PDA

View Full Version : Gathering Information about the Physical Intensities of Realworld Lights



jay3d
06-03-2007, 02:16 AM
Hi,

I found it important nowadays to achieve accurate lighting schemes in 3D renders especially in interiors (apart from accurate materials) so i started this thread hoping for more explanation between the community about how to convert those lighting colors and intensities to LightWave style of the latter (0% to 100% or above and what those values equals to real world measures like lumen ... etc.)

most important ones are the skylight and sun light simulation using those LightWave values style :)

Thanks!

Captain Obvious
06-03-2007, 04:57 AM
The light intensities in Lightwave are completely arbitrary.

MooseDog
06-03-2007, 06:50 AM
the color of the light though can be chosen along "real world" values by using the kelvin scale in the default lw color picker.

as to interior lighting, there's the whole issue of fall-off and intensity, but there now exists a cool workaround/solution in the form of an i.e.s. plug-in. here's the link w/in these forums:

http://www.newtek.com/forums/showthread.php?t=69154

Matt
06-03-2007, 06:55 AM
The light intensities in Lightwave are completely arbitrary.

Captain's right, until light intensities are in lumens or something 'real world', it's kinda pointless, at the moment lights are 0-100% of what? Box of frogs!

jay3d
06-03-2007, 09:52 AM
at the moment lights are 0-100% of what? Box of frogs!

I think there should be something "real" in that range of brightness 0% - 100%, just if there a way to calculate that formula of conversion between the brightness value of LightWave and the real lumens value like we say:

100% LW = 1xxx lumens for example

and then see how much the lumens value increases as we increase the brightness 1% by 1% then we can get *roughly* that magic formula (maybe it needs also some real world measurement and comparison with the rendered picture. :D

LAV
06-03-2007, 01:43 PM
I think there should be something "real" in that range of brightness 0% - 100%, just if there a way to calculate that formula of conversion between the brightness value of LightWave and the real lumens value like we say:

100% LW = 1xxx lumens for example

and then see how much the lumens value increases as we increase the brightness 1% by 1% then we can get *roughly* that magic formula (maybe it needs also some real world measurement and comparison with the rendered picture. :D


it would be beautiful if it were real. And maybe an original formula, written to achieve it, could force LW to satisfy your demands. But you understand that lumen and watt are absolute values while LW works with relative values. And then, what can be said of a light you can vary from 0% to 1000% or more? 1000% ??? :)
I seldom use real lights in interiors scene (I prefer a more 'impressionistic' kind of lightning) but I found that, for now, if you use Dialux (free and technically perfect) you can achieve whatever target you need to. Then you have to imitate its settings with LW. In a very coarse way.
Am I completely wrong?

Captain Obvious
06-03-2007, 01:44 PM
Well, light is completely, 100 % linear. If you have to set up a series of lights with varying intensities in a physically accurate manner (suppose you have ten lights in 10 cd/m2 increments, for example), just set them up with the same proportional values and you've achieved physical accuracy. The brightness of the rendered pixels is completely arbitrary anyway, because exposure is not predetermined. This is even true for cameras: The amount of exposure achieved on a film or sensor is not linear, and the conversion process makes it even more abstract.

There is no real truth in images. While a camera never actually lies, it is not an instrument of divine truth, either.

Complete physical accuracy in 3D rendering is, for all intents and purposes, impossible and irrelevant.

LAV
06-03-2007, 01:52 PM
Yes, I agree. But I suppose it would be interesting if I could simulate the light plan of a architectural space with my preferred 3d soft. I needed recently to set lights in a country house: everything was possible with lightwave (spaces, materials, objects), not the lights.

starbase1
06-04-2007, 03:24 AM
You are going to need HDRI everything to even come close...

Consider the awesome range of brightnesses you encounter in the real world, for example the Sun is magnitude -27, and the faintest star typically visible is 6.5. This is a logarithmic scale, where a difference of 5 magnitudes is a factor of 100 in brightness, thats a factor of about 5,000,000,000,000 by my estimate. 1-256 in gray scale is not much use for that.

Mind you, the power supply for a monitor that could diosplay something that bright would be pretty impressive in its own right, and at the other end you would have to sit in a darkened room for ten minutes to pick out the faint stuff!

starbase1
06-04-2007, 03:26 AM
On a more serious note, you also have to deal with human perception, which has a very powerful ability to correct for light casts.

Light at sunset is MUCH redder, but we correct for it. A common example in the other direction is how humans adapt to the lack of red underwater, and are often very disappointed at how drab and different coral or rocks look when brought into daylight.

Nick

Dave Jerrard
06-04-2007, 01:13 PM
Captain's right, until light intensities are in lumens or something 'real world', it's kinda pointless, at the moment lights are 0-100% of what?The maximum RGB values of an image. 100% would be a pure white - 255, 255, 255, or in floating point, 1.0. So, a pure white wall (255, 255, 255, 100% diffuse), lit by white light at 100% intensity will result in the wall being rendered at 255, 255, 255.

Naturally, in the real world, 100% diffuse surfaces (diffuse being a type of light reflection) do not exist - if they did, things would be extremely bright. A 100% diffuse piece of white paper would be as bright as the sun if you took it outside on a sunny day. It would be as bright as all the shining lights in a room if it was indoors. Obviously, this doesn't happen in reality, where we're not limited to 256 levels of brightness and lights all have an inverse square falloff.

Just think of the intensity of lights now as the percentage of a maximum brightness that can be correctly rendered without being over exposed (or clipped). If you're using inverse square falloff, you'll have more accurate real world lighting, because the light need to be brighter, the further it is away from a surface to light it with the same intensity. With the inverse falloff, you're not specifying the intensity of the light itself, but rather how intense it's effects are going to be at the nominal distance. Anything closer than this distance will receive more light, and anything further away will receive less, in an accurate method. Making the intensity higher, or increasing the nominal distance will both have the effect of making the light brighter. If the intensity is 100%, then anything within the nominal distance will likely be over exposed or clipped to a maximum of 255, unless you're saving out a HDR format, in which case, you'll still be able to see details above that intensity.

He Who Wouldn't Mind Some New Lighting Systems Though.

Lewis
06-04-2007, 01:35 PM
Hi Dave !

I have one more Question :)

What if we invlove falloff RANGE in to equation - how to calculate it then? If I set invers square2 falloff and intensity 100% but set falloff to 10meter it will be totally different than when falloff is set to 20m ? Is then light going to be 2x times brighter or more or what :)?

I ask that 'coz with new IES plugin we have we still don't have it controled like in 3DS MAX which reads LUMENS from IES file so you don't have intensity value when you load IES while in LW you still can adjust intensity AND fallof range which contradicts to IES idea from my understanding (since IES have all that SET) ?

Thanks

jay3d
06-05-2007, 04:42 AM
Originally Posted by Lewis
What if we invlove falloff RANGE in to equation - how to calculate it then? If I set invers square2 falloff and intensity 100% but set falloff to 10meter it will be totally different than when falloff is set to 20m ? Is then light going to be 2x times brighter or more or what ?

I think that the "Nominal Distance" acts as a light intensity multiplier in "Inverse Distance ^ 2" mode, it doesn't actually stop the light from shooting beyond that range.

for example if you rendered where light intensity = 100% and ID^2 range = 2 m is exactly the same when you set light intensity = 400% and ID^2 range = 1 m

so according to that you can simulate an area light using only radiosity by setting the light object material color to the light color and the luminosity to 4000% (and 0% to diffuse of course) and you will get a simulation of an area light with a setting of 100 intensity and Inverse Distance ^2 and range of 1 m

Lewis
06-05-2007, 04:55 AM
Yeah jay3d I know it's somehow that way but i want light to be set at 100% and just increase falloff and there is where problem occurs when using IES files - they are SET with all settings and yet in LW we still can tweak it so basically you don't know what is the REAL settings since it still gives you all options of falloff and it shouldn't work that way but yeah since LW IES is briliant HACK by Denis I think we need to wait for NT to implement IES in real lights properies :).

Iain
06-05-2007, 05:03 AM
I think that for anyone using LightWave, if it looks right, it is right.
If you need a more technical approach, currently you need to turn to the Maxwell or Fry render solutions but even they are only scientific approximations of physical light.

You can achieve incredibly accurate light simulations and still produce crap renders.
:devil:

jay3d
06-05-2007, 05:08 AM
exaclty!, I think what denis have to do currently to accomplish that is to make some channel modifier along with the IES node that will apply to the envelope to the "Range/Nominal Distance". but still one problem i'm currently investigating that what is the light intensity is to real world measures? and i think i can solve it through taking HDRI images by a camera and comparing the high dynamic range value with the actual stock light energy to get the constant between light intensities in light wave and real world measures ... :)

jay3d
06-05-2007, 05:09 AM
I think what denis have to do currently to accomplish that is to make some channel modifier along with the IES node that will apply to the envelope to the "Range/Nominal Distance". but still one problem i'm currently investigating that what is the light intensity is to real world measures? and i think i can solve it through taking HDRI images by a camera and comparing the high dynamic range value with the actual stock light energy to get the constant between light intensities in light wave and real world measures ...

jay3d
06-05-2007, 05:10 AM
sorry by mistake posted twice :foreheads

Iain
06-05-2007, 05:16 AM
I think what denis have to do currently to accomplish that is to make some channel modifier along with the IES node that will apply to the envelope to the "Range/Nominal Distance". but still one problem i'm currently investigating that what is the light intensity is to real world measures? and i think i can solve it through taking HDRI images by a camera and comparing the high dynamic range value with the actual stock light energy to get the constant between light intensities in light wave and real world measures ...

I'm not sure how useful that information would be unless LightWave's radiosity solution was a true simulation of physical light, which it's not.

jay3d
06-05-2007, 05:18 AM
If you need a more technical approach, currently you need to turn to the Maxwell or Fry render solutions but even they are only scientific approximations of physical light.

I know that Maxwell of Fryrender are "claimed" to be physically accurate but still there's some issues with Interior lights (electric lights) in Maxwell that being so dim even with physical presets, add to that Maxwell doesn't support IES!, and to simulate that through complex refractions is pointless and a rendering hog on an already sloooow renderer, so if what maxwell produces that is doable in lightwave with much more clear renderers (that i believe), but with little more tweaking of physical properties, instead of waiting hours and hours for a simple Cornell box render to clear it from noise???! :)

dpont
06-05-2007, 08:01 AM
We could say for Incandescent Lighting,
that 100% brightness (intensity) = 100 watts
in real world, and one 100 watts lamp = 1740 lumens, but
what it is annoying is that two 60 watts lamps = 1720 lumens.
Max Intensity lumens is in the IES Map info panel
(the "Lumen per Lamp" is exagerated by the manufacturer).

Denis.

jay3d
06-05-2007, 08:15 AM
We could say for Incandescent Lighting,
that 100% brightness (intensity) = 100 watts
in real world, and one 100 watts lamp = 1740 lumens, but
what it is annoying is that two 60 watts lamps = 1720 lumens.
Max Intensity lumens is in the IES Map info panel
(the "Lumen per Lamp" is exagerated by the manufacturer).

Denis.

Thanks Denis for the information! :thumbsup:

dpont
06-05-2007, 09:29 AM
A more extended table:
http://en.wikipedia.org/wiki/Luminous_efficacy

Denis.

Dave Jerrard
06-05-2007, 12:34 PM
What if we invlove falloff RANGE in to equation - how to calculate it then? If I set invers square2 falloff and intensity 100% but set falloff to 10meter it will be totally different than when falloff is set to 20m ? Is then light going to be 2x times brighter or more or what :)?It would actually be 4 times brighter at the 10 meter mark then. See below.

[/QUOTE]I ask that 'coz with new IES plugin we have we still don't have it controled like in 3DS MAX which reads LUMENS from IES file so you don't have intensity value when you load IES while in LW you still can adjust intensity AND fallof range which contradicts to IES idea from my understanding (since IES have all that SET) ?

Thanks[/QUOTE]It's not a linear falloff. The inverse falloff have a more abrupt drop in intensity closer to the light source and the drop become more gradual the further you are fromt the light. This is one of the main reasons why photography, film and TV studios try to place their lights farther fromt eh action - to even out the distribution of the light so it's not too bright at one side and too dark on the other.

With linear falloff, there is no light beyond the falloff range. That range marks the zero point - the point where the light has fallen to zero intensity, and there is no light beyond that, no matter how bright you make the light itself. At half that distance, the light will be half the intensity, at 1/3 the distance, it will be 2/3 intense, etc., because the falloff IS linear.

With the Inverse falloffs, the you're not setting a falloff range in the same sense you are with a linear falloff. Instead, this control defines the nominal distance, not a maximum range. Light still goes on forever, past this distance, but it continues to decrease in intensity. For Inverse Square falloff, the intensity is inversely proportional to the square of the distance from the light source. Each time you double the distance of an object to a light source, the amount of the light hitting it is reduced by 75%. If an object is recieving 100% intensity at 1 meter from the light (the light as an intensity of 100% with a nominal distance of 1 meter), it would only receive 25% intensity at two meters. At 4 meters, the light would only have a 6.25% (25% of 25%) intensity effect. At half a meter, it would be four times more light: 400%. This page has a good diagram of why this happens (http://hyperphysics.phy-astr.gsu.edu/hbase/forces/isq.html).


He Who Thinks It's Time To Reboot This Machine.

starbase1
06-05-2007, 12:42 PM
Incidentally, it's a bit specialist but there's a guy who has done a lot of work in calculating what the apparent colour of stars of differing spectral types would be to the human eye.

http://www.vendian.org/mncharity/dir3/starcolor/

It's very different from seeing a star, (i.e. a point source against a dark background).

Nick

jay3d
06-05-2007, 11:14 PM
Thanks Dave for the links, another important question is:

is radiosity in LW obeys that law as well?

Lewis
06-06-2007, 08:23 AM
Thanks Dave , I didn't express myself properly and i know that linear isn't same as square inverse but my main concern is somethign around what Denis explained. How to set EXACT amount of IES light since we still can tweak all light settings while we shouldn't need to tweak it when loading IES. But this what Denis said (100% ~ 100w) might help a bit so that i can calculate how much intensity i need to set for let's say 2000 lumens but falloff will still be mistery to set and get exact right IES light lenght (at spots with various filters) at render. Company for what i do freelancing regluary SELL lights and they do light simulations/calculations (Dialux is pretty nice free program for it but very limited in polygon count/type) so i need to be sure is that render light intensity right one like it would be in real life when lights are sold to client and mounted to buildings :).

Thanks for links they are helpful to understand all that technical "mumbo jumbo" little better :).

Elmar Moelzer
06-06-2007, 03:47 PM
Hello!
While it might be nice to have an option for "real world related lighting" of some sorts, it would IMHO not be practical in most circumstances.
As some have mentioned the light intensities in our environment vary greatly. E.g. did you know that the sun is more than 20 times as bright on a sunny day than it is on a cloudy day (same time of day)? This is a difference as in 5% and 100% light intensities...
We actually had that measured for our car headlight simulator, when we were desparately trying to get some relation between LWs light intensities and real world ones.
In order to make this work, we would not only need light intensity settings, but also exposure settings for cameras plus some kind of chip light- sensitivity for digital cameras (I think this varies too) and then in those the software corrects some stuff too, so really complicated I guess, but that might be doable with the real lens cameras, maybe...
Of course the diamater of the lens has an impact on that too (the bigger the lens the more light it gathers thats why lenses with a strong zoom get so big and wide at the end).
Human eyes are even more complicated to simulate since the eye does not only have an iris but the brain and the eye both do so much filtering as well and what we perceive as 100% white differs greatly depending on what we focus on.
E.g. you look at a white wall in bright daylight, right where the sun hits it and you would say that it is white. Now focus at a spot where only indirect light hits the wall, again looks white right? Now try to keep your eyes to focus on the directly lit part of the wall and try to see how dark the rest of the wall appears without loosing focus of the brightly lit part (difficult hu, but you can do it ;)). Pretty huge difference!
So what would you set your walls color to then? Maybe Pantone colors, but that might still not help with the "diffuse level" if you understand what I mean.
So surfaces would have to have some more real world- like settings too, as was mentioned before. I am not even sure what all the settings would be that would have to be simulated then. I am pretty sure though that things would not necessarily get any easier to setup.
Now, I am always for options and all that, so it might be cool to have optional settings that bring the LW lighting and other settings into some real world correlation, but I am kinda wondering how many people would actually use it in the end, since it is going to be really complicated and not very intuitive.
CU
Elmar

byte_fx
06-06-2007, 04:37 PM
Just a note about wattage vs. lumens.

In reality that depends on the manufacturer and type/model of bulb.

Clear vs. white coated vs. soft white for example.

But it does seem that LW 100% is pretty close to 100 watts equilivent. But this also seems to depend on surface texturing as well.

Bottom line - most renders don't need precise physical lighting accuracy as much as they need a well calibrated monitor for displaying.

Captain Obvious
06-06-2007, 04:38 PM
Just use Maxwell
There, I fixed your post for you. ;)




But it does seem that LW 100% is pretty close to 100 watts equilivent. But this also seems to depend on surface texturing as well.
No, it doesn't. I don't know where you got this idea. Exposures are completely arbitrary anyway.

byte_fx
06-06-2007, 05:45 PM
What part don't you see where I get it?

(man - talk about an awkward sentence)

Anyway - I get the idea that 100% is pretty close to 100 watts from comparing renders to real world counterparts.

Not saying 100% = 100 watts (which is subjective anyway) but that the ratio is close enough to use a starting point.

Again - in general. Highly dependant on type of LW used. And location/scale when using area lights.

If anyone wants to get picky over phsicallu accurate lighting then there are a whole universe of things to consider.

At a minimum - bu;b type and maker, bulb age (lumens and color shift as a bulb warms up and/or ages, voltage variations due to resistance in wiring to bulb/bulb contact in socket, etc.

At the very core every 3D app is arbitrary in this sense. It's the final result of hpw light affects surfaces within the app that determines any correlation between a virtual light and howit affects final output.

Which is why the ultimate test is how it looks on a properly calibrated monitor.

byte_fx

jay3d
06-06-2007, 05:49 PM
I'm gonna break this Maxwell-is-a-god-renderer idea with some images soon based on the studies here, IMHO that we should push NT to improve the renderer more by doing these kinds of threads, instead of saying: "OK NT, I'm going to beg NextLimit for a WORKING plugin, and beg them more in the future to support Nodes, Proc. Textures, ... etc., but i'm not going to support an already robust and supported renderer" that doesn't make any sense, because all the CG history, programmers tried to avoid using direct approach for raytracing, and they made several smarty coding to simulate what is actually can be simulated in other ways, not to bring an unbiased beast without any techniques of GI filtering, caching, ... etc., i think the future of renderers is not unbiased, it's biased with smart configurations and optimizations :)

byte_fx
06-06-2007, 06:08 PM
Haven't tried Maxwell.

But - yeah - I would like to see some major upgrades in LW's lighting model as well.

Actually the lighting engine has a major affect on the render engine's output.

Nearly everything I do is artificially lit - at the most there might be a narrow streak or two of sunlight coming thru an opening.

Dennis's plug is a step in the right direction and I'm glad to see it. Haven't tried it yet but it looks promising.

byte_fx

dballesg
06-07-2007, 02:38 AM
Hi,

Really interesting thread :)

I remember Jay Roth said that the lightning system was on the future development as well.

And for those that one to try a unbiased renderer without need to buy Maxwell or Fryrender, there is the free Indigo Renderer.

Someone with the nick jay3d on their forums said it is going to do a LW exporter to it.

And even allows network render! :) And they have a free tonemap utility! :)

Best regards,
David

Jure
06-07-2007, 02:54 AM
Hi!

I made some test to try and figure out how LW % are related to modo's W/sqm.

From my tests I can conclude that 1 W/sqm is approximately 1/3 of 100% intensity with inv. square falloff of 1m.

1 W/sqm =~ 32%
10 W/sqm =~ 320%
20 W/sqm =~ 640%
50 W/sqm =~1600%

So basic formula would be number (W/sqm)*32 =~ (LW intensity)%

Check attachments for examples.

Iain
06-07-2007, 03:15 AM
I think to compare LightWave's renderer or any general purpose engine to the likes of Maxwell is pointless. They are fundamentally different.
Maybe one day they will be more similar but at the moment that is fact.

What's wrong with having LW (and, optionally, FPrime or Kray) for fast, high quality, flexible rendering and having an MLT-based engine like Maxwell for the times it is required?
These software licenses cost nothing in terms of their relative return compared to other business investment.

LightWave's render engine is excellent but it's not Maxwell or Fry (really, it's not). Similarly, there are a million things you can do with LW that you cannot do with them.

We need to get away from this using non-Newtek software is a betrayal mentality.

Captain Obvious
06-07-2007, 03:31 AM
LightWave's render engine is excellent but it's not Maxwell or Fry (really, it's not). Similarly, there are a million things you can do with LW that you cannot do with them.
Like render things in under an hour! :)



Jure, keep in mind that A) modo defaults to a gamma correction of 1.6, and that the area lights in modo area size-dependant. A 2x2 meter area light in modo is four times as bright as a 1x1 meter one. This is not the case with Lightwave.

Jure
06-07-2007, 04:23 AM
Jure, keep in mind that A) modo defaults to a gamma correction of 1.6, and that the area lights in modo area size-dependant. A 2x2 meter area light in modo is four times as bright as a 1x1 meter one. This is not the case with Lightwave.

I tried to match both exactly.

I set gamma to 1.0 in modo.
Lights are spotlights (not area lights).

jay3d
06-07-2007, 05:19 AM
And for those that one to try a unbiased renderer without need to buy Maxwell or Fryrender, there is the free Indigo Renderer.

Someone with the nick jay3d on their forums said it is going to do a LW exporter to it.

And even allows network render! :) And they have a free tonemap utility! :)

Best regards,
David

Hehe, nice catch David!, however i'm the one who is doing this exporter!, but i'm having difficulties with Micro$oft Visual C++ Express 2005 because they deprecated most of the string functions in C.... Grrr .... so the library MiniXML 2.x doesn't compile anymore ... so i'm trying to use tinyXML library but it's a C++ one, so i'm trying also to find a way to call a C++ function from C without breaking plugin compatibility (ooohh tired of that) :D

jay3d
06-07-2007, 05:21 AM
Nice comparision Jure!, appreciate that! :thumbsup:

i will try to test that with realworld sun and sky :)

Captain Obvious
06-07-2007, 05:40 AM
The default directional light in modo has an intensity of 3 watts per srm2, and seems about as bright as the default directional light in Lightwave, with its 100% intensity. So, it seems logical to assume that 3 watts per srm2 is equal to 100% intensity in Lightwave. That matches well with your tests, Jure.

At 3 watts per srm2 and 100% diffuse (and 100% exposure), a directional light shining directly onto a surface will produce pure 255/255/255 white.

For luminosity, however, 1 watt per srm2 equals 255/255/255 white (provided that the luminosity colour is white, that is).

dpont
06-07-2007, 06:16 AM
Two parallel worlds,
Photometric or Radiometric?
That is the question,
http://en.wikipedia.org/wiki/Photometry_(optics)

Denis.

-EsHrA-
06-07-2007, 07:06 AM
two?!..

scientists allready discovered 13+ parallel universes.

:)


mlon

Mr Rid
08-12-2007, 07:28 PM
I'm not sure how useful that information would be unless LightWave's radiosity solution was a true simulation of physical light, which it's not.

Yes, I dont know what the math is (nor do I care) but I cant tell what LW radiosity is suppose to be approximating. An example you can try is like when I once tried to quick simulate my living room with one light to approximate one of those cheap floor lamps that bounces a 300w halogen equivalent light off a 100% diffuse ceiling (lite-200%, inverse- 250mm, 2 bounces). I had to turn the radiosity intensity up to 3000% to come close to what the level of bounce illumination is in my actual living room. But then you can imagine the amount of noise you get that 243 rays with 9-pass AA does not eliminate (not to mention the render time in 8.3 at the time). And the area for about six feet around the wall and ceiling near the lamp are unavoidably clipped (post correction with exr doesnt work). Radiosity also does not affect spec. I think there should at least be a selectable falloff value for radiosity bounce?

(48 rays, noAA)49227

Captain Obvious
08-12-2007, 07:49 PM
The trick is to use gamma correction!

Mr Rid
08-13-2007, 12:31 AM
The trick is to use gamma correction!

Had tried that too. Not giving a good result.

Dave Jerrard
08-13-2007, 04:13 AM
Yes, I dont know what the math is (nor do I care) but I cant tell what LW radiosity is suppose to be approximating. An example you can try is like when I once tried to quick simulate my living room with one light to approximate one of those cheap floor lamps that bounces a 300w halogen equivalent light off a 100% diffuse ceiling (lite-200%, inverse- 250mm, 2 bounces).
For one, there is no such thing as a 100% diffuse surface in the real world. For realism, you shouldn't be using it in 3D either, except in special circomstances. The same goes for Transparency and Reflectivity.
Two, for some reason you're equating a 300w light with a 200% intense LW light. Normally, you would keep the intensity about the same as the wattage. It's not entirely accurate, but it's generally a good rule of thumb if you consider a 200w bulb to be about twice as bright as a 100w one.
Three, you're not using inverse square falloff, so the light model is even more incorrect.

If done correctly, you will either have a very blown out area near the light, just as you would in a photograph exposed for the darker areas, or a darker image with the lightest areas just hovering around the clipping level (or higher if you're outputting hdr) by using a dimmer light (or less diffuse surfaces), simulating what you're get with a photograph exposed for the light area of the image.


I had to turn the radiosity intensity up to 3000% to come close to what the level of bounce illumination is in my actual living room. But then you can imagine the amount of noise you get that 243 rays with 9-pass AA does not eliminate (not to mention the render time in 8.3 at the time). Try it again in 9.2. The interpolated Monte Carlo or Final Gather are much faster now, and you can crank them to 24 bounces, with is four bounces higher than FPrime allows. Depending on the scene, you can get better render times with the LightWave render than you do with FPrime, but this is very scene dependent; I've seen it go either way.



And the area for about six feet around the wall and ceiling near the lamp are unavoidably clipped (post correction with exr doesnt work). As you would with a photograph exposed for the lower ranges. HDR only allows for a few extra stops of exposure over normal photographic ranges, making it just a bit wider than the RAW formats most digital SLR are capable of shooting, which itself add about 2-3 stops of extra range over 24bit color.



Radiosity also does not affect spec. I've noticed that. It should work with this. After all, you can get some great caustics out of it. Caustics, on the other hand, only respect transparency, reflection and specularity. And they usually look like crap.

He Who Would Love To Use Caustics More If They Didn't Swim From Frame To Frame.

Mr Rid
08-13-2007, 05:43 AM
For one, there is no such thing as a 100% diffuse surface in the real world. For realism, you shouldn't be using it in 3D either,...


I understand all this. As I said, initially it was a quick sim with just 100% diffuse and no other surface settings. But I have a more realistically surfaced version of the room that actually requires a radiosity intensity of 10,000% to approximate the bounce falloff in reality.


Two, for some reason you're equating a 300w light with a 200% intense LW light. ...

I was actually equating the intensity with what the real light looks like with the dimmer switch at a normal level which is never full tilt 300w. I could see the actual room just over the top of my computer monitor as I compared it to F9s.



Three, you're not using inverse square falloff, so the light model is even more incorrect.

I understand very well how the light falloffs work, and going for more realistic falloff with inverse square only means again having to crank the radiosity and/or light intensity higher to compensate. Adjusting falloff and diffuse a little more this way or that has little bearing on the necessity for an insane radiosity intensity in order to come close to a realistic bounce.



Try it again in 9.2. The interpolated Monte Carlo or Final Gather are much faster now, and you can crank them to 24 bounces,

This raises another issue I forgot to mention before- the more bounces, the more unrealistically saturated colors become. At 4 bounces, the red walls are turning the entire white ceiling to pink, although the actual ceiling is only very faintly shaded pinkish nearest to the walls. 24 bounces deeply saturates the whole room in a way that looks nothing like the actual room, and the intensity would have to be jacked up well above 10,000% (I didnt bother) as it also darkens the recesses much more than how the actual room appears.

No matter how you slice it, LW radiosity is just not bouncing realistically. It is yet another mysterious percentage value in LW that amounts to random guessing and counter-balancing for the user.

alifx
08-13-2007, 06:50 AM
IMO Mr Rid in every 3d software not only LW.. when you increase the radiosity bounces you should decrease the intensity of lights to get more ambient light and softer shading on surfaces, I you didn't do that with the lights you'll get much saturated colors on every surface.


I think you know that
and sorry if I misunderstand what you meant.

Dave Jerrard
08-13-2007, 10:19 AM
I was actually equating the intensity with what the real light looks like with the dimmer switch at a normal level which is never full tilt 300w. I could see the actual room just over the top of my computer monitor as I compared it to F9s.The human eye can see a much larger range of intensities than film or a monitor can show, so it's not a very good comparison for this. Your best bet would be to take a photo of the room and compare with that. "Your eyes can deceive you, don't trust them."






I understand very well how the light falloffs work, and going for more realistic falloff with inverse square only means again having to crank the radiosity and/or light intensity higher to compensate. Adjusting falloff and diffuse a little more this way or that has little bearing on the necessity for an insane radiosity intensity in order to come close to a realistic bounce.Inverse Square makes the light automatically brighter at the source than you get with just Inverse or Linear witht eh same falloff range. In the case of the two inverse modes, the intensity of the light that you set is how bright the light is AT the falloff, which for these types, is the Nominal Range. In your case, your light's intensity is 200% at 250mm from the light. It's brighter inside this distance and dimmer outside it. So, you're quite likely not even getting a full 100% intensity reaching the ceiling, or possibly, even the wall. If you're going to simulate the real world effects, then simulate the real world, and crank that intensity up. Or just mess with the HDR exposure and keep using a dimmer light, depending on the exposure you want.






This raises another issue I forgot to mention before- the more bounces, the more unrealistically saturated colors become. At 4 bounces, the red walls are turning the entire white ceiling to pink, although the actual ceiling is only very faintly shaded pinkish nearest to the walls. 24 bounces deeply saturates the whole room in a way that looks nothing like the actual room, and the intensity would have to be jacked up well above 10,000% (I didnt bother) as it also darkens the recesses much more than how the actual room appears. This is likely due to you overcranking the radiosity intensity, which is also cranking up the color saturation. If you crank the light itself, instead of the radiosity, it should tone down the saturation of the colors because now most of your light will be from the light, and not the nearly luminous surfaces around it.


No matter how you slice it, LW radiosity is just not bouncing realistically. It is yet another mysterious percentage value in LW that amounts to random guessing and counter-balancing for the user.I've never had any problems with it and I use it nearly all the time now. It works as I expect it to. You just need to use it right. If you're cranking it over about 500% then your main lighting is almost certainly off somewhere.


He Who Keeps His Radiosity Between 50% and 200% Most Of The Time.

Thomas M.
08-13-2007, 10:50 AM
As long as there isn't an iris to the camera you can open or close it's little use to talk about real world intensities.

Cheers
Thomas

Mr Rid
08-13-2007, 01:45 PM
IMO Mr Rid in every 3d software not only LW.. when you increase the radiosity bounces you should decrease the intensity of lights to get more ambient light and softer shading on surfaces, I you didn't do that with the lights you'll get much saturated colors on every surface.



Right, but if I turn down the light intensity, lets say by half, then I just have to crank the radiosity intensity up twice as much in order to maintain the luma and we are right back where we started. The counter balance game between diffusion, light and radiosity pecentages continues that are based on, 'If it looks right, then it it is right.' I dont know that anyone really knew what to base CG reflectance values on before Matrix Reloaded- http://www.virtualcinematography.org/publications/acrobat/BRDF-s2003.pdf

I keep waiting for these NASA/Lockheed, $700,000 reflectometer models to somehow trickle down into lowly, off-the-shelf development.

Mr Rid
08-13-2007, 05:28 PM
The human eye can see a much larger range of intensities than film or a monitor can show, so it's not a very good comparison for this. Your best bet would be to take a photo of the room and compare with that. "Your eyes can deceive you, don't trust them."

Right, I took photo and video at the time. Radiosity just isnt accurate. I had to fix the renders in post.



This is likely due to you overcranking the radiosity intensity, which is also cranking up the color saturation. If you crank the light itself, instead of the radiosity, it should tone down the saturation of the colors because now most of your light will be from the light, and not the nearly luminous surfaces around it.

I see six of one, half dozen the other. The color remains oversaturated, even with lower radiosity intensities.



I've never had any problems with it and I use it nearly all the time now. It works as I expect it to. You just need to use it right.


LW radiosity may seem very realistic on a monitor, but I see that the inaccuracies show up when you compare closely to an actual room with specifc colors as I was doing. I even used the exact pantone conversion for the paint color since the original reason for creating the living room model was to trial alt decor. But its kinda like how RealFlow liquids can look very convincing when you do everything right with say a brick dropping into a bucket of water. But then you look at video of the real thing and you suddenly see how it is way off.

Dave Jerrard
08-13-2007, 10:32 PM
As long as there isn't an iris to the camera you can open or close it's little use to talk about real world intensities. We don't have an iris, but we do have exposure tools in LightWave, so real world intensities are worth talking about. An iris isn't the only thing in a camera that controls the exposure; exposure time, sensitivity, and filters can also adjust the amount of light that appears in a photograph.


He Who Usually Uses Shutter Priority In His Photography.

Dave Jerrard
08-13-2007, 10:49 PM
Right, I took photo and video at the time. Radiosity just isnt accurate. I had to fix the renders in post.Since we don't have 100% physically accurate lights in Lightwave, this doesn't surprise me. You're going to have to tweak things in any app to get a perfect match




I see six of one, half dozen the other. The color remains oversaturated, even with lower radiosity intensities. Well, colors ARE going to get more saturated with each bounce; it's the nature of the beast. If white light picks up a 10% red tint on each bounce of a reddish surface, then each successive bounce is going to become redder and redder, in addition to it becoming dimmer and dimmer. This is very evident in a place like the Bradbury Building, where there's a lot of wood floors and walls. Under artificial light, which isn't strong enough to bounce from floor to floor appreciably, the wood looks just like that - wood. But when sunlight shines in through teh roof, the entire place takes on a very rich orange cast, making everything look more red than it really is. This is due to all the bounces of sunlight picking up a bit of the light brown wood color, bouncing to another wood surface and getting more brown in the bounced light. Unfortunately, in this photo, the camera's auto white balance compensated for much of the warm glow and the photo doesn't do the building justice. It's a lot warmer in person.

49278




LW radiosity may seem very realistic on a monitor, but I see that the inaccuracies show up when you compare closely to an actual room with specifc colors as I was doing. I even used the exact pantone conversion for the paint color since the original reason for creating the living room model was to trial alt decor. How accurate is your monitor. I've yet to see a straight pantone color on a monitor that's a perfect match for the same color on a Pantone swatch, no matter what app it's viewed in. I've tried doing pantones colors for logo work years ago, and it's nearly impossible to get it right. In the end, I had to just eyeball it to get it to look right, because the exact RGB values, as Photoshop converted them, were way off visually.



He Who Was Just Reminded That His TV Needs Calibration Again.

Iain
08-14-2007, 02:06 AM
I don't think anyone tries to pretend that LW does emulate real world lighting.
These claims are made by the makers of Maxwell and Fry so for the sake of a few hundred $, they're a great investment if that's what you want.

Even they do have to be tweaked, of course. It's all just code based on the study of physics at the - pink tinged - end of the day.

Mr Rid
08-14-2007, 05:11 AM
[QUOTE=Dave Jerrard]This is very evident in a place like the Bradbury Building,
49278

Yay, the Bradbury pics, more, more!