ABSTRACT
This preliminary study performed on a test sample shows for the first time the capability to perform 3D STIM-T experiments at ITN. Inefficiencies and inaccuracies in the experimental setup at the time of acquisition of the data here presented as in all data off-line analysis steps were clearly identified and are currently being improved. In fact, the STIM-T setup here described was abandoned and presently, a completely new sample stage is being installed.The sample rotation is now achieved through a motorized stage with a minimum angular displacement of 0.45◦ and positioning precision of less than 0.35◦. This stage accommodates the sample holder, consistingof a 2 axis limited tilt (of ±17◦ and ±20◦) precision goniometer, mounted on top of a XY-table with a ±5mm range and micrometer precision. Finally, the sample itself is placed on a needle clamped to the head tip of the goniometer setup. An alignment procedure aiming at a stable, non-precessing rotation has been devised and implemented. Using a micro-TV camera coupled to the motorized stage control computer the sample is brought to close alignment with the (vertical) rotation axis by repeatedly acting on tilt axes of the goniometer head. Next the sample is rotated by the motorized stage and the XY displacements used until no lateral motion is perceived of its geometric rotation centre line. Even without this new setup the digital corrections devised for the pre-processing of projections have been demonstrated as a necessary, extremely valuable and powerful tool, capable of coping with most of the mishaps of a non-optimal setup. Energy calibration, removal of noise and spikes from the projection data as well as vertical and horizontal alignments worked fine. Only the rotation centre correction was not so successful because of the significant sample precession and few projections taken in this exploratory experiment. As a result tomograms are poor in revealing sample details and are marked by a considerable amount of star-like and straight line artifacts, which were reduced by pre-processing the tomogram but not completely avoided without losing sample data aswell. Considering all these problems the 3D image reasonably reproduces the expected outer shape of the fibre and allows identifying the position of the SiCgrains, although reconstruction is far from perfect. Grains size and the small number of projections in the π angular range, made it difficult to achieve a better reconstruction. For the purpose of isolating sample elements and separately calibrate them as areal density, the background removal algorithm was of great
usefulness, making use of intermediate edges images, corresponding to different sample masses, and arriving to a final edges image, withreliable whole outer contour of the sample.
