For the three tested shapes, linear deflection 0.01 mm with angular deflection 0.3 rad gave a practical balance: the measured maximum deviations stayed below 0.01 mm, with output sizes that varied by shape. Treat this as a starting point to test, not a universal setting.

What was tested and how deviation was measured

The results.csv file records 45 STEP-to-STL exports: SIMPLE, MEDIUM, and COMPLEX, each tested across five linear deflections (0.1, 0.05, 0.01, 0.005, and 0.001 mm) and three angular deflections (0.5, 0.3, and 0.1 rad). The comparison images show the SIMPLE and COMPLEX shapes; the CSV also includes MEDIUM.

For each saved STL, 20,000 deterministic, area-weighted points were sampled across its triangles. The distance from each point to the corresponding source STEP boundary faces was measured; the CSV reports the mean and the largest sampled distance. The maximum is not a proven global maximum or the requested deflection value.

Results at the selected setting

Linear deflection: 0.01 mm; angular deflection: 0.3 rad

ShapeTrianglesFile size (KB)Largest sampled deviation (mm)Watertight
SIMPLE39619.40.0049true
MEDIUM5924289.30.0074false
COMPLEX386241886.00.0099true

Coarse and fine mesh comparisons

A simple cylinder shown as a coarse triangulated mesh beside a much finer mesh.
Simple-shape comparison: coarse and fine tessellations.
A complex STEP shape shown as a coarse triangulated mesh beside a fine mesh.
Complex-shape comparison: coarse and fine tessellations.

What the measurements support

  • At the tested setting, the three shapes produced different triangle counts, file sizes, and sampled deviations; results are shape-specific.
  • MEDIUM was not watertight in any of its 15 tested settings, so this was not specific to the selected setting. A separate topology diagnosis found zero-length boundary self-edges and overused mesh edges at outer-corner fillet junctions in the examined outputs.
  • At 0.01 mm linear deflection and tighter, changing angular deflection from 0.5 to 0.3 rad did not change triangle count for any shape; at 0.05–0.1 mm linear deflection, angular changes altered triangle counts.
  • For COMPLEX at 0.3 rad, changing linear deflection from 0.01 to 0.001 mm cut the largest sampled deviation about 9x (0.0099 → 0.0011 mm) while making the file about 10x larger (1,886 KB → 18,478 KB).
  • These measurements compare saved meshes with STEP surfaces. No parts were printed, so print quality, fit, strength, and manufacturing performance were not tested.

If your exported STL is not watertight, you can check it with STL Repair.