. Both LVDS drivers are compliant to the standards and operate at data rates up to gigabits-per-second.

The design of a battery-powered, portable headphone active noise control system with TI TMS320C5416 DSP is described. The preliminary implementation of the system on a C5416 DSK is also explained. The problems of fixed-point implementation are described and solutions are proposed. Sign-sign Fx-LMS algorithm with a dead-zone is introduced and used as the adaptation algorithm. Effective use of dynamic range to improve the accuracy in filtering operations is discussed. Details of the designed battery-powered DSP board are given and board software development process is explained. The DSK system and designed portable system is compared against two commercially available analog systems under three different types of noises / composition of tones, drill noise and propeller plane cabin noise. The results reveal that adaptive system has better overall performance.

Various techniques of optimizing the multiple class cross-entropy error function to train single hidden layer neural network classifiers with softmax output transfer functions are investigated on a real-world multispectral pixel-by-pixel classification problem that is of fundamental importance in remote sensing. These techniques include epoch-based and batch versions of backpropagation of gradient descent, PR-conjugate gradient and BFGS quasi-Newton errors. The method of choice depends upon the nature of the learning task and whether one wants to optimize learning for speed or generalization performance. It was found that, comparatively considered, gradient descent error backpropagation provided the best and most stable out-of-sample performance results across batch and epoch-based modes of operation. If the goal is to maximize learning speed and a sacrifice in generalisation is acceptable, then PR-conjugate gradient error backpropagation tends to be superior. If the training set is very large, stochastic epoch-based versions of local optimizers should be chosen utilizing a larger rather than a smaller epoch size to avoid inacceptable instabilities in the generalization results. (authors' abstract) / Series: Discussion Papers of the Institute for Economic Geography and GIScience

This thesis deals with the determination of the position of the solenoid core in real time based on the measured current. The reference position of the current is used for feedback control of the solenoid. For this issue, software tool Matlab/Simulink was used. For current and temperature measurements, PCB circuits were created. The whole project was carried out on the dSPACE platform.

Les actionneurs multi-axes sont de plus en plus prisés par les concepteurs de systèmes de nanopositionnement car ils permettent une réduction de l'espace occupé et de l'énergie consommée, une dextérité plus grande et une modularité avec peu de contraintes pour les applications. Certains de ces actionneurs et systèmes multi-axes sont cependant caractérisés par des oscillations mal-amorties qui compromettent de manière drastique leurs performances générales. Cette thèse concerne l'exploitation des techniques de commande en boucle-ouverte input-shaping classiquement utilisées pour amortir de manière sans capteurs les oscillations dans les systèmes mono-axes et les étendent pour qu'ils soient utilisables pour les systèmes multi-axes. Les résultats proposés dans la thèse qui sont des techniques input-shaping multivariables sont ensuite appliquées sur des actionneurs piézoélectriques classiquement dédiés pour les applications de nanopositionnement. / Multi-axes actuators are becoming more and more tempting to nanopositioning system designers as they enable them to save space, reduce energy consumption, increase dexterity and offer more modularity and freedom with fewer constraints to their applications. Some of these multi-axes actuators and systems exhibit however badly damped vibrations which strongly compromise their global performances. This thesis work exploits the advantages of the well-known feedforward input shaping techniques usually used to damp vibrations in monovariable (SISO) systems to present a new multivariable (MIMO) input shaping technique that can be used to damp vibrations in multi-axes systems. The approach that was used in this study is to extend a previous work that was done on multiple-input single-output (MISO) systems and generalize it for MIMO systems. The study demonstrates also the application of this newly developed technique on different piezoelectric actuators commonly used in nanopositioning systems.

Sigma-delta modulators (SDM) have come up as an attractive candidatefor analog-to-digital conversion in single chip front ends thanks to the continuousimproving performance. The major disadvantage is the limited bandwidthdue to the need of oversampling. Therefore, extending these convertersto broadband applications requires lowering the oversampling ratio (OSR) inorder. The aim of this thesis is the investigation on the topology and structureof sigma-delta modulators suitable for wideband applications, e.g. wireline orwireless communication system applications having a digital baseband aboutone to ten MHz.It has recently become very popular to feedforward the input signal inwideband sigma-delta modulators, so that the integrators only process quantizationerrors. The advantage being that the actual signal is not distorted byopamp and integrator nonlinearities. An improved feedforward 2-2 cascadedstructure is presented based on unity-gain signal transfer function (STF). Theimproved signal-to-noise-ratio (SNR) is obtained by optimizing zero placementof the noise transfer function (NTF) and adopting multi-bit quantizer.The proposed structure has low distortion across the entire input range.In high order single loop continuous-time (CT) sigma-delta modulator, excessloop delay may cause instability. Previous techniques in compensation ofinternal quantizer and feedback DAC delay are studied especially for the feedforwardstructure. Two alternative low power feedforward continuous-timesigma-delta modulators with excess loop delay compensation are proposed.Simulation based CT modulator synthesis from discrete time topologies isadopted to obtain the loop filter coefficients. Design examples are given toillustrate the proposed structure and synthesis methodology.Continuous time quadrature bandpass sigma-delta modulators (QBSDM)efficiently realize asymmetric noise-shaping due to its complex filtering embeddedin the loops. The effect of different feedback waveforms inside themodulator on the NTF of quadrature sigma-delta modulators is presented.An observation is made that a complex NTF can be realized by implementingthe loop as a cascade of complex integrators with a SCR feedback digital-toanalogconverter (DAC), which is desirable for its lower sensitivity to loopmismatch. The QBSDM design for different bandpass center frequencies relativeto the sampling frequency is illustrated.The last part of the thesis is devoted to the design of a wideband reconfigurablesigma-delta pipelined modulator, which consists of a 2-1-1 cascadedmodulator and a pipelined analog-to-digital convertor (ADC) as a multi-bitquantizer in the last stage. It is scalable for different bandwidth/resolutionapplication. The detail design is presented from system to circuit level. Theprototype chip is fabricated in TSMC 0.25um process and measured on thetest bench. The measurement results show that a SNR over 60dB is obtainedwith a sampling frequency of 70 MHz and an OSR of ten.

As part of Volvo Penta's initiative to further the development of predictive maintenance in their field test environments, this thesis compares neural networks in an effort to predict the occurrence of three common diagnostics trouble codes using field test data. To quantify the neural networks' performances for comparison a number of evaluation metrics were used. By training a multitude of differently configured feedforward neural networks with the processed field test data and evaluating the resulting models, it was found that the resulting models perform better than that of a baseline classifier. As such it is possible to use Volvo Penta's field test data along with neural networks to achieve predictive maintenance. It was also found that Long Short-Term Memory (LSTM) networks with methodically selected hyperparameters were able to predict the diagnostic trouble codes with the greatest performance among all the tested neural networks.