PDA

View Full Version : Light Falloff…



toma
12-19-2004, 09:48 PM
…While inverse distance^2 may be the "real way" light is falling off, this setting has two issues :

- shading is computed for objects that are far away even if the light's contribution is nearly null…

- surfaces that are near the light are over exposed…

…so I would like NT to add to inverse and inverse^2 a maximum distance range (this will speed up rendering) AND a maximum intensity setting to avoid over-exposition (this would also avoid random white pixels when rendering with radiosity, when some rays hit one of those "hyper exposed" area)…

cheers,
thomas

Karmacop
12-19-2004, 10:00 PM
A "maximum distance" would be good, especially in scenes with lots of lights, but I'd want it to be set by the user.

Surfaces near the light are over exposed because you haven't changed them since using linear lights. You can easily give a light flloff and not over expose close surfaces, you just have to change the light intensity.

toma
12-19-2004, 10:40 PM
no no no ! …changing the light intensity doesn't change the over exposure, because when using inverse distance the intensity is set for a nominal distance, so intensity decrease after that distance and increase before that distance in a exponential way, that means that the intensity at the source of the light is infinite no matter what intensity is set at the nominal distance… (well this is how I understand it ;) )

thomas

Panikos
12-20-2004, 03:43 AM
Linear Falloff lights emmit light to infinity even in subtle values.
To make rendering faster, you can exclude some objects from such lights as long as they dont influence them in a visible manner.

toma
12-20-2004, 04:04 AM
Linear Falloff lights emmit light to infinity even in subtle values.
To make rendering faster, you can exclude some objects from such lights as long as they dont influence them in a visible manner.

are you sure ?I think that the objects farther than the light's falloff are excluded for that light (or receive 0% intensity), on the opposite, inverse distance and inverse^2 lights are sending some intensity, even unnoticable, to every objects because they don't have a limited range…

cheers,
thomas…

toma
12-20-2004, 04:56 AM
this is a little test to illustrate my problem with the inverse^2 falloff when it comes to overexposition :

this is a linear falloff of 12 m with a point light (100%) :
http://vbulletin.newtek.com/attachment.php?attachmentid=17032&stc=1

the same with inverse^2 (nominal distance 12 m, light intensity 9%) :
http://vbulletin.newtek.com/attachment.php?attachmentid=17033&stc=1

the lighting is better in the far distance as the light intensity isn't clipped to 0 but decrease slower as it gets farther. between the nominal distance and the light, the intensity is growing faster as it gets closer to the light, and this is natural, except that when the shaded surface are really close to the light, they are way way (I mean way) over exposed, in this cases my little 9% point light manage to burn a pixel all the way up to 48 519 % (and I guess that this is not very natural :rolleyes: )
http://vbulletin.newtek.com/attachment.php?attachmentid=17035&stc=1

the problem gets even worse with radiosity, as some radiosity sample will hit such an area, and in this case, antialiasing can't do anything :
http://vbulletin.newtek.com/attachment.php?attachmentid=17034&stc=1

well, I hope NT makes it possible to clip the upper value when using inverse^2 distance falloff because I really prefer this type of falloff but it's some times a bit tricky to use it.

I just remember that G2 have such a feature, but I haven't used it since march 2004
(Fprime ;) )…

thomas.

Panikos
12-20-2004, 05:46 AM
are you sure ?I think that the objects farther than the light's falloff are excluded for that light (or receive 0% intensity), on the opposite, inverse distance and inverse^2 lights are sending some intensity, even unnoticable, to every objects because they don't have a limited range…

cheers,
thomas…

Heh

Exclusion implies Shadows too.
If you have a light with Inverse Falloff, and you have 1000 objects out of its "range"
Excluding this light from these objects will cause faster rendering.
If you have this Light with Shadows affecting everything, you will have the CPU spending time calculating 0.00001% shadows intensity for all your objects.

Exclusion involves :
- Shading
- Shadows

(Hint : There is a command that enables Shadow Exclude that is on by default.
I suggest you place it on the LW-GUI)

As far as intensity, you can use very low values considering the distances within your scene. Using FPrime in such application is an ideal case.

toma
12-20-2004, 06:02 AM
you are right about the exclusion behaviour, how ever on the intensity side, Fprime doesn't help… (well I guess I will have to use inverse, not inverse^2 :( )

cheers,
thomas

Pavlov
01-02-2005, 05:02 AM
Best thing would be, imho, a new custom falloff tool.
It should be a little graph editor (to control light's behavior in space, not time) with classic presets (inverse, inverse square, linear), but you should be able to edit that falloff directly manipulating and inserting points. It woul be the most powerful tool to control falloffs, since you could set a START and and END for the falloff, PRE-POST behaviors, and control decay with multiple points (this allows, i.e, to make intensity decrease, then increase again, then decrease again, etc).
Imho this kind of falloff tool should be put in every tool which already has decay options, also in modeling tools.

Paolo Zambrini

Lewis
01-02-2005, 05:54 AM
Hi Guys !

I remeber of one Protons little light tutorials and thing is that when you use inverse dinstance then LW is behaving differently. That dotted fallow line then isn't faloff distance anymore. Then it's expose area so you need to turn down fallow (mke it smaller) so that dotted line isn't touching any objects or you will get overexposed objects :). Crazy but that's how it works in inverse/inverse2 modes :).

cheers

toma
01-03-2005, 06:30 AM
Best thing would be, imho, a new custom falloff tool.
It should be a little graph editor (to control light's behavior in space, not time) with classic presets (inverse, inverse square, linear), but you should be able to edit that falloff directly manipulating and inserting points. It woul be the most powerful tool to control falloffs, since you could set a START and and END for the falloff, PRE-POST behaviors, and control decay with multiple points (this allows, i.e, to make intensity decrease, then increase again, then decrease again, etc).
Imho this kind of falloff tool should be put in every tool which already has decay options, also in modeling tools.

Paolo Zambrini

yes ! exactly what I mean :D

btw, happy new year !

cheers,
thomas

Kvaalen
01-03-2005, 08:41 AM
Best thing would be, imho, a new custom falloff tool.
It should be a little graph editor (to control light's behavior in space, not time) with classic presets (inverse, inverse square, linear), but you should be able to edit that falloff directly manipulating and inserting points. It woul be the most powerful tool to control falloffs, since you could set a START and and END for the falloff, PRE-POST behaviors, and control decay with multiple points (this allows, i.e, to make intensity decrease, then increase again, then decrease again, etc).
Imho this kind of falloff tool should be put in every tool which already has decay options, also in modeling tools.Funny... I had the same idea... :)