Skip to content

Latest commit

 

History

History
70 lines (65 loc) · 4.46 KB

FFT.md

File metadata and controls

70 lines (65 loc) · 4.46 KB
Raw

FFT

Use Multi-process to accelerate FFT from raw-data (sac/mseed) to fft-data (FFTData).

############################################
## 1.Set parameters 
############################################
cores = 2
flag = true
precompile = false
... 
# Set your own parameters here...
# Then run following code

############################################
# 2.run NoiseCC.FFT
############################################
using Distributed
addprocs(cores, topology = :master_worker)
@everywhere using NoiseCC

NoiseCC.FFT(
    # processes parameters
    cores,flag,precompile,
    # absolute path parameters
    FFTDIR,
    # select sac/mseed files or not
    select_file,
    # when select_file = ture, 'TMPTXT' works
    FILETXT,
    # when select_file = false, 'DATADIR' 'select_sta' 'STATXT' and 'ncomp' work.
    DATADIR,select_sta,STATXT,channel_regular,
    # instrument parameters
    rm_resp,respdir,
    # some control parameters
    input_fmt,time_norm,freq_norm,cc_method,
    # pre-processing parameters
    freqmin,freqmax,cc_len,cc_step,
    # criteria for data selection in time_norm
    factor_clip_std,factor_mute,time_half_win,
    # criteria for data selection in freq_norm
    freq_half_win,
)
  • cores: total number of processes. cores number should be less than data-file number.
  • flag: true of false; print computing time for debugging purpose.
  • precompile: true or false; precompile or not.
  • FFTDIR: dir to store FFTData.
  • select_file: true of false; whether specify sac/mseed files path.
  • FILETXT: sac/mseed files path.
  • DATADIR: absolute dir where sac/mseed files are stored.
  • select_sta: "no" or "sta"; fft only works for selected net-sta base on STATXT.
  • STATXT: station infomation including: network-name and station-name.
  • channel_regular: channel(regular expression) of sac/mseed data; we will use 'glob(comp,DATADIR)' to select data in 'DATADIR' path.
  • rm_resp: "no", "xml", "RESP", and "PZs"; select 'no' to not remove response and use "xml", "RESP", or "PZs" to remove response
  • respdir: path where resp files are located (required if rm_resp != "no")
  • input_fmt: "sac" or "mseed".
  • time_norm: "no", "one_bit", "clip", "mute", and "smooth"; "no" for no normalization in time domain.
  • freq_norm: "no", "whiten", "coherence", and "deconvolution"; "no" for no whitening, "whiten" for preserving only the phase component of the signal, "coherence" for smoothing absolute value of itself, and "deconvolution" for smoothing absolute value squared of itself.
  • cc_method: "CC" and "PCC"; FFT transformation depends on different cross-correlation methods, "CC" for pure cross correlation, and "PCC" for Phase cross-correlation, see [Ventosa et al., 2019] for "PCC".
  • freqmin: Float64; for 'bandpass!' of raw-data and "whiten" parameter
  • freqmax: Float64; for 'bandpass!' of raw-data and "whiten" parameter
  • cc_len: Int64; basic unit of data length for fft (sec)
  • cc_step: Int64; overlapping between each cc_len (sec)
  • factor_clip_std: Float64; 'clip' (default value), clipping the data whose std exceeds the 'factor_clip_std' to max/min mum value. || 中文: NoisePy与之对应的参数是 'max_over_std', 其默认值设置为10, 时间序列超过10倍均方差std的时间序列，把振幅设置成上下线（最大值），限制在一个范围内.
  • factor_mute: Float64; 'mute' (default value), mute the data whose amp exceed 'factor_mute' times the median to 0, for suppressing high-amplitude signals. || 中文: 时间序列超过3倍于波形的平均中值(使用 hibert 变换), 把振幅设置成0. median of envelope of A to find outliers.
  • time_half_win: Int64; 'smooth' (default value), the length(sample points) of the window for running mean normalization, using smooth(A, half_win) -> Smooth array A with half-window half_win (defaults to 3). || 中文: 将一段时间序列分成许多小的时窗，对每一个小的时窗内的波形进行滑动平均，NoisePy对应的参数是 "rma", 其窗口长度的默认值 smooth_N=10; 这也是bensen(2007)推荐的 time_norm 的方法.
  • freq_half_win: Int64; 'coherence' and 'deconvolution' (default value), the length(sample points) of the window for running mean normalization, using coherence!(F,20) or deconvolution!(F,20) -> Smooth array F with half-window half_win (defaults to 20). || 中文: 频率域内的滑动平均，NoisePy对应的参数是 "rma", 其窗口长度的默认值 smooth_N=10; NoisePy 中 time_norm and freq_norm 使用的参数是同一个，都是smooth_N