Le 19-02-27 à 03:55, Thomas de Groot a écrit :
> A small test image with the corrected settings. What appears to be the 
> horizon is in reality the semitransparent bounding box, the camera being 
> situated just a tiny bit above it. The mist strands do not touch that 
> upper boundary but be aware that their extension upward is controlled by 
> the warp{turbulence} of 0.4 for the y-direction. So, it is a bit tricky 
> to find the correct values for the density map without the strands 
> piercing the container.
> 
> Samples may be a bit low as there are some artifacts visible but I 
> wouldn't change it as in your scene they would be hidden by the water 
> texture.
> 
> I have used here a somewhat denser scattering colour and also scattering 
> mode 5 which I often prefer, here with excentricity 0.5.
> 
> For this small scene, render time was about 30 minutes.
> 

When using turbulence, a rule of thumb is to assume that the area gets 
enlarged by the turbulence value.

If you use turbulence 0.4, then, what was at 1 should be assumed to get 
displaced between 0.6 and 1.4, and up to 2 for turbulence 1.

Post a reply to this message

Am 27.02.2019 um 09:55 schrieb Thomas de Groot:
> A small test image with the corrected settings. What appears to be the 
> horizon is in reality the semitransparent bounding box, the camera being 
> situated just a tiny bit above it. The mist strands do not touch that 
> upper boundary but be aware that their extension upward is controlled by 
> the warp{turbulence} of 0.4 for the y-direction. So, it is a bit tricky 
> to find the correct values for the density map without the strands 
> piercing the container.
> 
> Samples may be a bit low as there are some artifacts visible but I 
> wouldn't change it as in your scene they would be hidden by the water 
> texture.
> 
> I have used here a somewhat denser scattering colour and also scattering 
> mode 5 which I often prefer, here with excentricity 0.5.
> 
> For this small scene, render time was about 30 minutes.
> 
Hello Thomas,

many, many thanks for your efforts and for solving this problem. I was 
on the wrong track. There is another issue with media and sharp edges I 
experienced playing around with the emitting media simulating the sun in 
this image and expected a similiar problem with the fog. But these two 
problems are different issues completely. As Alain confirms your results 
one has to consider the amount of turbulence while calculating the 
scaling of the media or the boundaries of the container object 
additionally.

I will depict the other issue ASAP. At the moment it is at 39% after 
some 5 hours of rendering. In this case the container cannot prove a 
problem since it is much larger than the media and a box and not a 
sphere. I fear it is one of the problems arising from the prescision 
settings within POV to deal with the trade off between precision and 
rendering time.

Best regards
Michael

Post a reply to this message

Am 27.02.2019 um 11:25 schrieb Paolo Gibellini:
> MichaelJF wrote on 22/02/2019 20:04:
>> Hi to the crowd,
>>
>> this image has still a lot of flaws (sun colour and reflections, the 
>> swan model itself, textures of the cattails) and may be limitations to 
>> media by POV. media seems to be handling color_maps or density_maps 
>> different than pigment or texture having no interpolation between the 
>> map entries but shap edges. This is what I derived in a first attempt. 
>> Rendering time was a little bit longer (3 weeks).
>>
>> Best regards
>> Michael
>>
> A beautiful and peacefulness image!
> Perhaps the cane thicket at bottom right are too bright. And it might be 
> nice to see birds flying in the distance.
> ;-)
> Paolo
Many thanks Paolo,
it is a first WIP to demonstrate my idea how one can model steam fog 
with POV. And as you can see from the discussion here it is far away 
from being ready. I will consider your idea of birds in the distance. I 
simply could copy them from another picture I made in the past where I 
had a formation of cranes in the far distance. But first I would prefer 
a more realistic sunrise with a better sky sphere and a better simulated 
sun.

Best regards
Michael

Post a reply to this message

On 27-2-2019 17:39, Alain wrote:
> Le 19-02-27 à 03:55, Thomas de Groot a écrit :
>> A small test image with the corrected settings. What appears to be the 
>> horizon is in reality the semitransparent bounding box, the camera 
>> being situated just a tiny bit above it. The mist strands do not touch 
>> that upper boundary but be aware that their extension upward is 
>> controlled by the warp{turbulence} of 0.4 for the y-direction. So, it 
>> is a bit tricky to find the correct values for the density map without 
>> the strands piercing the container.
>>
>> Samples may be a bit low as there are some artifacts visible but I 
>> wouldn't change it as in your scene they would be hidden by the water 
>> texture.
>>
>> I have used here a somewhat denser scattering colour and also 
>> scattering mode 5 which I often prefer, here with excentricity 0.5.
>>
>> For this small scene, render time was about 30 minutes.
>>
> 
> When using turbulence, a rule of thumb is to assume that the area gets 
> enlarged by the turbulence value.
> 
> If you use turbulence 0.4, then, what was at 1 should be assumed to get 
> displaced between 0.6 and 1.4, and up to 2 for turbulence 1.

Exactly! Thanks for the precision given. This was one of the reasons why 
I gave the hint to try to put the scaling /after/ the warp {turbulence}. 
I didn't pursue that path too far though.

-- 
Thomas

Post a reply to this message

I commented the steam fog approximation out, so it rendered a little bit 
faster. One sees two instances of the same object (it is the example 
from the tutorials shipped with POV but in a box {-2,2 } instead of a 
sphere). The only difference is

object { Sonne
    scale <500,500,500>*5
    translate MediaPos
}

object { Sonne
    translate <3,1,5>
}

with MediaPos=< 649.685959 , 1442.616855 , 39968.697311 >

Obviously I will not use it in my scene. It's just a riddle I 
encountered by chance. I played a little bit with intervals, samples, 
variance, confidence but without any success. It seems to be due to the 
scaling alone.

Best regards
Michael

Post a reply to this message

Attachments:
Download 'twoinstancesofaemittingmedia.jpg' (555 KB)

Preview of image 'twoinstancesofaemittingmedia.jpg'

On 28-2-2019 19:25, MichaelJF wrote:
> I commented the steam fog approximation out, so it rendered a little bit 
> faster. One sees two instances of the same object (it is the example 
> from the tutorials shipped with POV but in a box {-2,2 } instead of a 
> sphere). The only difference is
> 
> object { Sonne
>     scale <500,500,500>*5
>     translate MediaPos
> }
> 
> object { Sonne
>     translate <3,1,5>
> }
> 
> with MediaPos=< 649.685959 , 1442.616855 , 39968.697311 >
> 
> Obviously I will not use it in my scene. It's just a riddle I 
> encountered by chance. I played a little bit with intervals, samples, 
> variance, confidence but without any success. It seems to be due to the 
> scaling alone.
> 

My immediate reaction to this would be: "Did you take the scaling into 
account in the media build of Sonne?"

You need to do the following /essential/ operation, for the first 
instance, concerning the emission and, if applicable, scattering and 
absorption:

#local MediaScale = <500, 500, 500>*5;

sphere {<0,0,0>, 1 //or a box of course
   hollow
   pigment {rgbt 1}
   interior {
     media {
       samples 50
       emission <0.8, 0.8, 0.8>/MediaScale
       scattering {1, rgb <0.1, 0.1, 0.1>/MediaScale}	
       density {
	spherical
	density_map {
	  [0.20 rgb 0.0]
	  [1.00 rgb 1.0]
	}
       }
     }
   }
   scale MediaScale
}

This is explained in the docs btw in 2.3.7.5 Media and transformations.

I think that lies at the root of your problem.

-- 
Thomas

Post a reply to this message

Hello Thomas,

thanks again. May be I read this section some years ago but it did not 
came into my mind. The only problem was, that it led me to a wrong track 
regarding the container issues with the steam fog approximation. I will 
not use an emitting media as sun but will follow an approach by Johannes 
Eriksson (May-June competition of the IRTC in 2006). But this is a 
legacy scene and the conversion is not as simple as I expected.

Best regards
Michael

Post a reply to this message

Le 19-02-28 à 13:25, MichaelJF a écrit :
> I commented the steam fog approximation out, so it rendered a little bit 
> faster. One sees two instances of the same object (it is the example 
> from the tutorials shipped with POV but in a box {-2,2 } instead of a 
> sphere). The only difference is
> 
> object { Sonne
>     scale <500,500,500>*5
>     translate MediaPos
> }
> 
> object { Sonne
>     translate <3,1,5>
> }
> 
> with MediaPos=< 649.685959 , 1442.616855 , 39968.697311 >
> 
> Obviously I will not use it in my scene. It's just a riddle I 
> encountered by chance. I played a little bit with intervals, samples, 
> variance, confidence but without any success. It seems to be due to the 
> scaling alone.
> 
> Best regards
> Michael
Hints :

Don't change intervals from the default of 1. Higher intervals count = 
slower render and may introduce artifacts.

Variance and confidence are used for sampling methods 1 and 2, and are 
not used with method 3 (default).

If you use two value for samples, the second will be silently ignored 
when using method 3.

Post a reply to this message

> Hints :
> 
> Don't change intervals from the default of 1. Higher intervals count = 
> slower render and may introduce artifacts.
> 
> Variance and confidence are used for sampling methods 1 and 2, and are 
> not used with method 3 (default).
> 
> If you use two value for samples, the second will be silently ignored 
> when using method 3.

Many thanks, I think this issues should be stated in the docs more clearly.

Best regards
Michael

Post a reply to this message

Am 24.02.2019 um 20:43 schrieb MichaelJF:
> Am 23.02.2019 um 10:12 schrieb Jörg "Yadgar" Bleimann:
>> Hi(gh)!
>>
>> On 22.02.19 20:04, MichaelJF wrote:
>>> Hi to the crowd,
>>>
>>> this image has still a lot of flaws (sun colour and reflections, the 
>>> swan model itself, textures of the cattails) and may be limitations 
>>> to media by POV. media seems to be handling color_maps or 
>>> density_maps different than pigment or texture having no 
>>> interpolation between the map entries but shap edges. This is what I 
>>> derived in a first attempt. Rendering time was a little bit longer (3 
>>> weeks).
>>
>> Just my mustard ;-):
>>
>> The rising fog should be turbulated more, now it looks like small 
>> straight columns of mist... and what about using Chris Colefax's 
>> lensflare macro?
>>
>> See you in Khyberspace!
>>
>> Yadgar
> Hi Yagdar,
> the rising turbulence is just my problem (see the code I posted some 
> minutes earlier). I cannot see a reason to add lens flare to the image. 
> Especially since we have no conversion of the lens flare macros by chris 
> colefax to POV 3.7. They did not work with 3.7 so far. I played around 
> with them to solve this issue a while ago, but failed unfortunatelly.
> 
> Best regards
> Michael
After looking at a lot of photographies of sundowns and sunrises again I 
now understand your recommendation of the lens flare macros. But they 
are still resistant to my conversion efforts. To chance ambient by 
emission simply doesn't work here. If you study his disc approach it 
should work, but there seems to be a hidden secret I have not detected 
so far. With Jaime's lightsys this simple exchange works.

Best regards
Michael

Post a reply to this message