1s— 
S 
Ve 
-— 295 - 
The non-parallelism x, Ax and y, dy (de5) 
Three repeated observations in each extreme position of a parallax carriage, as 
described under III,1,d0,5, appear to be sufficient. 
Short periodic spindle errors (c.3) 
The observations are made e.ge for each second grating (thus for 0,2 mm incre- 
ments) on the input reference scale, The same gratings should be used in succes- 
sive routine tests. The measurements are repeated twice on each grating and per- 
formed for three full revolutions for each measuring spindle. 
The errors are represented graphically. 
CONCLUSION 
The proposed partial tests for the photogrammetric comparators illustrate the 
practical application of the general principles for the conducting of tests des- 
cribed in (4). These proposals do not pretend to present an ultimate solution to 
the problem, but rather to establish a basis for the further optimisation pro- 
cesse 
Although the detection of individual deterministic errors requires a special 
arrangement of partial tests, a general approach for the conduction seems to be 
possible, For this purpose, the error sources have been surveyed, analysed and 
classified into four principal groups. Instrumental error-sources which are 
physically interrelated belong to common groups. 
The partial tests must generally be carried out in a proper sequence, such that 
the error sources tested later do not affect the results of the preceeding tests. 
The conduction has been restricted to comparators provided with mechanical mea- 
suring standards. For the laboratory tests, only the objectives interesting for 
the users of comparators have been considered. 
The relation between the input structure and the sample size of the observations 
made for a test is, to some extent, flexible. 
Usually a choice can be made between larger input structures (many reference 
points) and fewer repetitions, and smaller input structures but increased number 
of repeated observations, The optimum proportion of both is generally not uni- 
guely defined, 
Tests based on smaller input structures and on many repeated observations result 
in a simple processing of the data. Hence, they are convenient for the routine 
tests. The number of the repeated observations required can be estimated by means 
of the statistical tools, 
Larger input structures in the tests facilitate refinements of the correction 
models, Thus, they are suitable for the laboratory tests, 
Although the scopes of the laboratory and routine tests are different, a proper