The dimensions, positioning and coating characteristics of the triangular projection panel are detailed in (Andersen et al., 2001; Andersen, 2004): a diffusing white paint manufactured by LMT allows to obtain an almost lambertian surface (perfectly diffusing), with only a 2.6% difference to the theoretical model.

The removal of screen covers, necessary to perform BRDF measurements, aims at leaving the incident beam path free, while the controlling of its shape is taken care of by the ellipses cut out from the metal sheet.

To minimize the blind zones, these screen covers must present elliptic shapes as well. Their exact geometry was determined following a similar procedure as for the metal sheet:

• First, their theoretical dimensions and positions were deduced by trigonometry on the basis of the intersection of a perfectly parallel beam (reaching the sample at different Qi angles) with the tilted detection surface (accounting for the shift between sample and detection screen base planes.

• Then, using on the results provided by the sample illumination analysis with the actual light source and on the metal sheet ellipses dimensions, adjusted horizontal and verti­cal axes for the screen ellipses were estimated, to which a 2 mm margin was added to avoid edge effects.

• After that, to determine the actual dimensions of the cut out covers, the thickness of the screen had to be taken into account; on the other hand, the covers insertion required a slant between the upper (external) and lower (internal) sides of the screen, chosen unique and equal to 20° to ease the screen manufacturing. To leave the beam’s pas­sage free through a screen of significant thickness, larger upper ellipses are required when the angle between the incoming beam and the screen plane increases (i. e. when I Q — Q0 I increases). The ellipses were thus adjusted accordingly, depending on each one’s incident tilt angle.

• Finally, as the above adjustment was only necessary for the ellipses half farthest from the Qj = ©0 direction, their vertical axes (and thus the blind zones) were reduced by re-centering them to open a passage for the actual beam only, still accounting for the screen thickness and a constant 20° slant.

The elliptic covers are held in place by small and strong permanent magnets inserted in the screen central piece. To achieve their removal and repositioning, a “permanent electro­magnet” (PEM) is used, i. e. a permanent magnet that can be deactivated by powering the surrounding coil. This PEM is mounted on a small wagon running on two rails parallel to the main axis of the screen thanks to an indented belt forming a closed loop. An additional on-board mechanism allows it to move up and down from approximately 3 cm, in order to extract and replace the covers. To ensure a reliable lifting, a mechanical “extractor” was added, using four screw-like pins that get inserted in four slots carved in each cover, shown on Figure 5(a); centering pins were added as well on protruding fingers to ensure a reliable positioning. An extra shift was implemented for the wagon movements to allow the extraction system to have a secure grip on the covers.

The limitations in the rails length made it impossible for this extractor to reach the tip cover. Its handling thus required an additional PEM device, together with some extra commands.

The wagon is driven by a stepping motor, controlled by a specific ISEL micro-controller with a RS-232 interface. A typical cycle of extraction, removal and replacement of a cover is sequenced as follows: [12]

• Wagon positioned out of the beam path and kept in place as long as needed to com­plete the image acquisition and processing phase;

• Wagon moved back above the open hole, PEM lowered, deactivated then lifted up empty, the cover being back in place.

Once the wagon movements were adequately calibrated to position it right above each cover, this new design was tested successfully with hundreds of random extractions at different screen inclinations.

The definitive screen panel is shown on Figure 5(b), where the wagon is in position to re­move the tip cover and where all other covers are missing. Figures 5(c) to 5(f) illustrate the sequence of events taking place when the projection screen obstructs the incident beam path.