S4 works on some models but shows no effect on others (e.g. tensile specimens)

[kevgeiger2]

S4 does work for me, but only on some models.
On other models – especially tensile test specimens (dogbone geometry) – I see no effect at all. The printed part looks identical to a normal Cura print.

The printer itself is not the issue. I fully re-trammed it, reduced play, and added extra mechanical stiffening. Prints are repeatable and stable, and S4 effects are visible on other, more complex models.

My impression is that S4 is highly dependent on geometry and resulting toolpaths.
Slender, symmetric models with mostly vertical and repetitive paths (like tensile specimens) seem to be barely affected, while more complex or asymmetric geometries show a clear deformation.

Is this expected behavior?
Does S4 require a certain amount of curvature, lateral path variation, or toolpath complexity to produce a noticeable effect?

Also, am I correct in assuming that S4 operates on toolpaths / motion (G-code) rather than modifying the STL geometry itself, meaning that Cura-generated paths for some models may simply be too trivial for S4 to have any visible influence?

That’s what I’m trying to understand.

[jyjblrd]

Could you give some before/after screenshot examples?

[kevgeiger2]

I think I’ve identified the reason now.

For the standard tensile specimen printed flat or upright, no supports are required and the geometry is already fully planar-printable. In this case, S4 appears to converge to the identity solution and applies no deformation.

When I rotate the same tensile specimen by approximately 45°, introducing significant overhangs, S4 produces a non-trivial result and the output looks clearly different.

This suggests that S4 only applies deformation when there is an actual overhang- or support-driven incentive, and otherwise correctly does “nothing”.

Here are images of the tensile specimen rotated by ~45°, where the effect becomes clearly visible.