Two options are available for inputting these numbers: either the keyword RIPVALS is defined and initialised by the user prior to executing CREATE/SPC2; or by default a period of 10 is assumed, together with the values 4, 5, 6, 3, 2, 7, 1, 4, 5, 0.
If a negative period or a period of value larger than 20 is specified, a message indicates to the user that a period of 10 or a period of value mod(period,20), respectively, is employed.
Three additional parameters are user-specifiable: the slope of the wave pattern generated (default 0.0); the phase of the pattern (an integer number n period meaning that the first pixel of the image will have the flux of the nth element of RIPVAL; defaulted to 1); and the dimension of the 1-dimensional image (default 128 pixels).
The table LINES.TBL must be copied into one's own directory for use. It is available in the MIDAS directory containing the source code of the GEOTEST context.
The spectrum generated consists of 660 pixels and repeats the pattern defined by the line positions three times. In the first 220 pixels, the line locations are used as read from the table. In the second 220 pixels, the lines are relocated on the edges of pixels (by subtracting 0.5 pixel units from the initial line positions). In the third set of 220 pixels, a user-specifiable offset is added in order to reposition the lines relative to pixel centers (the default is 0.645).
Should a line extend beyond the boundary of any of the 220-wide segments, then it is truncated at the boundary.
Two types of point spread functions (PSF) are available, namely boxes (default) and Gaussians. The full width of a box (default 1.0) or the full width at half maximum (FWHM) of a Gaussian (default 1.0) are user suppliable. If, for a box-like PSF, the width given is less than 0.001, then a delta function is used instead. For Gaussians, a FWHM below 0.2 pixel is not acceptable. In all cases, the peak value of the PSF is normalised to 1 unit. (Note that this implies that the actual appearance of the PSF may be quite different due to the effects of discretisation due to a finite step size.) Gaussians are evaluated every 1/100th of a pixel step size, (this is necessary in the case when most of the Gaussian is within a pixel), and truncated at 5 standard deviations from the mean.