National Taiwan University Dept Elect EngnCheng, Chia-LungChia-LungChengChang, Fan-RenFan-RenChangTu, Kun-YuanKun-YuanTu2006-11-142018-07-062006-11-142018-07-062005-04http://ntur.lib.ntu.edu.tw//handle/246246/200611150121247A low-cost highly accurate real-time GPS carrier phase disciplined oscillator system based on a single-frequency receiver is presented. In order to estimate the average frequency offsets of an oven-controlled crystal oscillator (OCXO) with respect to the GPS, the OCXO was connected to a modified GPS receiver to replace its original oscillator. Hence, the behavior of the OCXO was determined from the GPS carrier phase observations. The average frequency offsets of the OCXO with respect to the GPS could be estimated by performing difference operations on carrier phase observations of all satellites in view between two measurement epochs. To overcome the interference coming from the atmospheric delay, a real-time dynamic neural-wavelet forecasting filter was proposed. The parameters of the filter were obtained according to the results of a three-day experiment, in which the GPS carrier phase observations of a stand-alone configuration and a common-view configuration were compared. The compressed average frequency offsets were then used by the neural model predictive controller (MPC) for steering the OCXO via D/A converters. From our experiments, the normalized frequency offset of the disciplined OCXO could be improved from about two parts in 109 to about three parts in 1014, and the frequency stability (MDEV) could be improved from about eight parts in 1010 to about four parts in 1014 over 24 h.application/pdf289196 bytesapplication/pdfzh-TWAtmospheric delaydisciplined oscillatorneuralwaveletjournal articlehttp://ntur.lib.ntu.edu.tw/bitstream/246246/200611150121247/1/3616.pdf