PDA

View Full Version : Why LW lighting needs a re-write . . .



Matt
09-08-2003, 12:12 PM
I love LightWave but I'm fast coming to the conclusion that soft shadow lights in LightWave need a total re-write. They are slow, the quality is poor when scaled up and they don't even behave correctly.

Here are some tests to explain my frustration with them when using them in any interior setting.

I created a simple cube scene to test behavior of lights in LightWave. The translucent representation is merely to show the size, position and rotation of the light, the object has everything that would affect the rendering turned off.

Area Light Problems

http://www.creactive-design.co.uk/temp/area_light_problems.jpg

1
LightWave area lights don't emit light along their edges, this dark edge becomes a problem when trying to use area lights inside interiors as they give the look of removing light, surely some bleed would occur without having to resort to radiosity? Placing the light close to ceiling to simulate an panel light to hide the dark edge isn't possible either, as they need to be at least a metre away to have any sort of affect (see below).

2
If an area light is a true square why is the bloom almost circular?

3
Also why does the bloom stop at the edge of the cube? Shouldn't it be visible on the floor too?

When scaled up (x3 here) they become very grainy. The very fact we have an option called 'Shading Noise Reduction' admits that this is a problem. The only answer is to use more lights as oppose to scaling, but because they are slow with even one light this becomes a less attractive alternative, and you have the 'non-emitting' edge render problem to contend with too.

http://www.creactive-design.co.uk/temp/area_light_x3_size.jpg

If area lights are infinitely thin (which is why they don't emit light along their edges) then why when placed 2mm above the floor do they not effect it?

http://www.creactive-design.co.uk/temp/area_light_2mm_off_floor.jpg

Even at 20cm above the floor it has hardly any affect!

http://www.creactive-design.co.uk/temp/area_light_20cm_off_floor.jpg

Only at 1m off the floor does it start to behave as expected, despite the circular bloom.

http://www.creactive-design.co.uk/temp/area_light_1m_off_floor.jpg

Linear Light Problems

http://www.creactive-design.co.uk/temp/linear_light_problems.jpg

1
In a similar fashion to area lights, linear lights don't emit light from their ends, again this can cause problems with interior renderings.

2
Unexpected bloom shapes.

3
Again why does this bloom stop at the edge of the cube?

This linear light is 20cm off the floor, so why are the walls washed so brightly when they are further away than the floor?

http://www.creactive-design.co.uk/temp/linear_light_20cm_off_floor.jpg

Having lights that behave realistically is essential for interior renderings when radiosity isn't an option. In a commercial environment with real deadlines radiosity in LightWave just isn't viable, even with a small renderfarm. The reason why is because one frame can't be sent to multiple machines for rendering.

For people who use LightWave for visualisation of products and interiors, most of their output will be at print resolution, and print resolution with radiosity most probably won't render overnight. These are the same machines that are used during the day so tying them up rendering is a problem.

This is why I never use radiosity for interior renderings, and most of the time can't use any of the soft shadow lights.

When compared to renderers like Lightscape or FinalRender LightWave is very weak in these areas and needs addressing.

Matt

Ztreem
09-08-2003, 12:25 PM
I totally agree on that!!!

Nice work, by the way. :cool:

Earl
09-08-2003, 12:40 PM
Indeed, these among many other posted problems, are good arguements for a drastic re-write of the lighting/rendering system.

Of course, there will never be such a re-write, since backwards compatability is too important for some.

Matt
09-08-2003, 12:58 PM
so backward compatibility is more important than progress? I severely hope not!!!

Panikos
09-08-2003, 01:32 PM
LW's Area Lights and Linear Lights are internally composed of point lights.

Matt
09-08-2003, 01:35 PM
but with soft shadows.

wonder why point lights don't have soft shadows?

Elmar Moelzer
09-08-2003, 01:42 PM
This is why I am hoping for a new "Light"- plugin- class for LW.
There are many ways of doing lights and not all are suitable for all purposes. Imagine the amount of new light- types we could have with such a plugin- class.
I already have a few ideas too....
CU
Elmar

Earl
09-08-2003, 02:41 PM
I didn't mean to imply that I was one of those who held backward compatibility as more important than progress. ;) I for one say that if the need exists for a change, and that change would be for the better, then backward compatibility can take a backseat!

However, it's always nice if they find some way to implement new things (such as lights) while still being able to handle the old files properly. Perhaps what Elmar has suggested is the way to do it: allow a "Light" plugin class, making all light models to be able to be added as a light plugin. That way they could implement an all-new lighting system and then older files would simply translate over to access plugins which simulate (or act completely like) the old lights. This way everyone's happy. People with older scenes can still render them out with the same results (and same render times), while people who want to use newer and cooler light models can do so - all the while developers can release custom lights!

The ripper
09-08-2003, 03:17 PM
Very good thread Matt!!!
I always wanted to point these problems out but I always thought that would be hopeless.

I used Gaffer(Worley shader) in the past to solve some of these problems cause its soft shadows were good(slow though) and could be used on both type of LightWave lights(point, spot).

As It became kind of absolete and had some problems with lots of lights in a scene and some others such as shadows through transparent surfaces... I really thought Newtek was going to do something about it but I'm still waiting and I don't think we'll see something better quite soon.

I really enjoy Lightwave but I'm slowly thinking about choosing another 3D package for the future if they don't do something about it cause I do a lot of lighting works and I need something more effecient that what LW can offer now.

My 2 cents.

The ripper.

cavalos
09-08-2003, 06:08 PM
Improve the light thing NT...we all know you´re very busy right now on LW8 but the next release must be a "renderer upgrade release".

Best
Christian

marko
09-08-2003, 08:04 PM
history repeating ....

http://forums.newtek.com/discus/messages/2/10940.html?

takkun
09-08-2003, 08:59 PM
Originally posted by marko
history repeating ....

http://forums.newtek.com/discus/index.html Your link just goes to the front page of the old forum, did you mean to link this thread?

http://forums.newtek.com/discus/messages/2/34112.html?

Matt
09-09-2003, 02:22 AM
hehe! you can only keep asking! plus I think this shows the problems more clearly than the old thread!

like ripper I just want lightwave to be better.

we have a job on at the moment that we're working with another consultancy on, they're doing their renders in 3DSMax using radiosity, and it annoys me that for interior stuff lighting in LW lets me down, I can't use radiosity due to the render times, and due to the limitations of linear and area lights I can't create soft shadows in the interior without some other rendering anomaly.

The ripper
09-09-2003, 02:55 AM
Here's what I think to be another problem:
http://users.skynet.be/fa008785/AreaLight_Falloff_001.jpg

A: I set a 'linear' falloff for the area light so that It's not supposed to reach the cylinder object so no render anything further... falloff seems to work.

B: I set a 'Inverse Distance' falloff this time but as you can see, It doesn't take into account the falloff radius!... is this suppossed to be normal!?!

Extent
09-09-2003, 03:22 AM
in my experience with falloff using inverse is that it scales both directions. so the brightest point, the closest to the light, is not 100% but actually much greater than that, and it causes terrible blowouts (this was with point lights, but I assume it applies to all others as well) so your falloff range has to be changed, because a logrithmic falloff is so totaly different than a linear.

Having an arealight that gave off light along it's edges would basically be a soft edged point light tho.

You can get really great results if you dont limit yourself to "stock" lights. Play with custom softshadow rigs. I rarely use area lights and never use linear lights

Karmacop
09-09-2003, 03:44 AM
Yeah, I think inverse distance gives the light intensity at the falloff distance, it doesn't fall off completely at that distance like a linear light.

Earl
09-09-2003, 09:36 AM
When using Inverse Distance Falloff, the Range/Nominal Distance is the distance that the light will have 100% of the Light Intensity that you specified. Anything closer will increase the brightness exponentially. Anything further away will become darker exponentially.

Earl
09-09-2003, 09:40 AM
I remember at one time I suggested having the ability to specify the distance that you want the light to reach 0% of your specified intensity, and the distance that you want your light to be at 100% of your specified intensity. Any distance closer would still be at 100% of your specified intensity. The 'falloff' between 100% and 0% would be determined by a simple spline curve that was editable.

I think if I had that option, I wouldn't ever use any other falloff's.

facial deluxe
09-10-2003, 01:32 AM
200% agree, simple and efficient :)
Many other softwares have this solution.
Plus, we already have this kind of "double circle area" when using fog, if I'm no wrong....

Exper
09-10-2003, 04:38 AM
Maybe a crazy idea:
NT can implement an envelope with X=distance and Y=Intensity (or a gradient as well). ;)

This is a simple, fast and generic approach to create any kind of falloffs (realistic or not)!

Bye.

Exper
09-10-2003, 07:05 AM
Elmar...
I'm totally agree: LW absolutely needs a "Light"- plugin- class!
But I'm afraid... this one could require a complete re-write of the renderer! :(

Bye.

Elmar Moelzer
09-10-2003, 07:42 AM
I dont think that anything would require a complete rewrite.
Even though a major rework of the renderer might be a good idea anyway.
CU
Elmar

Exper
09-10-2003, 07:50 AM
Originally posted by Exper
Maybe a crazy idea:
NT can implement an envelope with X=distance and Y=Intensity (or a gradient as well). ;)

This is a simple, fast and generic approach to create any kind of falloffs (realistic or not)!

Bye. I'd like to see a comment about that one! ;)

Bye.

Elmar Moelzer
09-10-2003, 08:37 AM
Well, I would make an even better implementation of that and do a lightshader- plugin- class. This would allow for even weirder things, like say change of the light- color based upon the angle in which a surface is hit (ok, one can do that with a surfaceshader too, but setting that one up for all surfaces would be a pain).
There are a few other things that could be done with that, like textureable lights (imagine a volumetric texture on a light that would be cool, eh?). You get the idea?
CU
Elmar

Exper
09-10-2003, 09:19 AM
Elmar... you're right!
I'm totally agree... LW absolutely needs a "Light" plugin-class.
I'm thinking about new kind of light sources like: Rectangular Spots (parallel and conical) and Parallel Circular Spots (like in the old Impulse's Imagine software... did you ever use it?)!

But... I'd propose an envelope (and/or gradient) for falloffs as a fast thing to implement... so we can have something new in the forthcoming LW[8] (better than nothing). ;)

Bye.

Pavlov
09-14-2003, 05:58 PM
Hi,
i agree on Light-plugin class, it's nearly the same of the rendering engine plugin class that some (me for first) are asking.
Generally speaking, i'd like to get a large number of specific plugin classes: Antialias class, light class, rendering class, camera class, flare/volumetric class and so on.
Having a lot of "small" plugin classes would imho make developer's life easier (i.e. it would not be necessary to write a render engine just ot ge better antialias or new camera types).
I hope NT will consider seriosly these architecture/api questions.

Lighting is my day-bread... among all other limitations here mentioned, there's the limited range question: like pointed out, an envelope would be very useful. For example, some time a light needs to be in a place to get correct shadows, but light should become to act on objects just at a certain distance.
Generally, every option should be available for every light type; not having falloff and shadow map for distant lights has always been a pain.

Paolo Zambrini

Elmar Moelzer
09-14-2003, 06:16 PM
Pavlov, I might missunderstand what you are trying to say, but there is aleady a volumetric class in LW.
All other classes seem pretty reasonable to me.
CU
Elmar

jb_gfx
09-14-2003, 10:34 PM
There is some really interesting comments there. Btw Matt great listing and description of the weaks points in LW's lightning engine, you did a nice job.

Pavlov
09-15-2003, 01:57 AM
Originally posted by Elmar Moelzer
Pavlov, I might missunderstand what you are trying to say, but there is aleady a volumetric class in LW.
All other classes seem pretty reasonable to me.
CU
Elmar

You're right, but that class cannot handle volumetrics for Lights, just HV, fogs and other. No plugin can handle or modify what's into Light's property panel.
What is docked into any property panel is never handled by a plugin... :(
I hope we'll soon see that *everything* is handled by a plugin class...

Paolo Zambrini

Matt
09-16-2003, 07:54 AM
thanks jb_gfx! :)