At the very beginning of the calculation (especially of a Patterson function), the GraphEnt map may show a great deal of peaks, which later on disappear giving a map with essentially only the origin peak. Then new features start to emerge slowly, which more often than not, remain till the end of the calculation. This is an artifact of how the program contours the section that is plotted in the graphics window : GraphEnt will always plot with the first (dashed) contour at the mean, and then every 0.5 rmsd of the given map section (and not of the whole map). At the very beginning of the calculation, the GraphEnt map is almost uniform, but because the program contours the plot from the mean and every 0.5 rmsd (however small this may be compared with the mean), the graphics window will show peaks (which in reality are just slight modulations of an otherwise uniform map). As the calculation progresses the major features start appearing (which in the case of Patterson functions is the origin peak), and then as the data are being fitted more closely the finer detail starts building up. I will illustrate these events with a series of intermediate GraphEnt maps produced during the calculation of a difference Patterson projection. To show clearly the significance of the mean density, these graphs only show the density along a Harker line (v = 0.5). The data used for this example have been discussed in section 8.1 :
