Abstract:
The next generation wireless communication system has the goal of reducing
the power consumption, increasing the network capacity and global connectivity. Most of the signals generated, transmitted or received in wireless systems are generally analog in nature. An analog signal contains both information and energy. However, using the information carrying signal for energy harvesting may lead to loss of information and hence, may affect a system’s performance. Therefore, the objective of this thesis, “iDEG:integrated data and energy gathering for wireless systems" is to harvest the energy as well as information from the same signal without effecting the system performance for energy constrained applications. This implies, the designed system should recover the entire information from the partial signal, and hence the remaining signal can be used for the energy harvesting. In order to recover entire signal from partial information, we have examined different properties of the signal such as sparsity, low-rank and correlation in the data and for different applications such as wireless sensor network (WSN), millimeter wave (mmWave) communication and for wide sense stationary (WSS) signals. To accomplish the objective of efficient data gathering, We have first proposed a partial canonical identity (PCI) based compressive sensing (CS) framework, which randomly samples the observed signal at sub-Nyquist rate and improves the data recovery performance, under the sparsity condition in particular domains such as discrete cosine transform (DCT) and discrete Fourier transform (DFT). This PCI-CS framework reduces the computation cost, implementation complexity, energy losses and can also recover the missing data values. The performance of PCI-CS has been improved in noisy environment by proposing a robust two-stage algorithm, named as PCI-MF, which also utilizes the low-rank nature of the observed signal. The first stage of this algorithm utilizes PCI-CS to recover the sparsest solution and the rank of the data from the partial available information, which are jointly utilized in the second stage to de-noise the data in a matrix factorization framework. This algorithm has been compared with various conventional and state-of-the-art CS and matrix completion framework
in context of WSN and mmWave communication. This is due to the fact that a) WSN data is highly coherent in spatial and temporal domain, which results in double sparsity in the DCT domain and also reduction in rank of the data. Simulations are performed on two real datasets of Intel Lab and Data Sensing Lab. b) PCI-MF has also been proposed to estimate entire channel state information (CSI) from a few randomly varying noisy channel coefficients in mmWave massive multiple-input-multiple-output (MIMO) channel, where estimation of all channel coefficients is not practically feasible. The mmWave channel matrix is a low-rank matrix, which can be modeled as a two dimensional DFT form of a sparse matrix due to the directional beamforming. Simulations have been performed on two different datasets, where one dataset is generated in a real-world setting in the New York City. The next step of this thesis, also aligned with the title, is to develop an integrated data and energy gathering (iDEG) solution. The iDEG has been proposed for WSN, Analog-to-digital-converters (ADCs) and for correlated wide-sensestationary signals. We have proposed an iDEG framework for practical WSN. The sensor nodes deployed in WSNs generate an analog signal corresponding to the sensed parameters, which is sampled and digitized for further processing and transmission to the fusion center. The iDEG for WSN utilizes PCI based CS framework, which selects only a set of sensor nodes at every time point to transmit the data to the FC for recovering the entire data, while the rest of the nodes, which are not participating in data transmission are utilized to harvest the energy from the received analog signal. The performance of iDEG has been tested on a real WSN dataset from Intel Lab. Comparative results of iDEG with the conventional approaches highlight its efficacy. This work motivates us to rethink the sampling process of the ADC. A common ADC architecture is based on sample-and hold (S/H) circuits, where the analog signal is being tracked only for a fraction of the sampling period, and hence allow us to harvest energy in the remaining duration, developing an iDEG solution for ADCs. Therefore, we have proposed eSampling ADCs, which extends the structure of S/H ADCs without altering its data conversion procedure, while harvesting energy from the analog signal during the time periods where the signal is not being tracked. The amount of energy harvested can be increased by reducing the sampling rate, and hence we have also analyzed the tradeoff between the accuracy and the harvested energy. Our theoretical results also shows that eSampling 8-bit ADC acquiring bandlimited signal at Nyquist rate can harvest over 15 dB more energy than it consumes in the conversion procedure. To verify the feasibility of eSampling ADCs, we present a circuit-level design using standard complementary metal oxide semiconductor (CMOS) 65 nm technology. An eSampling 8-bit ADC which samples at 40 MHZ is designed on a Cadence Virtuoso platform. Our experimental study involving Nyquist rate sampling of bandlimited signals demonstrates that such ADCs are indeed capable of harvesting more energy than that spent during analog to- digital conversion, without effecting the accuracy. Finally, to validate eSampling in real-world scenario, a hardware setup has also been designed to harvest energy along with sampling at sensor node deployed for environment and health monitoring WSN application. The final objective for this thesis is to propose a joint sub-Nyquist eSampling and reconstruction based iDEG framework for multiple correlated stochastic signals by exploiting the general correlation without inheriting an inbuilt structure. This work exploits the correlation of multiple stochastic signals to improve the reconstruction accuracy at lower sampling rate, and hence increases the amount of harvested energy from the analog signals by exploiting eSampling method. We derive the achievable reconstruction error, maximum amount of energy harvested and the corresponding esampling system for arbitrary sampling rates andspectral structures by designing an optimal analog combining and reconstruction filter. The proposed system minimizes the error and maximizes the energy, when sampling below the Nyquist rate by preserving only the most dominant spatial eigenmodes aliased to each frequency. Our numerical results illustrate that joint esampling can achieve negligible reconstruction error at low sampling rates, and also allows the system to operate at zero power with up to 16-bits of quantization resolution.