Modifying a part currently on hold

[Ignacio Fernandez Galvan]

Hi all,

I've tried fixing the part 2515 (Wheel Hard-Plastic Large), which is currently on hold, but I've ended up rewriting some of it. There are still some things that could be fixed (I believe some quads in the tread are not quite planar, and there's some overlapping), but I think it looks pretty good now, except that I have no real piece to compare with.

Should I upload it now? As an update to the current 2515 or as a new submission?

[Michael Heidemann]

I have just downloaded your files.

If you change the header and say that all is your work, then it needs to be a new submission. But I think that will be deleted afterwards

Better way is to take a part that needs some correction and do the corrections. Then you insert a !HISTORY line with your name and what you did to this file getting better.

So to answer your question: Please restore the old header information and submit as 2515.dat update.

[Philippe Hurbain]

A huge improvement indeed! But there are still many issues...
- The subfile has many badly warped quads, you could run planarcheck tool (included in LETGUI) on it to automatically correct that.
- There are missing conditional lines, that can be auto-generated by Edger2 (also in LETGUI).
- edge of the subfile do not completely match 48 cyli in the middle of the track. Maybe (not sure) that this could be improved using Unificator tool on subpart after having temporarily placed the 48 cyli in the subfile.

Otherwise, yes, you can upload your files on PT, but even if you have rewritten most of the file, you must keep the original header with original author, and mark your contribution as an history line like Mikeheide did in previous header, for example:

Code:

0 Wheel Hard-Plastic Large
0 Name: 2515.dat
0 Author: John Boozer [jediknight219]
0 !LDRAW_ORG Unofficial_Part
0 !LICENSE Redistributable under CCAL version 2.0 : see CAreadme.txt

0 BFC CERTIFY CCW

0 !HISTORY 2007-11-17 [mikeheide] Fixed Bowtie Quads, BFC'ed
0 !HISTORY 2012-11-10 [jellby] Fixed gaps

Nice to see new names here

[Ignacio Fernandez Galvan]

Thanks for the info, I'll try to fix the remaining issues and restore the header

The warped quads are a pain, because the tread shape is all manual (almost no math involve, or at least I don't know what the math should be). A tool too fix them automatically would be great, but does LETGUI work on linux?

About the missing conditional lines, I didn't add them in concave sections, are they needed there too? Any other place where they're missing? (The quad non-planarity may be causing an issue with some lines at some orientations too, i guess.)

The mismatch between the subpart and the central cylinder may be difficult to solve satisfactorily, the vertices don't come at the same positions as those of the cylinder... Or do you mean the actual edge lines? Those I calculated with simple trigonometry, so they should be right within rounding error, unless I did some mistake (entirely possible)

[Michael Heidemann]

LETGui is "only" a grafik user interface for the very good tools made by Philo

As I wrote LETGui as a NET application it might work on Linux systems with the MONO Framework. I have never tested!

For the tools I can not give any information.

If you need help, please feel free to ask. We always like to have feedback for our apps.

[Philippe Hurbain]

The warped quads are a pain, because the tread shape is all manual (almost no math involve, or at least I don't know what the math should be). A tool too fix them automatically would be great, but does LETGUI work on linux?

No, LETGUI is Windows only. But it's only a GUI for a bunch of command line tools (many written by me...), and these tools are provided with source code. As they have a pretty simple structure, they should be relatively easy to recompile for Linux. Unfortunately my website is down at the moment because of a hosting issue, I hope it will be up again soon (tomorrow?). The link, when it's online again: http://philohome.com/ldraw.htm

About the missing conditional lines, I didn't add them in concave sections, are they needed there too? Any other place where they're missing? (The quad non-planarity may be causing an issue with some lines at some orientations too, i guess.)

Condlines in concave areas are indeed not required, but they are used by LDView to trigger smooth shading between facets. So they are a Good Thing ™
According to Edger2, there are 21 missing condlines missing in subfile on convex areas, and 37 total...
Edger2 must be recompiled to run on Linux, but on my website you'll aslo find Lee Gaiteri's Edger, written in Java, that should run directly. But it has less options/possibilities...

The mismatch between the subpart and the central cylinder may be difficult to solve satisfactorily, the vertices don't come at the same positions as those of the cylinder... Or do you mean the actual edge lines? Those I calculated with simple trigonometry, so they should be right within rounding error, unless I did some mistake (entirely possible)

Here's what I mean:
http://www.brickshelf.com/gallery/Philo/...515out.png
IMHO you'll get better results if you give up completely the cyli primitive, and create central rim from triangle/quads fitting exactly the subparts. My Coverer tool may help for this.

Edit: Cross post again with Mikeheide

[Ignacio Fernandez Galvan]

Condlines in concave areas are indeed not required, but they are used by LDView to trigger smooth shading between facets. So they are a Good Thing ™

I see. I'll add them everywhere then. Is this maybe the reason why the 3626ap01 face does not look smooth?

Here's what I mean:
http://www.brickshelf.com/gallery/Philo/...515out.png
IMHO you'll get better results if you give up completely the cyli primitive, and create central rim from triangle/quads fitting exactly the subparts. My Coverer tool may help for this.

Ah, that's a problem with the cyli primitive not being a real cylinder...

On the other hand, if don't use the primitive, then LDView (and others, I guess) won't be capable of replacing it with a real cylinder, Or, if I use the primitive but match the other polygons to the primitive's edges, then when a real cylinder is substituted the point's won't match. Am I right assuming from your answer that one should not use the primitives in these cases?

Same in the inner part, around the hub. I added edge lines following the real cylinder shape, not the 16-edged primitive. Should I do otherwise?

[Philippe Hurbain]

Is this maybe the reason why the 3626ap01 face does not look smooth?

No, it's another problem here: the multiple quads necessary to draw the pattern doesn't follow underlying shape geometry, that's why smooth shading doesn't work fine here. See discussions about texmap in this forum for a solution to this problem.

Ah, that's a problem with the cyli primitive not being a real cylinder...

Yes. When using primitives with adjacent quads, you must match underlying 16 or 48 segments structure. As you say, this is not perfect since there will be gaps if you substitute mathematical surfaces, but at least you get a perfect surface without substitution.

Am I right assuming from your answer that one should not use the primitives in these cases?

Since you have vertices in subpart that are placed in the middle of 48 primitive sides, result will always be imperfect (this creates T-junctions, see http://news.lugnet.com/cad/dat/parts/?n=6086). So I think that using quads/triangles would give better results here.

Same in the inner part, around the hub. I added edge lines following the real cylinder shape, not the 16-edged primitive. Should I do otherwise?

Edge lines should always follow exact geometry. Indeed, the hub needs some work too. I'd probably use a 48\4-4cyli for the outside, since you have 12 subparts around, and 48 happen to be divisible by 12 . You can match this hires cylinder to normal resolution primitives inside the hub using 48\4-4aring primitive.