Message Areas - "Poly count question"

Train-Sim.Com Message Areas

"Poly count question"

Email this topic to a friend
Printer-friendly version of this topic
Bookmark this topic (Registered users only)
Previous Topic | Next Topic Conferences MSTS Locomotive Design (Protected) Original message
clashley (191 posts)
Jan-18-02, 11:01 AM (EST) "Poly count question" I just ran across an interesting bit of information in this thread: http://www.train-sim.com/dcforum/DCForumID5/2031.html#11Specifically, it says the MSTS poly can be found by checking the "Normals" section in the .s file for a wagon/locomotive. Just for curiosity, I compared this number from my RS11.s file to the poly count reported by 3DStudioMax. The .s file says about 2,700, but 3DSM shows over 7,000. Fascinating...Just for grins and giggles, I decompiled the acela.s file, which should be about 3,500 polys, and checked the normals. The normals there are about 4,700.This raises an interesting question... Kuju recommends a poly limit of 3,500, although I know for a fact that it can be at least twice that without much problem. It would seem logical that the poly count should be derived from the .s file, since that's what MSTS actually uses. Does this mean that I can get away with even more detail than I previously thought? If it's a simple matter of keeping the normals to a manageable level, this opens a world of possibilities.Any thoughts on the subject?Cameron Lashley Alert Edit | Reply | Reply With Quote | Top
Table of Contents

Subject Author Message Date ID RE: Poly count question dnalot Jan-18-02 1 RE: Poly count question AlabamaRailfan Jan-18-02 2 RE: Poly count question troub Jan-18-02 3 RE: Poly count question svante Jan-18-02 4 RE: Poly count question troub Jan-18-02 5 RE: Poly count question troub Jan-18-02 6 RE: Poly profusion leads to confusi... Captain Bazza Jan-18-02 7 RE: Poly profusion leads to confusi... svante Jan-19-02 8 RE: Poly profusion leads to contusi... Captain Bazza Jan-20-02 9

Lobby | Topics | Previous Topic | Next Topic
Messages in this topic
dnalot (139 posts)
Jan-18-02, 11:40 AM (EST) 1. "RE: Poly count question" HiI just checked the SW1500.s file. It is the most detailed model I have (to date). The normal(s) were listed in the .S file at 5200. I have not had any complaints on performance issues. Can we go larger? Is it the detail in the .S file, or is it the textures that drag down performance.Mark Toland Remove | Alert Edit | Reply | Reply With Quote | Top
AlabamaRailfan (97 posts)
Jan-18-02, 11:56 AM (EST) 2. "RE: Poly count question" Yeah, this is interesting...TrainSim Modeler shows my C30-7 model as weighing in at 5021 polys after conversion. I checked the .s file and it shows 2522 normals....interesting. I wonder if we are missing something else?John D. Peterson
Webmaster - AlabamaRailfan.com
http://www.alabamarailfan.com(Insert cool logo here)
Remove | Alert Edit | Reply | Reply With Quote | Top
troub (200 posts)
Jan-18-02, 02:54 PM (EST) 3. "RE: Poly count question" The major flaw in the logic I see here is that the same programs that allegedly are reporting the wrong number of polygons are the same programs creating the .s files that list the "normals." Why would the program report an incorrect or inaccurate amount of polygons if it knows how many there REALLY are because it puts it in the .s file.It is important to realize that normals and polygons are two separate terms. That is, the normal defines ray that is perpendicular to the plane of a face. A face may contain many polys. Take a square, for example. One face, one normal, but it takes two triangles to draw.troub
MSTS: Illinois Central Remove | Alert Edit | Reply | Reply With Quote | Top
svante (419 posts)
Jan-18-02, 03:09 PM (EST) 4. "RE: Poly count question" I wonder if it's that simple. The stock box car (us2freight6) has 375 normals according to its s-file whereas my standalone ribside well cars, with far more details, only have 192 normals. And my European logging cars, with full load, have 431 normals (only 20% more than the stock box cars) even though they're extremely detailed and have an s-file that is almost 3.4 MB uncompressed. Remove | Alert Edit | Reply | Reply With Quote | Top
troub (200 posts)
Jan-18-02, 03:32 PM (EST) 5. "RE: Poly count question" LAST EDITED ON Jan-18-02 AT 03:32 PM (EST)It might have something to do with how the model is created. . .maybe if you "triangulate" your whole model before export (ie split all the faces into their component triangles), it might export each triangle as its own face with its own normal. This way, it might more than double the number of normals. Some people like to triangulate first, maybe to edit the triangles in wireframe or something, and maybe Kuju did this. 300-something polys (the normals you mentioned, if they came from a triangulated model, would represent the number of polys) doesn't sound very far off for those boxcars, as I recently went back to my old center-cupola caboose model and ran a new plugin for 3dCanvas that counts the number of triangles in the model, and I came up with around 400. I'm not happy with that caboose now (it was my first model), and now I know that I have a LOT of headroom to improve it. The point is, it was a simple model (just a little more complex than the default boxcar), and it had about 400-something polys.Long story short, my theory is that whether or not you triangulate your model might determine whether the normals reflect the number of polys or the number of faces in the model. I'm at work, so I can't try it with a model right now troub
MSTS: Illinois Central Remove | Alert Edit | Reply | Reply With Quote | Top
troub (200 posts)
Jan-18-02, 04:30 PM (EST) 6. "RE: Poly count question" LAST EDITED ON Jan-18-02 AT 04:33 PM (EST)Ok, I have managed a bit of testing with 3DCanvas. In 3DCanvas, converting all of your faces to triangles seems to make no difference to the number of normals in the .s file. I made a simple cube, 6 faces:

This has six square faces, and the .s file generated by 3DCanvas has 6 normals. Note however that the number of triangles to draw the object (the "polys") is 12.I triangulated the model:

and now it has 12 triangular faces, but the number of normals in the .s file generated from this model is still 6.Results might be different if you're using 3DSMax with conv3ds, though--3DCanvas exports .s files all by itself, so it may be smart enough to optimize the normals by recognizing that the pairs of triangular faces are on the same plane. Maybe someone with 3DSMax can try it, and report the results?troub
MSTS: Illinois Central Remove | Alert Edit | Reply | Reply With Quote | Top
Captain Bazza (245 posts)
Jan-18-02, 08:08 PM (EST) 7. "RE: Poly profusion leads to confusion" LAST EDITED ON Jan-18-02 AT 08:15 PM (EST)
There is no rulebook for modelling for the sim; you model what you get away with and it varies from model to model. Some unwritten Golden-Rules:
The sim will not accept all models - even if processed without apparent error with conv3ds. (For various reasons.)Remove hidden faces.Do not make any individual object mesh too complex. Make separate objects and LINK them to the MAIN object. The pic above shows a loco model which is composed of 16 objects (15 linked to main). The list of objects:
Main\
cabsides
chainhk
coal
couplings
firebox
headlits
levers
sandboxs
smkbxdr
vents
wheels1\rod01
wheels2\rod02
wheels3You can make a higher 3d detailed model using this technique. Without splitting off objects and linking them the model will not work in the sim.Be careful when texturing parts, I'm still working out the rules for this ;-(Don't let the computer WIN! Remove | Alert Edit | Reply | Reply With Quote | Top
svante (419 posts)
Jan-19-02, 01:46 AM (EST) 8. "RE: Poly profusion leads to confusion" My standalone ribside well cars (with just 192 normals) use almost 30 separate smaller objects linked to Main, and all polygons that can't be seen have been removed, so I've been using the technique you describe from the start. Remove | Alert Edit | Reply | Reply With Quote | Top
Captain Bazza (245 posts)
Jan-20-02, 02:18 AM (EST) 9. "RE: Poly profusion leads to contusions" That's interesting & useful - I've not tried linking so many objects to one mesh. Good to share the info with others, that's how they learn, just like me Don't let the computer WIN! Remove | Alert Edit | Reply | Reply With Quote | Top

Lock | Archive | Remove

Lobby | Topics | Previous Topic | Next Topic