Hi(gh)!

With my current house project ("Port Whatmough Residential Area") I 
encountered some strange artifacts when I tried to join two segments of 
a roof railing (a straight and a curved one) with a sphere.


having a radius of 3.5 units; as the railing generally is spaced 0.1 
units inwards from the roof's eged, the curved segment's (technically a 
section of a torus) radius is 3.6 units, therefore it's angular size is 


Although I've taken the resulting difference in the x direction (which 
calculates as cos(atan(0.1/3.6))*3.6) relative to the eastern end of the 
straight section into account, the spherical cornerpiece still does not 
fit seamlessly.

Also, along both railing sections, strange dots like those resulting 
from coincident surfaces are visible - but there is no concurrent 
surface at all!

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download '2011-09-30 port whatmough residential building #1, take 46 - extreme close-up of southern end of ground floor cutout rai' (16 KB) Download 'artifact_problem.jpg' (20 KB)

Preview of image '2011-09-30 port whatmough residential building #1, take 46 - extreme close-up of southern end of ground floor cutout rai'

Preview of image 'artifact_problem.jpg'

Le 30/09/2011 11:47, Jörg 'Yadgar' Bleimann a écrit :
> Also, along both railing sections, strange dots like those resulting
> from coincident surfaces are visible - but there is no concurrent
> surface at all!

What are you using for the railing sections ? spheresweep ?
hand made macro of sphere & cylinder ?

Post a reply to this message

Hi(gh)!

On 30.09.2011 13:20, Le_Forgeron wrote:

> What are you using for the railing sections ? spheresweep ?
> hand made macro of sphere&  cylinder ?

Just cylinders and (sections of) tori (spheres only where cylinders 
directly meet tori sections)... currently simply placed explicitly, 
later on I would like to program a macro to make my railing concept 
universally reusable (yes, also for the whole PoV-Ray community - I will 
place the macro in the povray.org object library)!

Attached is a view of another portion of the railing, showing how a 

radiosity (just radiosity { } within the global_settings statement - 
could it be that the artifacts are caused by that crude radiosity?) 
which might improve with future versions.

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download '2011-09-30 port whatmough residential building #1, take 48 - northwestern corner of ground floor roof.jpg' (52 KB)

Preview of image '2011-09-30 port whatmough residential building #1, take 48 - northwestern corner of ground floor roof.jpg'

Le 30/09/2011 13:56, Jörg 'Yadgar' Bleimann a écrit :

> Attached is a view of another portion of the railing, showing how a
> standard 90° corner is supposed to look like...

The old mechanic in me is asking: how the hell do you assemble that ?
There is no visible crew on the holding vertical poles.
The horizontal rail seems to be made of one piece.
And the vertical poles seems to be fixed directly in the cement.

For your cylinder/sphere/torus seams... the intersection of a torus is a
circle (which connect to a sphere of same radius) only for 3 well-known
cutting planes. All other intersections of a torus with a plane is a 3D
curve which does not match a circle.

For a point P on a torus:
there is the plane parallel to the plane of symmetry of the torus.
there is the plane perpendicular to the previous plane and passing at
the centre of the torus.
and there is the plane for the Villarceaux circles. (also going via the
centre of the torus, with a slant)

Post a reply to this message

> Hi(gh)!
>
> With my current house project ("Port Whatmough Residential Area") I
> encountered some strange artifacts when I tried to join two segments of
> a roof railing (a straight and a curved one) with a sphere.
>

> having a radius of 3.5 units; as the railing generally is spaced 0.1
> units inwards from the roof's eged, the curved segment's (technically a
> section of a torus) radius is 3.6 units, therefore it's angular size is

>
> Although I've taken the resulting difference in the x direction (which
> calculates as cos(atan(0.1/3.6))*3.6) relative to the eastern end of the
> straight section into account, the spherical cornerpiece still does not
> fit seamlessly.

I think that should be:

     cos(asin(0.1/3.6))*3.6
   = sqrt(sqr(3.6)-sqr(0.1))

In addition, are you sure you cut the cylinder and torus properly?

Post a reply to this message

Le 30/09/2011 20:21, clipka nous fit lire :
> Am 30.09.2011 11:47, schrieb Jörg 'Yadgar' Bleimann:
>> Hi(gh)!
>>
>> With my current house project ("Port Whatmough Residential Area") I
>> encountered some strange artifacts when I tried to join two segments of
>> a roof railing (a straight and a curved one) with a sphere.
>>
>> The curved segment follows a rounded 90° cutout of the ground floor
>> having a radius of 3.5 units; as the railing generally is spaced 0.1
>> units inwards from the roof's eged, the curved segment's (technically a
>> section of a torus) radius is 3.6 units, therefore it's angular size is
>> slightly less than 90° (see second attached image).
>>
>> Although I've taken the resulting difference in the x direction (which
>> calculates as cos(atan(0.1/3.6))*3.6) relative to the eastern end of the
>> straight section into account, the spherical cornerpiece still does not
>> fit seamlessly.
> 
> I think that should be:
> 
>     cos(asin(0.1/3.6))*3.6
>   = sqrt(sqr(3.6)-sqr(0.1))
> 
> In addition, are you sure you cut the cylinder and torus properly?

If, as I now think, he used a translation of the clipping plane, the
answer is no. (a rotation should have been used)

Post a reply to this message

Hi(gh)!

Am 30.09.2011 21:53, schrieb Le_Forgeron:

>> In addition, are you sure you cut the cylinder and torus properly?
>
> If, as I now think, he used a translation of the clipping plane, the
> answer is no. (a rotation should have been used)

The cylinder is not cut at all, but defined according to the length of 
the straight rail piece, with equal y and z coordinates respectively at 
beginning and end, so there is no distortion. The torus is defined in 
default position, i. e. parallel to the x-z plane and cut by two boxes 
rotated only around the y axis, so the cut planes are always circles, 
not ellipses!

See you in Khyberspace!

Yadgar

Post a reply to this message

Hi(gh)!

> I think that should be:
>
> cos(asin(0.1/3.6))*3.6
> = sqrt(sqr(3.6)-sqr(0.1))

With

<4.9+(3.6-sqrt(sqrt(3.6)-sqrt(0.1))), sh+rfh/(mr+1), 4.1>

as the sphere's coordinates rather than

<4.9+(3.6-cos(atan(0.1/3.6))*3.6), sh+rfh/(mr+1), 4.1>

the result is even worse - see attached image!

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download 'railing_corner.png' (19 KB)

Preview of image 'railing_corner.png'

> Hi(gh)!
>
>> I think that should be:
>>
>> cos(asin(0.1/3.6))*3.6
>> = sqrt(sqr(3.6)-sqr(0.1))
>
> With
>
> <4.9+(3.6-sqrt(sqrt(3.6)-sqrt(0.1))), sh+rfh/(mr+1), 4.1>
> [...]
> the result is even worse - see attached image!

Of course: You mixed up sqrt and sqr :-P

My bad though, because sqr doesn't exist as a predefined formula in 
POV-Ray; so it should be:

   sqrt(3.6*3.6 - 0.1*0.1)

Post a reply to this message

Hi(gh)!

Am 01.10.2011 03:19, schrieb clipka:
> Of course: You mixed up sqrt and sqr :-P
>
> My bad though, because sqr doesn't exist as a predefined formula in
> POV-Ray; so it should be:
>
>    sqrt(3.6*3.6 - 0.1*0.1)

Now the result is the same like with my original version (see attachment)...

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download 'whatmough_residential.png' (29 KB)

Preview of image 'whatmough_residential.png'