The problem of isometry for threshold-based sampling such as integrate-and-fire or send-on-delta is addressed. While for uniform sampling the Parseval theorem provides isometry and makes the Euclidean metric canonical, there is no analogy for threshold-based sampling. The relaxation of the isometric postulate to quasi-isometry, however, allows the discovery of the underlying metric structure of threshold-based sampling. This paper characterizes this metric structure, making Hermann Weyl's discrepancy measure canonical for threshold-based sampling. The usefulness of the approach is demonstrated by means of a novel adaptive threshold-based sampling scheme, which, applied on electrocardiography signals, allows a substantial reduction of the number of samples while maintaining nearly the same SNR after reconstruction.
In stochastic computing (SC), a real-valued number is represented by a stochastic bit stream, encoding its value in the probability of obtaining a one. This leads to a significantly lower hardware effort for various functions and provides a higher tolerance to errors (e.g., bit flips) compared to binary radix representation. The implementation of a stochastic max/min function is important for many areas where SC has been successfully applied, such as image processing or machine learning (e.g., max pooling in neural networks). In this work, we propose a novel shift-register-based architecture for a stochastic max/min function. We show that the proposed circuit has significantly higher accuracy than state-of-the-art architectures for uncorrelated bit streams at comparable hardware costs. Moreover, we analytically proof the correctness of the proposed circuit and provide a new error analysis, based on the individual bits of the stochastic streams. Interestingly, the analysis reveals that for a certain practical bit stream length a finite optimal shift register length exists and it allows to determine the optimal length.
The importance of online system identification (SI) in power electronics is ever increasing. It enables the tracking of system parameters, which in turn can be used for online controller tuning. Hence, SI is a key element for improving a converter's dynamic performance, stability and reliability. In this paper, a novel state-space-based SI approach utilizing the step-adaptive approximate least squares (SALS) estimation algorithm with observation matrix randomization is proposed. The presented concept yields an accurate state-space model of the converter while simultaneously achieving a fast convergence rate and low computational complexity. Consequently, the estimated state-space model is utilized to automatically tune a full state feedback (FSF) controller. This results in an improved converter performance in terms of overshoots, undershoots and settling times. The proposed concept is verified by a prototype system comprising a two-phase buck converter and a field-programmable gate array (FPGA). The provided measurement results highlight the effectiveness and benefits of the presented method over state of the art algorithms as well as z-domain estimation. It is shown that the number of required estimation iterations is more than halved in comparison with state of the art SI approaches, while accuracy is improved.