| In the information age,with the rapid development of technologies such as mobile communication and satellite navigation,people have higher requirements for the stability and accuracy of clock frequency sources.Currently,due to its good short-term stability and low cost,the application field of oven controlled crystal oscillator(OCXO)is constantly expanding.However,OCXOs are affected by various factors such as increasing working time,changes in environmental temperature,and internal process aging,which can cause gradual drift changes in the output frequency and result in poor long-term stability.This restricts their application in high-precision communication fields.To address this issue,this thesis begins research on synchronization algorithms for OCXO.The synchronization algorithm for OCXO mainly consists of two stages: time synchronization stage and timekeeping stage.This thesis proposes an improved Kalman filter-based time synchronization algorithm for the time synchronization stage,which utilizes GPS pulse per second(1PPS)to synchronize the OCXO and improve its frequency accuracy.For the clock timekeeping stage,a gate-controlled recurrent neural network-based clock timekeeping algorithm is proposed,which enables the OCXO to maintain high stability even after losing GPS 1PPS signals.The aim of this thesis is to design high-precision time synchronization algorithms and high stability clock timekeeping algorithms to achieve synchronization and meet the performance requirements of 5G communication systems.Firstly,this thesis introduces the physical structure of OCXO,and the factors such as aging and temperature that mainly affect its output frequency.Furthermore,the GPS time synchronization principle and time synchronization errors are analyzed.In the time synchronization stage,an improved Kalman filter-based time synchronization algorithm is proposed,and combined with PID control to tame the GPS 1PPS to OCXO.This not only enables the removal of clock phase differences through singular value decomposition and smoothing,but also reduces the influence of GPS 1PPS signal phase fluctuations or other factors on OCXO time synchronization.The phase difference is then input into an analog-to-digital converter to obtain a voltage adjustment value,which is compensated to the OCXO to achieve frequency correction.This ensures that the OCXO clock signal maintains the same standard as the GPS clock signal after calibration,effectively compensating for its accuracy disadvantage.Next,in the timekeeping stage when the system loses the GPS 1PPS time reference,the output frequency of the OCXO is affected by various factors,including crystal aging and changes in environmental temperature.Due to the effects of these factors,the output frequency of the OCXO may drift,thereby affecting the output frequency accuracy of the system.To address this issue,this thesis proposes a frequency prediction model for the OCXO,which is based on a gated recurrent unit neural network.The model can simultaneously consider both temperature and aging factors and utilize the excellent adaptive and nonlinear generalization ability of neural networks to learn the frequency change pattern of the OCXO in the timekeeping stage.This can predict the frequency drift of the OCXO in the timekeeping stage and compensate for the output frequency of the OCXO using an appropriate calculation method to optimize the long-term stability of the OCXO.Finally,a synchronous clock hardware system was designed and a testing platform was set up to test the OCXO time synchronization and timekeeping algorithms.The simulation results showed that the accuracy of the synchronous timing algorithm can reach about 5ns,which meets the time synchronization standard requirements within5 G communication systems.The hardware test data also verified that this algorithm achieved high-precision time synchronization.After losing the reference clock signal,based on the OCXO related data recorded in the timekeeping algorithm,the GRU neural network timekeeping algorithm was selected and compared with several other methods.The experimental results showed that the GRU network has higher prediction accuracy and faster operation speed.After accurately predicting the frequency offset caused by temperature and aging of the OCXO and compensating the frequency offset to the OCXO,the hardware test results showed that the timekeeping accuracy is within-1.408μs/12 h,which meets the timekeeping indicator requirements within 5G communication systems ±1.5μs/12h,and the data reflects that this algorithm also achieves high stability timekeeping.Therefore,the high-precision time synchronization and high-stability timekeeping algorithms designed in this thesis can further improve the security and reliability of 5G communication systems. |
