So Im working on a new scene and were doing some rendering tests in Superfly and I meshed a bit around with the settings and especially the Diffuse bounces settings. Which accordingly to tool tip description control the indirect diffuse light. I have always kept this on 2 usually. But I did a bit of testing and was wondering if anyone can give some examples of why you would use it? and under which circumstances that would justify the extra render time. Because based on the following tests Im not really sure what exactly the huge benefit is from using it.

The test scene is a WIP scene using Superfly materials and all images are rendered using the same settings besides those mentioned under the images.

Test 1:

(No clamp filters have been used in this image)

Its pretty obvious how much noise is in the left image, purely due to the 2 diffuse bounces. However whats more concerning I think is that the render time is doubled.

Test 2:

(No clamp filters have been used in this image)

To increase quality of the render its common to increase the pixel samples as well. However again this will greatly increase render times as well. But again the left image contains a lot more noise than the one on the right. However the greatest visual difference between the images from what I can see, is the green painted iron shelf on the left side of the image. Which seems a bit more flat in the right one compared to the left one. However whether that's due to the noise rather than the diffuse setting, is hard to tell I think.

Test 3:

(Uses clamp filters 1 in all of them)

This shot uses the light from the HDRI map shining through the roof which is the only light source. And it seems like there is a very minor difference in lighting, if you look very careful. However again the render time is much higher than the non diffuse one.

Have anyone done any testing on this and in which cases these diffuse bounces should be used? Because personally for me, double render time based on the visual differences in these test images is not really justified to why this shouldn't be kept at zero at all times.

Would really like to know what experiences others have done in regards to this.

We probably first need to understand what that setting does, there is a difference between diffuse mat and diffuse light and I've seen descriptions that suggest either. The other thing, is the bounce count absolute or relative to the overall count, ie do the spec/trans bounces remain the same or are they boosted to keep the number of samples constant. Not surprised the fireflies increase as diffuse mats tend to flatten the light response. Did you look at the log to see what impact the changes had on the ray count?

The way I understand it, it controls the indirect lighting, so any light that bounces off other objects. So direct light hitting a wall for instant would then bounce to other objects or maybe a hidden light sources like light from underneath a door etc, so its basically what controls the global illumination as I see it.

The total amount of samples are the same (At least when looking at the render window, if thats what you mean), which I would expect as its controlled by the pixel samples as far as I know. As to what goes on behind the scene, I assume that there is an increase since each ray would have to calculate these additional diffuse bounces.

I think whats confuses me, is as you say, the amount of firefly this single setting adds, however Im not sure why I wouldn't be surprised to why it adds so many? Because logically the diffuse bounce light would have to be dimmer than the direct light source, but still it adds lots of very bright fireflies, so personally I would actually assume the opposite to be true as more effort is spend on calculating it. Obviously it makes sense that it would add more fireflies than if this setting is set to 0 which based on the description means its not being calculated, but in that case I would expect a huge difference in lighting quality.

So I did another test, where I added 3 light emitting spheres and a single non emitting sphere at the top to receive the diffuse bounces. Also I cranked up all the settings to 16 bounces for everything, except for the pixel samples which is still 15. But looking at the two images its very difficult to see a difference I think, I would assume with all the surfaces in the broken concrete, the effect of the indirect lighting would be easily seen. But honestly the only way I can see a difference is when I quickly click between the images. Which shows that the image using diffuse bounces is slightly lighter than the one without.

Test sphere

Zooming in on the sphere at the top, shows the effect of the indirect lighting a bit more clearly. In this case I have disabled the HDRI map so the only light is coming from the 3 spheres.

So the effect can be seen on the top of the sphere where light from the roof bounces unto it.

So it have an effect, but I thought of feel like something is not 100% correct here with this setting, it might be my tests that are not good enough to clearly show the effect of it. But taking into account that leaving this setting above 0 will double render times with very little to show for it. As it can be seen in the small thumbnails where the HDRI have been turned back on again, the effect of the indirect lighting is very easily removed, if just a slight amount of light interferes with it.

I don't know if it makes any difference, maybe it does, are you using the original Material Room PoserSurface for Superfly, or converting textures to either PhysicalSurface or CycleSurface nodes?

I use PhysicalSurface 99% of the time, which is the same as CycleSurface, its just Cycles nodes combined into an easy to use setup, which is very appealing to me, The only time I use the actual CycleSurface are for some specific materials where I think its slightly easier to use. PoserSurface I only use for Firefly.

The materials are created for PBR.

I don't know if its one of those Superfly things :D So I tried changing some things, so for instant setting minimum bounces to 250000 and maximum bounces 500000, Diffuse bounces and Glossy bounces to 150000 each. Based on the tooltip the higher the settings the longer the render time. Now with these insanely high settings, I would assume that the render would never be finished, since I have a fairly low end computer. But instead it have near to no effect on render times, if any. As if they are completely ignored. The diffuse bounces have an effect as shown above, but whether its 1 or 150000 is of no importance apparently.

I know what the min and max transparent bounces does, and why there is even a min setting for this im not sure, but the rest is a bit of a mystery to be honest and can't seem to find any explanation to what exactly they do, are there any logic to these settings or how they work, if at all?

Here is a quick test from me. Camera inside a shed with a window, no lights, purely lit by EZDome. The shed uses the standard PoserSurface in the material room. Rendered using Superfly. They are straight jpg exports from PPro11 at 100% quality. The only difference between the two pictures is the first one had diffuse bounces set to 0, the second set to 10 (highlighted in red in my render settings pic).

Using my gpu (GTX 980ti) the 0 diffuse bounce render took 159.24 seconds.

The 10 diffuse bounce render took 411.63 seconds.

To me this shows the difference between 0 and 10 is and for me worth the extra time.

I suppose it also matters how much light you have in a scene. If there's enough that the scene doesn't need too much light bouncing around (for example in your test 1 and test 2 in your first post) you can probably get away with fewer bounces. The images I posted show that increased diffuse bounces may be more preferable if there is only a single light that doesn't give a direct light to the scene.

I also just noticed that there is almost no reflection of the cup in the shed window with no diffuse bounces, yet there it is with 10. That may be another thing to factor into your bounces number.

Why so complicated? Here's a simple test. A box, a sphere, a spot light.

zero bounces:

one bounce:

two bounces:

sixteen bounces:

I see what you mean, in my examples from what I can figure out, its simply because there is to little light to really show the effect clearly and that the HDR lighting can cause the effect to not really show that well. But excellent examples you made.

I don't see much difference past 6 bounces in the stuff I'm doing.

Diffuse bounces are very misunderstood. So are the bounces in general. Before you can explain bounces, you have to understand what they actually are and how they are calculated.

The camera sends out a ray and then it hits something for the first time, that is not a bounce. That only starts the bounce counters, no matter what it hits.

Consider this example.

Lets go over some possible bounce scenarios with these settings.

Camera, diffuse (start counters on next hit), glossy (1st bounce, glossy), diffuse (2nd bounce, diffuse). Max of one diffuse reached, terminate ray. Ray cancelled in two bounces

Camera, glossy (start counters on next hit), transparent (14 times/bounces), Transparency can override max bounces, and is the only render setting that can, glossy (15th bounce), diffuse (terminate ray, max diffuse reached) Ray cancelled in 16 bounces, which had nothing to due with the 16 max trans bounces, or the Max Bounce setting because of the transparency over ride.

Camera, glossy, glossy, glossy, diffuse, glossy, glossy, transmission, transmission, transmission (terminate ray, reached 8 Max Bounces)

Camera, diffuse (start counters on next hit), 16 transparency, (terminate ray, max trans reached). Which is where black spots in trans hair come from if the Max Transparency setting is to low.

Render setting have more to do with lighting and materials in the scene than anything. As you can see from the examples, the numbers in the bounce settings is conditional to what it is going to hit in the scene.

If you have no specular and reflection in the scene, say a cartoon render, glossy will have virtually no effect on the render at all. Diffuse is going to make a huge difference, depending on the lighting.

But if that scenario is reversed, and glossy is set to low with mirrors in the scene, some things may not show correctly in the render.

The best settings are really the minimum you actually need to properly render the scene, based on the scene.

And that is different, for every scene.

On top of these settings, you also have the other render settings that can massively affect the information and direction of each bounce. Refractive and Reflective Caustics, Filter Glossy, and Clamp sample settings. 

If you are creative with the light path node, you can add far more conditions to the actual shaders themselves. An example of this is terminating a max transparency termination to transparent instead of the default of zero (black).

It would appear I responded to someone that deleted a post.....

The information is still good even if that poster was more interested in the links he posted than and answers