in figures 11.6, 11.7, and 11,8. The negative areas are better defined. The 

 ridge along the vertical axis is weakened although there is still a trace of 

 column noise. For Data Set 2A, the covariance surface actually becomes 

 slightly negative on the vertical axis. The covariance surfaces still need 

 some minor corrections, but they will not be too difficult to make. 

 The spectra for the reduced data sets 



The spectra for Data Sets 2A and 3C (in terms of variance) and the sum 

 of the values for the two (in terms of E value) are shown in figures 11.16, 

 11.17, and 11. 18, There are no negative values I The numbers at the grid 

 intersections should be divided by lO'* to put them in units of (ft) . The con- 

 tours are labeled in units of (ft) , and as mentioned before they should be 

 interpreted as the integral over the spectrum on a square of the same size as 

 the grid of the plotted numbers. 



These spectra definitely show the effects of column noise. There is a 

 strong ridge along the horizontal axis of the spectral coordinate system. The 

 spectrum for Data Set 3C shows a decrease in the effect of curvature in 

 producing high values at the origin. 



In general the above two spectra appear consistent with each other. The 

 0.0100 and 0.0050 contours are in roughly the same positions on the two 

 spectra. The peak in the spectrum for Data Set 2A has a value of 

 0.2052 (ft)2 whereas the corresponding value in the spectrum for Data Set 

 3C is 0.0797. The ratio of 0.2052 to 0.0797 is equal to 2.57. 



For Data Set 2A, the number of degrees of freedom is given by equation 



(11.8). 



185 



