Skip to content

0.0.15 "October"

Compare
Choose a tag to compare
@iannesbitt iannesbitt released this 10 Jul 17:42
· 206 commits to master since this release

This version of readgssi (called October) adds many improvements and bug fixes. This release adds support for the 5103 and 350HS antennas, fixes several bugs, outputs better figures, and contains improves horizontal and vertical filtering routines. Added the ability to timezero differently for multi-channel files. Figure changes include: 300 dpi (instead of 150), programmable zoom with the -e x1,x2,z2,z1 flag, a tight bounding box for figure exports, and a patch that calculates two-way time properly.

Version 0.0.15 "October" continues the naming scheme based on names of mountain peaks in the Williamstown, Massachusetts area.

changes since 0.0.14

  • added support for the 5103 400 MHz antenna, and an antenna called CUSTOM in the header, for which center frequency is not known but is likely adjustable
  • added support for the 350 HS hyperstacking antenna
  • fixed a bug that created an extra matplotlib axis
  • added rounding to epsilon_r value on y-axis depth-type display
  • added a patch for #12 to improve axis sizing and tight_layout() handling
  • changed background removal (BGR) function to operate similar to RADAN boxcar function
    • this change means that you can no longer simply specify -r for full-width background removal. new nomenclature for full-width BGR flag is -r 0. -r 100 will apply a 100-trace rolling boxcar BGR (after normalization and stacking)
    • this is done with the scipy.ndimage.filters.uniform_filter1d() function and means that scipy is now explicitly required
    • function will automatically filter out low-frequency washout noise caused by bgr, using obspy's implementation of scipy's butterworth bandpass from 0.1*f to 10*f, where f is antenna frequency. nyquist is automatically accounted for by obspy.
    • BGR value is displayed in title if set > 0, or as "full" if BGR is full-width (i.e. -r 0)
  • added zerophase vertical triangular FIR filter, and changed default filter from butterworth to triangular to attempt to match that in RADAN. butterworth is no longer used (but kept for posterity).
    • this function uses the scipy.signal.firwin() filter design and scipy.signal.lfilter() implementation
  • added ability to suppress plot title using -T or --titleoff from the command line or title=False in main python function
  • changed ns_per_zsample in header to read data shape from header rather than array size
  • added a bug fix for filtering that caused the sample rate to be set too high
  • significantly improved triangular bandpass filter, so that it nearly matches RADAN
  • when Z axis value is set to "samples", the zeroed samples are taken into account. so, if the array had 150 samples sliced off, the plot axis will now start at 150 instead of 0
  • added per-channel handling of timezero. you can pass a list with -Z 233,150,92 (or zero=[233,150,92,None] in python). defaults to -Z 2 (zero=[2,None,None,None])
  • added programmable zoom with -e left,right,up,down flag and list
  • cleaned up some dangling floats
  • updated help text and README help text
  • added automatic tight bbox axes for savefig
  • changed from 150 to 300 dpi for easier poster printing measurements
  • fixed calculation for two-way time (previously per #17, calculation erroneously displayed one-way time)

Changes to previous versions can be found in the changelog.