Squeezing out better SU model performance

hebrides

This got started somewhere else and I though rather than continuing to tresspass in another thread I would start my own.

I am finding an unusual way to look at SU models to check the soundness of the model. It is in an effort to produce models that are cleaner, easier for SU to read, perform without unusual behavior in any renderer, can be moved to other modelers and formats and retain their integrity and will verify as solid in a CAD application. These files are as well "watertight" by default.

If you have a need to get this possessed with improving your models or if you are packaging your models for universal application I suggest you give this a try.

You need a CAD or Solid modeler and an app or service that will convert .skps to .stl. to experiment this way.

What you see below are images from my solid modeler.

The first mesh is a skp file that has been inspected.
It is a mess.

It was built this way:

half profile by hand
copied inplace and mirrored
Additional rectangle added to extend the bottom of the profile
new profile pulled along original

The second mesh is a skp file made with modified profile that was complete. (no additional rectangle required)
It is very clean, slight file size improvement, 13k to 10k

Copied inplace and mirrored
removed face
healed
pulled to length

As you can see SU has had to work much harder at defining the first shape than it did the second.
There is no doubt that the overhead required to carry the first model in SU is larger than the second.

When this is multiplied say to 100 copies there is a real difference in the performance between the first model and the second.

The last image is the same model one corner swept manually with follow-me (looks clean) on the second corner I set the path by highlighting the lines and clicked (it's a mess)

Although this type of process is time consuming It is providing many advantages in my environment.

Have fun!!

solo

Facinating stuff,

I understand the logic as far as optimising a model goes, but does this not absolutely limit ones modeling options?
As any and every ruby used will 'mess' the mesh, organic/freeform modeling would be a catastrophe.

So unless you are planning to use a 3d printer, what besides cleaner models (which do not seem to bother the render apps I use) will the benefits be?

hebrides

Good question. The models that I have been building have simple to moderate meshes. The list of advantages this method provides is growing. I'm sure there is a point of no return where the difficulties that have to be overcome forgo the effort. I haven't found this to be the case yet. I will continue to build this way for certain components. As I continue I will let you know if I hit a wall with a complex mesh.

The advantages so far:
smaller files
ability to carry more instances comfortably
faster navigation
0 rendering apparations
very clean import and export

edit: I have no need for 3d printing. It just happens that the STL is the most accurate itermediary file type.

solo

hebrides wrote in another thread relating to this one:

@unknownuser said:

I took considerable time (many hours) first of all to experiment with model mesh. I did my best to explain. Solo ignored. I took from that that he didn't think my advise was appropriate. That would suggest that I should not continue in his thread. I posted in another thread for those who have an interest in further optimization of their models. Funny, Solo came out of the woodwork. As I suspected, he dismissed the advise immediately.

I clearly never dismissed your idea/method, I just questioned it's practicality.

jeff hammond

@hebrides said:

If you have a need to get this possessed with improving your models or if you are packaging your models for universal application I suggest you give this a try.

maybe i'm missing something by not being aware of the other thread that led up to this.. but, what exactly are you suggesting i try?
at least by the examples posted here in this thread, it seems like it would be harder to draw the messed up version in SU than the clean one..

from this thread:
http://www.sketchucation.com/forums/scf/viewtopic.php?f=79&t=17109&start=30
you can see that i'm interested in drawing more cleanly but i fail to see how what you're saying is going to help me.

remus

i think the point hebrides is trying to make is that some of the native SU tools (follow me, push pull etc.) dont produce clean geometry (although you have to export the SU model in to a solid modeller to see the unclean geometry.)

As a remedy to this situation, its being suggested that you draw geometry only with the drawing tools (line, rectangle, arc etc.) and dont use follow-me , rubys etc.

Personally i think the time penalty incurred during modelling is far to high to account for any marginal increase in performance that might be gained through this method.

Hebrides, another thought: i wonder if theres something happening on export to mess up the models?

jeff hammond

@remus said:

i think the point hebrides is trying to make is that some of the native SU tools (follow me, push pull etc.) dont produce clean geometry (although you have to export the SU model in to a solid modeller to see the unclean geometry.)

oh, ok.. i don't use and solid modelers so i don't know about that..
you're talking about something other than turning on hidden lines in SU which is where my main concern is at... with hidden lines on, i like everything to be clean and logical because it makes it way easier/more accurate to actually build said model than when vertexes/intersections are at random points..

other than that, i really don't care..
next....

remus

Yes, we're not talking about simply turning on hidden geometry.