model are then measured in all grid points. From these figures a stand-
ard altitude error is calculated and accepted as a measurement of the
accuracy obtained. Unfortunately this figure is not a true standard
error saying something about standard errors to be found in practical
work, and the reasons for this are as follows.
The residual y-parallaxes were not eliminated i. e. the grid model was
not formed by relative orientation,
'The test measurements were made for a single special plane, but in
practice, models do not coinside with this plane, especially those of hilly
terrain,
'The test base used in some cases will give rise to an overlap less than
50 %.
From the above it is evident that the whole model volume ought to
be more realistically tested, and standard errors calculated from such a
test.
[n short, the common procedure is:
Simple measurements combined with trials to eliminate errors one
after the other. Finally a great number of grid measurements used only
to obtain standard errors.
A much better method is to start with complete grid measurements,
and from these calculate the required corrections at each adjustment
point. After introducing these corrections it is sufficient to check in a
limited way to ensure that no mistakes have occured. From the grid
measurements, it is possible to calculate standard errors to be expected
in the instrument after adjustment.
Simple calculations, using the final check figures, will show whether
the calculated accuracy has been obtained.
The calculations involved in the method sketched are of considerable
magnitude, and can therefore only be economically processed by means
of an automatic computer. A program of these calculations has been
designed to suit the Swedish electronic computer FACIT EDB in
Stockholm. At present the program contains formulae for 17 adjust-
ment points i. e. 17 unknown and their standard errors for the Wild
A 7, but formulae for any instrument may be added without difficulty.
Dy means of a set of parameters the program is prepared for any group
of measured grid points. Experience will indicate the most favourable
group of measurements. As well as for mechanical instrument correc-
tion, the results may be used as mentioned in the first paragraph. It is
also possible to solve the problem with arbitrarily chosen fixed adjust-
ment points. This is convenient if some adjustment points are correct.
After grid measurements the instrument is used in a normal way and
corrections are made later when the results have been returned from the
automatic computer,
In [1] Hothmer has studied the effect of errors in gimbal axes but
according to this report the results were inconclusive. His investigation
3