172
In general the conclusion can be made that the mean error in this
manipulation will not exceed o,oi mm if carefully executed.
The adjustment of a photograph in the plate-holder.
To estimate the accuracy of this manipulation, 5 points were marked
on a photograph according to the scheme of fig. 3. The centre-point
was brought 20 times to coincidence with the mark on the fixed glass-
plate. To fix the result of each centring directions were measured
to the points 1 to 4 by which the co-ordinates of the rotation-centre D
could be computed with respect to the axis through the four points.
The discrepances with respect to the average value are a measure for
the accuracy.
The average values for the co-ordinates of D were 12 and 36
microns. These amounts show that the centring of the fixed glass-plate
was in this case rather bad. The mean errors in the setting of the photo
graph were:
mx 5,0 micr, my 5,9 micr.
Conclusion.
Since both above mentioned centrings are independent it can be
assumed that the mean error in the position of the chosen measuring
centre with respect to the rotation-centre may be computed to
m J/ 102 62 ca 12 micr.
To give an idea of the influence of this amount it will be compared
with the accuracy of the computed data for the estimation of the
isocentre or the plumb-point.
The distances from these points to the principal point are resp.
1/2 and f.i.
The mean square errors in these distances are therefore
w,2 f2 in? and ;«22 2 in?.
Assuming that 10 cm and that mx m» 12 micr, there
follows
nii 1,3 cgr, resp. nii 0,7 cgr.
In general can be said that the accuracy of the estimation of tilts
has not yet reached this height so that in this respect the radial-trian-
gulator furnishes no substantial contribution to the inaccuracy of the
triangulation.
The accuracy of the X-movement.
In the above mentioned treatment it is assumed that the displace
ment of the plate-carriers is rectilinear and gives no rise to secundary
rotations. These rotations, if they occur, don't express themselves in
the circle-readings and are therefore dangerous sources of errors.
For the estimation of these errors a mounting was chosen as out
lined in fig. 4.
