after the sun has set ...

[lightning2911]

i wanted to try a night time version of my interior set.

i add 2 lamps (sphere) and 2 meshes that act as meshlight in the background. then i turn all sun and sky settings off.

below is what i get with different FG (16, 32, 128) settings. my questions now are why is there light under the lampshade mesh (its only open on the upper side) and how do i get rid of the white dots? it does not seem to be in the photons settings. i also tried to raise the lamps samples from 8 to 16. anything else i can try?

[AirChrome09]

In this topic there's the same problem
http://www.yafaray.org/community/forum/ ... =13&t=2631

I'm not sure but I think that the light behind the lamps is because:
1- The lamp reflects the light that the wood behind reflects and so on. (I have similar lamps in my living room but there isn't this problem; maybe because they are attached on the wall.
2- The lamp object is too thin.

These are suppositions made from what I can see from the images you posted.

Note: don't take this like an expert suggestion...

[Samo]

I believe it is the lamp type and position that its producing that kind of dots, for instance the spherelight running into walls and meshes.

First use an rectangular arealight, second be sure the arealight is not in contact with any surface, and keep it as separated from the wall as possible for a more efficient sampling.

Also your modelling workflow should follow realistic patterns, avoid single-face meshes, extrude them all.

[lightning2911]

thx for the tips ... will try them out

[lightning2911]

i have replaced the original (yaf sphere) lamps in the front with blender area lights and without much change. now i have replaced them with little cubes as meshlights. when i have the cubes higher than the lampshadeboxes i can see them glow. once i place them inside the lampshades there does not seem to be any light going up. the lightshades are "correctly" modelled boxes with thickness and opening upwards.

what surprises me is that the meshlights in the back work that well. is it just the lampshades?

the (yaf) sphere lights worked fine except for the light under the lampshade but i am pretty sure that the radius touched the mesh of the shade and the wall.

here is the file http://www.papermodels.at/testfile2.blend

these are pics i get:

no lamp activated (foreground)


lamp activated and seeable


lamp activated and submerged

[Samo]

do not use mesh lights, use arealights

[lightning2911]

this is what i get with arealights



what i dont understand is: why is there no light shining out upwards where the arealight is pointing? why behind the closed shade onto the wall?!?

as you can see in the blend file nothing is touching something.

and why do the meshlights behind print without problems?!?

i guess i am missing something little but important to make it work ...

[stoneage]

It is the lighting.
I deleted the three lamps and some of the meshlights, moved the remaining meshlights away from the objects and placed an area light in the centre of the space, and it looks like this:-

You really can't have lights in contact with geometry at all, or it will create noise.

testfile2.7z

(844.84 KiB) Downloaded 397 times

[lightning2911]

so i cannot get a view how it would look with certain lights?

maybe i am doing something wrong but as far as i can see all the lights are not in touch with geometry.

[IkariShinji]

Here's my try - wanted to keep the mood:

Test render

Replaced all meshlights in the back with area lights. Replaced the area lights in the foreground lamps with spot lights. Still some fireflies, though. Rendered with a tweaked version of your 'photon' setup. Had to crank up the settings quite a bit...

Here's the .blend:

testfile2.zip

Corresponding .blend

(1.05 MiB) Downloaded 427 times

[Samo]

background meshlights are producing most of the white dots IMO

[stoneage]

so i cannot get a view how it would look with certain lights?

maybe i am doing something wrong but as far as i can see all the lights are not in touch with geometry.

You can, you just might have to do it a little differently. For Blender lights it is the round part that needs to be free from the geometry(probably why light was showing underneath the shades), and meshlights can be inefficient if you want to use them as a main light source(I just realised the meshlights normals were pointing the wrong way too, which would have been making it worse).

One way would be to render without the light shades, then composite them in afterwards, another option is to replace all the meshlights with area lights and make them visible.

Just flipping normals on the meshlights and making sure the round part of the area light is not in contact with anything will help. It probably won't make an acceptable render though as there will likely still be some noise from the meshlights.
You could try increasing samples for the meshlights and for the integration method you are using, and use higher AA samples and passes to clean it up, but that will significantly increase render times.

EDIT - Here is a render with all the meshlights replaced with area lights:-

[Samo]

The problem is not meshlights, the problem is so much geometry so close to light sources. Some very bright paths are bouncing on geometry close to light sources and the image sampling is not able to domesticate those bright paths.

Adding to that problem is the fact that I believe arealights probably use a differerent and more efficient light sampling that mesh lights.

In fact, in your first examples, those crazy bright paths in secondary bounces are able to get into primary rays as well.

For instance in this example I have to use a very contundent sampling strategy, with as many as 1024 FG samples, to get rid of very bright high-contrast FG samples on geometry close to light sources

1024 FG samples x 3 AA samples

[lightning2911]

thanks all for the great input. i will try to work this all into my file and post my results.

[condar]

The problem is not meshlights, the problem is so much geometry so close to light sources. Some very bright paths are bouncing on geometry close to light sources and the image sampling is not able to domesticate those bright paths.

Adding to that problem is the fact that I believe arealights probably use a differerent and more efficient light sampling that mesh lights.

In fact, in your first examples, those crazy bright paths in secondary bounces are able to get into primary rays as well.

For instance in this example I have to use a very contundent sampling strategy, with as many as 1024 FG samples, to get rid of very bright high-contrast FG samples on geometry close to light sources

1024 FG samples x 3 AA samples

maybe a 128 FGsamplesx10 passes strategy?