This is the main directory devoted to illustrating prediction error
filtering (deconvolution).  If you are new to SU, start by reading
"The SU User's Manual" and the README in $CWPROOT/src/demos.

The X and PS demos pop up several examples of deconvolution.
	The demo scripts are meant to be READ.  Then you can
	clone what you need for your application.

NOTE: The files deconparams3 and deconparams4 are needed in these
	demos, please do not erase them!  The same is true of the
	C program traces.c that is used to create some traces with
	synthetic reverberations for these demos.
	   
For real examples, see Decontest, Mig, Nmo, Nmostack, Pef in the
	$CWPROOT/src/su/examples subdirectory.

Demos
--------------------------------
XDecon1
PSDecon1

We make some simple data: 8 traces with 2-way reverberations and
a weak tilting reflector hidden underneath the reverberations.  The
spike data is convolved with the minimum phase wavelet specified below.
Then a tiny bit of noise is added--tiny because: (1) the definition
of signal to noise is stringent--based on the biggest amplitude
on the trace and (2) these are single traces, usually you'd have
a cmp gather and the ultimate goal of stacking to combat noise.

In this demo, we do spiking decon first and then do gapped decon
(prediction error filtering).  You are to assume that the decon
parameters are estimated from the autocorrelation shown in frame two.

Notice the emergence of the "hidden" reflector after prediction error
filtering!

Construction of the minimum phase wavelet with Mathematica:
(2-z)(3+z)(3-2z)(4+z)(4-2z)(4+3z)//(CoefficientList[#,z])&
    {1152, -384, -904, 288, 174, -34, -12}
--------------------------------
XDecon2
PSDecon2

Same as in demo1, except here gapped decon comes before
spiking decon.  Again, you are to assume that the decon
parameters are estimated from the autocorrelation shown in
frame two.
--------------------------------
XDecon3
PSDecon3

We use the same data described in the first demo.  But here
we illustrate how to set up a series of tests to determine the
optimal gap for prediction error filtering.
The parameter values used in the successive tests are recorded in
the file deconparams3 and also echoed to the screen.
--------------------------------
XDecon4
PSDecon4

We use the same data described in the first demo.  But here
we illustrate how to set up a series of tests to determine the
optimal length for spiking deconvolution.
The parameter values used in the successive tests are recorded in
the file deconparams4 and also echoed to the screen.

Jack--07/95
---------------------------------------------------------------
XShape
PSshape

We use the data described in the demos above. Here we apply 
sushape to convert the minimum phase data into spike data.
	
---------------------------------------------------------------
Relevant self-docs:   supef, sushape

John Stockwell	 | jstockwe@mines.edu
Center for Wave Phenomena (The Home of Seismic Un*x)
Colorado School of Mines
Golden, CO 80401 | http://www.cwp.mines.edu/cwpcodes
voice: (303) 273-3049  |    fax: (303) 273-3478.
