PDA

View Full Version : Question about lighting with luminous polygons



unstable
08-26-2016, 10:46 AM
I saw this thread under General Support http://forums.newtek.com/showthread.php?151225-How-to-make-a-studio-backdrop-landscape-scene and found Matt Gorner's tutorial interesting. I downloaded his scene and lowered the passes from 96 to 6 and the walls of the room still looked good. I swapped out his rectangular lights with ones I created and they rendered out the same. However, when I created small point type lights, no matter what I do, I can't get the corner of the room to appear smooth. I've tried upping the luminosity, setting the passes back to 96, and even added more lights, but nothing seems to work. Is there no way to accomplish smooth walls using small luminous polygons?

I attached some renders showing the original, then the original using my rectangular lights, then the original with the original rectangular lights and my smaller lights, then finally with just the smaller lights. Notice the difference in smoothness in the corner of the room.

RebelHill
08-26-2016, 11:35 AM
Increase the number of GI rays... a lot.

Danner
08-26-2016, 12:09 PM
With radiosity the more variation in contrast you have (by that I mean small and overly bright luminous polygons in an otherwise dark room) the harder it is to smooth things out. you could cheat by dropping the intensity of the small lights and adding a large polygon, unseen by camera, unseen by rays (so it doesn' show up in reflections) and non shadow casting if you have any lights. You'll get a similar result but much faster, cleaner and with more control.

Surrealist.
08-26-2016, 01:20 PM
Those splotches are coming from your interpolated settings. Minimum Pixel Spacing and Maximum Pixel Spacing

http://www.except.nl/lightwave/RadiosityGuide95/RadiosityDiagram3.jpg

This guide is a good read:

http://www.except.nl/lightwave/RadiosityGuide96/index.htm

What is likely happening is that the smaller lights are not entering enough rays into the scene to be evaluated smoothly so you get the blotches.

What RH said is correct because that increases the chance for rays to hit and be evaluated. The other solution mentioned is to add more light. Both solutions result in more rays to be evaluated.

You can also adjust the MPS settings.

But in my opinion the best solution is to uncheck interpolated and go brute force.

If you do that you have more than one place to adjust for grainy images. You can adjust the number of Rays in GI, the number of shading and lighting samples as well as Adaptive Sampling in the camera tab. Between those three you can tweak it to get a decent result without as much of a render hit. Generally you will get a faster render if you go light on samples (Shading, Lighting and Rays Per Evaluation) and heavy on your Adaptive Sampling in the Camera tab. Lover numbers like .01 give you more AS. But you can play with these and find what works best for your situation.

unstable
08-27-2016, 06:36 AM
Thanks for all your suggestions. What RH and Surrealist say makes sense except where image 3 comes in. That one is where I add my little lights with the big lights which rendered fine until adding the little lights. I think pic 3 rules out Danners suggestion as well. But I will try what you all suggest to see what happens. Thanks much

unstable
08-27-2016, 07:56 AM
I don't know why I would ever question RH or Surrealist, I should know better. :bowdown: You were correct and I was amazed!! I had to read through that article several times and will probably have to do so several more times before I really understand it, but I increased the MPS to 30 min and 60 max. I was blown away not only by the fact that the wall went smooth, but by the fact the render time went from 1:40 minutes down to 14 seconds. :boogiedow Wow!! I've attached the results. The first is at MPS 30 - 60 and the second is 60 -90, but I reduced the floor blur as well. I don't think the increase of the MPS did anything for the second one. I think this is very good information to have though. Thanks again.

RebelHill
08-27-2016, 08:18 AM
Increasing the MPS in your specific case works because there are no small scale details in the scene. Add a bunch of small geometric details (like greebles/whatevs) whose "occupation of the frame" is smaller measured in pixels, and you'll see very quickly that its no solution in all situations. Coming up against this, you'll find that the main responsibility falls to the ray numbers.

Surrealist.
08-27-2016, 08:25 AM
EDIT: Posted at the same hime this ^ and this :

Just remember that that setting is fine for smooth walls. But in practical applications where you have a lot of detail with geometry textures, bump maps and so on, you will loose detail. It will also not animate without artifacts unless you make a cache of the radiosity. So basically what you have done is made low quality settings. I suggested it more or less so you'd understand what is happening. It is faking the result by smoothing over the area. With more detail in the scene this does not work. you have to shorten the range and add more samples. Or better (assuming you have a decent computer and LW 2015) you can go brute force method and keep detail. The settings are much more easy to understand and you can tweak a good render time. It is pretty much like using Arnold or new Renderman RIS or Cycles. Not exactly but the same basic idea. You are controlling quality with samples. And since you are doing that and not faking it across regions you will have less artifacts when rendering animations.

unstable
08-27-2016, 03:39 PM
Well of course there would have to be more to it. :rolleyes: I do have a decent computer (i7-4960) and I am using LW 2015. So, my question to you now is to make sure I understand which 'rays' Rh refers to and Surrealist's comment about sampling. I've attached a screen shot of the areas I believe we're discussing. When Surrealist says increase samples, I'm assuming you're referring to the Maximum Samples (Red circle). But when RH says 'rays' I think he means the rays per evaluation (blue square) and not the Ray Recursion Limit (yellow square). Am I correct in regards to my understanding?

Also, just wanted to ask Surrealist if he could be more specific about what he means when he says 'brute force'. I've seen you state this in several threads and while I'm sure it makes sense to many, I have no idea what you mean. Sorry.

RebelHill
08-27-2016, 05:12 PM
well... GI rays... so rays on the GI panel, which equals samples (2 terms for the same thing).

Prince Charming
08-27-2016, 05:19 PM
Well of course there would have to be more to it. :rolleyes: I do have a decent computer (i7-4960) and I am using LW 2015. So, my question to you now is to make sure I understand which 'rays' Rh refers to and Surrealist's comment about sampling. I've attached a screen shot of the areas I believe we're discussing. When Surrealist says increase samples, I'm assuming you're referring to the Maximum Samples (Red circle). But when RH says 'rays' I think he means the rays per evaluation (blue square) and not the Ray Recursion Limit (yellow square). Am I correct in regards to my understanding?

Also, just wanted to ask Surrealist if he could be more specific about what he means when he says 'brute force'. I've seen you state this in several threads and while I'm sure it makes sense to many, I have no idea what you mean. Sorry.

Brute force would be turning gi to monte carlo and turning off "interpolated"... and using as many samples as necessary to get noise free solution. Personally, in LW, I think your best bet is using regular lights with geometry reflectors. That will give you the best result fastest IMO.

Surrealist.
08-27-2016, 06:12 PM
Well of course there would have to be more to it. :rolleyes: I do have a decent computer (i7-4960) and I am using LW 2015. So, my question to you now is to make sure I understand which 'rays' Rh refers to and Surrealist's comment about sampling. I've attached a screen shot of the areas I believe we're discussing. When Surrealist says increase samples, I'm assuming you're referring to the Maximum Samples (Red circle). But when RH says 'rays' I think he means the rays per evaluation (blue square) and not the Ray Recursion Limit (yellow square). Am I correct in regards to my understanding?

Also, just wanted to ask Surrealist if he could be more specific about what he means when he says 'brute force'. I've seen you state this in several threads and while I'm sure it makes sense to many, I have no idea what you mean. Sorry.

Sorry, assuming you knew those areas.

First leave everything at default.

Rays Per Evaluation.

This, in the Radiosity Pannel, It is the setting to how many rays are cast into the scene to be evaluated. You can experiment with this setting but it does not have to be real high, even using Interpolated off. (which turns LW GI into what is referred to as a Brute Force renderer. This means it is completely ray dependent and not using zones to smooth from one to the other. Try from 16-64. If you have more details, maybe higher. But use this judiciously. Used to be the only real way to reduce noise. But I have noticed the sampling ioon 2015 to work completely differently then earlier versions and you can use Adaptive Sampling to help.

Render Panel:

Lighting Samples
Shading Samples

Another way to reduce noise. Lighting Samples reduce a lot of noise (even with GI) and won't take as long to render.

Camera Panel

Adaptive Sampling. Lower this number to get more samples into the scene. Play with it, you'll see what it does. Default is .01 which is more than enough for more situations. Raise the number to speed the render or for testing. For testing I often use .5.

Hope this helps.

Prince Charming
08-27-2016, 06:49 PM
Another way to reduce noise. Lighting Samples reduce a lot of noise (even with GI) and won't take as long to render.

This is what I was getting at.... Even when using "brute force" gi it is still better to use actual lights in LW. I would still use reflectors that are unseen by gi for reflections, but its definitely better to use lights. You can get cleaner renders in less time with lower settings.

The only time I would say that you are better off using only gi and no lights is if you are using an hdr with importance sampling turned on, but for any interior shots with small light sources that are very directional its better to use lights.

Surrealist.
08-28-2016, 10:19 AM
Agreed. I have some quick examples here I did, just showing the subtle difference between Area Lights, AO and GI.

Another one I use a lot is the DP Dome light.

Lots of tools to play with.

rustythe1
08-28-2016, 11:16 AM
shiny stuff is also a big problem here, turn on cache radiosity for the render as always, then just make all the surfaces flat grey, then render, then put all your surfaces back, turn the cache back to locked and the render will be much cleaner, obviously you have to repeat this process if you move anything.

Surrealist.
08-28-2016, 12:55 PM
I forgot to put the link:

http://forums.newtek.com/showthread.php?151149-Shadows-Problem