POV-Ray : Newsgroups : povray.binaries.images : Progress on anniversary scene Server Time
13 Oct 2024 01:18:58 EDT (-0400)
  Progress on anniversary scene (Message 1 to 8 of 8)  
From: Cousin Ricky
Subject: Progress on anniversary scene
Date: 10 Feb 2024 16:36:02
Message: <65c7ec42@news.povray.org>
Of the 4 unfinished elements that were part of my POV-Ray 30th
anniversary tribute scene, 3 have been published: DeskLamp, RC3Wood, and
ShapeOfTheDay.

Unfortunately, there are a couple of lighting problems that showed up in
the interaction between DeskLamp and ShapeOfTheDay.  Most obviously, the
caustics are out of control.  More subtly, the light bulb emission is
too low; this can be seen as the dark spot in the reflection of the lamp
at the lower right corner of the wood block.  I have not figured out
what the problems are.

Image povray_30th_anniversary.20230605.jpg was rendered right after I
completed RC3Wood, and that is when the caustics bug first manifested.
This bug has diminished the usefulness of ShapeOfTheDay with refractive
and, to a lesser extent, metallic objects, but there are workarounds.
The easiest is to use one of the outdoor settings, which don't use DeskLamp.

An indoor workaround for image povray_30th_anniversary.20240207.jpg was
possible because I rendered it with separate photons and radiosity
passes.  For the photons pass, I used Declare=sotd_Quality=0, which
de-implements the lamp fixture, then re-enabled the lamp for the
radiosity passes.

The new image applied my light bleeding technique.  It is noticeably
darker than the 2021 render; I do not yet know why this is.


Post a reply to this message


Attachments:
Download 'povray_30th_anniversary.20230605.jpg' (423 KB) Download 'povray_30th_anniversary.20240207.jpg' (261 KB)

Preview of image 'povray_30th_anniversary.20230605.jpg'
povray_30th_anniversary.20230605.jpg

Preview of image 'povray_30th_anniversary.20240207.jpg'
povray_30th_anniversary.20240207.jpg


 

From: Alain Martel
Subject: Re: Progress on anniversary scene
Date: 11 Feb 2024 10:29:21
Message: <65c8e7d1$1@news.povray.org>
Le 2024-02-10 à 16:35, Cousin Ricky a écrit :
> Of the 4 unfinished elements that were part of my POV-Ray 30th
> anniversary tribute scene, 3 have been published: DeskLamp, RC3Wood, and
> ShapeOfTheDay.
> 
> Unfortunately, there are a couple of lighting problems that showed up in
> the interaction between DeskLamp and ShapeOfTheDay.  Most obviously, the
> caustics are out of control.  More subtly, the light bulb emission is
> too low; this can be seen as the dark spot in the reflection of the lamp
> at the lower right corner of the wood block.  I have not figured out
> what the problems are.
> 
> Image povray_30th_anniversary.20230605.jpg was rendered right after I
> completed RC3Wood, and that is when the caustics bug first manifested.
> This bug has diminished the usefulness of ShapeOfTheDay with refractive
> and, to a lesser extent, metallic objects, but there are workarounds.
> The easiest is to use one of the outdoor settings, which don't use DeskLamp.
> 

In povray_30th_anniversary.20230605.jpg, my guess would be not enough 
photons being shot. If you use a count, I suggest switching to spacing.
When using count, remember that it's spread across all light_source and 
all target objects. ALL non-shadowless light will emit photons unless 
they have this photons block :
photon{refraction off reflection off}
as the default is :
photon{refraction on reflection on}

You may use a non unit value for the target attribute of an object. That 
will be used as a spacing multiplier.
target 0.2 will DIVIDE the spacing by 5 (meaning 25 times more photons), 
while target 2 will double the spacing (meaning 4 times less photons), 
used when shooting photons at that object.

Maybe also increase the jittering of the photons. It looks like there is 
none.

The dark area in the reflection of the desk lamp look credible. I often 
see something similar IRL. You may try to give the interior of the shade 
an finish that include a moderate amount of emission. That should make 
the darker area less obvious.


Post a reply to this message

From: Cousin Ricky
Subject: Re: Progress on anniversary scene
Date: 19 Feb 2024 12:20:41
Message: <65d38de9@news.povray.org>
On 2024-02-11 11:29 (-4), Alain Martel wrote:
> 
> In povray_30th_anniversary.20230605.jpg, my guess would be not enough
> photons being shot. If you use a count, I suggest switching to spacing.
> When using count, remember that it's spread across all light_source and
> all target objects. ALL non-shadowless light will emit photons unless
> they have this photons block :
> photon{refraction off reflection off}
> as the default is :
> photon{refraction on reflection on}

I stopped using photons count many years ago, probably in 2006, when I
bought a computer with a decent amount of memory.  Both images used
photons spacing, and they used the same spacing value.

> You may use a non unit value for the target attribute of an object. That
> will be used as a spacing multiplier.
> target 0.2 will DIVIDE the spacing by 5 (meaning 25 times more photons),
> while target 2 will double the spacing (meaning 4 times less photons),
> used when shooting photons at that object.

That is just side-stepping the issue, because the '30' object uses the
identical photons block in both images.  All that changed was the light
source.

> Maybe also increase the jittering of the photons. It looks like there is
> none.

Jitter is turned on permanently, and as recommended in the POV-Ray
reference, I have not changed the default value.  (I neglected to
mention this in the DeskLamp user manual.)  Are you attributing the
discrete rainbow stripes to non-jittering?  That is most likely due to
imperfect anti-aliasing (interpolate 2) in the image map used to define
the glass '30'.

To be clear, the changes between these images, including the one I
posted in in 2021, were in the include files, not in the scene
definition itself.  The bottom line is that there is something about the
lamp fixture that causes the caustics to go haywire.

> The dark area in the reflection of the desk lamp look credible. I often
> see something similar IRL. You may try to give the interior of the shade
> an finish that include a moderate amount of emission. That should make
> the darker area less obvious.

An emitting object cannot be darker than the surface it is
illuminating--in this case, the interior of the lamp hood.  The problem
is that the emission on the light bulb object is too low.  I will not
put an emission on the lamp hood interior, because it does not emit
anything.


Post a reply to this message

From: Cousin Ricky
Subject: Re: Progress on anniversary scene
Date: 19 Feb 2024 12:34:00
Message: <65d39108$1@news.povray.org>
On 2024-02-19 13:20 (-4), Cousin Ricky wrote:
> On 2024-02-11 11:29 (-4), Alain Martel wrote:
>>
>> Maybe also increase the jittering of the photons. It looks like there is
>> none.
> 
> Jitter is turned on permanently, and as recommended in the POV-Ray
> reference, I have not changed the default value.  (I neglected to
> mention this in the DeskLamp user manual.)  Are you attributing the
> discrete rainbow stripes to non-jittering?  That is most likely due to
> imperfect anti-aliasing (interpolate 2) in the image map used to define
> the glass '30'.

I just realized I confused photons jitter with area light jitter.  But
since I have not used the jitter keyword in the photons global_settings
block, the default carries.  Since the same photons global settings were
used in both images, this cannot be the problem anyway.


Post a reply to this message

From: Mr
Subject: Re: Progress on anniversary scene
Date: 18 Mar 2024 04:55:00
Message: <web.65f800e41e38ca9316086ed06830a892@news.povray.org>
Cousin Ricky <ric### [at] yahoocom> wrote:
> On 2024-02-19 13:20 (-4), Cousin Ricky wrote:
> > On 2024-02-11 11:29 (-4), Alain Martel wrote:
> >>
> >> Maybe also increase the jittering of the photons. It looks like there is
> >> none.
> >
> > Jitter is turned on permanently, and as recommended in the POV-Ray
> > reference, I have not changed the default value.  (I neglected to
> > mention this in the DeskLamp user manual.)  Are you attributing the
> > discrete rainbow stripes to non-jittering?  That is most likely due to
> > imperfect anti-aliasing (interpolate 2) in the image map used to define
> > the glass '30'.
>
> I just realized I confused photons jitter with area light jitter.  But
> since I have not used the jitter keyword in the photons global_settings
> block, the default carries.  Since the same photons global settings were
> used in both images, this cannot be the problem anyway.

Hi,
First, area lights can't generate proper caustics in any renderer, the caustics
generator has be lit with a single point of emission, and excluded from others,
If soft shadows are needed the area light can probably still be used as shadow
only with some complementary exclusions and many options off.

Then, among other things, some of these will just as well be too obvious for
some advanced user as you to have forgotten about, but still, and not saying
features should be turned off, ...but may be investigated just in case...

* Did you use importance sampling?
* Did you use *any* HDR image?
* Did you use *any* fresnel outside reflection block?
* Did you use conserve_energy for all reflective materials?
* Did you use albedo keyword?
* Did you use some big light related macros like Lightsys or spectral?
* Did you use any distortion of camera such as a normal block?
* Did you use Newton thin film coating (iridescence)?
* Did you use double_illuminate?
* Did you use backside diffuse?


Post a reply to this message

From: Cousin Ricky
Subject: Re: Progress on anniversary scene
Date: 18 Mar 2024 09:14:04
Message: <65f83e1c$1@news.povray.org>
On 2024-03-18 04:52 (-4), Mr wrote:
> 
> Then, among other things, some of these will just as well be too obvious for
> some advanced user as you to have forgotten about, but still, and not saying
> features should be turned off, ...but may be investigated just in case...
> 
> * Did you use importance sampling?

Yes, in the updated scene only.

> * Did you use *any* HDR image?

No.

> * Did you use *any* fresnel outside reflection block?

Yes, on the floor (both the 2021 scene and updated scene), the magenta
checks on the hand plane (both scenes), the wooden parts of the hand
plane (updated scene only), and the light bulbs (updated scene only).

> * Did you use conserve_energy for all reflective materials?

Yes, on non-metallic materials only, in both scenes.

> * Did you use albedo keyword?

For specular albedo, yes, in both scenes; for diffuse albedo, no.

> * Did you use some big light related macros like Lightsys or spectral?

No.

> * Did you use any distortion of camera such as a normal block?

No.

> * Did you use Newton thin film coating (iridescence)?

No.

> * Did you use double_illuminate?

No.

> * Did you use backside diffuse?

No.


Post a reply to this message

From: Mr
Subject: Re: Progress on anniversary scene
Date: 19 Mar 2024 08:50:00
Message: <web.65f989ab1e38ca9316086ed06830a892@news.povray.org>
> > * Did you use importance sampling?
>
> Yes, in the updated scene only.
>
Then did you check by putting explicit importance values everywhere that your
samples may no longer get lost on far off  or less valuable areas?


> > * Did you use *any* fresnel outside reflection block?
> > * Did you use albedo keyword?

Since both of these are newer features, does deactivating them everywhere solve
the excessive light?


> > * Did you use conserve_energy for all reflective materials?
>
> Yes, on non-metallic materials only, in both scenes.
>

Strange, I would use it primarily on metallic materials, am I completely wrong
in this usual workflow assumption? or is yours a result of some other
constraint?


Post a reply to this message

From: Cousin Ricky
Subject: Re: Progress on anniversary scene
Date: 20 Aug 2024 19:55:00
Message: <web.66c52b851e38ca9360e0cc3d949c357d@news.povray.org>
Cousin Ricky <ric### [at] yahoocom> wrote:
> Of the 4 unfinished elements that were part of my POV-Ray 30th
> anniversary tribute scene, 3 have been published: DeskLamp, RC3Wood, and
> ShapeOfTheDay.
>
> Unfortunately, there are a couple of lighting problems that showed up in
> the interaction between DeskLamp and ShapeOfTheDay.  Most obviously, the
> caustics are out of control.

I found out what the problem is: photons are broken with true inverse squared
lighting.  (See my post from June 19 in p.beta-test.)  The DeskLamp module
automatically took advantage of true inverse squared attenuation for POV-Ray
3.8, something that my render rig did not do prior to incorporating DeskLamp.

DeskLamp 2.0.1 has true inverse squared attenuation turned off by default; see
p.object-collection.

> The new image applied my light bleeding technique.  It is noticeably
> darker than the 2021 render; I do not yet know why this is.

My best guess is the implementation of the light source.  The hood of the
finished DeskLamp physically occludes the light source, while the earlier place
holder lamp merely simulated the occlusion with a spotlight.  The earlier
spotlight underestimated the extent of the occlusion, so while the direct
illumination is the same for both setups, the global illumination is lower for
the DeskLamp.


Post a reply to this message

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.