FPGA Power Modeling And Hotspot Analysis

Because of the rapid development of Integrated circuit in its scale, speed and technology, IC is currently facing unprecedented challenges—reliability problem caused by IC's power and heat. As one of IC's primary branches, FPGA is widely used in the field of aerospace which is highly sensitive to IC's reliability, so it is also facing the severe challenges of power and heat issues and the challenge is more prominent. In order to solve this problem effectively, this paper studies and discusses the power consumption and hotspot issues of FPGAs.Firstly, we discuss the power consumption issue. A Hierarchical strategy is applied to establish the power model for FPGA's function modules, with which we can effectively evaluate the power consumption value of FPGA's every function module. The details of establishing FPGA's power model are as follows:1. A hierarchical strategy is adopted to divide the internal resources of FPGA, then the basic cell of power consumption model is obtained;2. To establish the power model for each basic cell. The key to establishing the power model is to extract the functional relationship between power consumption and its factors. In this paper, we use 3 algorithms to extract the functional relationship between power consumption and its factors. They are linear dimension reduction algorithm, nonlinear dimension reduction algorithm and regression algorithm. At the same time, we analyze their advantages, disadvantages and application rules so that we can prefer a good algorithm for each cell, regarding to the precision and efficiency of the power model;3. To design a hardware verification system based on STM32, by verifying and correcting our model we can improve the accuracy of the model.Secondly, it comes to the hotspot issues of FPGA. This paper introduces a RC network thermal model, which is based on the principle of thermoelectric dual. In this principle, people draw an analogy between thermal resistance and resistance, thermal capacitance and capacitance, the atmospheric environment and ground, heat source and current source, temperature difference and voltage difference. So thermal problem is turned into electrical problem, thus, we can calculate the chip temperature quickly and accurately, and get the temperature distribution image with ease. With temperature distribution image, the hotspot of the chip is located. With the above theory, this paper provides two circuit examples based on FPGA. The following is the related work:1. The steps of thermal analysis for the specific circuit are introduced in details;2. Hot Spot is used to calculate and evaluate the temperature of the functional circuit. The evaluation of the static and dynamic temperature distribution of the die and the package components is realized;3. According to the evaluation results, we propose some methods of improving the evaluation accuracy in Hot Spot;4. We address the contribution hotspot analysis make to power optimization and low power design.Finally, we conclude all the work. The power consumption model is verified by hardware verification system and its accuracy is less than 10%. The model can be used to evaluate not only the total power of FPGA system, but also each module's power of FPGA. So it can be well used for hotspot analysis for FPGA, thus providing a strong basis for the reliability analysis of FPGA.