摘要
从实际的日长变化 (ΔLOD)序列中提取出的若干谐波信号之和构成一个模拟ΔLOD序列 ,分别用多级滤波器和巴特沃斯、切比雪夫、椭圆等数字递归滤波器进行滤波 ,分离其中的季节性、年际和十年尺度的变化分量 ,然后与实际模拟信号作比较 ,考察滤波后信号的端部畸变情况。结果表明 ,在我们所感兴趣的频率范围内 ,在所比较的 4种滤波器中 ,无论是高通 (分离季节性变化分量 )滤波 ,还是低通 (分离十年尺度变化分量 )滤波 ,以Vondrak平滑算法为基础的多级滤波器都具有最小的端部畸变率 ;对于高通和低通滤波而言 ,畸变较小的分别是巴特沃斯和切比雪夫滤波器 ;当用作带通滤波以分离年际变化分量时 ,直接用模拟数据序列滤波 ,以切比雪夫滤波器的端部畸变最小 ;增加统计延拓数据后再滤波 ,以巴特沃斯和椭圆滤波器的端部畸变最小。无论采用那种滤波器 ,在数据序列两端增加统计延拓数据 ,都会减小滤波结果的端部畸变。
By using some commonly used filters, such as MULTI-STAGE, ELLIPSE, CHEBYSHEV, and BUTTERWORTH, seasonal, interannual, and decadal components are separated from a synthetic LOD series which was composed of 11 periodic terms extracted from a real LOD series during 1962～1998. And then comparisons of the filtered results with the original components have been made to investigate the deformations induced by the different filters at the two ends of the data series. Some results are shown in figure 1 and the tables 2～4. It is clear by the comparison that the MULTI-STAGE filter has least deformation rate at the two ends of series either it is used for high or low pass filtering, the BUTTERWORTH (for high pass filtering) and CHEBYSHEV (for low pass filtering) have less deformation among the others. For the band pass filtering, however, the CHEBYSHEV filter has least deformation. After adding some extended data obtained by fitting the original data series as a linear statistical model, comparison of the filtered results show that the BUTTERWORTH and ELLIPSE filters have least deformations for band pass filtering. The ends deformation induced by filtering will be smaller for any filters after adding of some statistically extended data to the original series.
出处
《中国科学院上海天文台年刊》
2001年第1期50-56,共7页
Annals Shanghai Astronomical Observatory Chinese Academy of Sciences
基金
国家自然科学基金 (编号 1983 3 0 3 0
19973 0 11)
上海市科学技术发展基金 (编号JC14 0 12 )资助
