How to avoid lumps when working with subdivided mesh
triangle
I am trying to make a car chassis. However, I often run into the problem of the mesh appearing lumpy or has dents in it.
No matter how much I try to adjust the vertices around to make it more smooth, it always shows up. It usually occurs at a quad corner points or "diamond" polygons.
I am using blender.
Replies
I think @FrankPolygon 's sketchbook is a great place to dig into. https://polycount.com/discussion/221392/sketchbook-frank-polygon/p1
start simpler, subdivide, work from there. you are cutting in too many things at the density you have right now.
So in order to avoid lumps I should add more geometry? I thought it was the opposite. Less geometry means less vertices to manage, vertices that could be moved the wrong way and cause lumps.
Less geometry is better, in general, when starting to learn how subdivision works.
Remember three points make a curve in modelling after adding smoothing (subdivice, opensubdiv, turbosmooth etc)
;)
https://3dtotal.com/tutorials/t/making-of-lotus-elise-mk-1-razvan-maftei-vehile-car-lotus-wheel
I am trying to make a car chassis. However, I often run into the problem of the mesh appearing lumpy or has dents in it.
No matter how much I try to adjust the vertices around to make it more smooth, it always shows up. It usually occurs at a quad corner point.
In addition generally following sentiment, offered by others.
Sometime ago, stumbled across insight that clarified in my own mind an optimal polygonal workflow, creating Class A (Reflective Observable) surfaces that personally remains a simple guiding set of vehicle modeling principles - i.e.
Uniformly spaced edge flow sounds like a good idea, but for car models there's alot of areas that need hard edges, such as the divider between the front seats and passenger seats. In order to get a proper hard corner on a subdivided mesh, you need one edge going vertical and another going horizontal.
Which causes non uniformed edge flow like this.
You can try merging edge loops in order to avoid unnecessary edge loops spilling off into other parts of the mesh, but then your making triangles or diamond quads, which usually causes shading issues.
Alot of images of car models on the web appear to have evenly spaced edgeflow, but they seemingly only exist to look good in wireframe, and I wonder if they are really needed.
Are all of these edges really needed in these relatively flat areas of the mesh?
My only explanation is that they are there so once the car is separated into parts, they can perfectly align with other parts of the mesh that need that density.
I bet Frank will be tempted to work up some relevant examples. But frankly (!) it's all there in his sketchbook already if you take the time to read, experiment, and absorb the knowledge feast that's being filleted and fricasseed by our fearless Frank.
Everyone has offered really solid advice. Think that Neox, Eric and sacboy have summarized the key points of the issue and provided a lot of good links to additional resources on the topic. Just wanted to quickly address a couple of different points brought up in the discussion.
Subdivision modeling tends to be an iterative process. So, there's reasons to both keep things simple and use more geometry. Blocking out the major forms is a lot easier to do when there's less geometry to manage but adding small details, to surfaces with compound curves, can require using more geometry to provide support around the shape transitions.
Sometimes the block out process can be simplified by combining adjacent panels but arbitrarily modeling an entire car body as a contiguous quad mesh tends to introduce a lot of unnecessary complexity. The panel lines provide a natural break in the shapes. Which can be used to reduce loops in a non-visible area or separate sections of the mesh so the local density can be increased to support small surface details.
The examples below show what this process could look like. Parts of the mesh may be merged or split for different reasons and during different stages of the block out. Most of the time this decision will come down to balancing accuracy and editability.
All of the reflective surfaces and subtle forms make automotive body work a challenge to model. Creating a base mesh with relatively consistent loop spacing is generally considered best practice because it helps prevent artifacts in the edge highlights and surface reflections. There's differing schools of thought on other workflow topics. E.g. rink-wrapping a guide shape, using creases instead of support loops and how much subdivision should be applied before detailing. So, it's probably worth considering what the model will be used for and read up on the current best practices for that particular discipline.
When in doubt, a flow check with different materials and a reflection map can help highlight any problem areas. Here's a link to a detailed write-up on how to do a flow check. https://www.ebalstudios.com/blog/3d-modeling-artifacts-flow-check-reflection-map As mentioned in the article: there are some inherent and unavoidable limitations to the surface quality of polygonal models. While the flow check is generally a good indicator of surface quality, it's also important to focus on the final results and avoid making unfounded assumptions based solely on a model's wire-frame.
Below is a comparison of a base mesh with contiguous panel lines and a base mesh with split panel lines. Each type of material preview shows a different aspect of the subdivided mesh. A material with a wide highlight roll-off is good for catching most types of subdivision artifacts but more reflective materials provide a deeper picture of the surface's overall quality.
A contiguous, low density mesh is fine for blocking out larger shapes but often lacks the amount of geometry required to support complex shape intersections on curved surfaces. These small, poorly supported surface details can interrupt the existing edge spacing and this tends to reduce the overall quality of the surface. Sometimes it is possible to compensate for undesired smoothing deformation by adjusting the shape of the base mesh but this tends to introduce other flow issues that are still visible in surface reflections.
The comparison below shows how the overall surface quality from the block out is preserved by applying one level of subdivision before adding smaller surface details. Increasing the mesh density does reduce the editability of the larger shapes but this isn't an issue since they were finalized in the previous block out.
If the project does require a contiguous mesh then any components that were split off during the block out or detailing processes can be subdivided to match and merged with the adjacent panels. Below is an example that shows how applying the subdivision level for detailing also locks in the existing curvature of the shapes and reduces the overall visibility of flow changes between panels that have been merged.
It's also worth mentioning that there's a few different ways to reduce edge loops without interrupting the flow of the mesh in a visible area. Here's a link to another write-up that looks at several different ways of handling support loops that intersect curved shape transitions. https://polycount.com/discussion/comment/2769713/#Comment_2769713
Too bad.. if @FrankPolygon had split this up in different posts i could have liked every one of them 😍. (So make just one or two simple follow ups 😉.)