All posts related to V2
By feynman
#376018
Hej,

I have a real world object, lit with an emitter (LED) inside a glass cylinder (inside regular, outside sand-blasted) and at SL 19 obtain a lot of noise; likely a caustics issue. There is only one emitter (sphere hidden from camera) in the scene and an HDRI environment. I have searched in other threads but only found suggestions related to architectural visualisation. What am I doing wrong here?

Thanks in advance!

http://bayimg.com/naFiCaafL
User avatar
By tom
#376088
If I get it right, you have an invisible emitter. So yes, you should make it Hidden from Relections/Refractions instead of making it Hidden from Camera. Because, your emitter is enclosed in a glass shell which still reveals it on reflections/refractions. Besides, why do you need this invisible emitter? It should be OK without it as it's not there in the real model. Noise is as expected because rough glass requires higher SL to clear up.
By feynman
#376091
Thanks Tom; the emitter sphere is hidden from view inside the model, that's what I meant; meaning if the glass shape is removed, it cannot be seen from most camera angles as it is inside that metal tube below the glass. That is what I found so surprising: the sphere, although hidden in the model AND hidden from... selected in Object Parameters > Appearance seems to interact with the glass.

I read somewhere on the forum that at SL 21, such noise will disappear? Or is that not true? I would have thought, compared to all these complex architectural visualisations with lots of emitters and glass most users are working with, my scene would be one to laugh at, in terms of there hardly being any complexity at all... well, one learns as one goes along : )
User avatar
By eric nixon
#376100
Light is reflecting off metal tube? is metal material very simple (not too smooth, not too bright), is emitter very simple. Are artifacts caused by coincident faces in lamp geometry?

maybe upload it.
By feynman
#376101
eric nixon wrote:Light is reflecting off metal tube? is metal material very simple (not too smooth, not too bright), is emitter very simple. Are artifacts caused by coincident faces in lamp geometry?

maybe upload it.
It's an IES/EULUMDAT emitter, so the light is emitted upwards only and cannot interact with that tube. But, thanks, I do a pack and go and upload it to a dropbox folder here https://www.dropbox.com/sh/ro9z6hvm4takife/bzs4-AtBYr
User avatar
By eric nixon
#376109
Ok, quick observations;

The ies file is a 'wallwasher' shape so it emits light upto 180 degrees, but off to one side, try without ies, maybe use a flat octagon pointing up.

The emitter geo has around 2500 poly's, try one with around 50, and use a sphere with even topology, not hundreds of stretched triangles at the poles.

The orientation of the emitter looks wrong, but I cant tell which way the ies is mapped, due to locked uv's, so I cant say for sure without testing further.

The glass geo needs to be remodelled/re-exported, the geo at the fillet is all wrong. (holes in the mesh)

There is no tube around the emitter? (the metal one you mentioned)

I would avoid force fresnel on glass (minor issue) and definately avoid additive mxm's like the white plastic.

In general it looks like your workflow isnt working so well, I would consider avoiding maxwell studio IMO. You need to see more the actual polygons esp when making lights, which is a tricky subject.

2hr sl20 1 machine (3.6ghz x12) added metal tube around a flat octagon emitter. Turned of 'on-the-fly' disp on the cord, if you want disp use pretess, its MUCH quicker.

Image

The main problems stem from importing cad data with bad mesh topology, IGS format is unusable, but things like STP convert well to poly's if saved in WRL. I use cad data from grabcad frequently, so I wouldnt change your modelling workflow, just the way it gets exported.
By feynman
#376112
eric nixon wrote:Ok, quick observations;

The ies file is a 'wallwasher' shape so it emits light upto 180 degrees, but off to one side, try without ies, maybe use a flat octagon pointing up.

The emitter geo has around 2500 poly's, try one with around 50, and use a sphere with even topology, not hundreds of stretched triangles at the poles.

The orientation of the emitter looks wrong, but I cant tell which way the ies is mapped, due to locked uv's, so I cant say for sure without testing further.

The glass geo needs to be remodelled/re-exported, the geo at the fillet is all wrong. (holes in the mesh)

There is no tube around the emitter? (the metal one you mentioned)

I would avoid force fresnel on glass (minor issue) and definately avoid additive mxm's like the white plastic.

In general it looks like your workflow isnt working so well, I would consider avoiding maxwell studio IMO. You need to see more the actual polygons esp when making lights, which is a tricky subject.

2hr sl20 1 machine (3.6ghz x12) added metal tube around a flat octagon emitter. Turned of 'on-the-fly' disp on the cord, if you want disp use pretess, its MUCH quicker.

The main problems stem from importing cad data with bad mesh topology, IGS format is unusable, but things like STP convert well to poly's if saved in WRL. I use cad data from grabcad frequently, so I wouldnt change your modelling workflow, just the way it gets exported.
Thanks for checking this. I used a default sphere from the library; I have now replaced with a low polygon sphere. The IES emitter was measured with an integrating sphere http://www.uv-groebel.com/pmu_ulbricht.php and is to the highest standard achievable. The LED it represents has an illuminance output cone of 120° with a fairly strong central bias (upwards). The blue IES dots in the viewport show that it is oriented correctly. I have turned off "force fresnel" on the etched glass MXM now. I re-exported the glass part from a solid modeling software and checked with JavaView (math. analysis software) that the integrity of the mesh is ok. Unfortunately, the issue persists. It is, as if the emitter is "shining through" whatever surface is in its way (the perforated thick "top" disk with its central 30 mm hole above the emitter). I deleted all unnecessary objects and materials and uploaded that new version to the same Dropbox location as before, if you don't mind looking at it one more time.

We have to use Studio, because for standard industrial design software such as Catia, Alias, Creo or NX, there are no plug-ins available. And the meshes output are usually good, because rapid prototyping machines require a triangulated mesh of this kind to work, too. Actually, in industrial design, where one has to constantly work with 2-3 CAD softwares, sometimes more, in parallel, having a standalone render package such as Studio is quite good, because then one has to deal with only one consistent render interface.
User avatar
By eric nixon
#376113
Yes the ies was actually symetrical, it was only the thumbnail indicated side-throw.
And the meshes output are usually good
But all the meshes in this file had the same problem?, ofcourse it only really matters for transparent geo and emitters.

I think that you need learn a 3D package so you can see what the polygons are actually doing, like I said cad data can usually be converted but it needs to be checked manually. And I also feel that you're better off learning auto-didactically, just get stuck in and experiment.
By feynman
#382381
Just to finish this old thread off, the oddest thing occurred: On the actual object, the light "floats" as the rendering shows, even though the emitter is buried http://3.bp.blogspot.com/-puTr7Sr1Kzk/U ... lights.jpg

Must be a most peculiar reflection/refraction having to do with the glass shape. Well, at least that tells me that rendering an IES emitter with Maxwell does actually work realistically, which is important when developing luminaires ; )
Will there be a Maxwell Render 6 ?

Let's be realistic. What's left of NL is only milk[…]