3D Scanner Match - two synthetic passes and rigid registration Original CC0 1.0 example. Coordinates are millimeters. No physical scan or noise. Pass A covers x=0..60 mm, pass B x=40..100 mm, y=0..20 mm, z=0 or 6 mm. The overlap is x=40..60 mm. Each pass has 682 points; the combined CSV keeps all 1,364 points with A/B labels. Overlap duplicates are retained, not welded. These are point coordinates, not a fused, watertight or CAD-ready mesh. Pass B is transformed by a known +8-degree global Z rotation and [3,-2,1] mm translation. Six named, noncoplanar common landmarks supply exact point pairs. The script reloads exported CSV coordinates and fits a rigid transform by SVD. There is no correspondence search, ICP, outlier rejection or scanner calibration. A second result deliberately applies +1-degree Z rotation and [1,0,0] mm shift after registration. It shows why a separate end-to-end check is needed. The left control is point A-0000 at [0,0,0] mm in A; the right is point B-0660 at [100,0,0] mm in B's original frame. Original span is 100 mm. Both fit residual and span are reported. Reproduce in a fresh output folder: python reproduce.py new-registration-result-001 Checked with Python 3.11 and NumPy 2.3.3; install NumPy separately if needed. Original inputs, combined coordinates, transforms, report and script are included. All output paths are new; the script never cleans existing files. Numerical residuals approaching zero are expected for this exact noiseless fixture. They are not an achievable accuracy claim for a real scanner. A low landmark fit residual alone does not prove correct coverage, units, local feature quality or long-span error on real data. Do not use the fixture's span-error value as a product benchmark. Symmetry, wrong correspondences, surface changes, noise and deforming parts are not tested. No GUI or CloudCompare run was made.