Maybe this AI-generated description can help? I'm not working on the Maya code so I haven't checked what that code actually is doing.
Based on standard NURBS tessellation terminology, these options can be understood as follows:
| Sample Type |
What it does |
Typical Result |
| Entire Domain |
Treats the entire U/V parameter domain as a single region and distributes samples across it. |
Most uniform sampling, but may miss local detail in areas with uneven parameterization. |
| G1 Segments |
Splits the surface at G1 (tangent continuity) boundaries and samples each smooth patch independently. |
Better preservation of creases, fillets, and tangent breaks. |
| Per Span (default) |
Samples each NURBS span separately (a span is the region between knots). |
Good balance of accuracy and performance; generally the safest choice. |
| Knot Values |
Places samples directly at knot locations in the NURBS definition. |
More faithfully follows the original NURBS structure, especially where knot density varies. |
| Adaptively |
Adds or removes samples based on curvature/error metrics rather than a fixed distribution. |
More polygons in curved areas, fewer in flat areas; often the most efficient for complex shapes. |
When to try a different Sample Type
If you're seeing:
- Faceting on curved surfaces → try Adaptively or Knot Values
- Problems near trims, fillets, or continuity breaks → try G1 Segments
- Uneven tessellation density → try Per Span or Adaptively
- Unexpected artifacts from CAD imports → try Knot Values, since it follows the NURBS knot structure more closely
// Stephen Blair
// Arnold Renderer Support