GI question (radiuses) ?

[fabw]

Hi to all,

 

I am new on your forum and I am not so long user of mental and max.

 

I make a lot of tests now to understand how photons behaves in different sittuations. I render GI + FG.

I noticed that when I decrease photons radius form 20 cm to 5 cm my scene (interior) become brighter - thats fine. I understand it is because photons dynamics is getting better.

 

But what is bad..

 

I lost shadows,information in shadows, contrast in shadows, chairs become floating ...

 

Why is that ? Do I need to compensate this brighter image with exposer control?

Do I need more FG rays? What to do?

[maciejwypych]

If you decrease the photon radius and leave the same number of photons they may not cover the area hence the insufficient information in shadows.

 

You can add ambient occlusion so you get more details or increase number of photons.

 

It is actually more complicated and complex than that. Read through some GI theory and tutorials.

 

Without seeing your scene and settings it's hard to tell exactly what's the best solution to your problem.

[Crazy Homeless Guy]

Turn off FG when you are setting up your GI. It will allow you to see the individual photons in the scene. You will be able to see the amount of GI coverage, and have it is effecting detail.

[fabw]

I think amount of photons is not a problem. I have full coverage of photon in the scene ( i use 2.000.000, 5 cm photons on 20 m2).

 

As I said when I decrease radius form 20 cm to 5 cm overall scene is getting much brighter. It cause that I lost contrast and everything is floating over a floor.

 

So I need to compensate brightness by exposure control? Or make sth else ? More Fg? AO? Is AO a way to go?

 

Is it a good idea to add AO in materials with very big distance like 3m.

[Crazy Homeless Guy]

If I understand correctly, ....when you have the photon radius set to 20cm, you have full coverage. But, when you decrease the photon radius to 5cm, you do not have full coverage. And you will not have full coverage at 5cm unless you add more photons.

 

Basically, you are making the photons a fraction of the size by decreasing their radius, so it is going to make more photons to fill the scene because they are smaller.

 

If you turn off FG, and look at you Photon solution at both 20cm and 5cm you should be able to clearly see the problem.

 

AO is more preference than anything else, and it deals with adding extra detail and shading. It is not a compensation for GI, it can only make the GI look better one applied.

[fabw]

No. You get me wrong. With radius 20cm i use 500.000 photons. With radius 5 I use 2.000.000 photons. Coverage is very good with both situations but changing (decreasing) radius have affect on brightness of the scene ( do you know this?) and then I lost shadows and contrast in shadows.

I mean useing small GI radiuses and adding FG for me loosing sense or maybe i am wrong. But my experiences say better to use Gi radiuses about 20-40 cm when you use FG but you dont have a good dynamic of the light than.

 

Just looking for answer...

[Justin Hunt]

Open the message window

Turn the photon radius off

Calculate the photon map

 

In the message window read off what mentalray decided was a good radius to use. That will give you a good idea as to what your radius should be.

 

Basically the photon radius should just overlap each other to get good coverage. Then increase the max photons per sample to get more detail.

 

FG will do two things, 1- smooth out the photon solution and two add detail and skylight. If your fg interpolation is too high it will "wash out" detail like contact shadows. Increasing the Rays per fg sample will add detail.

 

A simple answer to the exposure control, try it if it works for you use it

[Justin Hunt]

Here is what is really happening when you change the photon radius

 

As you can see @ 50mm there is alot more lighting information than @ 10mm.

 

The reason the 10mm image is brighter is that the contrast between sampled areas and non-sampled areas is big, so there smoothing is going from super bright to super dark resulting in a brighter image. Whereas the 50mm image has less contrast between sampled and non-sampled areas, the smoothing has more mid-tones to work with so is averaged more evenly.

 

The reason the 10mm image is loosing the contact shadows isn't because the image is brighter, its because there is less information to work with.

 

Conclusion, if the photon radius is too small the less accurate/ pleasing the image will be.

 

Side note, this is a 5mX5mX2.5m box, light by a mrsun and Sky, Suns photon tagrget radius is just slightly larger than the opening. Mentalray used a radius of 1576mm.

 

jhv

[fabw]

Thanks Justin. A lot of good informations.

 

So is it a right way to increase a numbers of photons in a scene when decreasing radiuses?

 

"Conclusion, if the photon radius is too small the less accurate/ pleasing the image will be."

 

So there is no sense to decrease Gi radiuses too much?

[Justin Hunt]

yes if you decrease the radius then you must increase the number of photons to compensate. In my test I had to go up to 2million photons to get an equivalent result.

 

It also took much longer to render.

 

You should be trying to find a balance of Photons, radius, lighting detail and render time. Smaller radii and many photons dont necessarily give a better result than a larger radii and fewer photons.

 

jhv

[fabw]

"Open the message window

Turn the photon radius off

Calculate the photon map

 

In the message window read off what mentalray decided was a good radius to use. That will give you a good idea as to what your radius should be.

 

Basically the photon radius should just overlap each other to get good coverage. Then increase the max photons per sample to get more detail."

 

I am not sure if its correct because max sets Gi radiuses size at 1/10 of the scene by default.

[Justin Hunt]

yes 1/10, which is fine if the bounds of the scene are pretty much what your view is, but not so good if its much bigger.

 

I like you first let mentalray decide, then use its value as a guide to set my own.

 

jhv

[Cesar R]

how many photons is TOO many photons? I am going to post a scene shortly but w/out FG I find that I need 90,000 photos per light for the scene to light nice an clean. Also, 1200 photons per sample.

[Cesar R]

How can I find out what size sampling rad. is 3dsmax using when the Maximun Sampling rad is unchecked?

I know that I can measure my scene and take that dimension *.10 but is there a way to know the EXCACT radius used in a recently completed render?

[Justin Hunt]

how many photons is TOO many photons? I am going to post a scene shortly but w/out FG I find that I need 90,000 photos per light for the scene to light nice an clean. Also, 1200 photons per sample.

 

it depends on the scene and the number of lights being used, the more lights, then the lower the number of photons. I ususally start @ 20000 for testing and then go up from there. There is no set formula but there is a point where adding more photons gives no more detail.

 

How can I find out what size sampling rad. is 3dsmax using when the Maximun Sampling rad is unchecked?

I know that I can measure my scene and take that dimension *.10 but is there a way to know the EXCACT radius used in a recently completed render?

 

Open the render message window , Render-> message window.

 

Start rendering and once the FG calculations start, stop the render. In the message window there will be a note about the photon radius

[Cesar R]

I cant find any other way to pull up the messenger window than to wait for an error to occur. I am on Max Design 2011.

In regardst to the photons, what is highest you have gone on a scene?

 

I need to make a thread on my WIP / learning file.

[Justin Hunt]

go to the menu at the top, Rendering -> Render Message Window

 

The higherst, 8 million, but in the end I used 200 000. I wanted to see how high I could go and other than taking a really long time to calculate and even longer to render I found I got similar results using 200 000 in much less time.

 

If you like/ are able to send me the file and I can take a look. Other wise post your settings and a render

 

jhv

[Cesar R]

haha thanks, no idea how I missed that.

I am not doing any real work at the moment but rather just trying to understand MR a little better. I am going to post some settings with some detailed settings.