PDA

View Full Version : Output color space set to RGB - why renders are faster?



federicosibella
10-13-2015, 02:13 AM
Good morning,

I'm an italian vfx student and I'm actually following a Lightwave course. Honestly I'm more into digital compositing rather than 3D, but since it's a generalist course I've to dive into Lightwave. It scares me! A lot, but the keypoint of my first thread is all the mess around color spaces. My teacher explained that from version 11.5 the contrast's computations for adaptive sampling (no idea what do all these words mean, just refering what I've written down in my notes!) is calculated in RGB space and not in linear space as before. Here I didn't get what does "contrast's computations" mean and above all: Are All the other color informations calculated in RGB too? I thought the internal calculations of Lightwave were in linear color space! So confused about this! Another point pretty unclear is the relations between this innovation (Output Color space set to RGB) and the threshold value for camera sampling. Why should I reduce the threshold? Or even better: What the heck does the threshold do?

I should know more about this stuff because it woun't get better in future lessons

Thanks for your kindness and time, I beg your pardon for my poor english.

Regards

Federico

RebelHill
10-13-2015, 05:25 AM
https://www.youtube.com/watch?v=UgNB9tZWUmM

federicosibella
10-13-2015, 07:42 AM
Thanks so much for this great reply, wanna follow the entire tuts! :D

federicosibella
10-13-2015, 09:55 AM
I've followed the class about sampling and I've another doubt about why Dome Light (a sampled light by the way!) is behaving so strangely when it comes to cast ray-traced shadows over the surface of the model. It's like a patch-work of dark areas and I've no idea how to fix it! Even increasing the light samples under the light properties doesn't change that much! I come from C4D, mainly used for motiongraphics and it seems I've to restart from scratch my 3D graphic knowledge!

130320

stiff paper
10-13-2015, 10:28 AM
I've followed the class about sampling and I've another doubt about why Dome Light...
Those polys are completely black. I don't think that's caused by the light itself. When a poly renders like that it often means that the shape of the poly is too complicated to be resolved for lighting calculations.

In Modeler, try selecting one of the black polys and View>Hide Selected to hide everything else. If it's a complex shape, then it should be made up of many quads all joined together, not just one big complex shape. You can use Split Polygons (CTRL L) to cut up a complex shape into individual quads. (Select your polys, switch to points selection mode, select two points, CTRL L, deselect points, select two more points, CTRL L... and so on.)

Alternately, you could try just triangulating the poly (CTRL T). If the poly is very complex, it might get it wrong. Use Split Polygons to simplify it a little then try again.

Or, of course, it could be something else. It's hard to tell without examining the actual model.

federicosibella
10-13-2015, 10:57 AM
@Cardboard thanks so much for the explanation, the fact is that if I open it in C4D or Maya I don't have that issue, I can light the model easily! I tried to focus just on those polys, but they don't seem so complicated... well, there are huge polygons that cover small ones whit reversed normals...but why bad polys are well rapresented in other softwares?

stiff paper
10-13-2015, 11:05 AM
It might be because the C4D and Maya basic renderers are hybrid scanline/raytrace renderers? (I don't know, but I do know that a scanline renderer will handle non-planar polys that a fully raytraced renderer will be unable to render.) The LW renderer is fully raytraced and has a fairly low tolerance for tricky polygons (there's a new LW renderer coming in LW 2016).

Is that model downloaded from somewhere or did you make it? (If you made it... well done...)

Hold on. I'll P.M. you...

10-13-2015, 11:10 AM
That's not the best of places to start with that question: why isn't this a problem in other software?

The better is: why does other software let me make bad geometry and still work with it?

Cardboard had an answer that would get you where you want in LW. Because in LW, greater than four point patches/faces/polygons don't light well on occasion.

As to other fixes, lwcad has tools to make those kinds of polys into quads or tris.

Perhaps the next render engine will allow us to render 'misbehaving' polygons.

spherical
10-13-2015, 02:27 PM
That's not the best of places to start with that question: why isn't this a problem in other software?

The better is: why does other software let me make bad geometry and still work with it?

Exactly right. Much the same as I have cited regarding 3D slicers that will happily slice a completely Borked model and produce garbage when printed or milled, but the users of slicers that are smart enough to recognize a crappy model and balk a slicing it, complain that the latter is "broken" and file bug reports. Completely backward.


Perhaps the next render engine will allow us to render 'misbehaving' polygons.

Please do not ask for this. It's a slippery slope when applications are allowed to be "forgiving". Learn how to model correctly. It not only makes for more efficient renders but also applies to external operations like 3D printing, CNC milling and interchange with other applications in the pipeline. Garbage in, garbage out, garbage back in.

10-15-2015, 04:21 AM
Spherical,
For me, you just hit the 'ding-ding-ding(x5)' bell for me.

No, I don't want that kind of engine and you are right, don't even make it sound like a request.

Just got two 3d printers in and indeed, there are behaviors that have to be changed, for me, to get good prints out of these things. I fear that the older stl converter in 11.63 is part of my problem but the errors as reported by netfabb are all my doing in the modeling.



Federicosibella,
Did any of Cardboard's suggestion work for you? Curious as to your progress.

As an aside, there is a Cardboard on another site that is a whiz with cnc stuff and the like. FPVLab is a space I've seen him in. You wouldn't be the same? That would be an odd but cool thing were you. lol

Well, there goes my more random thoughts for the day.
:)
Robert

spherical
10-15-2015, 01:13 PM
Spherical,
For me, you just hit the 'ding-ding-ding(x5)' bell for me.

No, I don't want that kind of engine and you are right, don't even make it sound like a request.

:)


Just got two 3d printers in and indeed, there are behaviors that have to be changed, for me, to get good prints out of these things. I fear that the older stl converter in 11.63 is part of my problem but the errors as reported by netfabb are all my doing in the modeling.

I made a suggestion in the 3D printers thread. 11.6 STL exporter is fine in all of the versions. Never had a problem with them. But, I manually choose the triangulation method employed on the quads before using the converter. Sometimes, the default tri operation will produce extremely long, thin tris; which aren't good for slicers. Look for the Fast Triple Fan and Fast Triple Traverse in the Multiply > Subdivision > More... dropdown. Choose the one that looks the most even across the surfaces, then run the STL converter. I usually save what I call a "build" file and run this operation on it; preserving the quads in the original for easier modification later on.

Be careful with Netfabb. It tends to find stuff that isn't broken and fix it; thereby introducing errors in areas that had none. I never mess with it anymore and instead do the operation suggested in the 3D Printers thread.

gerardstrada
10-17-2015, 06:00 PM
First page of this thread (http://forums.newtek.com/showthread.php?119485-Why-is-the-quot-Adaptive-Sampling-quot-pass-so-slow) might help to understand what contrast threshold is, how it affects AA and why Adaptive Sampling is faster now.



Gerardo

stiff paper
10-19-2015, 08:18 AM
Did any of Cardboard's suggestion work for you?
I took a look at the model. It originally came from a C4D site and all of the black areas on the render were "inside out" boxes. Because they were sealed boxes, Unify wouldn't flip them. The easiest solution was to make all the surfaces double sided, which worked. C4D must make surfaces double sided by default.


As an aside, there is a Cardboard on another site that is a whiz with cnc stuff and the like. FPVLab is a space I've seen him in. You wouldn't be the same? That would be an odd but cool thing were you. lol
No, sorry. That isn't me. This is the only place where I'm Cardboard and I'm not actually anything else anywhere else. That's me... I am the mystery meat in the gas station sandwich.

I have no idea what that means...

Hmm. I can't work out if I'm happy or dejected that there's somebody else in the world who's so disdainful of the idea of picking the "coolest" and most impressive online name that they also chose Cardboard.

I'm going to have to change my name to Soil or Lint or something...

federicosibella
10-19-2015, 11:22 AM
You guys are really impressive! It's such a new world for me that I feel overwhelmed by the amount of informations! Wow... my comfort-zone lays around color keying, premultiplication and all that stuff, but I have to do some efforts and start learning! No excuses... like the ones I tell myself when workout time comes! @Cardboard has been so helpful, I really appreciated his suggestions! By the way I agree with you as far as "learning how to model" instead of relying on permissive softwares. Since I'm new in this community I take advantage of your expertise and I ask you how C4D is considered inside the cg community? I don't like who categorize applications, for instance digital compositors are all into nuke, but my mentor told me that Fusion or Shake are far more powerful and stable than nuke, but the market is oriented towards that product, so the artists have to follow this rule! And if you have time which are the greatest advantages of LW? I need some motivational reasons to jump out the comfort-zone mentioned before! :D

Thanks!

stiff paper
10-19-2015, 03:38 PM
C4D is well respected. It isn't used much (if at all) for Visual Effects work, and it's used very little for archvis, but it completely owns motion graphics work, and it deserves to. Nothing else even competes.

Personally? C4D is a nice piece of software. It's rock solid and reliable (compare it, for example, to Modo, which crashes five times a day). Its negatives are that it's very expensive, Maxon seem to be failing to develop it properly, and you can only really find motion graphics work if you use it as your main app.

You should be able to do almost anything you want to in 3D using C4D. But... expensive.

I think you might have been given some unreliable advice about Nuke, Fusion and Shake. Shake is now very old, and even back when it was still being developed it was never as powerful (or as good) as Nuke. The reason Shake became popular was that Nuke was not available to buy at that point, as it was Digital Domain's in-house compositing program. Fusion has suffered from a lack of development for several years now and will need a lot of time and effort putting into development before it can catch up to Nuke. Even so, Fusion is still very capable and for free it's an amazing bargain.

Nuke is the VFX industry standard because it's simply the very best at what it does.

LW's greatest advantage is that it's cheap, and very capable if you know how to use it.