I've been trying to learn basic modeling since about Feb of this year in order to build and texture some farm implements in Blender to use in Farming Simulator 2015. The impetus is that most modders rarely texture their work and the results are an eyesore in-game. Actual modeling of high poly things is going well and is the easiest part. I'd like to eventually understand how to mirror low poly objects along with their normals and AO, but I'm not able to reproduce the basic examples I'm finding in tutorials here or apply them to my own models.
My progress starts and ends with adding edge loops to a cube primitive, using a displace modifier to create a cage with full smoothing to achieve averaged normals, and baking out a "high poly" cube with beveled edges. I get the expected result. A low poly that cube looks great, beveled edges, no seams.
I've read the "making sense of hard edges, uvs...etc" and a couple of EarthQuake's posts where he demonstrates the value of edge loops on cylindrical objects many times.
When I go to bake a cylindrical object using a smoothed cage and edge loops as described in those threads, this is my result:
That's not at all what I expected. I thought we went to all that work creating edge loops and increasing our vertex count moderately in order to specifically eliminate this exact issue according to those threads I mentioned.
Further, once I remove all the extra loops and just use a basic hectagonal cylinder and bake without a cage, the result is perfect.
Here I don't understand why I went to the effort of putting in edge loops and what happened for them to result in the problem that they're supposed to solve. Actually, I don't understand how this result is possible without skewing. The explanations made sense in that the loops allowed the effects of normals at the edges to be separated from those along the length. I don't know how much of this is a blender thing and where my misunderstanding is.
I tried other times to use the averaged normal setup using edge loops and I can't get any reasonable result that way. Here I tried using edge loops and a cage again. The hex nut area is shaded flat and the threads sticking out are a smooth group on the low poly and I've got UV seams at those hard edges. The nut bakes with no seams but the cylindrical thread portion is a distorted mess with artifacts.
In one of the posts, there is a section on Destructive Baking workflows and I think I used that specifically to come up with a method that works. I'm not just a little frustrated by the fact that what I found to work seems to be explicitly what the author, an expert, recommends against.
If I take that nut and threads and bake it in three steps I get a great result. Bake the nut with a cage, move the forward bolt face UV island out of the UV space, bake the threads without a cage, then finally bake the bolt face.
That seems like too many steps, even if only by one. So my conclusion is that in the future, this should be modeled as two intersecting meshes just to simplify the baking. I don't know if that's right.
The very last thing I tried failing at was understanding how to avoid seams in the basic beveled cube model by putting UV seams on all the edges and adding significant "padding" between the UV islands. This is explained in numerous posts across the internet as a quick solution to seams on the basic beveled cube problem. No matter the amount of "padding" I put using separate UV islands, I still get seams on my model.
I gave up last month because it felt like nothing I read or do seems to work as described so I can't figure out a workflow to finish making the few parts I want because I've still got to learn how to combine all these objects in one UV and develop a workflow for making changes without having to re-bake everything if there's one piece or even just a face that I'd like to adjust. My project may not be big enough for this to seem like a huge concern and of course no one but me is going to see it, but it's the principle of the matter of learning a new subject correctly and having a workflow that's fundamentally sound in order to continue.