PDA

View Full Version : Nodes Look Fine ... But Not in F9 ...



JBT27
03-10-2012, 04:41 AM
I've added the bump outputs from a Hetero Terrain and Ridged MultiFractal to get some landscape detail, controlling the Bump Amplitude of both with Slope Gradients.

In VPR and Fprime, I'm getting something like I want and expect ... but in F9, it looks crude and nothing like.

I've tried several different cameras - it's not that.

I'd appeciate any ideas - clearly VPR and Fprime are interpreting it correctly, or as I would expect at least, but the main renderer is not ...

Thanks.

Julian.

JBT27
03-10-2012, 07:26 AM
Well, I give up ... for now.

There's five hours down the drain.

Hybrid MultiFractal behaves the same way - nice crisp bumps in VPR and Fprime, garbage in F9/F10.

I have tried Denis's Scalar Bump, which from a procedural is generating no bumps ... not sure about that, but probably misunderstood the use of the node ... vector scaling with normals from Spot Info. Nothing is giving me the result I see in VPR that I posted above ... except VPR and Fprime ... one of those frustrating wasted mornings/days ... :mad:

Julian.

JBT27
03-10-2012, 08:47 AM
At its simplest, you plug a bump output into the bump input of the Surface node ... don't you? And it just works:

http://www.youtube.com/watch?v=LHQqo-Hqk2M

Watching this from 2010, everything seems fine and as expected.

So one after another, I tried each of the 3D Textures' bump outputs plugged into the Surface bump input.

All display fine in VPR, but these render as garbage in F9/F10:

fBm Noise
Wood
Wood2
Turbulent Noise
Turbulence
Hybrid MultiFractal
Marble
MultiFractal
Ridged MultiFractal
Bricks
FBM
Hetero Terrain
Honeycomb

Again, VPR and Fprime are interpreting these correctly; I vaguely recall flagging this up when nodes first appeared in LW, and being told that bumps behave differently in nodal than they do in layered, so some such. Unless I'm missing something, I'd say this looks bust, or at the very least, unfinished ...

Julian.

bobakabob
03-10-2012, 09:04 AM
Which version of LW are you using? Have you tried not using nodes by comparison? Any samples you could post?

JBT27
03-10-2012, 09:14 AM
Using LW11, Build 2238.

Using bumps in the 'old' layered way shows matching results in VPR/Fprime and F9/F10 - the procedurals listed work OK as layers used the 'old' way.

That said, I tried using the Bump Layer node - got the same disparate results.

I can post a simple example - a ball is enough to show the difference - the lunar surface is 18Mb and something I'm doing for a friend, so disinclined to post that ... but I can run up another or similar.

Oddly enough, I just ran through some of the IFW2 nodes, bump output straight into the Surface bump input - works fine - same in F9 as VPR/Fprime.

Julian.

dwburman
03-10-2012, 10:47 AM
I did a little test here. I ran the bump out from from a few different 3D Procedural nodes into the bump in on the surface. I looked at them both in VPR and an F9 Render. I did have to change the color space on the viewport to get it to match VPR, but the level of bump detail was the same.

JBT27
03-10-2012, 10:54 AM
So my skewed bump renders are a colour space issue? Hmm ... I did faff with that briefly but got nowhere. I'm using the default install, so what did you do exactly to make it happen? Knowing that will make my day ... seriously, been a bad day!

Julian.

JBT27
03-10-2012, 10:58 AM
Hmm ... I'm changing display colour space in the viewer (for F9 renders), but the bump detail is still badly off.

Julian.

dwburman
03-10-2012, 11:15 AM
I don't know why the bump works here and not on your system. It is very strange.

I thought that maybe it was a surface sampling setting, but I changed it from the default of 8 to 1 and it didn't look different.

Have you tried trashing your config files (after backing up your current ones)?

JBT27
03-10-2012, 11:18 AM
Yep, it's very odd. I haven't done the config files trashing, but I'll try that now - might as well - it might even be that simple!

Julian.

JBT27
03-10-2012, 12:15 PM
RJJ of IFW shaders fame just emailed offering to take a look at a scene, and pinned down the problem ... I was upgrading my IFW sub, so much appreciated on that.

The problem was scaling - big objects. The lunar landscape is 22km square, and the sphere I was trying was 15km (I didn't realise that, but thank goodness because it pinned down the problem).

So some of the 3D Textures flake out on objects with a big size - should have remembered that one. It doesn't explain why VPR and Fprime displays them correctly though ... so something to be wary of there.

Also IFW2 Nodes do not have this problem, but then as RJJ pointed out, they have a better maximum range for noise textures!

Just going to confirm all this for myself, but that seems to be the answer ... keep it small ... or RJJ's suggestion to scale everything 10:1 or higher, presumably the node values. Scaling down is possibly more manageable and keeps the numbers small.

Julian.

JBT27
03-10-2012, 01:42 PM
Just another thought on this, given what RJJ's observation was. I scaled the test ball down, to 100m, along with the surface scale ... F9 still bears little resemblance to the VPR - better, but artefacted and mushy. I don't at what point these scaling issues kick-in ... seemingly quite low, and it's not consistent - some nodes are fine, others are not.

Not the first time I have had these troubles with nodal bumps, and that crosses from 9.x to 10.x and now 11.0 - not universal, but troublesome.

10.1 showed the same problem earlier, and that is referencing wholly separate configs.

The IFW2 nodes, again, are delivering though, so that's something.

Julian.

dwburman
03-10-2012, 02:07 PM
ah. good to know. Hopefully that will be improved on some how. I suppose you could render in FPRime if you had to.

JBT27
03-10-2012, 02:12 PM
Yes, though in general there is a problem here that needs addressing, as clearly VPR is not struggling with those big numbers ... neither is Fprime.

That said, frustrating day though it has been, I have been reminded that the LW scaling issue, which has been there all along, can bite one in the proverbial still, but that there are ways around it, and indeed the fine IFW2 collection, which is not troubled by this problem at all so it seems.

Julian.