80,000 average photons per light?

[RevitGary]

I have read that 80,000 average photons per light is a good setting.

 

When would one change that number and why?

 

If you have one sunlight and 2 sky portals in an interior is 80,000 photons per light a good setting? If you have one sun and 2 sky portals and a dozen spots in the interior is 80,00 still a good setting?

 

Is there a rule on how much this setting should be based on how many lights you have in the scene?

 

I also typical use 500 photons per sample. Is this a good idea?

I have experimented and it doesnt seem to affect much with the image quality or rendering time so 500 seems to work.

 

One more .... I let max chose the maximum sampling radius. Is there ever a reason not to let max choose this size and to overide it?

 

Thanks!!!

[Justin Hunt]

The actual number of photons shot is dependant on the lights intensity, ie the brighter the light the more photons shot.

 

If you open the message window it will tell you exactly how many photons each light is shooting. From there you can either increase and decrease the number of photons each light shoots.

 

jhv

[RevitGary]

OK.... But this is a GI setting not a light setting?

[Crazy Homeless Guy]

I think photons per samples are meaningless if using FG for the smoothing. Which is the preferred way to do it. They were used when you tried to smooth the lighting solution with only GI, and not FG.

 

Also, I am an advocate to using the manual over ride settings to control the energy and photons per light, rather than a global catch all. I think it creates faster rendering times, and gives you a lot more control over how the light interacts with the scene.

[RevitGary]

ok.... so how many photons per light is a good rule of thumb for an interior?

would you use a different number for the sun....say 80,000 and another one for interior spots?

[mattclinch]

completely depends how many lights are in your scene...

[Crazy Homeless Guy]

You could try turning off FG, and rendering the image to make sure the photons are overlapping.

 

If using the average number of photons per light, then pay attention to the photon sampling radius, the number of photons and the intensity. If using the manual way, pay attention to the photon radius, the light intensity, the energy, and the number of photons generated.

 

The first image attached has 3 lights, with each generating 10,000 photons and a 1" sample radius. You can see how the photons do not overlap. This is bad. If I turn on FG, the scene will be blotchy.

 

The second image has 60,000 photons and a 1" sample radius. You can still see some areas that photons are not reaching. I could simply increase the photons sampling radius to maybe 3", and it might cover everything. Or I could increase the number of photons, while still using the 1" sampling radius.

 

I chose to increase the number of photons, which is what is shown in image 3. In this image I am using 120,000 photons per light with a 1" radius. The tighter you keep your radius, the more detail you should be able to get out of the solution. However, in large spaces it is common to use a 12" or higher sampling radius. It is all about a speed vs quality trade off.

 

The 3rd image has sufficient coverage in my opinion. So now I am going to turn back on FG, and if my settings are right on that, I should get a nice smooth image.

 

This scene is a common scene that is posted around the internet. I got it from a Jeff Patton thread, but I think he got it from someone else. The lights that I currently have set do not produce the best shadows and such, so ignore that, and only look at the smoothness of the lighting solution.

 

Not sure if that helps, but maybe it gives you an idea of how photons are working according to their settings. As you can see, the number of photons is highly dependent on the scene, and number of lights in that scene.

 

Also, when working in the manual mode, the color of your light is separate from the color of the photons. So you will need to adjust the color of the photons manually also. Maybe the manual method is a bit much to jump into, but it will give you ultimate control, and the best trade off between quality and speed at the end of the day.

[thorsten hartmann]

hi

 

for me is the number of photon important. I use min. 1.000.000 (low quality) to max. 10.000.000 (high Quality) Photon for all Lights in Scene, that is enough.

 

example for max. 10.000.000 Photon:

a ) when i have 10 Light = 1.000.000 Photon per Light

b ) when i have 100 Light = 100.000 Photon per Light

c ) when i have 1000 Light = 10.000 Photon per Light

d ) ....

Edited November 29, 2009 by thorsten hartmann

[RevitGary]

Well..... it is becoming clearer now.

 

Crazy, it seems like in your solution, it might be overkill when using it with final gather. I am very concerned about rendering times. I guess i will need to find a balance. My scenes dont have anywhere near the photon coverage you are showing. I need to do some testing.

 

 

Hartman, good answer this is the kind of thing I was looking for.

How does the sun and sky portals fit into this equation? Would the sun emit the same number of photons as an interior spot? I would think not. Same question with sky portals where do they fit into this equation?

 

thanks!!!

[thorsten hartmann]

hi

 

this is a good question, but difficult to answer. The priority of the number of photons depends on the light intensity and area.

 

Indoor....

Day: max. Photonnumber: 50%Sun + 50%(IES & Portal Light).

 

Evening: max. Photonnumber: 25%Sun + 75%(IES & Portal Light).

 

Night: max. Photonnumber: 0%Sun + 100%(IES & Portal Light).

 

Outdoor....

Day: max. Photonnumber: 75%Sun + 25%(IES & Portal Light).

 

Evening: max. Photonnumber: 50%Sun + 50%(IES & Portal Light).

 

Night: max. Photonnumber: 0%Sun + 100%(IES & Portal Light).

Edited November 29, 2009 by thorsten hartmann

[Crazy Homeless Guy]

hi

 

this is a good question, but difficult to answer. The priority of the number of photons depends on the light intensity.

 

Indoor....

Day: max. Photonnumber: 50%Sun + 50%(IES & Portal Light).

 

Evening: max. Photonnumber: 25%Sun + 75%(IES & Portal Light).

 

Night: max. Photonnumber: 0%Sun + 100%(IES & Portal Light).

 

Outdoor....

Day: max. Photonnumber: 75%Sun + 25%(IES & Portal Light).

 

Evening: max. Photonnumber: 50%Sun + 50%(IES & Portal Light).

 

Night: max. Photonnumber: 0%Sun + 100%(IES & Portal Light).

 

Are these actual percentages, as supplied by Mental Images?

[thorsten hartmann]

hi travis,

 

no this values comes from my work with mental ray.

[RevitGary]

Ok but what if you have a daylight indoor scene with a Sun, 3 sky portals,

and 20 spot lights?

 

thanks again!!!!

[Crazy Homeless Guy]

Crazy, it seems like in your solution, it might be overkill when using it with final gather. I am very concerned about rendering times. I guess i will need to find a balance. My scenes dont have anywhere near the photon coverage you are showing. I need to do some testing.

 

My solution doesn't really have anything to do with FG. When using Photons, FG is used to smooth the results. It does other things, but that is its main resoponsibility when using photons.

 

I simply turn off FG so I can actually see the photons before they are smoothed. It lets me know whether I am using to high of a number of photons, or not enough photons.

 

The results below are both rendered with FG and GI, in automatic mode, not manual.

 

One is using 20,000 photons per light, with a 3" sampling radius, and the second is using 400,000 photons per light with a .25" sampling radius. The one with the lower photons is not as nice, but they are both fairly close.

 

Anyway, It might be worth turning FG off, and hitting render just so you can see what is happening when you adjust the numbers. I think it makes understanding what the photons are doing easier.

[RevitGary]

I have experimented with turning FG off and using a grey material overide with all windows on a frozen layer. I am doing some more now.

 

What kind of render times did you have on the 2 examples you show?

[Crazy Homeless Guy]

What kind of render times did you have on the 2 examples you show?

 

I am away from home on a not very powerful laptop. 32bit / 2gb ram / Core2Duo at 2ghz. I was also rendering with a AA of Box 1/4 - 4, and a resolution of 640 x 480.

 

With those settings, my 20,000 photon test timed like this:

RC 0.3 info : wallclock 0:00:59.64 for rendering

MSG 0.0 info : Total non-render time: 6.1875 s.

MSG 0.0 info : Total geometry translation time: 2.3125 s.

MSG 0.0 info : Total object data translation time: 0.0313 s.

 

Switching to Box 1 - 16, which I typically use for production:

RC 0.3 info : wallclock 0:01:58.20 for rendering

MSG 0.0 info : Total non-render time: 5.5625 s.

MSG 0.0 info : Total geometry translation time: 2.4063 s.

MSG 0.0 info : Total object data translation time: 0.0000 s.

 

When I switched to the 1-16 my render time doubled. This should be expected. Typically I experience the biggest hits on render time to be material glossy samples, and AA/Sampling. If I had actual materials set up int his scene, the render times could be quite a bit different.

 

Also, I have area shadows in this scene, which also slow things down.

 

If I were working on my production machine, I would expect this image to render a min of 4 or 5 times faster.

 

The 400,000 photon test times:

RC 0.3 info : wallclock 0:02:12.69 for rendering

MSG 0.0 info : Total non-render time: 5.2656 s.

MSG 0.0 info : Total geometry translation time: 2.2813 s.

MSG 0.0 info : Total object data translation time: 0.0156 s.

 

The time for calculating lighting is oddly appears to be less with the higher settings, which I didn't expect. It did go down every time I rendered, so the scene may be caching things.

 

But again, this is a really simple scene with no materials. It is very common for me to up to a couple of hundred lights in 1 scene, with materials. It is at these times when using to many photons can speed things down quite a bit.

Edited November 30, 2009 by Crazy Homeless Guy