Data Quality

The two images agree in three areas:

• The central region (out to ring 14) where the surface is well adjusted. Figure 10 is the trace along the panel boundary between rings 6 and 7, clockwise from the dish high-point, from the DAY and NIGHT images. The pixel/pixel correspondance is excellent.

  Figure 10: A comparison of a portion the DAY and NIGHT data from the inner part of the reflector. This is the scan along the boundary between rings 6 and 7. AZ=0 is the high point of the dish.

  

• Ring 14 stands out as a problem; it is depressed by 1 to 1.5 mm over most of the antenna, probably more in the top left quadrant (sectors 1-6).

  In the first three quadrants (clockwise from the top) ring 14 is depressed by about 1.5 mm. In the last quadrant the depression is larger. Visual inspection of the area near az=-45 degrees showed a step of 6 mm between rings 14 and 15; and 2 mm between rings 13 and 14, in excellent agreement with the images. An abrupt transition of this magnitude will produce Gibbs effect ripples in the phase, as is observed.

• The outer rings (15-17) have substantially higher rms than the inner section.

  Here again the top left hand quadrant is worst affected. A scan in azimuth, averaging rings 16 and 17 (figure 11) indicates a step of several mm near az = -45 degrees.

  Figure 11: This is a scan around the antenna surface, averaging the surface error of the outer two rings.

  

The two images disagree in two areas:

• In the central region (rings 1-13) there are a number of circular local extrema which align with the ring boundaries. These are maxima in the NIGHT image, and minima in the DAY image. This point is shown more clearly in figures 12 and 13 where we have computed the azimuthal average over a number of annuli concentric with the map centre. The panel boundaries are shown as dotted lines. It is clear that the panel mid-points are raised above the panel edges during the day, and lowered at night.

  The rms error contributed by this effect is $\sigma \sim 0.25mm$.

  It is also clear that the radial profile defined by the panel boundaries is the same for both DAY and NIGHT images to within 0.1mm (rms). The profiles defined by the panel mid-points differ by 0.5mm.

  There is a natural explanation for this panel movement: it is thermal in origin, due to the differential expansion between the aluminium panels and the steel backup structure. (See, eg, Christiansen and Hogbom, 1985, p. 52).

  If this thermal hypothesis is confirmed, then one might expect the effect to have a zero-point defined at the time the panel is locked to its mounting studs. This could imply that the surface could be optimised for night-time (low temperature) operation.

  Figure 12: The azimuthally-averaged surface error radial profile for the May29 (day) observation. The vertical dotted lines mark the panel boundaries. Note that the panel mid-points are all higher than the panel edges

  

  Figure 13: The azimuthally-averaged surface error radial profile for the June12 (night) observation. The vertical dotted lines mark the panel boundaries. Note that the panel mid-points are all lower than the panel edges

  

• There is a large scale deformation of the surface, very roughly astigmatic in appearance in the DAY image, with the top and bottom sections are raised relative to the mean surface. At night there is a smaller effect aligned along the elevation axis.

  It is not clear whether this is a genuine mechanical effect (the surface really is deformed), or an artefact of the analysis. However, the better quality of the calibrations, as well as the cleaner apprearance of the NIGHT amplitude image suggest that the NIGHT image should be used as the basis for the panel adjustment settings.

  The panel-to-panel displacements are little affected by these deformations - that is to say, one can still make meaningful statements about the panels once the deformations are recognised. Great caution is nonetheless advisable.