Research And Design Of High Precision Timekeeping System Based On Neural Network

Crystal oscillator as a frequency source,in communications,satellite navigation,range,combat command,power,finance,intelligent manufacturing and other fields have a wide range of applications,involving physics,measurement,machinery and other disciplines.In general,when there is an external reference clock source,the crystal is tame,when the reference clock source is not available,it is necessary to use frequency compensation technology to maintain the accuracy and stability of the local crystal output frequency.In this thesis,how to compensate the output frequency offset of the local crystal oscillator after the timekeeping system loses the reference clock source is researched.Firstly,the development history,application and research status of crystal oscillators are expounded,as can be seen from the research contents of many researchers,the current research on high precision timekeeping system mainly focuses on how to minimize the error of frequency prediction,and finally compensates for the frequency offset of crystal oscillators.Then it introduces the relevant characteristics and working principle of crystal oscillator,analyzes the main factors affecting output frequency stability,expounds the causes of the poor stability of crystal oscillator,and obtains the method of compensation frequency offset by analyzing the voltage control characteristics.On this basis,considering that temperature and aging are the main factors leading to frequency offset,a neural network-based crystal frequency prediction algorithm is proposed.The algorithm uses the previous measured values of crystal oscillator frequency to train the neural network model,characterizes the law of crystal oscillator frequency drift,and improves the accuracy of crystal oscillator frequency prediction.Finally,according to the voltage control characteristics of the crystal oscillator,a high-precision timekeeping system design scheme based on CPLD and MCU chips is proposed,and the timekeeping system is designed using a voltage-controlled crystal oscillator with low output frequency accuracy.The system converts the frequency offset value predicted by the neural network model into a compensation voltage value and acts on the voltage control end of the crystal oscillator,improving the stability of the output frequency of the crystal oscillator and achieving high-precision timekeeping functions.At the same time,a test environment was set up and the system performance was tested.The test results show that the output frequency stability of the crystal oscillator compensated by this method is improved,and the method of using neural network to predict the frequency deviation is effective.Accurate frequency source provides a strong guarantee.