Abstract:The ring oscillator (RO), as an FPGA-based temperature sensor, has been widely applied in the field of temperature detection due to its advantages of simple structure, low cost, and ease of integration. However, the temperature measurement accuracy of ring oscillators is susceptible to multiple factors, including the number of inverters, inverter layout, oscillation frequency, sampling duration, sampling interval, and cooling time, which are critical design and operational parameters. Therefore, optimizing these parameters to enhance measurement accuracy holds significant research importance. This paper systematically analyzes the impact of these key parameters on the temperature measurement performance of the ring oscillator using the control variable method. Firstly, experimental studies on the influence of different inverter counts on oscillation frequency and temperature error reveal that increasing the number of inverters decreases the oscillation frequency. Further experiments demonstrate that optimizing the inverter count to 40~48 achieves the best measurement accuracy and resolution. Additionally, this paper performs an in-depth analysis of the inverter layout, finding that the delay caused by interconnections between left and right slices within the same configurable logic block (CLB) is significantly greater than that of interconnections across CLBs. Through layout optimization and the selection of specific configurations, the delay can be effectively increased, thus improving measurement accuracy. By comparing various parameter combinations, such as sampling duration, sampling interval, and cooling time, the optimal system parameter configuration is proposed. The experimental validation under the optimal parameter combination shows that the temperature error can be reduced by at least 0.5 ℃.In the environment of 25 ℃~85 ℃, compared with the comparative parameter combination, the average temperature error has decreased from 2.0 ℃ to 1.2 ℃, which is a reduction of 0.7 ℃.Furthermore, at temperatures above 65℃, the temperature error remains consistently controlled within ±1℃.The final results demonstrate that the parameter optimization method proposed in this paper significantly enhances the temperature measurement accuracy of the ring oscillator, providing strong support for the design and application of FPGA-based temperature sensors.