Go to the list of seismic processes.
Go to SIOSEIS introduction.
SIOSEIS was used aboard the R/V Ewing during the SIGMA experiment
from 25 August 1996 to 10 October 1996, Steve Holbrook, Chief Scientist.
My favorite pictures:
Brash ice Iceberg Iceberg Greenland
David Reid's photo album
All digital pictures courtesy of David Reid, R/V Ewing 2nd Engineer
photos of R/V Maurice Ewing
Seismic Source Signature
The SIOSEIS onboard processing had several objectives:
1) Provide quality control so that acquisition errors could
be corrected as soon as possible.
2) Provide a stacked section that could be used for geologic
interpretation so the shooting program could be modified
if something "cool" turned up.
3) Provide a copy of the data on DAT tape to take home to SIO.
The 3480s would go to WHOI, the DLC (Danish Lithosphere
Center) would read and reformat the 3480s aboard the Ewing
using ProMAX.
4) Steps 1 and 2 should be done as quickly as possible (at least
as fast as the data was collected by the DMS-2000).
Dr. Graham Kent had made his HP 9000/715 available for the
"realtime" processing along with an 3480 tape drive with stack
loader and a DDS-2 format DAT tape drive. The stack loader was not
used because the HP was able to keep up with the brute stack
processing and inserting a new tape into the drive is faster
without the stack loader.
It was also decided that each 3480 should be in a unique file on
the DATs so that later extraction for the SIO mass store would be
simplified. The SEG-Y files on DAT were written with the SEG-Y
standard violated by using IEEE floating point rather than IBM
floating point. It was also decided that decimating the data from
2ms to 4ms without anti-alias filter was acceptable.
Unfortunately, the DLC's 3480 tape drive was damaged during
shipment and ProMAX was unable to read a test tape using the
3480 attached to the shipboard Sun computer. The shipboard
3480 could not be attached to the DLC Sun because of differences
in computer buses. It was decided that the DLC would read the
DAT tapes written by SIOSEIS.
The first line shot turned out to have a miriad of problems,
one of which was an unbalanced streamer. It was deemed that
a static correction for the streamer depth would help the
stack since part of the streamer was floating and part of it
was 20m deep. The streamer depth is available in the trace 0
of each shot, but neither ProMAX nor SIOSEIS extracted it. To
further complicate the problem, ProMAX ignores trace 0 totally.
Some mechanism for getting depth into a ProMAX and SIOSEIS
form had to be established before tape copying could commence.
To accomplish these goals, modifications were made to SIOSEIS:
1) Parameter OFFLINE was added to SEGDIN to automatically
eject the 3480 after it has been read.
2) Parameter REWIND was added to DISKOA to make "circular"
disk files (useful for realtime QC).
3) Had SEGDIN set a signal on each new tape so that process
OUTPUT would start a new SEG-Y file on every new 3480 tape.
4) Modified the tape change routine to write only 1 EOF when
OUTPUT parameter REWIND is used.
5) Parameter TR0 was added to SEGDIN to write the LDGO/DIGICON
trace 0 to disk since ProMAX ignores the SEG-Y trace 0.
This disk write was removed after 7 and 8 below were completed.
6) Made OUTPUT write trace 0 before each trace 1 when SEGDIN
parameter TR0 is used.
7) Add parameters BGP and CGP (bird-group pairs and compass-
group pairs) to GEOM to define where the birds are on the
streamer (varies on each streamer deployment).
8) Write the "receiver elevation" in the SEG-Y trace header
when BGP is used.
The extraction of the streamer depth uses a subroutine by John
Hopper that extracts the bird ids and depths and then given the
ranges, interpolates and/or extrapolates to the other groups
of the streamer. SIOSEIS uses the SEG-Y trace header location
"receiver elevation" for the streamer depth.
Program list.ewing was modified to dump ALL the different
"sections" of trace 0.
Process SEGDIN was modified to recover from various acquistion
failures:
1) Shots with a zero shot time are dropped. These shots
have a shot number of 0 and are not in the navigation file.
2) SQTP errors cause a shot to be terminated early and do not
have a terminating filemark. SEGDIN previously did not
honor the early traces of the following shot. SEGDIN now
recognizes a new shot without the filemark.
3) A "streamer rebuild" error causes traces 2-12 to have
large crazy trace numbers. SEGDIN previously would skip
to the next shot when this ocurred. SEGDIN was modified
to detect this situation and ignore only the bad traces.
4) SLIC errors cause the LDGO "trace 0 section 11"
not to be updated, resulting in the shot number and shot
times to remain the same until the system is restarted.
This usually results in more than 20 shots with the same
shot number and shot times. The data are fine, as is the
navigation file. SEGDIN was modified to detect this
situation and increment the shot number, though the header
shot time remains constant.
The "opcode error" does not appear to have an effect on the data.
SIOSEIS processing was divided into parts:
1) The brute stack with SEG-Y DAT tape output immediatedly
after reading the SEG-D 3480 tapes. Every 15th shot is
written to a"circular" file.
2) A plot of the current shot in the circular file.
3) A plot of the stacked data every two hours on 8.5x11 paper and
taped to the previous 2 hour plot.
4) An fk migration and plot of the stacked data every four hours.
5) Final plot on the NovaJet.
The brute stack runs on the HP 715 in 15 minutes and is designed to:
- Read SEG-D 20 bit binary tapes and decimate to 4 mils.
- Write every 15th shot to a "circular" file for QC (quality control)
- Retrieve streamer depth and assign streamer geometry
- Write a SEG-Y shot tape with trace 0 (tape copy)
- Trace gather according to common midpoint
- NMO and MUTE based on Hydrosweep water depth
- Stack
- Despike
- Disk output for other use
A subtle processing trick was used in the NMO process in order to
laterally vary the NMO velocity function. Earlier work in the area
showed that a velocity function could be designed for the area.
The velocity varied with water depth. Graham Kent had modified NMO
such that the parameter vtrkwb (velocity track water bottom) modifies
the normal SIOSEIS spatial interpolation by changing the RP number
into the water depth determined by Hydrosweep. The velocity function
was described by water depth so that the spatial interpolation is
done by water depth. SEGDIN has always put the Hydrosweep depth in
the SEG-Y trace header.
The shot plot script generates a Sun rasterfile with the latest shot
from the "circular" shot file. The raster file is displayed on the
screen using xloadimage and is converted to PostScipt so it can be sent
to a printer. With filter and gain, this script runs quickly on
the HP without a noticable impact on the stack script.
The stack plot script generates a Sun rasterfile of 2 hours of data and
sends it to the screen using xloadimage. The rasterfile is also
converted to PostScript and can be sent to a printer. These 8.5x11 plots
were then taped together for a fast permanent display. With filter and
gain, this script runs quickly on the HP without a noticable impact on
the stack script.
The fk script performs FK filtering and migration of 2048 traces
(2000 stacked traces and 48 pad traces), each trace having 2048 samples
(8 seconds). Successive FK segments overlapped by 50 traces so that
the segments would appear seamless when spliced together. The
2048x2048 FK processing took 3-4 minutes on the HP 715 and could be
done on the same computer as the stack while the stack is waiting for
a new input tape from the acquisition system.
The final plot script generates a plot of several FK migrated segments
for the NovaJet plotter. The NovaJet plotter is similar to to HP
DesignJet plotters, so the PLOT parameter nibs 2859 can be used. The
resulting SIOSEIS rasterfile can be translated to an HP-RTL file by using
program SIO2HP. The RTL file is then sent to the Ewing Sun computer
(hess) with the NovaJet plotter using the binary mode of ftp. On hess,
the plot is sent to to NovaJet using the Unix cat command:
cat RTL-file > /dev/bpp0.
Miscellaneous stuff:
OUTPUT:
1) I did not use the auto tape change feature of OUTPUT because:
a) A SEG-Y file will be split, causing the 3480 to be in
two pieces.
b) It's slow because SIOSEIS does not allow the shot to be
broken, so it has to reread the traces written on the current
tape, then backup again before writing the file mark.
c) I found the end of the DAT tapes seem to have lots of write
errors before finding the EOT marker. Rewrites appear to be
VERY slow.
d) There are 5 minutes or more between the end of input tape
and when the DMS-2000 has another 3480 ready, doing a manual
rewind and load of the new DAT tape is efficient. The down
side is that OUTPUT doesn't close the tape with a file mark,
which some systems/programs may choke on. ( I think sioseis
now treats any tape error as an EOF - the "new" SCSI tape
drivers return a -1 for ALL errors, so I can't distinquish
between them anyway!)
Shallow water tricks:
1) Do an "inner" mute (a tail mute of the inside or close traces)
to get rid of the deeper portion of the record that is overwhelmed
by ship noise. SMUTE has ADDWB YES, use it. SEGDIN now converts
the Hydrosweep water bottom depth to water bottom time.
2) Use a VERY severe "outer" mute. For "reatime" purposes, in 400m
water, I used the NMO stretc parameter of .06. That means any
sample with nmo greater than 60 mils will be muted out.
Go to the list of seismic processes.
Go to SIOSEIS introduction.