Also the results of the connections on grids confirm the preceding conclusions: they show, 
however, the presence of other residual systematic errors in the formation of the model. 
The torsion presented by both strips on grids and the general curvature, particularly 
evident in the second strip, are due to the systematic errors in x and in co, found in the single 
models. 
On the contrary, the scale variation, particularly evident in the first strip, and the error in 
height, that gradually increases along the strips and appears in both of them, clearly denounce the 
presence of a systematic convergence error that cannot be seen in the results of the first part of 
the research. However, this error cannot be larger than io", which we can deduce by ascribing 
the whole distortion in X of the first strip only to the systematic error of convergence. 
On the other hand, we must point out that the total closure errors, which include the effects 
both of an eventual disorientation of the first model and of the systematic and accidental errors 
in the connection, are very small if compared to the length of the connection itself. Moreover, the 
errors of both strips present so regular a behaviour that we are led to suppose that an adjustment 
with interpolation formulae of II degree would reduce them to very small values. 
What has been deduced so far from the analysis of the strips on grids, can also be deduced 
from the study of the behaviours of the orientation elements in the single models. In theory, they 
should all be null; on the contrary, in practice, they behave as shown in figures 3 and 4. 
We must point out, besides the presence of a systematic error of A<p (decreasing value of b z ), 
a remarkable systematic difference in the behaviour of the instrument in the position of «base 
in » and « base out ». 
The angular elements have variations that, on the average, are larger than i c . Of course, these 
variations remarkably influence the orientation elements b x and b z , The cause of these variations 
is not very clear, but it is probably related to the systematic measure errors of the plate-coordinates: 
for instance, to eventual errors in the measure screw. The variations of the transversal parallaxes 
producing in the model a <p variation of i c , are about 5 ¡um. If we consider that in the AP/C 
the parallaxes are obtained from the comparison of two coordinates, we can deduce that the 
coordinates themselves are affected by an error smaller than 5 ¡um, as guaranteed by the 
building firm. In fact, the screws of the AP/C are not provided with a correcting device for 
the periodic errors. 
In short, these experiments testify that the AP/C has given very good results, which can be 
compared with those of a precision plotting apparatus. 
3. - The tests on real strips have been executed using the photographic material of Commis 
sion B (O.E.E.P.E.), available at the Centro of Milan upon permission of the President of the 
Commission, that is strip 2.6.4. of the Reichenbach Polygon, taken at the relative flying height of 
1.200 m, with a Wild RC 7 camera, on plates, focal length 100.35 mm > size 150 X 150 mm 2 . The 
ground flown over is very uneven with height differences that, within each model, are larger than 
20 % of the flying altitude. In the strip area 61 check points marked on the ground have been 
uniformly distributed; furthermore, there are several other natural points of known height. 
Strip 2.6.4. has been triangulated twice with the same procedure that may be outlined in 
the following points. 
Interior orientation. - We have obtained it by measuring the coordinates of the four fiducial 
marks represented by holes placed nearly in the center of the four sides of the plates; in fact, the 
principal point of each plate, to which all the plate coordinates are referred, is that point whose 
X coordinate is the mean of the X coordinates of the two fiducial marks placed on the two plate 
sides normal to the direction of the X axis and the Y is the mean of the Y coordinates of the 
other two fiducial marks. No principal point correction or objective lens distortion have been 
introduced into the computation program. In fact, both the correction of the principal point and 
the distortion curve given by Wild are very small; moreover, preceding experiments on the same 
material let us suppose that the errors in the determination of the distortion be of the same entity 
as the average measured distortion. 
9