Question on "Plates without Corner" (2450/30503/26601)

[Niklas Buchmann]

I was working on the new Plate 2x2 without Corner (26601) using the exisiting larger plates 2450 and 30503 as a starting point.

I noticed that the existing parts are designed around the cutouts on the 45-degree corner in such a way that some editors, LDCad among them, show holes in the surface of the part. Is this a problem with LDCad or the existing parts? Are they still viable as a base for the new 26601 part?

I attached my current design to this post.

[Philippe Hurbain]

The problems you mention are probably caused by T-junctions (see http://wiki.ldraw.org/index.php?title=Av...Draw_parts). Could you post a screenshot showing the problem to confirm the diagnostic?

Edit: I had a close look at 2450 and 30503 in LDView and LDCad, but noticed no display artifacts (despite these parts do contain T-junctions)

[Niklas Buchmann]

The problems you mention are probably caused by T-junctions (see http://wiki.ldraw.org/index.php?title=Av...Draw_parts). Could you post a screenshot showing the problem to confirm the diagnostic?

Edit: I had a close look at 2450 and 30503 in LDView and LDCad, but noticed no display artifacts (despite these parts do contain T-junctions)

2450:
   
30503:
   
   

[Magnus Forsberg]

Sorry, but this part is already present at the PT



http://www.ldraw.org/cgi-bin/ptdetail.cgi?s=26601

I've made a new primitive to it. Be sure to download it too.

[Magnus Forsberg]

It looks to me like you have a rounding problem. Only two decimals might not be enough here.
Did you copy the prims from the other parts?

   

Turn on "Wireframe" and zoom in really close (by changing "View mode") to the corner, and you'll see that the lines don't coincide into a single vertice.

[Niklas Buchmann]

It looks to me like you have a rounding problem. Only two decimals might not be enough here.
Did you copy the prims from the other parts?



Turn on "Wireframe" and zoom in really close (by changing "View mode") to the corner, and you'll see that the lines don't coincide into a single vertice.

Those are screenshots of the existing official files (first from 2450, second and third from 30503). The errors were the reason I was asking what's acceptable on a new file and what's not.

Magnus' 26601 version on the other hand looks clean even in LDCad, somehow I missed it last week.

[Philippe Hurbain]

Magnus is right, the problem is mainly a rounding issue. T-junctions could be a problem, but in the screenshot below there is none...
Two digits (as used in 30503) is grossly insufficient, especially for transformation matrices of rotated primitives. 3 digits for coordinates / 4 digits for matrices coefficient is the bare minimum to get good results, and I generally use 4 and 5 digits to make sure.