Our ability to harness the power of X-ray radiation has led to many important scientific discoveries. Two common applications for X-rays are crystallography and imaging through objects (tomography). Facility design is a critical component in optimizing a system’s performance for an application and insuring operator safety. In the first part of this dissertation, the design, construction, and performance of an adaptive X-ray research facility is discussed, with emphasis placed on development as a future X-ray interferometry system. Performance is evaluated from thermal measurements, measuring characteristics of detectors, measuring characteristics of the X-ray generator, and by demonstrating successful diffraction and reflectivity experiments.One of the fastest growing fields in metrology is part measurement using industrial X-ray computed tomography (CT). The ability to calibrate the system for accurate measurements to date has relied on physical reference objects. A novel, direct silicon bonded reference object for X-ray CT was developed and is discussed in the second part of the dissertation. This reference object is constructed by bonding four etched silicon pieces from a single wafer. Within each reference object are two sets of patterns grouped into linear and curved features, with features ranging in size from 5 μm to 1200 μm. The standalone silicon object measures 4 × 4 × 2 mm3, or 6 × 6 × 4 mm3 when encapsulated in a silica and epoxy protective shell. X-ray images of the reference object were taken, reconstructed into a 3D model, and features were extracted. A different part, cut from the same wafer, was measured by coherence scanning interferometry (CSI) to perform a simple comparison study to CT data. Two features were extracted from the CT and CSI data, one with an edge constructed by a cubic polynomial (asymmetric cubic), and the other having a smooth, continuous curved edge (chain-link). CSI measurements were within 0.7 μm of the nominal design, while CT measurements were up to 3.1 μm from nominal. Two algorithms were used to measure the cubic curve, both programmed to minimize the tangent-normal residuals. One algorithm calculated best fit coefficients and the other used the nominal coefficients. The 1σ deviations for the CSI data were with 0.12 μm and 0.14 μm respectively, and for the CT data were 0.32 μm and 0.39 μm respectively. With the chain-link feature, circles were best fit to the outer edges. Similar results were seen with 1σ deviations of 0.13 μm for the CSI data and 0.4 μm for the CT data. Use of a high-resolution X-ray CT system with smaller voxel sizes should improve the CT results.