What is causing this stretched shading on the mesh and how can I fix it?

Question

I am modeling a Z GT500 and I have this weird stretched shading/reflection on the rear panel:

Despite the topology is somewhat fine. Can looptools fix this or is there any other way?

I include a .blend file if you want to take a closer look: https://drive.google.com/file/d/1-0mVlcdVFdSZYZEO0RaS-2kbSpCSOgdB/view?usp=sharing

The rest of the topology isn't the best but that's not the problem now.

Thanks in advance for your help!

Answer 1

Try adding Subdivision Surface modifier. Probably your topology doesn't allow it though… Your question is a good example of why Subdivision Surface or NURBS workflows are powerful.

This might not be a real (or full) answer to your problem, but since I'm adding some images I might as well post it as one.

Such shading issues are typically caused by not enough smoothness. Consider the following shape:

I created it by starting with a mesh circle of a very high resolution (to hide the fact each circle is actually a polygon, let's pretend the screen resolution is infinite and it's an actual curve) and n-gon filling. Then I inset the circle, remove the center face, remove left half of it, and extrude the cut parts to the left. Now, other than the sharp $90°$ corners on the very left part of the image, you would say the orange outline is smooth, right? And you would be correct. But to a similar degree you would also be correct if you said those sharp corners are smooth, as ridiculous as it sounds.

1. The 1st derivative of smoothness here is the continuity of the outline: it's not a dashed line - - - -, it's continuous. Whichever fragment of the outline you choose, regardless how much you zoom on it, you won't find a gap.
2. The 2nd derivative of smoothness is continuity of angles: here the corners fail, because once the outer path reaches the top-left corner, there's a sudden jump from $90°$ to $180°$ normal, then from $180°$ to $270°$ then a long smooth inner path leads to the bottom-left corner, smoothly transitioning this $270°$ angle back to $90°$, and then again you have two sudden angle jumps $90°$ → $180°$ → $270°$. But other than that the outline is smooth, and if we applied mirroring on this shape (so it looks like a pill) the inner and outer outlines would be completely (2nd derivative) smooth.
3. But now consider the outline near the green/blue boundary. Is it smooth? The 3rd derivative of smoothness is continuity of the change of angles. The blue area has a constant change of $180°$ per $π$ meters or $1°$ per ~$17$ mm. A little bit faster (per mm) change on the inner side as the radius of the bend is smaller there, but on both sides separately it's consistent, constant. So the blue part is perfectly 3rd derivative smooth. The green part, not counting the sharp corners, is also perfectly 3rd derivative smooth, because the angle doesn't change at all, and $0$ is constant. The problem is the connection, when the rate (outside) jumps suddenly from $1°$/~$17$ mm to $0$.

To further clarify, here's a screenshot from a cool video game Railbound:

The blue junction (the rounded part) joins with the straight, white rails in the same way the blue and green parts join on the previous image. But this is not how real railways are actually shaped: it's not enough to smoothly change velocity (if a train moving at 50km/h rotated 90° within less than a second, people inside would simply die), you also want to smoothly change acceleration, otherwise people inside will get surprised by the forces acting on them suddenly changing and will fall over. As smooth as this transition seems:

When you go from green to blue (or vice-versa), suddenly the centrifugal force (dis)appears completely. It's very similar with reflections, and those are actually often used to test the smoothness of objects. You can test it here:

Rounded cap on Curve with Geo Nodes

If you download my .blend , add before group output "realize instances" → "merge by distance" → "subdivision surface", in shader make the material metallic, decrease roughness, and disconnect the normal calculation (which is there for optimization purposes to use instances), you will see how the Catmull-Clark algorithm improves smoothness to the next derivative, improving the reflection change: