We propose a method to determine the required performances of the positioning mechanics of the optical elements of a
beamline. Generally, when designing and specifying a beamline, one assumes that the position and orientations of the
optical elements should be aligned to its ideal position. For this, one would generally require six degrees of freedom per
optical element. However, this number is reduced due to symmetries (e.g. a flat mirror does not care about yaw).
Generally, one ends up by motorizing many axes, with high resolution and a large motion range. On the other hand, the
diagnostics available at a beamline provide much less variables than the available motions. Moreover, the actual
parameters that one wants to optimize are reduced to a very few. These are basically, spot size and size at the sample,
flux, and spectral resolution. The result is that many configurations of the beamline are actually equivalent, and therefore
indistinguishable from the ideal alignment in terms of performance.We propose a method in which the effect of
misalignment of each one of the degrees of freedom of the beamline is scanned by ray tracing. This allows building a
linear system in which one can identify and select the best set of motions to control the relevant parameters of the beam.
Once the model is built it provides the required optical pseudomotors as well as the requirements in alignment and
manufacturing, for all the motions, as well as the range, resolution and repeatability of the motorized axes.
We present the continuous scan operation of the ALBA-NOM as a working mode that allows obtaining low noise in
short time, as well as high accuracy measurements. In the traditional step-scan operation, the position of the probe beam
is kept fixed while many data points of autocollimator are averaged for noise reduction. This operation mode is very
safe, as one has a perfect correspondence between mirror position and measured angle, but it is time inefficient, as it
disregards all the data values acquired during motion, and basically averages data values taken under identical
conditions. On the other hand, continuous scan is less safe in terms of correspondence between mirror position and slope,
especially for NOM systems for which the autocollimator does not accept an electronic trigger. Nevertheless, it is
possible to perform independent acquisitions of the autocollimator and of the linear stage data during a scan, and
synchronize signals a posteriori. This solves the main problem of continuous scan with a NOM. Continuous scan
operation for performing measurements is very efficient for noise reduction per unit time, as it allows integrating every
single data value taken by the autocollimator. In addition, it opens the possibility of introducing pitch variations of the
mirror between scans. This allows obtaining many independent datasets that can be combined using error suppression
techniques to reduce not just noise but systematic errors too. In this paper we report the methods and the main results.