Modelers—especially Trek modelers—often find themselves needing to etch (shield) grids into their hull surfaces. This isn't unique to Trek models, of course; hull panel lines everywhere and the following applies equally to any sort of inset line on a non-planar surface.

A common approach, which ultimately leads to problems, is to use the existing geometry as a basis for the grid layout. Here we have an eighth of a sphere, which we'll use for demonstration throughout this tutorial. I have colorized it with an alternating red/blue checker pattern to indicate different "hull panels" that I want to etch onto its surface.

What to do from here? A common, but incorrect, approach is to simply extrude along each group's normal. The reason this fails is that it leaves you with wedge-shaped "troughs," rather than nice rectilinear insets.

Note: Many extrusions in this tutorial are highly exaggerated to illustrate what's going on!

The first step on the road to correcting this is to add an inset for each panel. The exact tool will vary with your package (for Blender, it's Inset Faces, for LightWave it's Bevel or Multi-Shift), but the idea is the same. When using inset, you should inset each panel group by half the total gridline thickness. Personally, I like to have grids that are about 5cm wide, so I'd inset 2.5cm, but the exact amounts are up to you. Here's what our octantsphere looks like after insetting.

Tip: You can select all of one particular type of material group (depicted here as blue or red) and inset them in the same operation. This is, in fact, part of why they are grouped in this checkerboard pattern!

Tip: It's also a good idea to give your insets an entirely different material at this juncture for easier selection.

Your instinct now might be to negative extrude your newly minted grid lines and call it a day. While the typical extrusion depth for gridlines of this size will let you get away with it, look what happens when you exaggerate the extrusion depth.

Whoa! Why are our grids so fat now?! The reason has to do with Normals. Every vertex, and by association every face, has a Normal that can be thought of as a ray indicating which way the vertex or face is "facing." When you have a vertex that is shared by multiple faces, its Normal is an average of the associated face normals. (There's actually some duality to this; typically, it is the face that stores Normal data and the vertex normals are derived from this, but not always!) When you extrude as we have here, you're extruding along the Normals computed from our insets and our original topology lines, which were spherical. It's not very different from simply scaling up a sphere, in fact.

Note: Your exact method of extrusion may vary. In Blender, Extrude Region with an offset of 0, and then Shrink/Fatten with an offset of the desired depth will produce the results you see here, and is my recommended approach for this sort of detailing.

How do we fix this? A first step is to remove the edge loops in the middle of our gridlines, created when we inset the different panel groups. This will get rid of most of this "distortion" (it's not really distortion; it's doing exactly what it should based on the local topology).

Edge loops removed from un-extruded surface.

Resulting extruded surface.

As you can see, the grid lines where we didn't have a grid intersection now look correct. The intersection grids are still fattened, though. Why is this? It's again an issue of pre-existing surface topology and the Normals along which we're extruding our faces. These intersection points are not flat, while our other grid lines are flat.

What makes them different? When we removed the interior edge loops from our pre-extruded insets, we made the grid surface planar. Rather than having an edge loop that cut through the middle, creating non-co-planar (this isn't the same as non-planar, which describes a single surface that is made of two triangles with unequal face normals) surfaces, we have flat surfaces. Because of the 0-offset extrude executed earlier, we removed any association they had with their neighboring polygons, and so they extruded only their face normal without distortion.

However, the intersection polygons are a different story! They have topology information coming from the all of the incoming gridlines: latitudinal and longitutidnal. Whereas the non-intersecting portions of the gridlines have only the neighboring two polygons to influence their vertex normals, the intersections have to juggle vertex normals from four sources, which result in normals that no longer conform to the rigid extrusion angle we want. This image might help make things clearer:

So, how do we fix this? By ensuring that intersections are co-planar with incoming gridlines. This means we can't actually use our underlying geometry's edgeloops as the basis for our grid insets, because whenever we do, we are going to run into a situation where the grid intersections aren't co-planar. Using longitudinal intersections, will get longitudinal distortion, and vice versa for latitudinal intersections.

Here's our octantsphere with completely new lines manually cut in for the grids.

Here's a close up, showing where the panels end vs. where the original geometry lines existed.

Here's our octantsphere with cleaned-up insets.

And finally, the resulting extrusion, free of distortion!

From here, it doesn't hurt to apply a nice micro-bevel to the various edges you've created (each panel surface, and the corner edge of each panel's trough; putting bevels in the troughs is generally a waste of polygons).

The resulting bevel, after a bit of cleanup.

And voila!

As a final note, I use the term "extrusion" a lot throughout the tutorial. In Blender, you actually want to do an extrude with 0 offset, followed by a Shrink/Fatten with the desired offset. Extrude by itself will extrude along the entire selected group's vector, which won't at all do what you want. Shrink/Fatten moves polyons along their individual normals while preserving their vertex relationship. The reason to extrude with 0 offset first is to create new polygons to move around, without disrupting the original borders.

Suggested Further Study