points in the standard positions; they have been generated with 
e - 4 ym. The scale of the photographs is 1:10.000, the principle dis- 
tance is 153.210 mm and there is a 60% forward overlap and 20% sidelap. 
The base length is approximately 900 m. 
The network is indicated by a thick line in the figure at the block 
perimeter. The length of each side is approximately 900 m. At all sta- 
tions, directions have been generated with Or - l mgon; for all sides, 
distance measurements were generated with 9i 2 1.5 cm/km. The horizontal 
coordinates of the stations were computed with respect to a computa- 
tional base (S-base) which consists of two corner points (see /3/, /11/, 
/12/). The maximum standard ellipses of the computed coordinates occur 
at the other corners. The semi-major axes are approximately 7 cm 
(compare /13/). The reliability of this network is poor (compare /13/). 
For the distance measurements, boundary values up to 1.40 m are found 
for a = 0.001 and B = 0.80 (see /2/). The effect of such errors on the 
coordinates is in the same order of magnitude. 
Horizontal control points for the bundle block were taken from this net- 
work at a distance of twice the base length. There are height chains at 
the beginning and end of the strips and across the middle of the block, 
with points at a distance of 2-base length. In the block adjustment, the 
horizontal and vertical controls were considered as not stochastic. 
The adjustment of the block gave an estimated 6 = 4.33 m. This is 
somewhat larger than the original % because the ground control is taken 
as not stochastic in the computations. In this generated block, check 
values were available for all computed coordinates, from which the root 
mean square errors were computed for the residuals of the x, y and z- 
coordinates, see figure 2. Because separation was made for points at the 
perimeter and points inside the block, the results in terrain scale were 
(in cm) 
Paz "y uy, Mey 
perimeter 7.0 7.7 13.8 7.4 
inside 5.6 6.2 12.1 5.9 
The PATB program has the option of self-calibration, for which we used 
the Ebner set of parameters /6/ (figure 3). In this adjustment, self- 
calibration was not used, but the program can still compute test values 
for the parameters. These are equal to the parameter value that the 
parameter would have if self-calibrated divided by its standard 
deviation. For the first two parameters, bi and bo; these values were 
0.05 and -0.17, respectively. This means that in a test with o - 0.001, 
the parameters are not significantly different from 0.00, so that self- 
calibration is indeed not required in this case. 
3. THE EFFECT OF A DISTORTED NETWORK 
The network was adjusted again, but with an observational error intro- 
duced in the distance measured for the side indicated by an arrow in 
figure 1. The magnitude of the error is 75 cm, which gives a test value 
= 1.85 when data snooping is used. Thus when testing is done witha = 
- 113 -