Although not always seen in low S/N data, there are two patterns (unfortunately not fixed) which are always present. The first, and least obvious, of these arises because of interference between the read beam and an FOC digital timing waveform which has a 4 pixel period. It appears in the data as a vertical striped pattern. Although very coherent with regard to orientation and frequency, this pattern is inconsistent from image to image.
Second, and more often seen, is an approximately sinusoidal pattern which
produces a striping orientated at about 45 to the frame axes (see Fig.
3). This pattern has a period of
3.35 pixels (F/96) and is believed to
originate from a Moiré effect between a TV tube grid and the diode array on
the target. The peak deviations of this sinusoid can be as much as 10%or
more and it seems to intensify when countrates are in the nonlinear regime. In
fact, it can be used as a quick way of recognizing the presence of significant
nonlinearity in an image since it does not appear to be present in very low
count rate images (<0.01 counts/pixel/sec).
One final pattern effect which should be mentioned, although it is not a true
pattern, is the effect which we call the ``thumbprint''. This feature is often
seen by observers in the low S/N background of their geometrically
corrected data. In appearance, it looks like a fingerprint and has a relatively
low spatial frequency, typically with a period of 20-30 pixels.
This is not a true pattern since it does not represent background variations.
Instead, what is seen are areas of higher and lower variance in the Poissonian
noise characteristics of the background. This comes about as a result of the resampling algorithm used in the geometric correction. Essentially, if an output pixel maps directly to the center of a pixel in the input image then little or no smoothing is involved. However, when they map to points between pixels in the input the output value is a weighted average of the set of four input pixels involved (in order to ensure flux conservation). The result then is a mild smoothing, and thus a small reduction in the variance. The pattern is identical in all images which use the same geometric correction file.