Using yulewalker.m with Autocorrelation Input

The difference between time signals and autocorrelation in section 6.12.1.1 and here is that the input would be an autocorrelation instead of a seismic signal. We use the BSU program bxcr to compute an autocorrelation data set in BSEGY format. Fore example, using the same data as above, we issue the following commands from an xterm within the directory with the data set of interest.


bxcr twave.seg twave.seg 0.0 0.5 0.1
bstk bxcrtwav.seg


The first command cross correlates the data set twave.seg with itself (ie. an autocorrelation). The gate is 0 to 0.5 seconds, for a maximum correlation lag of 0.1 seconds. This is followed by a stacking or averaging of all the autocorrelations into a single signal with program bstk. This averaged autocorrelation is then replicated so that the input and output BSEGY files have the same number of traces. In generating a spectrum, it does not matter which one of the traces from file bstkbxcr.seg we use. They are all the same. In Figure 51 we again use trace 30, but could have used any other equally well.

Figure 51: (A) Picked portion of autocorrelation. Sets spectrum order at 156. (B) Input file from bstk of bxcr. (C) Plot of the selected trace 30. (D) All pole amplitude spectrum.
\includegraphics[scale=0.9]{FigureW}

In both of these examples of Yule-Walker spectra, the Octave programs are computing amplitude spectra (not power spectra). The square root is taken of the power spectrum before plotting.