A Research On Low Power Consumption And High Reliability Of Digital Feedback Circuits

Information theory is the fundamental tool to study the information flow system. The information entropy is the measure of the average uncertainty in the random variable, and the mutual information is the measure of the dependence between two random variables. As the mutual information, which is often used to characterize the processing capability of the communication systems, is closely related to the channel capacity and the circuit power, we use the information theory to analyze the digital circuits in this thesis to improve their reliability and effectiveness, and to provide a theoretical basis for the design of digital circuits.According to our probability model, we discuss the variation trend of information entropy, mutual information and channel capacity, and study the supply voltage and circuit power from the view point of information theory. Based on this idea, we analyze the typical digital circuits, the basic gate level circuits, MRF circuits with high reliability and low power consumption, and the stochastic computing circuits.We discuss the variation trend of mutural information and its influence factors of the basic gate level circuits to provide a theoretical basis. We analyze the supply voltage of MRF circuits to explain why MRF circuits can work well in ultra-low supply environment. We also study the calculate length and the power consumption of the stochastic computing circuits to provide the theoretical basis for stochastic circuits designs.As the mutual information is commonly used to characterize the processing capability in the communication systemsand is closely related to the channel capacity and the circuit power, we analyze the mutual information and fault tolerance features of the abstract circuits to obtain some fundamental properties. By dividing the circuits into symmetric and asymmetric modules, we discuss the relationship between the mutual information and the correct probability or the asymmetric factor, which can be used for digital circuit designs. We also analyze the influence of feedback structure on mutual information. We prove that circuits with feedback can increase the correct probability, decrease the asymmetric factor, and finally lead to an increase of mutual information. We provide the general research method on the mutual information for digital circuit systems, which can be a theoretical basis for digital circuit designs and large scale integrated circuit analysis.Low-power technology is one of the key issues for the digital circuit design, especially for battery-operated, portable system. Decreasing supply voltage is one of the most efficient ways to reduce the energy consumption. But the traditional CMOS devices cannot perform stability function in ultra-low supply voltage application. MRF based circuits can provide under 10-5 BER performance in ultra-low supply environment, which is much better than traditional CMOS circuits. However, there lacks theoretical work focusing on the analysis for reliability and supply voltage in MRF circuit designs, which becomes a bottleneck in MRF-based VLSI design. Therefore, we analyze MRF circuit based on information theory. We not only prove that the MRF circuits have lower supply voltage than that of traditional circuits, but also give a reasonable explanation for this situation. The analysis can provide theory support for MRF-based low power and high reliable circuit designs.Stochastic computation attracts a wide attention because of its simple implementation structure, low circuit latency and flexible configuration, we analyze the stochastic computing circuits based on the information theory. We firstly analyze the noise existing in the stochastic computing circuits and its influence to the channel capacity based on our mathe matical modeling. Then we analyze the bound of the power consumption or calculated length of the stochastic computing circuits according to the shannon theorem. This analysis can provide a theoretical basis for the application of the stochastic computation.We provide the general analysis for digital circuits based on the information theory. We have discussed variation trend of mutual information and its influence factors for the basic gate level circuits. We also have analyzed the bound of supply voltage, circuit power, and circuit energy for two typical application circuits: MRF based circuits and stochastic computing circuits. The work can provide a theoretical basis for the digital circuit designs.