64 
three dimensional orthogonal co-ordinate transformation of one side to the 
other. This procedure was repeated three times giving four different sets of co 
ordinates of the steel balls. These sets were transformed to the same system and 
the average of the four sets was used in the following computations. 
The floating markkas set on the top of the visible part of the steel ball. It is 
to be noted that the identification of this point was very difficult, because of 
the reflection of light on the ball surface, and dust in the holes. It became evi 
dent that this influenced the accuracy of the co-ordinates, when it was disco 
vered that the standard error of unit weight of the performed resections was 
highly significantly different from the expected value, see chapter 8.3.3. As 
the settings were on the tops of the balls, all co-ordinates were reduced to give 
the position of the centres of the balls. This is probably the point which best 
corresponds to the measured image point in the roentgenograms taken of the 
test object. 
From repeated measurements of the test object we estimate the precision (the 
lower limit of the accuracy), as defined by the standard deviation, of the final 
co-ordinates to be 4 ¡urn. 
Fig. 19. 
Stereoscopic pair of the stereo-microscope mounted on the A 7. The object to be measured 
is placed on the table below the microscope. The table is fixed to the base carriage of the 
A 7, and moved in space by the hand wheels and the foot disc.