Normal woes

SixGhost

Resource contributor
#1
Here's my annoyance:



I cut out both gear doors and they are symmetrical, but since each door is a different object, I inevitably get a hard edge right at centerline. The lower fuselage being convex doesn't help to alleviate the effect.

It's really disrupting the flow of the mesh IMHO and AFAIK we can't make use of the normals modifier in Max.

Now, what would happen if I did something like this?



I would extend every door with a piece of geometry from the other door so not to have a hard edge at centerline and keep the curvature of the lower fuselage, then use a multimaterial composed by my fuselage base material applied to the faces that form the actual door and a fully transparent material applied to the faces that form the extension. Finally map and texture the corresponding faces in totally different positions.

Has anybody tried something similar? Would it work? Am I a paranoic nitpicker?
 

Paul Domingue

Resource contributor
#2
You need a central row of polygons that is perfectly horizontal prior to separating the two halves of the gear doors, then slice that row making your division. Then the two halves will join perfectly.
 
#3
Once you throw textures on it, it'll look fine, especially because the edges of panels are generally a spot where grime collects, which will obfuscate the seam. Bear in mind that they probably look that way in real life, too. In fact, in real life, there's probably a small gap between panels creating a seam that's even more visible. I wouldn't waste much time worrying about it.

EDIT: I just checked a photo and not only is there a visible seam, but you can see how the light affects the panels differently than the surrounding areas. I think that you're trying to fix something that ain't broke. I'm sure the end result will be great as-is. Remember that real airplanes are very imperfect.
 
Last edited:

SixGhost

Resource contributor
#4
Yes Paul, that would indeed fix it but at this stage I have no intentions to change the topology.

Erick, you're right when you say that textures will conceal it and that planes are full of imperfections. I took this as an example, however there may be different instances where this effect is rather disturbing depending on ones topology.

BTW, looks like the general consensus is that I'm paranoic!:D:D Can't say I'm not!
 

n4gix

Resource contributor
#5
Luca, learn about "Smoothing Groups". This is how one would properly handle such a situation as you've described.
 

SixGhost

Resource contributor
#6
I'm afraid Bill smoothing groups would do nothing. I may have the same SG throughout both doors but since they're two different objects the seam between them would always be there!
 

Paul Domingue

Resource contributor
#7
As long as the two separate edges join each other at any angle of degree greater than zero no smoothing groups will correct this. It isn't until the edges are joined that smoothing will take effect and that is not possible in this situation.
Luca, from what I have seen of your model, it shouldn't take you very long to correct the polygons to achieve what you want. Once you split the two halves add a small, 0.001m negative extrusion to both door edges where they meet and you will have smoothly curved, joining doors with a beautiful seam.
 

SixGhost

Resource contributor
#8
Paul, it's not so much about the effort rather than the consequencies. I have antennas and holes that would have to be modified along that edge and I'm not really thrilled to touch mesh density. That said, it's doable, no doubt about it. Still, this is just an example of the problem. When I'll have to cut the doors for the main gear (curved doors on curved mesh) the same issue will present again and I won't be able to mitigate it easily with just topology. That's why I was asking if my theoretical solution would work in your opinion. I guess I'll try and see when the time comes.:)
 

n4gix

Resource contributor
#9
Not the same smoothing groups, but a blend of two different SGs on the polys along the seam.
One door is SG1, the other door is SG2. Along the poly seam on both doors use SG1&2.
 

SixGhost

Resource contributor
#10
Mmh...I think you skipped the part where I say I have 2 doors, so 2 different objects. You can't blend SG's between 2 different objects. Bill, have another one on me!:rotfl::rotfl:
 

n4gix

Resource contributor
#11
Ah well, it was just a thought. OTOH, we've made a virtue from a necessity by simply making the edges of the doors just a bit "dirty" to mitigate any visible seam issues:
 

Paul Domingue

Resource contributor
#12
Some pictures on the point I was trying to make. Identical cylinders both 24 sides with the same size doors cut. If doors are cut through polys rather then at edges you end up with smooth blending regardless of the smoothing groups used since smoothing groups do not extend across separate parts. Image one with doors open, 2 with them closed and the rest is self explanatory.



 
#13
On the same principle, you could simply increase the resolution of the door. The edges are probably slightly rounded in profile as they bend in to form the sides and insides. I think that was your original point, no? It's a good idea.
 

Paul Domingue

Resource contributor
#14
On the same principle, you could simply increase the resolution of the door. The edges are probably slightly rounded in profile as they bend in to form the sides and insides. I think that was your original point, no? It's a good idea.
Even increasing the number of sides to the cylinder, which I assume you mean by resolution, will not eliminate the noticeable edge as long as two opposing polys reside on different angular planes.
Another way to say this is that the circumference of the cylinder is measured not by its outer most vertices but through the center of all the outer polygons. The more sides you add to the cylinder the closer to the true circumference you get but that is not practical for our modelling needs.
 

n4gix

Resource contributor
#17
Some pictures on the point I was trying to make. Identical cylinders both 24 sides with the same size doors cut. If doors are cut through polys rather then at edges you end up with smooth blending regardless of the smoothing groups used since smoothing groups do not extend across separate parts. Image one with doors open, 2 with them closed and the rest is self explanatory.
Paul, thanks for providing a most excellent mini-tutorial on how to easily solve this problem. You are a true :wizard: !
 
#18
Even increasing the number of sides to the cylinder, which I assume you mean by resolution, will not eliminate the noticeable edge as long as two opposing polys reside on different angular planes.
Another way to say this is that the circumference of the cylinder is measured not by its outer most vertices but through the center of all the outer polygons. The more sides you add to the cylinder the closer to the true circumference you get but that is not practical for our modelling needs.
I'm not talking about increasing the number of sides; I'm talking about increasing the resolution of the entire door, meaning the edges, too. In real life, there isn't a monomolecular hard edge where the door face stops and the door's depth begins. The edge is going to be fairly rounded. So you build the door, including its inner portions, then chamfer the edges a couple of times and smooth it. Now, whatever crease you see will just look like whatever crease you can see on the actual airplane.

TL;DR: build the rest of the door and round off the edges.
 
Top