Applying a filter

For a more time-series oriented example, how about convolving a dataset with a filter? First create the filter by entering a list of numbers into a text file (using your favorite editor). Call it ``filterfile''. Then use the SEPlib utility ``Dd'' to convert that ASCII list into a SEPlib floating-point data file:

Dd in=filterfile ein=0 esize=4 > filter.H

Note that in this example we specified where the actual data was by specifying ``in=filterfile'' directly on the command line. The command-line parameter ``ein=0'' tells Dd that the input is ASCII. Dd then creates an output SEPlib history file with esize=4 (meaning floating-point binary data) and n1 equal to the number of entries found in ``filterfile''.

Having created a history file for the filter, we can apply it to the data in Txx.H using the program Filter:

Filter < Txx.H filter=filter.H > Txx_Filtered.H
Wiggle < Txx_Filtered.H par=plotpar | Tube

Here the file filter.H is an ``auxiliary input file''. (Note if the data file pointed to by the history file filter.H contained just one entry, a one, then Txx_Filtered.H would be an unmodified copy of Txx.H.)

Auxiliary files are used by many programs (both for input and output) that require more inputs and outputs than the standard one-file-in one-file-out SEPlib structure allows. Usually one or more axes of the auxiliary files correspond to axes in the input or output files. For our example, axis 1 in the input history file Txx.H corresponds to axis 1 in filter.H; thus they should have the same sampling rate, and the parameter d1 in both files should ideally match. As it turns out Filter doesn't check for this. Some programs do, however; and in some SEPlib programs the standard parameters (n1, n2, n3, d1, d2, d3, o1, o2, o3) found in the auxiliary files may be required and significant.