Dark glass in light bulb (I read before looking for answer)

[horacio]

Hi, I'm new here and testing Indigo since a week ago (and I'm really happy with it). I would hate to sound repeated, so I have looked for all around the forum and read all related topics to don't come and to do my question without trying to solve it by myself.

I'm testing lights, so I modeled a light bulb (all parts, including the filament), but in black environment I only can get dark glass or, at least, not showing all reflections that it should have. Besides, the light does not seem bright enough to illuminate the environmentm and it only lights in front of it, and the rest is too dark. I attach some shots with different tracing methods, and then, the description of procedures:

At day:




In dark environment:




About Glass bulb:
1.- It is a volumetric closed object.
2.- It has a real 0,5 mm of depth.
3.- It has asigned "Thin glass" (the default preset), with 54 of reflection and 100 of power (no transparent, no emitter selected).
4.- The original material in SU is white and has 20% of opacity.

About filament:
1.- It's an emitter, blackbody, 5000 of temperature, 0.0001 gain, 200 power (I have tested high values to get more brightness) and 17.5 efficacy.
2.- The original SU material is a sort of beige.

About procedure:
1.- I'm using Reinhard tonemapping with default values.
2.- In day using sun-sky, and in black environment, using SU background and black selected.
3.- In first shots (both day and night), using Bidi-MLT and in last ones using Bidi-PT. (I got Bidi-MLT, because give me transparency on shadow of bulb in day and in bulb itself at black). Other tracing methods give me no transparency in bulb and it shadows, as showed in second shots.

System:
- Windows 7, 64 bits
- SU 6
- Indigo 3.0.17
- Skindigo 3.0.14

Renders done with Bidi-MLT cooked for more than 8 hours, then I supose in dark environment will not get more transparency (or more brightness) on glass than what is already seeing there, I'm right?. So, what I'm missing?. Or is a filament a too small object to give more powerful light as in real life is?

Thanks for patience (for reading many times similar questions) and for help, if provided.

HORACIO

[Pibuz]

Hi Horacio!
In my opinion, your basic trouble resides in the fact that you should no longer use the Thin Glass preset (I bump Whaat to remove that from the drop-down list, if he is reading this): I have been warned by Lycium about keeping using it, for it can give non-real results (like yours).

You say that you have modeled an enclosed volume as a glass bulb: why didn't you use the default specular shader then? TG was only used to simplify calculations when dealing with bidimensional glass geometries (such as windows in architectural viz, basically): with curved surfaces is generally better to use the specular shader because it gives correct caustics, while TG doesn't. On the other hand, specular shader wont work if your glass is not an enclosed volume, so make sure it is actually closed. Since bidimensional glass surfaces don't exist in real-life, the TG shader can progressively be removed.

Morevoer, make sure you check the "glass acceleration" feature in the "advanced" section: this should make the calculations more easy to Indigo.

Since you're dealing with a complex light/material situation, I'd suggest you give up the GPU acceleration dropping the PT method, favouring the BiDir MLT. I have to say that PT method has recently been enhanced a lot even in difficult light situations, so there should be no harm in giving it a shot anyway.

[horacio]

Thanks, Pibuz, for your quick response. I understand than Thin Glass is a preset (an easy one to lead to confusion) to be assigned to single bidimensional geometries, but that was the only preset that allowed me to see a little bit of light through the glass.

I had tested also with Specular and Glossy Transparent (as default presets) and have done many test with different values, different IOR's, etc., and all what I get is something like this:



Or is that it still need maaaaaaany hours more to may see transparent glass?.

In this second picture I rendered the model with glass removed, and here you can see that the filament is supported on a sort of "Y" of glass, and as you can see it is correctly rendered, and shows a perfect transparency:



Anyway, there is still the problem of low brightnes in environment light and as seen in second picture light in sorrounding environment is still too low. So, actually I have two problems: dark glass and low brightness. Second one seems more simple, and more about correct assingment of materials and values than light itself, but I cannot figure it out.

[Pibuz]

Regarding the low brightness...man: you're lighting a lightbulb in a big space, I wouldn't expect it to be lit as in a full sun shade Maybe you've set the power correctly, that can be true, but as a photographer would do in real life, you'll have to change some camera parameters to see some significant lighting modification. You can do that directly while rendering: switch to Camera tonemapping and look for the ISO value, and increase it. The scene will become more bright, you'll see.

Second theme is the reaction of your glass lightbulb: I'd exclude the glossy transparent shader, for it's to make blurry glasses, stick to the plain specular shader instead. It may be rendering wrong for two reasons:
1. That is not a perfectly close geometry: check it carefully; you should try to render a simpler geometry made of glass, such as a cube or a sphere, to see if your workflow is correct;
2. The opacity is too high. Since you're using a preset, you should know that when you select it, the SU material's opacity automatically switches to a non-zero value (it might be around 20). The opacity value controls the amount of colour which is absorbed by the glass volume: check the base colour of the SketchUp material and make it a very light grey, and set the opacity value to 5-6, that should work.

[horacio]

You were right about camera settings for brightness on the environment. I wasn't taking care enough to that. Even of course I wasn't expecting a "sun light" in night, , but was meaning the common light environment on a small room of 3x4 mts. with a lightbulb of 100 watts, like now:



About transparency I'm afraid still not getting good results. I test a simple cube (normals facing outside), and of course, as expected, the result was a consistent and heavy (that I imagine, ) glass cube, and the result is this:



And then tried making it a thin-walled object (interior side with normals towards interior, and exterior with normals to exterior), with panes of 0.7 mm thick, and result was this:



Finally I returned to my original rounded bulb, which I'm sure it's closed geometry, and changed base color of material to a 204 RGB based, and give it 5 of opacity, and result is this:



I tried too turning opacity to 1, and result is the same. Of course, one more time I tested changing many values to see if some diference were noticed, and nothing.

Here's a 3 position based screenshot of model:



Currently I'm rendering bulb light with no walls around, and result is almost completly darkness.

Thanks, Pibuz, for help provided.

[horacio]

Sorry for double post. But finally, have reached desired results. Apparently it was definitely a mix of camera settings and materials settings best suitable for the case, specially for night shot, which was my original problem with transparency:

- Finally I modeled a single-sided object, not volumetric (the glass bulb itself) and asigned "Thin Glass" at its defult settings. The result was perfect for day and night.

Screenshots with the new best modeled light bulb:




Perhaps, "Thin Glass" is not as bad in some cases, Pibuz. Anyway, thanks a lot for your help.

[Pibuz]

Glad you solved this! It's nothing that I'm a fan of TG, or anything: I suggested you as I was told by Lycium I suppose the developer discourages the use of that shader since it can give non physical accurate results, and we all know how much realism and physical accuracy is important to Glare
BTW, I'm happy to know that the good old TG shader, though dismissed in the Indigo architecture, still does work right, and in some cases helps resolving issues!

[Whaat]

Good to see that you got your desired result using TG. I do think it is possible to create your result without using TG and have it 'physically correct'. I am not sure why your attempts were not successful.

I have been strongly 'encouraged' to remove the TG material from SkIndigo but I myself find it useful in many situations. One thing to be careful of is that it can cause excessive fireflies in some situations.

In the future, I will probably remove it from the material types list but possibly include it as a material preset instead.

[CTZn]

Hi whaat, on this topic I think we could ask about the TG status again because 3.2 shipped with "enhancements for null materials", off the top of my head.

I am thinking now that the width of the bulb was maybe under a certain limit, that would have explained why only a few rays made it out without being terminated. But the box situation shows otherwise.

This could be verified by scaling the scene up 10x, just in case. By default, the minimal thickness for a medium is 0.2mm (above is safer on my behalf).

[Headroom]

IIRC the parameter that may be in question here is the "ray_origin_nudge_distance". It's explained in the Indigo Technical Reference Document.

I am not sure, however, if it is a parameter that can be accessed through any of out exporters.

[CTZn]

The infamous ROND ! That was quite hypothetical though.

This or an issue in modeling leading to this situation, or... the best way to learn it is to share the skp stripped to the essential so users can inspact it for you.

Clearly the two last renders are in favor of the "thin glass" preset.