Can someone explain what's causing the splotchyness and how to fix it? I saw the same phenomenon in PP2012 shortly before upgrading to 2014, so I know it's not 14's fault.  It has to be something I'm doing wrong.

I get the same result whenever I add one or more point lights to any scene and set them to Inverse Square attenuation, regardless of what kind of set - boutique, living room, garage, etc - I'm using them in.  I've already checked the following:

The light is not inside the Foyer's geometry.

There are no image maps plugged into the light's surface.

The effect does not appear when IDL is switched off, or when I use Inverse Linear or Constant attenuation.

GC is enabled at 2.2.

I tried deleting all but one of the point lights and BB's sphere, and still got the splotchyness when I switched the remaining light's attenuation to Inverse Square.

So.  Thoughts?  Ideas?  Suggestions?  I've set Reality 3 aside until a more stable version is released - see this thread for details - and need to be able to use Inverse Square attenuation.

Thanks.

U need to increase your IDL quality.

Laurie

What's a good setting for draft renders?

I would say:

 - increase bounces, as nature itself applies an infinite number of them, and the setting is just a cut-off speeding up render time somewhat at the cost of creating artifacts. Poser applies the amount of bounces it needs, adhering to an upper limit as set. By applying the Render FireFly option from the Scripts menu, you can adjust the IDL bounces separately.

 - reduce the lighting levels, as the splotches are apparent cases of overlit areas. Poser handles the lighting internally the correct way (as it uses EXR format) but when presenting the result, levels are clipped.

 - gradually increase quality settings, as higher levels come with longer rendertimes. You have to find yourself some best balances for your drafts and finals, for this scene.

Hope this helps

Quote - This is the Dream Home Foyer and Living Room by Redhouse Studios. I'm using one Infinite light for the Sun, Bagginsbill's Environment Sphere to give it something to bounce off of, and two Point lights for the recessed ceiling lights above the entryway. All three lights are set to cast Raytraced shadows.

Can someone explain what's causing the splotchyness and how to fix it? I saw the same phenomenon in PP2012 shortly before upgrading to 2014, so I know it's not 14's fault.  It has to be something I'm doing wrong.

I get the same result whenever I add one or more point lights to any scene and set them to Inverse Square attenuation, regardless of what kind of set - boutique, living room, garage, etc - I'm using them in.  I've already checked the following:

The light is not inside the Foyer's geometry.

There are no image maps plugged into the light's surface.

The effect does not appear when IDL is switched off, or when I use Inverse Linear or Constant attenuation.

GC is enabled at 2.2.

I tried deleting all but one of the point lights and BB's sphere, and still got the splotchyness when I switched the remaining light's attenuation to Inverse Square.

So.  Thoughts?  Ideas?  Suggestions?  I've set Reality 3 aside until a more stable version is released - see this thread for details - and need to be able to use Inverse Square attenuation.

Thanks.

that is rather pretty though. lol rather a magical effect.

This is a test render I just completed, using the same Foyer with only one Point light at 80% intensity (it was also at 80% in the first test) with Inverse Square attenuation and GC enabled at 2.20.  For the Render Settings, I moved the Auto Settings slider to the line marked Final, switched to Manual settings, clicked Acquire from Auto, and manually ramped the IDL quality up to 40 (the first test was at 7).

Quote - that is rather pretty though. lol rather a magical effect.

Mmm... yes... a shame it's not what I want.

Point lights shine up onto the ceiling as well as down into the room. If they are in fixtures, they shine onto the fixture surface.

These surfaces that are close obey the inverse square law, which for point sources means that the intensity not only decreases with increasing distance, but also increases with decreasing distance - squared.

So - imagine you set the light to 80% intensity - that's the luminance at 1 Poser Native Unit, or roughly 103 inches (or 96 if you still believe in the Poser 5 measurement system. No matter...) Let's keep the math simple and call it 100 inches.

At 50 inches, the intensity is 4x 80% = 320%.

At 25 inches, the intensity is 16x 80% = 1280%

At 10 inches, the intensity is 100x 80% = 8000% (Already pretty intense but we're not done)

At 2 inches, the intensity is 2500x * 80% = 200,000% !!!!!

So - if you have those point lights 2 inches away from the ceiling, parts of the ceiling are brighter than the arc of a welding torch.

Now - part 2:

IDL sends out rays to sample the lighting from the world and decide what light is reaching every point in the scene.

It can't send out as many rays as real life, so it sends out anywhere from a couple dozen to a couple thousand from each sample area, depending on your IDL quality settings.

Now you have these few ultra hot spots on the ceiling, and the rest is much darker. Most of the time, none of your few-dozen sample rays hit a hot spot and it looks normal. Once in a while, a single ray hits one of these 2500 x brighter spots and it overwhelms the other samples - you got a hot spot on the wall.

Solution: (Assuming you don't accept the get-another-renderer solution)

1. Use spotlights instead of point lights so they only shine down, not up. Requires some tweaking as the falloff is not set up as mimicing real lights by default. I'm talking about the angular falloff here. Real spotlights have some angular falloff, but not like Poser default spot lights do.

2. Move the point lights away from the ceiling a bit more.

3. Put something black between the point lights and the ceiling (just in the places that are really close.) If you happen to "see" these objects because they are in the cameras view, then make them invisible to camera. IDL will "see" them, but the camera won't.

Quote - 2) Move the point lights away from the ceiling a bit more.

Did you use Inverse Square attenuation on the point lights for your lamps in this thread?  Where were your lights positioned?

BTW - I tried that idea, and moved my Point light about 10 inches below the ceiling fixture.  No dice.

The ceiling point light was about 10 inches below the ceiling. Maybe I had it on inverse linear. The other two were inside the table lamps.

Quote - The other two were inside the table lamps.

How did you to avoid the splotching I'm getting?

I just checked - all my lamps were on inverse linear after all. With inverse square, the close proximity to IDL objects creates the splotches in my scene as well.

Inverse square is the more correct one, but only if Poser implemented a minimum distance cutoff clamp value. This would fix the problem. Real lamps are not point sources, and they do not become infinitely bright - inside a certain radius you are no longer looking at a single point but a gigantic bulb (from the point of view of a nearby fly).

I don't know why they left out this simple solution: The exponent should approach -2 (inverse square) at larger distances, but approach 0 (constant) at smaller distances, with a suitable blend in between. This would neatly approximate the reality of a light bulb, while still using a pointer source geometry simplification.

Quote - I just checked - all my lamps were on inverse linear after all. With inverse square, the close proximity to IDL objects creates the splotches in my scene as well.

Ah-hah.  So, our favorite Poser guru got the same thing I did.

Quote - Inverse square is the more correct one, but only if Poser implemented a minimum distance cutoff clamp value...  I don't know why they left out this simple solution...

Any idea how Cage's scene came out so smoothly lit?

Perhaps it helps to alter the direct lights into high intensity IDL lights (objects with high ambient values). When using panels like that (to mimic photographers softboxes) I noticed a rather constant lighting level at short distance, and about inverse square at large distance - like mentioned by BB above.

BTW I found IDL to become splotchy as well at low lighting levels in an uneven lighting ditribution. Like using mentioned softboxes but no light from a dome, and the panels at a larger distance from the objects in the scene and/or rather small compared to those objects. Growing the panels and putting them close to the objects helped a lot (as it does in reallife photography as well!!).

In other words: we might have to undertake a rather serious investigation to the splotchiness of IDL in general. Time to re-open Muppets Lab...

You made me search. The correct equation is 1 / (a + bx + cx^2) and the derivation of the coefficients a, b, and c can be deduced from simple geometric models and assuming that the distance, x, is with respect to a non-point light, but something having an actual volume and surface.

For a sphere light, see this beautiful demonstration:

http://imdoingitwrong.wordpress.com/2011/01/31/light-attenuation/

Quote - For a sphere light, see this beautiful demonstration:

http://imdoingitwrong.wordpress.com/2011/01/31/light-attenuation/

That is pretty nice.  It'd be even better if we could achieve that effect in Poser.

just for your info, I re-did the math myself.

For a disk-shaped lightsource (as a pointlight with some size at a distance), with radius R lighting a sensor at distance d, the light captured by the sensor is directly proportional to { 1 - 1 / sqrt(1+ R^2/d^2) }.
Just fragmenting the disk into tiny inverse square pointlights, and integrating over the disk.

Close to the disk, lighting levels become constant and far from the disk they become inverse square:

d = 0.1R, L = 0,90
d = 0.2R, L = 0,60
d =    1R, L = 0,30
d =    2R, L = 0,10
d =  10R, L = 0,005
d =  20R, L = 0,0012
d = 100R, L = 0,00005

Similar math holds for calculating lighting levels in streets having large shopping windows shining on the pavement. Constant near the shops, inverse square at large distance, and about inverse linear in between. That's why software like Poser offers all these three options for light attenuation, btw.

Now the issue: why is Poser IDL producing splotches, and what to do about it?

I thought I already explained why above. When Inverse Square is chosen, objects that are close to the light are lit at levels approaching a welding torch. Not in real life, but because Poser models point lights as points, not spheres.

So, is Inverse Square + IDL broken?  Should I submit a bug report to SM?

I showed them this in Poser 8. They know it does this. This phenomenon was explained to me by Stefan Werner. They know exactly why it happens. They expect you to deal with it. I don't know why they don't just put the spherical formula in.

Also, if the light shader could have one more kind of node - a node that gives the distance to the light, then we could implement any distance based falloff we want using math nodes.

We do have the position of whatever is being lit, and I used to that to create inverse square falloff lights back in Poser 5, using math nodes. But you had to type in the coordinates of the light into one of the nodes every time you moved it. If we even just had an LP (light position) then combined with P I could do it.

Cage's OP in the other thread I linked to asked another question about Inverse Square.  Now I wonder if the image he included actually used Inverse Linear and he just didn't realize the difference.  If this issue dates back to Poser 8 and earlier, the light in his desk lamp would not have appeared that smooth otherwise, unless of course he knows enough about Poser to have been able to "deal with it" himself.

How did they get such a smooth result?

You can change the start and end distances on a point light...doesn't that help some? Maybe make the start distance farther than the point light is from the ceiling/table/lampshade/lamp/etc? I don't have time to test that theory right now. Either way, I'd imagine you'd still get the spotchy shadowing, which I've had problems with in the past too, but seems to soften them when I jack up the IDL quality...

Laurie

Quote - You can change the start and end distances on a point light...doesn't that help some?

Those parameters only apply to the linear falloff mode. Falloff begins at the start distance and the light goes to 0 at the end distance.

They should have used the start distance for the inverse falloffs, treating it as the radius of the light sphere. Since it defaults to 0, the default would be a traditional point light, but for those situations where it matters to adjust, you'd adjust the radius. Too bad.

Perhaps they didn't use IDL in those pictures? Perhaps there is no environment or, at least, no geometry right next to the pointlight's location.

Inverse Square Fall-off is not completely "broken" as you put it, but rather it fails quite badly in certain particular circumstances (as I remember pointing out to you when you asked the same question elsewhere last week). If you keep those circumstances in mind and avoid them or put in very simple measures to counteract the problems of those circumstances, everything's fine.

The attached picture is lit using only an ISF pointlight and IDL.

The pointlight is set to 40% intensity and is positioned less than an inch below the slightly off-white ceiling. How did I avoid the speckles? A square positioned between the light and the ceiling, and set to pure black so that none of the ceiling that's in the light's "ISF hot zone" is available to bounce that light back into the scene. I did a render without the black square and, sure enough, the speckles appeared.

Aww...that's too bad :( I like the look of inverse square better than linear, but I do end up using linear most of the time. I only use inverse square for my render box (visible in preview but not when rendered and has light emitter checked), where the side walls are so large compared to the universe that the light never reaches them and then use a spot (inverse linear) and emitters (very dim) in the ceiling. Seems to cancel everything out well enough. LOL

Laurie

Quote - > Quote - How did they get such a smooth result?

 

Perhaps they didn't use IDL in those pictures? Perhaps there is no environment or, at least, no geometry right next to the pointlight's location.

 

Inverse Square Fall-off is not completely "broken" as you put it, but rather it fails quite badly in certain particular circumstances (as I remember pointing out to you when you asked the same question elsewhere last week). If you keep those circumstances in mind and avoid them or put in very simple measures to counteract the problems of those circumstances, everything's fine.

 

The attached picture is lit using only an ISF pointlight and IDL.

The pointlight is set to 40% intensity and is positioned less than an inch below the slightly off-white ceiling. How did I avoid the speckles? A square positioned between the light and the ceiling, and set to pure black so that none of the ceiling that's in the light's "ISF hot zone" is available to bounce that light back into the scene. I did a render without the black square and, sure enough, the speckles appeared.

Does it still work if you make the black square "invisible to camera"? Was just thinking in this instance that there'd be no way to render any part of a ceiling tho I guess most renders it's usually hidden ;).

Quote - Does it still work if you make the black square "invisible to camera"? Was just thinking in this instance that there'd be no way to render any part of a ceiling tho I guess most renders it's usually hidden ;).

Yes, it does. I've just tried and it works exactly as in the original.

If I point the camera at the light, and the black square, it looks kind of odd how the light is now lighting up the ceiling with a bit missing, but in real life that patch of darkness would be covered by the light bulb, fluorescent tube, lampshade, etc. That oddness is only visible because Poser lights don't show up as actual objects, but yeah... "Visible in camera" switched off definitely doesn't bring back the splotches.

Quoting myself from my first reply up top:

Quote - 3) Put something black between the point lights and the ceiling (just in the places that are really close.) If you happen to "see" these objects because they are in the cameras view, then make them invisible to camera. IDL will "see" them, but the camera won't.

I have a new 4th solution. The source of the problem is the light is over excited inside a small radius, but the primary symptom is that the nearby object is glowing at like 30000 intensity.

If you can't fix the source, address the symptom instead.

So - make sure it can't do that. Change the material on the surface. Don't use the built-in diffuse and specular. Instead, build shaders as I do, using nodes assembled and plugged into Alternate_Diffuse.

Just before the assemblage goes into the Alternate_Diffuse, connect an additional Color_Math node. Set it to Min. Set the second value to white and plug in a math node. Set the math node to something reasonable, like 10. Now when the surface brightness gets above 10, it will just stop at 10. Problem solved.

this is the Lighting Report from Muppets Lab :)

I was able to reproduce the splotches, they occur under very specific conditions. Deviating from those conditions in any way smoothes the lighting results, and make the splotches go.

To make splotches I had to:

 - use point lights, with inverse square attenuation, at high intensity, nearby a wall, ceiling or other object in the scene. This setup creates extreme hotspots on the nearby object.

 - use IDL rendering at reduced quality settings. This propagates the hotspots around but fails to smooth them out.

So do note that it's not the light itself, but the hotspot it creates which causes the issue under lowQ IDL rendering. Of course, the issue can be avoided by

• not using IDL rendering
• not using point lights
• not using inverse square attenuation
• not using high intensity

but usually this conflicts with the scene you're building. When the pointlight is not in view of the camera (a bit behind, at the ceiling or wall) you will profit from making a big hole in there, so there is nothing to cast the hotspot. Ultimately, ramping up the quality settings help you out at the cost of extra render time.

LowQ is: bounces 2, irr caching 10, IDL quality 10, pixel samples 3. See image

HighQ is: bounces max (12, or up using the D3D FireFly render in scripts menu), irr caching 90, IDL quality 90, pixel samples 30. Just play with small renders to find the best quality / rendertime settings before going big. Note that when bounces go up, (ambient) lighting levels go up seriously so you have to adjust your light intensities accordingly.

I did not find that setting the Dist Start to anything made a difference (sorry Laurie), it still created the hotspots. Spotlights do not create hotspots, inverse linear attenuation seems to create a minor hotspot but not intense enough to create splotches at LowQ IDL rendering.

I keep seeing information that I gave in first post being given as "new". I'm getting irritated.

notes on the scene:

• you can see the room, the pillars, the bed and the ball shaped guy in it :-)
• you can "see" the pointlight nearby the ceiling, at least the hotspot is clear
• the wall behind the camera is missing, the scene is surrounded by an EnvSphere and this way outdoor light shines in.
• outdoor light is one infinite light, you can see its shadowing on the back wall
• plus the EnvSphere which has all Diffuse and Specular blackened out, and has a full white ambient instead
• other materials are full white diffuse/specular, the simple way.

Quote - Quoting myself from my first reply up top:

I was going to refer LaurieA to that response, but I'm the sort of person who likes to test the truth of anything I say before I definitely say it, and since I'd never tried "Visible in camera" switched off on an ISF "fix" square, I wanted to see what would actually happen. Once I knew with absolute 100% certainty that it still worked, then I was happy to respond with an affirmative!

On the subject of ISF lights, I used to use your ISF shader set-up before Poser 8/PP2010 came along with in-built ISF. It was slightly more work to get the co-ordinates for the light position but it was definitely worth the extra few seconds compared to standard Poser lights at the time. The thing that I was looking into doing was changing it so that instead of always having an exponent of 2 (which is unrealistic if we're being really pedantic), it could be changed to have some other exponent in there. I was just about to start playing about with that idea when PoserPro 2010 came along and it got abandoned. I'm wondering if it's worth bringing that idea back up again, since neither pure inverse linear nor pure inverse square fall-off models any feasible light source in the real world.

Quote - I'm wondering if it's worth bringing that idea back up again, since neither pure inverse linear nor pure inverse square fall-off models any feasible light source in the real world

Easy enough to do with matmatic.

I'm making it now.

from the hiQ settings downwards to find what really matters:

• pixel samples from 30 to 3 - no splotches
• AND quality from 90 to 10 - splotches !!
• pixel samples back to 30 - still splotches but far less and softer
• irr caching from 90 to 10 - even less and softer splotches, but more irregulaties in corners (wall/floor connections)
• all three down - soft spotches and irregularities
• bounces down from 12 to 2 - see first test image
• irr caching up to 90 again - strong splotches
• IDL quality up to 90 instead (caching to 10) - very very mild splotches and irregularities
• quality up, pixels up (chaching down, bounces down) - even milder splotches and irregularities

so... it's not just one of the settings, it's all four or them. Cashing is the nasty one, if it goes up splotches increase but irregularities decrease. Then you need quality and pixels up to decrease splotches again, and bounces to finish things off. 

and I do like BB's "light absorbing wall paint" too, as a good alternative (instead of drilling holes in the ceiling).

I just experimented a lot with a light shader.

I made it pay attention to a virtual radius. I used the radius based equation given above. It doesn't look any different than a straight inverse square, except within 3 radii. For a light bulb, which you usually don't even look at, this is a moot point.

So the real benefit is not in inverse something-not-squared. The benefit is clamping the value so it doesn't go into the millions. As I demonstrated already this is easily done with a modified wall shader. Note you don't do this on all walls - just the surfaces that are very close to a light source.

I also made the light shader self-clamping like the wall paint shader. The effect is similar, but due to angle-of-incidence, it has some problems, creating a region approaching the hotspot that is unnaturally clamped. The visual results are better if the wall absorbs the light instead of making the light no generate the light.

Also, the wall shader technique doesn't need adjustment if you move the light. So I'd say do the wall shader and skip the complications of a light shader.

Quote - Here is the light material that script generates.

Thanks for that!

Just downloaded it and trying it out now. I take it that PM:Max Value is the clamping value you mentioned, yes? I'm only asking because your previous ISF shader had a node called "Unit_Distance_V1" that controlled the distance at which the strength of the light was at the value as entered in Intensity.

Yes max value is the clamping value. In this version I did not provide a unit distance - I just assumed one PNU so that the overall illumination would match what a Poser ISF light does. You can set the intensity outside the shader as normal.

Good points.  I actually considered them myself sometime after posting that screenshot, but by then it was too late to delete the post.

Quote - Inverse Square Fall-off is not completely "broken" as you put it, but rather it fails quite badly in certain particular circumstances (as I remember pointing out to you when you asked the same question elsewhere last week).

If it can't work correctly without some kind of a "patch job" by the user, then it's broken, and no one using Poser for the first time will be any the wiser about how to fix it without reading a thread like this one.  Nowhere in the Reference Manual does it say anything like, "By the way, Inverse Square doesn't work well with IDL enabled in some situations, so here are a few work-arounds you can try."

BTW - Sorry about cross-posting, I was just hoping one of our Poser gurus here at Rendo might have some answers.

Quote - 3) Put something black between the point lights and the ceiling (just in the places that are really close.) If you happen to "see" these objects because they are in the cameras view, then make them invisible to camera. IDL will "see" them, but the camera won't.

I was hoping not to have to try this, as it creates unwanted addtional work to position the square, especially if you have several point lights in a scenel using Inverse Square.  Nevertheless, it did work for the single light in my test scene (see screenshot), so thanks for the suggestion.

Unfortunately, I don't imagine it would work well with things like desk lamps, so your light shader would suffice in those circumstances.

As BB mentioned, the difference between a full inverse sqaure attenuation (1/d^2) and the correct formula (for a real bulb with radius R) I derived in an earlier post {1- 1/sqrt(1+R^2/d^2)} is noticeable in the first say 3 lightbulb-radii only. Here is the graph, one with a linear scale, one with a log scale.

As you can see one should avoid placing the bulb near an object with 1 radius, which - in real life - is impossible anyway as the bulb sets a lightsource at such a distance in the first place.