Thread: Problem with merging large files

Started: 2017-02-08 03:37:40
Last activity: 2017-02-09 03:58:47
Topics: SAC Help
Lukas Janku
2017-02-08 03:37:40
Hi everyone

I am having problem (probably) related to some maximum length of the seismic data within Sac.

I have been dealing with some really crappy seismic data with messed-up timing. 3 Traces from the same sensor. The data had frequent gaps in them and for some reason, the gaps are occasionally of different lengths for different components. I have about 1 month of continuous record (100 Hz sampling). My timing correction included:

1. Merging all the files from one component together (have a computer with large RAM).
2. Use some of the recorded earthquakes where I know the p-wave arrival for getting the timing of my 1-month block right.
3. Splitting the month-long file into day-long files using a script.

The problem I get is when I use an earthquake at the beginning of the 1-month block (consistent for different components), at the end of my correction process my 3 components are slightly shifted with respect to each other near the end of the 1-month block.

Does anyone has an experience with similar issue? The only reason I can think off is the different length of the gaps in the data, combined with huge amount of merged data, can make some internal variable 'overflow' at different time for different components... But that's just an impression.

Thanks a lot

Lukas Janku
University of Canterbury

This email may be confidential and subject to legal privilege, it may
not reflect the views of the University of Canterbury, and it is not
guaranteed to be virus free. If you are not an intended recipient,
please notify the sender immediately and erase all copies of the message
and any attachments.

Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
information.

  • Brian Savage
    2017-02-08 15:46:43
    Lukas,

    I would suspect that you are running up against the precision of the floating point value used for timing in SAC. See the attached plot. It shows the maximum time value (y-axis) without loss of precision for a variety of sampling rates (x-axis). Anything above the line with the triangles will not have sufficient precision.

    Your data is samples at 100 Hz (10^-2), and the maximum time you can represent without precision loss is a couple of days, but not a month.

    I guess I am not surprised you are seeing shifts at the end of the month. Does this shift occur within the plots or does it show up in another way ?

    Brian

    On Feb 7, 2017, at 2:38 PM, Lukas Janku <lukas.janku<at>pg.canterbury.ac.nz> wrote:

    Hi everyone

    I am having problem (probably) related to some maximum length of the seismic data within Sac.

    I have been dealing with some really crappy seismic data with messed-up timing. 3 Traces from the same sensor. The data had frequent gaps in them and for some reason, the gaps are occasionally of different lengths for different components. I have about 1 month of continuous record (100 Hz sampling). My timing correction included:

    1. Merging all the files from one component together (have a computer with large RAM).
    2. Use some of the recorded earthquakes where I know the p-wave arrival for getting the timing of my 1-month block right.
    3. Splitting the month-long file into day-long files using a script.

    The problem I get is when I use an earthquake at the beginning of the 1-month block (consistent for different components), at the end of my correction process my 3 components are slightly shifted with respect to each other near the end of the 1-month block.

    Does anyone has an experience with similar issue? The only reason I can think off is the different length of the gaps in the data, combined with huge amount of merged data, can make some internal variable 'overflow' at different time for different components... But that's just an impression.

    Thanks a lot

    Lukas Janku
    University of Canterbury
    This email may be confidential and subject to legal privilege, it may
    not reflect the views of the University of Canterbury, and it is not
    guaranteed to be virus free. If you are not an intended recipient,
    please notify the sender immediately and erase all copies of the message
    and any attachments.

    Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
    information.

    ----------------------
    SAC Help (http://ds.iris.edu/message-center/topic/sac-help/)

    Sent from the IRIS Message Center (http://ds.iris.edu/message-center/)
    Update subscription preferences at http://ds.iris.edu/account/profile/


    Attachments
    • Lukas Janku
      2017-02-09 03:58:47
      Hi Brian

      Thank you for your reply. I was suspecting something like that. The shift shows up when I use the function CUT on the 'timed' signal blocks - I use a script to extract 10 minutes signal blocks based on a table with local earthquake times. In general the shift gets stronger the further you are from the timing event, which would suggest the float precision error accumulates over multiple blocks.

      Gonna try to work around it by calculating the time shift and then using script to merge and split the files in shorter chunks.

      Thanks

      Lukas
      ________________________________
      From: sac-help-bounce<at>iris.washington.edu [sac-help-bounce<at>iris.washington.edu] on behalf of Brian Savage [savage<at>uri.edu]
      Sent: 09 February 2017 01:47
      To: SAC Help
      Subject: Re: [sac-help] Problem with merging large files

      Lukas,

      I would suspect that you are running up against the precision of the floating point value used for timing in SAC. See the attached plot. It shows the maximum time value (y-axis) without loss of precision for a variety of sampling rates (x-axis). Anything above the line with the triangles will not have sufficient precision.

      Your data is samples at 100 Hz (10^-2), and the maximum time you can represent without precision loss is a couple of days, but not a month.

      I guess I am not surprised you are seeing shifts at the end of the month. Does this shift occur within the plots or does it show up in another way ?

      Brian

      On Feb 7, 2017, at 2:38 PM, Lukas Janku <lukas.janku<at>pg.canterbury.ac.nz<lukas.janku<at>pg.canterbury.ac.nz>> wrote:

      Hi everyone

      I am having problem (probably) related to some maximum length of the seismic data within Sac.

      I have been dealing with some really crappy seismic data with messed-up timing. 3 Traces from the same sensor. The data had frequent gaps in them and for some reason, the gaps are occasionally of different lengths for different components. I have about 1 month of continuous record (100 Hz sampling). My timing correction included:

      1. Merging all the files from one component together (have a computer with large RAM).
      2. Use some of the recorded earthquakes where I know the p-wave arrival for getting the timing of my 1-month block right.
      3. Splitting the month-long file into day-long files using a script.

      The problem I get is when I use an earthquake at the beginning of the 1-month block (consistent for different components), at the end of my correction process my 3 components are slightly shifted with respect to each other near the end of the 1-month block.

      Does anyone has an experience with similar issue? The only reason I can think off is the different length of the gaps in the data, combined with huge amount of merged data, can make some internal variable 'overflow' at different time for different components... But that's just an impression.

      Thanks a lot

      Lukas Janku
      University of Canterbury

      This email may be confidential and subject to legal privilege, it may
      not reflect the views of the University of Canterbury, and it is not
      guaranteed to be virus free. If you are not an intended recipient,
      please notify the sender immediately and erase all copies of the message
      and any attachments.

      Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
      information.


      ----------------------
      SAC Help (http://ds.iris.edu/message-center/topic/sac-help/)

      Sent from the IRIS Message Center (http://ds.iris.edu/message-center/)
      Update subscription preferences at http://ds.iris.edu/account/profile/


      This email may be confidential and subject to legal privilege, it may
      not reflect the views of the University of Canterbury, and it is not
      guaranteed to be virus free. If you are not an intended recipient,
      please notify the sender immediately and erase all copies of the message
      and any attachments.

      Please refer to http://www.canterbury.ac.nz/emaildisclaimer for more
      information.

08:21:03 v.9a3bf08e