Hi Tracers!

Finally I started with my first real big project - a giant colonist
spaceship providing room for more than 400,000 people in 352 cylindrical
habitation modules arranged in octagonal metal frameworks around a
7-mile-long central body containing, among other facilities, the thrust
engines and large spherical incubators for hatching genetically
engineered terraforming pioneer lifeforms. The habitation modules are
grouped in clusters of eight, each module has 16 levels, each level 10
(or 11) apartment sections.

In the image, you see the inner parlor (later, I will add some railing
along its inner edge, and of course an elevator system connecting the 16

position of the module cylinder.  The area lights are x, y, 1, 4,
adaptive 1, no jitter. But I just can't imagine why the lighting is so
patchy close to the lamp! It can't be a question of brilliance as I set
the value for the walls fairly low, to 0.1 - but what is it then?

Hoping for hints...

... see you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download 'raumschiff.jpg' (28 KB)

Preview of image 'raumschiff.jpg'

Has a coincident surface look to it, tried a regular point source yet?  Then
again, looking closer, the wall surface seems to have a normal applied, too
small scale and deep perhaps if so.
The door near the light on the right doesn't have a window or portal or
whatever that is.  Misplaced object possibly?
Anyway, if the area light dimensions come into contact with a object surface
that might not be such a good thing anyway, although I can't recall it being
a problem myself.  At least not if used as a looks_like.
Makes me wonder if it could be the 1,4 array with a x for width the 1 might
be confusing the ray tracer if in contact with a surface.  But the looks of
this rendering overall sure has that coincident surface appearance, noting
the flooring and ceiling parts are likewise speckled.

Bob H.

Post a reply to this message

Like Bob says, looks like a coinsidence surface problem to 
me. 

--
Cheers
Steve              email mailto:ste### [at] zeroppsuklinuxnet

%HAV-A-NICEDAY Error not enough coffee  0 pps. 

web http://www.zeropps.uklinux.net/

or  http://start.at/zero-pps

  3:21pm  up 126 days, 16:11,  2 users,  load average: 1.01, 1.09, 1.06

Post a reply to this message

"Bob H." schrieb:

> Has a coincident surface look to it, tried a regular point source yet?  Then
> again, looking closer, the wall surface seems to have a normal applied, too
> small scale and deep perhaps if so.
> The door near the light on the right doesn't have a window or portal or
> whatever that is.  Misplaced object possibly?
> Anyway, if the area light dimensions come into contact with a object surface
> that might not be such a good thing anyway, although I can't recall it being
> a problem myself.  At least not if used as a looks_like.
> Makes me wonder if it could be the 1,4 array with a x for width the 1 might
> be confusing the ray tracer if in contact with a surface.  But the looks of
> this rendering overall sure has that coincident surface appearance, noting
> the flooring and ceiling parts are likewise speckled.
>
> Bob H.

Hi Tracers!

"Bob H." schrieb:

> Has a coincident surface look to it, tried a regular point source yet?

Yes, meanwhile, see attachment! But the only change I see is that the shadows
in the dim upper floor are less differentiated, whereas the spottiness remains!

>  Then
> again, looking closer, the wall surface seems to have a normal applied, too
> small scale and deep perhaps if so.

Yes, it has a normal, but its bump_size is quite small, i. e. 0.075, so this
can't account for the spottiness!

> The door near the light on the right doesn't have a window or portal or
> whatever that is.

No, the small round door windows have variable transparency and phong (I'm
already considering animations...), so I set up an array containing the
transparency and phong values for each door window on the respective floor,
they are meant to change within 3 seconds (= 45 frames) from full opacity to
full transparency and vice versa.

> Misplaced object possibly?
> Anyway, if the area light dimensions come into contact with a object surface
> that might not be such a good thing anyway, although I can't recall it being
> a problem myself.  At least not if used as a looks_like.

That was a problem I encountered before; originally, the centers of the
cylindrical lamp bodies were placed at a radius of 5 from the center of the
whole cylindrical habitation module, while the inner surface of the round wall
also was at 5 (i. e. a cylinder with radius 5 cut from an originally 5.1
cylinder - the remaining wall is 0.1 units thick). But then the image got
completely messed up with bizarre coincident surfaces effects, so then I placed

the lamp bodies instead at 4.99.

You should know that the from whole lamp body a large box is subtracted to
prevent it from peeking out of the outer side of the round wall (its radius is
0.15, while the wall is only 0.1 thick). As I did that only after I added caps
and belts, I assigned no texture to the box... could THAT be the reason??

I just looked up the code again and found that the box goes from z=0.01 to
z=0.2... with the center of this slightly-more-than-half-cylinder at 4.99, its
cut side is still in direct contact with the wall. Now I changed the z value of

the box' lower left front corner to 0.001 and started again...

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download 'spaceship_corridor_point_lights.jpg' (20 KB)

Preview of image 'spaceship_corridor_point_lights.jpg'

"Yadgar" <j.b### [at] ndhnet> wrote in message
news:3B2200DF.B10D830A@ndh.net...
>
> That was a problem I encountered before; originally, the centers of the
> cylindrical lamp bodies were placed at a radius of 5 from the center of
the
> whole cylindrical habitation module, while the inner surface of the round
wall
> also was at 5 (i. e. a cylinder with radius 5 cut from an originally 5.1
> cylinder - the remaining wall is 0.1 units thick). But then the image got
> completely messed up with bizarre coincident surfaces effects, so then I
placed
> the lamp bodies instead at 4.99.

Although probably not a factor here you need to be aware that small numbers,
such as those of thousandths or less, can be trouble for POV-Ray if it
concerns an objects size.  Sphere for example will degrade at a mere <0.02,
box next, and eventually cylinder.  The cone primitive proves to be sturdy
at most any scale.
This doesn't apply to translated surfaces, lights, far as I know anyway.
Just thought I should mention it since you are working with fairly small
scales already.

> You should know that the from whole lamp body a large box is subtracted to
> prevent it from peeking out of the outer side of the round wall (its
radius is
> 0.15, while the wall is only 0.1 thick). As I did that only after I added
caps
> and belts, I assigned no texture to the box... could THAT be the reason??

If a CSG difference is used then yes, it matters that the cutaway object
isn't textured after the previous object to be cut is textured.  If
clipped_by is used it wouldn't matter.  An untextured object is blank, or
rather appears opaque black, so to leave it textureless (i.e. without the
parent CSG object texture) is sure to be conspicious.

> I just looked up the code again and found that the box goes from z=0.01 to
> z=0.2... with the center of this slightly-more-than-half-cylinder at 4.99,
its
> cut side is still in direct contact with the wall. Now I changed the z
value of
>
> the box' lower left front corner to 0.001 and started again...

Check into using clipped_by I think might be the answer if you're using
difference now.
This rendering still looks a lot like coincident surfaces but have you tried
using more sources for a higher sampling in those area lights? 2x4 or 2x8
might be needed, it could be that 1x4 causing the problem in part or whole.

Bob H.

Post a reply to this message

Hi Tracers!

"Bob H." schrieb:

> If a CSG difference is used then yes, it matters that the cutaway object
> isn't textured after the previous object to be cut is textured.  If
> clipped_by is used it wouldn't matter.  An untextured object is blank, or
> rather appears opaque black, so to leave it textureless (i.e. without the
> parent CSG object texture) is sure to be conspicious.
>
> Check into using clipped_by I think might be the answer if you're using
> difference now.
> This rendering still looks a lot like coincident surfaces but have you tried
> using more sources for a higher sampling in those area lights? 2x4 or 2x8
> might be needed, it could be that 1x4 causing the problem in part or whole.
>
> Bob H.

Now I've tried clipped_by (in the lamp body object within looks_like) AND 2x8
lights - and here's the result!
Gradually, I begin to ask myself how bad these things could go wrong... :-(((

The two pale blue spotted rectangles separated by a narrow strip of dark wall
obviosly are the box I liked to have cut from the lamp body - but are you sure
that clipped_by really substitutes difference rather than intersection (the PoV
Manual describes it as equivalent to intersection!)?

See you in Khyberspace!

Yadgar

Post a reply to this message

Attachments:
Download 'spaceship_area_light_2x8_clipped_by.jpg' (28 KB)

Preview of image 'spaceship_area_light_2x8_clipped_by.jpg'

"Yadgar" <j.b### [at] ndhnet> wrote in message
news:3B27A02C.C7CFE4D3@ndh.net...
>
> Now I've tried clipped_by (in the lamp body object within looks_like) AND
2x8
> lights - and here's the result!
---snip---
> that clipped_by really substitutes difference rather than intersection
(the PoV
> Manual describes it as equivalent to intersection!)?

Yeah, you've taken a backward step there  :-(
clipped_by doesn't leave any surfaces where the parent object is, just opens
it up like a empty can instead of a solid one.  That was the idea I had in
mind anyway.  Maybe I should have mentioned to 'inverse' it (the clipping
object) or something, thought you'd have gotten the idea yourself how it
worked.
Looks like that's what happened anyway, that you left the part you wanted
gone.

Bob H.

Post a reply to this message