PDA

View Full Version : Lights,inverse square falloff etc.



silverlw
02-12-2005, 12:21 PM
I have tried hard to find some information about inverse square falloff and how to use it but nada.
All lights should have square falloff enabled and i have found some kelvinvalues to get the colortemperatures right for different kind of lights but now my question:
How strong should the value in lw be for a 60 watt lamp and how far should i set the falloff? Would be extremly gratefull if someone could enlighten me.

Integrity
02-12-2005, 04:24 PM
This can get complicated...I don't know everything about it but I guess I'll share what I know.

According to Wikipedia.org a 60 watt lightbulb would be 2800 K, which is why it appears yellowish-orangish...if your looking from outside into someone's house.

As for the falloff...as far as I know technically speaking in the real world the falloff would be infinite, it would never end. In Lightwave I think they added a end value for convenience and artistic control.

I hope this helps.

toby
02-12-2005, 05:00 PM
What I was taught is that the falloff setting, except for linear, is the point at which the light begins to fall off, not where it stops, because it is infinite. In other words, you would set the falloff value to be the same as the radius of your light bulb.

silverlw
02-13-2005, 05:36 AM
Thanx for your quick answers and theories. I have a slight different theory to offer after some (alot) testing. Please knock me down if you know this theory is wrong but here it is anyway:
When Inverse square falloff is enabled the "range/nominal distance" doesnt do anything at all. (Stupid of Newtek to have the setting enabled but there might be for some purpose i dont know about.)
You get exactly the same result by raising or lowering "light intensity" because the falloff is always constant = square falloff. The candlelight,spotlight, atom bomb or the sun, it doesnt matter, the falloff all behave the same way except for their lightintensitys. I still wonder how intense a 60 watt lightbulb is and how it relates to other lights. :)

Captain Obvious
02-13-2005, 07:13 AM
After some testing, the nominell distance does make a difference with inverse square falloff. It does seem to be as toby says.

I can't really figure out what to use, though. Inverse square with nom. dist set to the area the lightbulb or whatever seems to make some kind of sense.

silverlw
02-13-2005, 07:42 AM
So youre saying you cant reproduce the exact same result by just changing it's intensities?

Captain Obvious
02-13-2005, 08:53 AM
So youre saying you cant reproduce the exact same result by just changing it's intensities?

Reproduce the same result as what? What do you mean? :confused:

silverlw
02-13-2005, 09:58 AM
you have found that "the nominell distance does make a difference" but i say you can get exactly the same effect just by altering it's intensity. When you increase "distance value" the light goes brighter right? You get exactly the same look by increasing intensity instead or am i wrong?

Lynx3d
02-13-2005, 11:51 AM
Yea Nominal Distance has basically the same effect as intensity.
It just makes setup more intuitive, if your object is 10m away from the light source and you set nominal distance to 10m and intensity to 100% it gets illuminated as if the light source had no falloff and 100% intensity.
That's the whole trick of it.

And for the 60W...you can't translate that into LW because you'd need physical models for everything, camera exposure time, light instensity in e.g. lumen, gamma control etc. but LW is not that heavily physically based, and actually i doubt it would make work any easier.

Captain Obvious
02-13-2005, 12:43 PM
Well, no, you can't do the same exact thing by changing the intensity. Like toby said, the nominell distance is where the falloff starts. Another theory could be that the nominell distance is the distance at which the intensity is measured (ie, 100% intensity at 1 meter is the same exact setting as 25% intensity at 2 meters). Or would that be 50%? I can't recall how falloff works. Argh.

Lynx3d
02-13-2005, 12:56 PM
From what i observed falloff definiltely doesn't start or end anywhere (except for linear i guess, never use it), before nominal distance intensity increases and beyond it decreases.

Captain Obvious
02-13-2005, 01:25 PM
From what i observed falloff definiltely doesn't start or end anywhere (except for linear i guess, never use it), before nominal distance intensity increases and beyond it decreases.

Well, that supports my second theory that the nom. dist. is at what distance the intensity is measured.

toby
02-13-2005, 02:08 PM
Lynx said what I was thinking - I'm curious why you want the light to be so technically accurate?

Now that I think about it... we hardly ever use any falloff on lights where I work. We light things to look good, and the distance from the light usually makes no difference - unless the light source in-camera, and then we just use whatever setting looks good - are you doing architectural work?

Jure
02-13-2005, 02:18 PM
Well, that supports my second theory that the nom. dist. is at what distance the intensity is measured.
It is indeed correct what Lynx3d said. Changing Nominal distance or Intensity has the same effect. You can try it for yourself. Set distance to 100m and intensity to 100% and compare it to distance of 1m and intensity of 1000000%. You will get exactly same result. For example a 100% diffuse material at a distance of 1m from light will turn completely white when rendered with light at 100% and nominal distance of 1m. Object closer will still appear as white but since LW stores data in FP values the intensity of light on that object is acctualy higher than 100%. You can check that in ImageViewer FP by moving mouse over an image and reading pixel values at top.

Captain Obvious
02-13-2005, 02:30 PM
Just how high is LW's dynamic range, anyway? I heard it's 96-bit, but is that including the alpha channel, or is that added afterwards, so to speak? It does make sense that it's plus alpha, since then it'd use standard 32-bit floats (32 x 3 = 96)... Which should, theoretically, allow it to go as high as it wanted.

With a 32-bit float, you have 22, I think, bits of fraction data, which is a whole lot more data than a standard 8-bit integer has (ie, there are more than 256 values between 0 and 1). Even if they used 0 as black and 100.000 as 'white' (ie, 255 255 255 on the display), they could have literally billions of hues between that, and still have brightness a billion times brighter than white. It seems, well, a bit overkill, basically. Maybe I'm in error?

[/OT] ;)

Lynx3d
02-13-2005, 03:23 PM
It's 32bit floats, so 128bits with Alpha (just look at what image view FP says in the thitle ;) )

Theoretically that's plenty to adjust, but raising exposure quickly causes a very grainy image (without radiosity and such)...not sure why, think i'll have to try against Yafray or another renderer with HDR image output.
You have to be very carefully to not waste FP precision in a rendering pipeline.

toby
02-13-2005, 03:29 PM
Theoretically that's plenty to adjust, but raising exposure quickly causes a very grainy image

Could your graininess be caused by dithering? ( Effects/Processing tab/Dither intensity ) turn that puppy off!

Lynx3d
02-14-2005, 01:14 AM
Oh, it indeed makes a difference, but without it just gives massive banding wich looks several times worse...
So i don't know where LW wastes all that precision...under-exposure is impossible to fix somehow.

Elmar Moelzer
02-14-2005, 02:20 AM
Hello!
An inverse^2 falloff theoretically never ends. At some point the light will simply reach such a low strength, that it can be considered ineffective.
The Nominal distance is the distance to the lightsource at which the light has the strength set in the Light Intensity- setting. Below that distance the light it will actually get brighter until it reaches the lightsource.

LWs renderer is 96 bit FP, that means 32bit per channel (RGB). It also has 32bit Alpha and 32bit z- buffer, everything together is 160 bit.
CU
Elmar

toby
02-14-2005, 02:56 PM
Thanks Elmar, but I'm confused - why would you need 32 bit Alpha and Z buf, and where are the extra bits in the rgb for the high dynamic range?