Lighting difference

[RALPH_PINEL]

Has anyone one had this issue

I have a room lit by a skylight. I have vray lights in the windows set to portal.
I can render the room with irradiance map and light cache image attached.
If I switch to PRI. brute force, SEC. brute force the image comes out much darker.

Can anyone help me understand what might cause this?

Also can anyone suggest which method would be the fastest

For 500 frames the camera tracks across the room. The curtains in the nearest window are blown in the breeze.
I was planing to use the IRmap and light cache, hide the curtains, bake the light, then render with brute force on the curtain material.

Settings for the IRmap multi Incremental min.-3, max.0. subD.200, sample.30 Lightcach flythrough subD 3000. this was taking 2 hours on a 24 core 3 ghz machine for the light pass. Ouch!

This prompted me the look at brute force. DMC setting - global subdivisions multiplier 0.001 noise threshold to 0.005 adaptive dmc - min 1 max 30. This is when I ran into the light differance issue. The render looked about 6 stops darker.

[joconnell]

It's your secondary bounce causing the major difference. Brute force is as it's name suggests, doing a bounce of light for every single pixel in the image, so it's very very accurate but also far more accurate than Irmap and light cache. When you set brute force to be your secondary global illumination, you get an option of how many times you want vray to bounce the light before stopping. As a general rule, light will get weaker and weaker each time it bounces off a surface unless it's pure white or a perfect mirror (not likely) so each bounce of the light tends to contribute less and less brightness each time. The default of 3 light bounces in the brute force engine is a decent amount but as you say it's a bit darker looking than if you'd have used light cache for the secondary bounce. Light cache has no options for how many bounces it will do, it's hard coded to be 100 bounces. It's not calculating a bounce for every pixel in your final image though, it's doing a smaller amount of samples spread over the image and then averaging them together in the end to give you a bright and smooth result. With light cache, it will bounce light around until it reaches bounce 100, hits another light cache sample where it will stop, or just stops when it's contributing only a tiny difference to the overall result. The short answer is it's being less accurate and doing less work than brute force is, so it's able to handle 100 bounces with ease where using the per pixel, brute force method would take much longer.

You're totally right on the method you're thinking too - Baked irmap and LC with the curtains hidden, then unhide the curtains and turn off "use irradiance map" on the material applied to them.

[RALPH_PINEL]

Thanx for the reply, When I tried brute force i had set the bounces to 100. but it still came out darker. As you can imagine when I saw how much darker it was I stopped the render.
I was depressed by how long IRmap & light cache was taking. I was hoping to find some magic solution to get it rendered. It's a scene for my friends short film. No money for me or a render farm.

[joconnell]

Hmm, normally light cache is pretty damn quick - normally between a few seconds and a minute depending on the scene for me. Might be something else in the scene causing a problem? Things like co-planar faces and the such can often make light cache sad.

[svetlozar.draganov]

If I switch to PRI. brute force, SEC. brute force the image comes out much darker.
Also can anyone suggest which method would be the fastest

Settings for the IRmap multi Incremental min.-3, max.0. subD.200, sample.30 Lightcach flythrough subD 3000. this was taking 2 hours on a 24 core 3 ghz machine for the light pass. Ouch!

This prompted me the look at brute force. DMC setting - global subdivisions multiplier 0.001 noise threshold to 0.005 adaptive dmc - min 1 max 30.

1.IM+LC is usually much faster approach than BF+BF - Does this scene involves animated objects/lights ?
2.LC as a secondary engine generates much more bounces that BF and it is perfectly normal to have slightly darker result with Brute Force.
3. Would it be possible to give us more information about what approach you are using for IM+LC rendering ? Two hours sounds too much for IM+LC rendering.
4.If it is possible - please send this scene for investigation. We could provide out FTP if needed.
5.Could you please also attach a few examples if BF+BF rendering ?

[RALPH_PINEL]

Hey Svetlozar cheers for the help.
The scene is 500 frames.
The camera tracks across the room.
There is a cigaret still smoldering in the ash tray. This will be rendered separately and composited in nuke.
The window is broken and the curtain is blowing in the wind.
My approach was IM + LC, I will hide the moving curtain and BF that in final pass.
The setting I'm using are in the first comment.
The LC takes an expected amount of time it's the IM that is taking the time.
I'm going to merge all the elements into a new file in case something has got screwed up.
I will attach a section of the BF + BF
I will collect the scene and send to you via Hightail. Who should i send the file to?
I will do this tomorrow.
Thanks Ralph

[Neilg]

Switch off reflections & glossy effects while rendering the IR map. There's nothing in that scene that should take very long though.

[svetlozar.draganov]

I will collect the scene and send to you via Hightail. Who should i send the file to?

Please send it to support@chaosgroup.com and include a link to this thread.
If it is possible please include a few already calculated frames with IM+LC and BF+BF approach for comparison.

[svetlozar.draganov]

The reason which cause darker results in BF+BF approach is activated Store With Irradiance Map option.
This option has been turned on for all light in the scene and since Irradiance Map engine is not used the illumination is incorrect.
If you disable it it will get a very similar illumination even with BF+BF approach.

[RALPH_PINEL]

cheers Svetlozar, Any tips to speed up the IM + LC this is the method I was planing to use.
Cheers again.

[svetlozar.draganov]

In that scene the best approach according to me would be to use IM + LC approach since all the objects in the scene are static - only the camera is moving.
Here you could find the complete tutorial about how to set up fly-trough animations:
http://help.chaosgroup.com/vray/help...ials_imap2.htm

If you have any moving objects (except animated lights) you could hide them in the pre-calc stage , turn OFF Use Irradiance Map checkbox from Materials Options and unhide them for the final-rendering stage.

Since the camera is moving very slowly I think that you could render only single IM for each 100 frames of the animation.
Furthermore the IM / LC and AA settings are far to high - my suggestion is to switch all those settings to their default values do a test render 5-10 frames and increase them only if needed.
The slowdown in IM-calculation is due to Detail Enhancement option which means that Brutte Force will be used instead of IM, and 200 Subdivisions which is quite high value.

[RALPH_PINEL]

Cheers Svetlozar