Go to the list of seismic processes.      Go to SIOSEIS introduction.

EW9914 example

Cruise EW9914 aboard the R/V Maurice Ewing was in the Lau Basin with Chief Scientists Alistair Harding, Graham Kent, and John Collins.

Brute stacks were displayed on the Atlantek plotter immediately (minutes) after EACH 3490E tape was ejected by the Syntron recording system. This rapid processing scheme allowed the scientists to find magma chambers beneath the ridges and modify the cruise plan accordingly.

    There were three onboard processing objectives:
  1. Near real-time brute stack.
  2. Tape copy and reformat.
  3. Quality control.

The Syntron recording system wrote 145 shots onto IBM 3490E tapes. With a 10 second recording length, a 4 millisecond sample interval, 480 recording channels and a 37.5 meter shooting interval, tapes were filled in 45 minutes. The tape copy and reformat took 13 minutes and the brute stack took 22 minutes, leaving 10 minutes left over.

Immediately after completion on the Syntron, each 3490E tape was taken (sneakernet) to the Sun SparcStation 10, "heezen", that has a 3490E tape drive. Each tape was read twice; once for the brute stack and once to copy and reformat to SEG-Y onto DLT tapes. Doing the tape copy and stack in separate scripts allows greater flexibility in changing stack parameters and restarting the stack script.

The stack script ultimately generates a plotfile that can be plotted on the Atlantek plotter while the stack script is still running. Thus, the stacked seismic line can be viewed while it is being collected. No need to wait until the end of the line!

The brute stack script does not use all 480 traces (6 km) because the SparcStation 10 is not fast enough to perform both tasks (reformat and stack) before the next tape comes from the Syntron system. Initially we used alternate traces from each shot as if the hydrophone spacing was 25m rather than 12.5m. While we used all 6km of the streamer, the far range traces contributed to excessive "stretch" due to out-of-plane scatterers and incorrect velocities. (Process segdin parameters ftr 2 ltr 480 trinc 2). It was decided to use every trace of the near 3km of the streamer in the stack (process segdin parameters ftr 241 ltr 480 trinc 1 (remember 480 is closest to the ship)). This also changed the subsurface spacing to 6.25m.

Program atlantek was run on the Sun Enterprise Server (named "grampus") that has the Atlantek thermal printer attached as: /dev/ihcp0. The sioseis plot file generated by the stack script on SparcStation "heezen" is written to a disk that is NFS mounted by "grampus". Program "Atlantek" continuously checks for the plotfile being appended and plots the new information as soon as it is written on "heezen". The program doesn't know when the plotfile is really complete, so it must be terminated with a control-c. The program should be restarted everytime a new line is started by the stack script because the plot filename changes.

The stack script also writes every 50th shot into a circular disk file in /ldata/real-time/shots. It also writes every 50th cmp gather and every 50th moved-out/muted gather to disk so they may be quality control checked.

One of the keys to doing "real-time" stacks is having a velocity function ahead of time and using the nmo parameter "vtrkwb" which allows the velocity function to vary based on the water depth inserted by the Hydrosweep system. Without "vtrkwb", the velocity function is spatially varied according to the SEG-Y shot/rp number; with the "vtrkwb" parameter, "fno" refers to the water bottom depth.

The mute, after nmo, always starts from the Hydrosweep water bottom also.

The very low filter passband of 5x40 caused some wrap around effects in the plot, so the fft length was increased by processing more data. Initially process segdin used parameters: stime 2 secs 4. This made the filter fft 1024 points long with only 24 points padded. The segdin parameters were changed to: stime 1 secs 5, so that the fft was 2048 points long with a 778 point pad.

The brute stack script has several features:

The copy script reformats the SEG-D 3490E tapes onto SEG-Y DLT tapes. Each 3490E tape is written as an SEG-Y file (process output parameter newfile), even though the SEG-Y standard does not allow multiple files on a tape. The output tape further violates the SEG-Y standard by writing in IEEE format.

Both the brute stack and copy scripts utilize segdin parameter OFFLINE, which rewinds and ejects the input SEG-D tape after it has been read. This enables the "operator" to know that a new tape should be "mounted". After the new tape is entered, a file named "in" is created with the tape unit number in it. SEGDIN waits for the creation of this file and starts reading the new tape. SEGDIN then deletes the file "in" so that it waits at the end of tape for yet another tape.

The shot plot script displays the "latest" shot from the circular shot file written by the stack script.

Using the Sun CDE environment on "heezen" simplified switching between the processing tasks. One CDE screen was set up for the tape copy/reformat, one for the stack, one for the QC plots, and one for my personal stuff (e-mail and /opt/NSCPcom46/netscape).

The reformat screen had two windows; one running the script and one for controlling the tape ready file "in". Both windows belonged to "heezen" and were in directory /ldata/realtime/reformat.

The stack screen had three windows; one running the script, one for controlling the tape ready file "in", and one for running the Atlantek plotter program.

I added two files to the stack and reformat directories for controlling the input tape file "in" so that the watch standers did not have to know how to use an editor. File "go" had the input tape unit "43" in it. File "stop" had a "-1" in it. When a new tape was ready the watch could simply type: cp go in The "in" file with a -1 is the normal sioseis termination flag and processes gather and stack flush all buffers.

The reformat script was not terminated with the "-1", rather it was control-c ed when sioseis asked for a new output tape.

I added the following lines to my .cshrc to simplify processing:

set path=($path /export/sioseis/bin )
alias xl "xloadimage -r 90 sunfil &"
alias sd "/export/sioseis/doc/siodoc"
alias s "suntops < sunfil > psfil"

The streamer depths were not placed into the SEG-Y headers during the reformat step because the streamer was towed at a very uniform depth. Processes SEGDIN automatically converts the SEG-D extended trace header into a SEG-Y trace 0 with trace id 28.

User discussion.

Go to the list of seismic processes.      Go to SIOSEIS introduction.