PDA

View Full Version : Dielectric issues ?



Pavlov
05-23-2013, 10:40 AM
hi all,
anyone here gets nice results with dielectric ? on thick items i always get a poor result, with weak reflections and no fresnel.
Im mean, glass is nearly not transparent and very reflective at glancing angles, dielectrinc node alone seems to produce less dynamic results... kinda an air bubble. Absorption is nice but it affects color filtering, but how do you control fresnel or empower reflections ?

thanks,
Paolo

Andy Webb
05-23-2013, 12:00 PM
Yes I agree, I have thought for some time that the reflections on the Dielectric node were very weak.

It always seems to work on a sphere much better, but flat surfaces just do not seem correct to me, at least not what I would have expected.

I've gone back to creating some glass the old way, it seems to give a much more acceptable result.

dee
05-23-2013, 02:11 PM
hi all,
how do you control fresnel or empower reflections ?

For stronger reflection use Material Booster from db&w tools.

3dworks
05-24-2013, 01:39 AM
For stronger reflection use Material Booster from db&w tools.

i second this - works perfectly for architectural purposes! ;-)

djwaterman
05-24-2013, 02:25 AM
I'm pretty sure Dielectric needs some work, or a better tutorial, or an alternative node that is nicer to use with more control. I've been using it but it still feels like it should be better.

3DGFXStudios
05-24-2013, 03:05 AM
I use it all the time and get very good results with it. However a bit more control would be nice. A "reflection booster" for example.

Pavlov
05-24-2013, 03:09 AM
Reflection booster is here:
http://www.db-w.com/products/dbwtools/docs
nice tool, you can plug a material here, boost any aspect of it, and output again to Material.
Anyway it's not enough, i insist that Fresnel effect in Dielectric is too weak (if there's any), and anyway i need to control it.
Paolo

Jim M
05-24-2013, 06:11 AM
Well you can split the dieletric and multiply the reflection by any modifier, so you could use anther fresel or an inverse fresnerl or whatever you required.

Andy Webb
05-24-2013, 06:19 AM
I must remember to use Material Boost next time I use Dielectric. :D

Pavlov
05-24-2013, 07:03 AM
Jim - ok but... isnt dielectric supposed to provide realistic glass and transparent mediums ?
so why should i workaround it... imho it's up to NT to code something which works without too many fixes :)

Paolo

jwiede
05-24-2013, 07:12 PM
Jim - ok but... isnt dielectric supposed to provide realistic glass and transparent mediums ?
so why should i workaround it... imho it's up to NT to code something which works without too many fixes :)
Precisely so. If Dielectric's goal is to easily provide realistic glass-type materials, yet requires significant tweaking and modification to achieve acceptable "realistic glass"-like results, it doesn't appear to be successful at that set goal.

Tobian
05-25-2013, 07:51 AM
Just a simple question here guys: of you who uses Linear Workflow? Just curious, because I always used to find that true Fresnel values for reflection used to look weak, and I'd over-egg them, before I moved to Linear Workflow?

I do NOT like the way absorbtion works with Dielectric (an absorbtion of 2.... 2 Ducks? 2 taj Mahals? 2 metres?, nope not metres...) but that's a separate issue. I do wish Dielectric had more control and more features, such as adding thin film from the car paint shader, and a dye and forward scatter feature :)

Lewis
05-25-2013, 11:53 AM
Oh yes , definitely I'd like Dielectric to get more options for boosting reflections, spec and possible coloring without adding too much of tinting which absorption do and you can't get some colors with absorpition 'coz if you type in high values like 500 or so of something (whatevert that number is ;)) you get some colorshifts that totally change actual color you might be wanting. It might be techically correct (I don't argue on that) but it's too limiting so if you can't change this dielectric then make "Dielectric Pro" with all the tweaks/options :). If we can have 3 type of skin shaders why not more Dielectricts too ;). Also tehcnically correct in Biased renderering engine with no phisically correct lights like LW is not really a great description or feature to mention too often, if we can fake it to look good then so be it even if is not so much technically correct :).

3dworks
05-27-2013, 06:27 AM
Oh yes , definitely I'd like Dielectric to get more options for boosting reflections, spec and possible coloring without adding too much of tinting which absorption do and you can't get some colors with absorpition 'coz if you type in high values like 500 or so of something (whatevert that number is ;)) you get some colorshifts that totally change actual color you might be wanting. It might be techically correct (I don't argue on that) but it's too limiting so if you can't change this dielectric then make "Dielectric Pro" with all the tweaks/options :). If we can have 3 type of skin shaders why not more Dielectricts too ;). Also tehcnically correct in Biased renderering engine with no phisically correct lights like LW is not really a great description or feature to mention too often, if we can fake it to look good then so be it even if is not so much technically correct :).

100% agree!

allabulle
05-27-2013, 10:09 AM
me too.

Tobian
05-27-2013, 10:18 AM
So guys.. the Linear Colour Space thing...?

Pavlov
05-27-2013, 11:25 AM
Tobian, i always use Linear Workflow, sRGB mode :)
Paolo

Thomas Leitner
05-28-2013, 02:59 AM
Jim - ok but... isnt dielectric supposed to provide realistic glass and transparent mediums ?
so why should i workaround it... imho it's up to NT to code something which works without too many fixes :)


Precisely so. If Dielectric's goal is to easily provide realistic glass-type materials, yet requires significant tweaking and modification to achieve acceptable "realistic glass"-like results, it doesn't appear to be successful at that set goal.

Please show us your problems with dielectric. And tell us how you can make itīs settings easier. For normal glass you have to deal with four settings: color, absorption, ior and roughness. Maybe the way NT handles absorption can be done otherwise. But the results are acceptable "realistic glass"-like. We did some comparisons with photos of real objects and dielectric matched the look quite well (only LWs caustics doesnīt work very well, but thatīs another story).

ciao
Thomas

Jim M
05-28-2013, 03:08 AM
Jim - ok but... isnt dielectric supposed to provide realistic glass and transparent mediums ?
so why should i workaround it... imho it's up to NT to code something which works without too many fixes :)

I concur, but at the same time would suggest that Dielectric node does provide realistic glass.

Andy Webb
05-28-2013, 03:40 AM
I concur, but at the same time would suggest that Dielectric node does provide realistic glass.

I would agree that in certain environments, like a studio set-up, Dielectric used as glass works a treat.

However I've had so many other situations where I've had to revert to the old methods just to get better reflections.

I've had situations where a close-up render of say a wine glass, looks great, but put in the environment I need it in and all I get is its shadow,
reflections become almost none existent.

At times I get the feeling Dielectric looses its reflectivity the further away you get from the object, certainly at lowish light levels.

Either way, more control over reflections would be more than welcomed. :D

lardbros
05-28-2013, 05:57 AM
I always thought the Dielectric node was incorrect too... that is until I actually looked at a glass in a normal environment that is :D

Have a look at a glass in your kitchen tonight... there will be other aspects that make it look more real, aside from the reflections.
We're so used to seeing glass photographed for adverts, that our in-built assumption is that it reflects more than it actually does. We've been tricked by adverts and the like.

I was very surprised how dim my glass was at home, and actually... it's a very difficult subject to light in the real-world too... therefore it's tricky in LW as well :D

Andy Webb
05-28-2013, 06:17 AM
Real glass is probably less reflective than might be thought, that's for sure, and it may be that Dielectric is 100% accurate (though I need convincing), however for those situations where accuracy is less important than the effect, more control would be a great help not a hindrance. :D

Lewis
05-28-2013, 06:22 AM
I did say that I'm not arguing technical correctness but we need more options even if is not correct esp. 'coz LW render/lights aren't technically correct anyway so it might be hard to get some results in that mix/combo :). Afterall complete 3D is "fake" so why not give us more options :).

Andy Webb
05-28-2013, 06:29 AM
I did say that I'm not arguing technical correctness but we need more options even if is not correct esp. 'coz LW render/lights aren't technically correct anyway so it might be hard to get some results in that mix/combo :). Afterall complete 3D is "fake" so why not give us more options :).

Exactly :thumbsup:

On the subject of lights, isn't it about time we had technically correct lights? (probably should be a different thread). :)

Lewis
05-28-2013, 06:31 AM
Exactly :thumbsup:

On the subject of lights, isn't it about time we had technically correct lights? (probably should be a different thread). :)

Hehe if you ask me it was time years go but they always push it to "next time...." :).

Tobian
05-28-2013, 07:52 AM
Yes, I think Dielectric is largely correct, other than a couple of issues, but it's a lack of advanced features which is the problem (and not even just cheating!)

Currently with LW we have 'classic' 'hand-cooked' and 'materials' - Classic is invariably 'wrong' and difficult to make right, but people like that. With materials we have fixed functions, but they are hard to mix up, and there's also the issue of BSDF integration (optimised calculation of the shaders, as a unit) Which means that any tricks with material mixing or blending result in horrendous slowdowns in speed. This is largely the case with hand-cooked node materials too, as they have no BSDF integration at all. The Materials have also got a number of tools which aren't replecated elsewhere (like correct thin-film in Car paint) which would be pretty in any of the other shaders too. In all honestly I would rather they have basic energy conserving materials that perform correctly and more tweak-able hand-cooked ones with energy conservation built in, and more power features to do boosts.

The problem with making a fully PBRT system is that the whole pipeline needs to be addressed. If you have 'realistic' Lights you'll also need exposure control and colour temperature tools. Spec also really needs to be addressed. It's still needed, because getting a clean soft Spec hit from a very small distant ultra luminous source is problematic without very complex importance sampling (aka the dreaded spotchies/fireflies). Which is the problem unbiased renderers face all the time. One problem I have with Specular, especially on low turbulence surfaces / high glossy, is that if you say have a inv square light with a small radius and a large distance, you should get a bright clean, small, spec dot. What you get is a 2% bright fuzzy blob. Specular reflections get smaller, not dimmer. The softer the reflection you have, the more this ameliorates into a correct solution (because a very small dot spreads out into a larger fuzzier blob) but it's not really right. This is something I'd like looked into.

The thing with realism is it sucks quite often :D It's boring and or dark. Adding in cheats to help more often than not cause breaks elsewhere. This is one of the revelations of LWF, in that often you need to cheat a LOT less using that, because you get decent lighting and reflections working all nicely together with it!

Pavlov
05-28-2013, 07:52 AM
guys, it's not that is looks correct or incorrect... it doesnt have any fresnel effect :)
Reflection and transparency are pratically the same at any angle. It just CANT be correct, imho.

Paolo

dee
05-28-2013, 07:59 AM
I always thought the Dielectric node was incorrect too... that is until I actually looked at a glass in a normal environment that is :D

Have a look at a glass in your kitchen tonight... there will be other aspects that make it look more real, aside from the reflections.
We're so used to seeing glass photographed for adverts, that our in-built assumption is that it reflects more than it actually does. We've been tricked by adverts and the like.

I was very surprised how dim my glass was at home, and actually... it's a very difficult subject to light in the real-world too... therefore it's tricky in LW as well :D

Problem is, clients don't want boring realistic glass, they want this shiny outstanding stuff from adverts. :)

Tobian
05-28-2013, 08:15 AM
Pavlov, not sure what you're doing wrong, but it does have Fresnel effect, and it's quite apparent.

Dee, you tweak your lighting, and luminous geometry which is just as easy! :)

lardbros
05-28-2013, 08:37 AM
Problem is, clients don't want boring realistic glass, they want this shiny outstanding stuff from adverts. :)

Ha... completely agree!! :)

Just was surprised by my own observations after playing with the dielectric, and looking at the real thing.


I would want more control over my dieletric nodes too... in fact, it feels almost too simple, and I rarely use it because of this. :D

Tobian
05-28-2013, 08:58 AM
Ok quick example...

Simple Delta material with a Fresnel Function input. On the left is an inversed square falloff area light set to have a range of 1m and a intensity of 500. On the right a 1m card with luminosity of 500 and no diffuse.

The illumination on both sides is approximately the same, but the spec hit is VERY weak from the 'specular' component. In the second render I moved them both to 10m away. Again very similar lighting, but again the 'spec' is very dim. 3rd pic, I ramped up both to 2500 intensity (which caused blotching issues for the radiosity). again the lighting is similar, but the spec hit is way higher.

As a final test, I made the material the vanilla Dielectric shader, with a bit of roughness, and you can clearly see the Fresnel falloff, in both spec and reflection.

The only way to 'fix' the weak spec is by pairing a spec only light, but that can make everything slow!

Matt
05-31-2013, 02:28 PM
What you have to understand with Dielectric (or any of the physically based materials) is that they respond better to more physically based lighting. Inverse squared falloffs, correct power of illumination in the scene etc.

If you want to go outside of physical correctness, you can use the layered material editor.

Lewis
05-31-2013, 02:32 PM
What you have to understand with Dielectric (or any of the physically based materials) is that they respond better to more physically based lighting. Inverse squared falloffs, correct power of illumination in the scene etc.

If you want to go outside of physical correctness, you can use the layered material editor.

Ok that's cool but how do we setup Phisically correct lighting in LW when we don't have real world Light units nor Sky Light etc..f ? Can you share some scenes with us Matt to describe differences.

Thanks

Sanchon
05-31-2013, 02:45 PM
Right. Physically correct object and environment is the key to achieve realistic glass, windows. Each sheet of window glass have 4 mm of thickness. Window have two sheet of glass with 16mm empty space between them. Realistic sky have much more bright intensity than other parts of environment - use good HDR spherical map with luminosity 500-800% for reflections only.

114633

Use "partial internal reflections" for dielectric node when glass object have right thickness dimensions - works best with two sheet of glass. Add very little normal distortion. Don't use "exclude from vStack". Dielectric refraction index = 1.2, 1.3 works best for me in this situation.

Dim interior with bright curtains or something else and additional cylindrical map with tree line outside building can help to achieve interesting reflections. Tree line and other scene elements reflections should be much dimmer than HDR sky reflection - like realistic windows glass.

Matt
05-31-2013, 02:55 PM
Ok that's cool but how do we setup Phisically correct lighting in LW when we don't have real world Light units nor Sky Light etc..f ? Can you share some scenes with us Matt to describe differences.

Thanks

Inverse squared falloff on a light behaves how REAL lights work, the intensity values of course in LW are LW units, so you pump them to what looks good for what you're doing. By skylight are you meaning physically based sky model? You could use HDRI, SkyTracer 2 (with no clouds) or a gradient / image mapped to a Dome Light, LW has a lot of choices for natural looking lighting.

Another tip, if you want stronger reflections at the expense of physical accuracy, you can increase the refraction index, that will boost the side reflections more. Of course it means the RI of the glass will not be accurate, but it works.

Matt
05-31-2013, 03:00 PM
Also, look at real life, you don't always get those bright 'spec' looking highlights:

http://www.rachaelbriana.com/wp-content/gallery/abstract/wine-glass-reflection.jpg

http://eirikso.com/wp-photos/White-Wine-Reflection2.jpg

http://image.shutterstock.com/display_pic_with_logo/77952/77952,1268170972,3/stock-photo-studio-shooting-glass-bottle-isolated-on-white-48326290.jpg

http://image.shutterstock.com/display_pic_with_logo/85050/85050,1274790681,18/stock-photo-ketchup-in-glass-bottle-studio-isolated-53845849.jpg

It's all about lighting.

Tobian
05-31-2013, 06:26 PM
Matt. I do understand physically accurate surfacing. My images show how the inverse squared lights match quite well luminous geometry, with MC rad... ONLY for Diffuse, but not the specular hit. Spec for inverse squared lights is clearly wrong... Not sure how else I can explain it but with images?

Matt
05-31-2013, 06:26 PM
First image, Absorbtion 1.0, Refraction index 1.5.

114639

Second image, Absorbtion 1.0, Refraction index 2.0.

114638

Third image, Absorbtion 75, Refraction index 1.5.

114637

Scene file.

114636

Andy Webb
06-01-2013, 05:10 AM
Well Matt, you've sort of highlighted the very thing I mentioned earlier, in studio lighting conditions Dielectric works great.

I, on the other hand, have occasions where I may have for instance, a large banqueting scene, in a large space with stage lights, down lights, banners, all sorts in the room and the
table glassware just disappears due to the weak reflections.

I try to use Inverse Falloff ^2 as much as I can, it lights everything fine but doesn't have much effect on the Dielectric material.

If I can find an example I'll post it.

Even if it is correct, I would still prefer to have the option available to change things, the advantage of Dielectric over the old surfacing method is
that it is unnecessary to have an air surface, so to use the old method is much more time consuming.

Tobian
06-01-2013, 06:34 AM
Well inverse squared lighting has almost no effect on Dielectric because it doesn't have any diffuse component, and as I explained above, Spec hits from all lights are weirdly weak. If you want to use any 'lights' with the materials, make a clone of the light, and make it Spec only, and make it about 10-12x as bright. (or use a paired bounce card)

Studio lighting scenarios like that one are a really great way to get attractive lighting.. Of course there are no... lights.. just luminous geometry :) Materials+HDR+Luminous geometry+inv square falloff lights for diffuse work great, just not the spec part.

Celshader
06-01-2013, 11:56 AM
...the advantage of Dielectric over the old surfacing method is
that it is unnecessary to have an air surface, so to use the old method is much more time consuming.

No surface has required "air polys" since 9.3 introduced the Volume Stack. The Volume Stack keeps track of rays entering/exiting refractive surfaces, not Dielectric. As long as you have an enclosed surface (ex: a box or ball instead of an infinitely thin sheet of polygons), and as long as that surface is not double-sided or excluded from the VStack, you will not need air polys.

What Dielectric offers over "rolling your own" surfacing is a physically correct model. I do love what Dielectric does:

https://dl.dropboxusercontent.com/u/9264841/forums/beta/_Reference/Renders/Marble-Cracked-Cropped.jpg (http://forums.newtek.com/showthread.php?121339-Lost-these)

https://dl.dropboxusercontent.com/u/9264841/forums/beta/_Reference/Renders/motion_theory-tanqueray_3_glasses_ad_small.jpg


http://vimeo.com/40413253

Andy Webb
06-01-2013, 12:03 PM
I seem to have missed that one!! :ohmy:

I thought the Volume Stack only affected the Dielectric node :(

Celshader
06-01-2013, 12:08 PM
I seem to have missed that one!! :ohmy:

I thought the Volume Stack only affected the Dielectric node :(

"Exclude from VStack" defaults to "on" when loading older models with surfacing created prior to 9.3. It was meant to soften the transition from the "old" method of using air polys, since everyone and his sister had been using air polys for refractive surfaces since the 1990's and had a ton of legacy assets that relied on the older technique. However, it's possible that the transition was TOO soft -- few artists noticed that 9.3+ surfaces no longer required the weight of "air polys" in order to get accurate refractions in a scene.

It wasn't just you -- I heard about a guy at R&H who sincerely believed that "Exclude from VStack" had something to do with volumetric lighting. :ohmy:

DonJMyers
06-01-2013, 12:30 PM
Once again celshader saves the day!

stiff paper
06-01-2013, 01:38 PM
Well... okay. It's only embarrassment... so what the hey...

It's been a long, long time since A-Level Physics, but... doesn't the inverse square law only apply to a point source? That is, for the specific case where individual photons (let's not have that discussion) are emitting from a single point and therefore are, in effect, spreading out highly predictably with the distance traveled from the source. Or, in other words, fewer photons per fixed area for increases in distance.

I've always been under the impression that the more coherent or focussed light is, then the less that an inverse square falloff holds true. Mainly because the photons (ahem) aren't spreading out at all in the same way. I could be very wrong. Hence potential embarrassment.

But coherent and focussed would include any spotlight (and lasers, of course).

Lewis
06-01-2013, 02:23 PM
That's great test scene Matt BUT now go and try to make Bright fully colored/tinted ORANGE with that setup/Dielectric only NODE

Like this:
http://fc00.deviantart.net/fs7/i/2005/164/b/e/Torus_Knot___Orange_Glass_by_Hafunui.jpg
or this:
http://www.google.hr/imgres?imgurl=http%3A%2F%2Fimg1.etsystatic.com%2F0 00%2F0%2F6164032%2Fil_fullxfull.330127729.jpg&imgrefurl=http%3A%2F%2Fwww.etsy.com%2Flisting%2F97 950511%2Fvintage-orange-glass-bowl-glass-dish&docid=8Ko7aXZfG0Gn3M&tbnid=bl3Oxq4nLdMibM&w=1360&h=1154&ei=8VWqUYOtFMn24QSLxYDYBg&ved=0CAcQxiAwBQ&iact=ricl

In your scene try this: Go to color of Delectric, pick Yellowish/Orange color (255/100/0) then try various absoprtion levels. At 10 - it's too weak and not orange enough, if you add more absoprtion at Absorption 30 it'll become pure RED (WTF?) so you can't really get pure vivid Orange. Happens with some other colors too, they all seem to shift to some color when you add more abosrption and on low absorption level they arne't "strong" enough.

lardbros
06-02-2013, 01:48 PM
Nice find there Lewis... never realised that before.

Just tried quickly to reproduce your error, and it seems that to get the orange you need to fiddle with the hue until it's pretty much yellow. The higher the absorption, the less control you have in the hue.

It gets to a point when Absorption is at 16, and in order to make it Orange the Hue has to be on 041, which is actually pretty yellow... if I punch the absorption up even more, it ends up red again. Very odd indeed... and most definitely wonky!!



Oh Matt, nice photos of glass... also very nice caustics!!! ;) Hint, hint!

Lewis
06-02-2013, 04:22 PM
Nice find there Lewis... never realised that before.

Just tried quickly to reproduce your error, and it seems that to get the orange you need to fiddle with the hue until it's pretty much yellow. The higher the absorption, the less control you have in the hue.

It gets to a point when Absorption is at 16, and in order to make it Orange the Hue has to be on 041, which is actually pretty yellow... if I punch the absorption up even more, it ends up red again. Very odd indeed... and most definitely wonky!!


Well yes you can fiddle with it and get somethign OKish eventually but if is phisically accurate why does Absoprtion completely changes some colors (it's not just orange try soem other colors and you'll get some totally weird color shiftws with high absorption levels). Why does 255/0/0 work fine as RED on all Absoprtion levels be it 1 or 10 or 50 or 250 and doesnt work same for Orange (just single example, not isloated case/only color with this problem) and already at 20 it's completely shifted/tinted differently ?

lardbros
06-02-2013, 04:57 PM
I was completely agreeing with you Lewis :D
It's not working as you'd expect, so hopefully Newtek can sort this out for the bugfix?! :)

Pavlov
06-02-2013, 06:55 PM
I noticed some strangeness, precisely:
- flat glass sheet - i.e. stair parapet. Glass border in most cases looks darker, Dielectrc show it bright in some places, even if there's nothing to refract/reflect.
- I tried to simulate plexiglass chair, properly modelled , i.e. http://www.100matrimoni.it/uploads/kartellluisghost.jpg and results are not 100% convincing, you get too little reflection with correct parameters (plexi has IOR 1.25). Also, getting dark edges where plastic is thicker was tricky at best. At last i got something i like, but i had to use material multiplier to boost reflection otherwise it looked like a soap bubble... lighting is correctl sunsky only and a pair of inv2 falloff ara into lights.
114675

Paolo

Andy Webb
06-03-2013, 03:27 AM
I noticed some strangeness, precisely:
- flat glass sheet - i.e. stair parapet. Glass border in most cases looks darker, Dielectrc show it bright in some places, even if there's nothing to refract/reflect.
- I tried to simulate plexiglass chair, properly modelled , i.e. http://www.100matrimoni.it/uploads/kartellluisghost.jpg and results are not 100% convincing, you get too little reflection with correct parameters (plexi has IOR 1.25). Also, getting dark edges where plastic is thicker was tricky at best. At last i got something i like, but i had to use material multiplier to boost reflection otherwise it looked like a soap bubble... lighting is correctl sunsky only and a pair of inv2 falloff ara into lights.
Paolo

This is also my point, Dielectric in these sort of scenes just doesn't reflect enough.

Andy Webb
06-03-2013, 03:55 AM
I've hunted around for an example and found this very old partially done scene, not the best of lighting I grant you but it does show how Dielectric doesn't reflect enough.

I'm only referring to the win glasses here.

It has crossed my mind that lack of reflections could be because of the use of the VStack, no back faces to pick up reflections, we are only seeing reflections on the front face.

Anyway, I've rather overdone the boosting and the scene certainly isn't great.

As you can see from the shadows, there is a strong back light, the photograph on the end is from a recent wedding, and that also has a strong back light from a window but
the reflections are there.

Also, is it me or is there not much Fresnel on the wine glasses?

114683

Pavlov
06-03-2013, 04:20 AM
no, it's not you. As i said at the beginning, fresnel and reflectons are weaker than natural glass, imho.

Paolo

Andy Webb
06-03-2013, 04:58 AM
I do have other examples but I can't post them :(

Danner
06-03-2013, 06:33 AM
To get the dark edges on flat glass panes, like the ones on tables and modern showers I just change it to a different surface, I make it non transparent, saturated darkish green and give it some reflection, it renders faster and I can control it separately. I know it's a hack, but it's one of those that are win win.

Tobian
06-03-2013, 09:53 AM
Be aware if you're multiplying the reflection channel, then it's going to push it outside of the Energy conservation space, so that's not the best method for boosting reflections (if faster to set up).

I think the confusing issue here is that we don't really want to 'boost' the reflection so much as add another reflective layer, or customise the Fresnel function. Most plastics are finished with a secondary laquer, so that boosts their reflectivity, but you need to be careful not to have it reflect more than 100% (Andy, your reflections look REALLY wrong in the second image). The thing about the reflectivity is we need the ability to alter the facing reflectivity, not ALL of the reflectivity, because at glancing angles the reflectivity should not exceed 100%. If you have to do that, then you need to re-look at your scene as it's not right. It has to be said not all materials have the same Fresnel falloff per value: different materials have differing polarisation axes, which can alter the shape of the curve. Ideally it would be good if we had a custom database of them, or the ability to modify it ourselves. Also a lot of glass has a metallic layer coating, either to be decorative, or in the case of window glass, to stop so much sunlight penetrating inside (all glass buildings can get too hot!) This again pumps up the reflectivity, and lowers the transmission, which is what it's for. It's just expensive to add a second reflection shader to LW materials, without BSDF integration (or just cheating and altering the Fresnel to be custom)

Dielectric is fine as a glass and simple water shader, but it's missing a few things, it needs a customisable reflection, and scattering. It's the scattering function which makes plastics look so colourful: Light particles reflecting off dye particles in the plastic, not just light being absorbed by the material. I think it's why Absorbtion may not be 'wrong' so much as it's only half the story. But I do agree Lewis, It's very hard to control and get the right colour of glass, especially for thin glass. I tend to just make a refraction shader with a colour tint instead, as it's less fuss than making the dielectric work properly.

Pavlov, your image looks gorgeous btw! :)

Andy, I do want to point out that in your reference image, there's clearly some kind of studio lighting, or very high contrast bright areas in the scene, because the highlights are very bright. Your render doesn't look like it has any HDR elements to provide such luminous reflections. Even in an interior, you have to understand the contrast ratio difference between interior and exterior illumination means that a bright window would count as a bright source, by the time you raise the contrast in the dark interior. Because LW doesn't have exposure control or realistic light sources, people tend to just normalise the lighting too much in LW. Really you should be having your sun a LOT brighter than interior lights, then fiddling with the exposure, and tone-mapping the results. I suspect a lot of the problem with that scene is because the other materials aren't energy conserving, and the glass.. is. Really you should be making the rest energy conserving, and you should get better results. I could take a look at the scene if you want me to make any suggestions on how to improve the appearance?

As another note, often glass has a dark green edge. Quite often this is actually just green paint, in the real world, faking it, to get that powerful of an FeO absorption look.

Andy Webb
06-03-2013, 10:25 AM
I did point out that the boost was too much, but I was trying to show that there was plenty to reflect it just wasn't showing up.

Also it is an interior with a brightly lit stage area, no windows or sunlight. It's far from being the best lighting for this but I just dragged out an old set-up to show the issue.

Tobian
06-03-2013, 11:00 AM
Yah I am just pointing out other things you could do to help with the issue. Generally, a lack of contrast ratio in materials and environment, and non-linear workflow make reflections in PBRT materials look weaker.

As a note about the inverse power law, yes, Cardboard, you're right that the falloff would be based on a nominal zero dimensional light, (not that such things exist in nature). The falloff in lighting is not because the energy loses energy, but because of the increase in surface area relative to the emission object's surface area, so there are less photons to go around, the further you are from an emissive object. It's only really an approximation, as, yes, in some special cases, if the photons are concentrated over a very small area, then you will have less falloff, and in the case of a Laser near zero, by design. That's why with IES lights the web is measured, and not just assumed to be inverse squared falloff.

Matt
06-03-2013, 11:20 AM
@Lewis - Of course Absorbtion will affect the colour.

Matt
06-03-2013, 11:33 AM
Like this:
http://fc00.deviantart.net/fs7/i/2005/164/b/e/Torus_Knot___Orange_Glass_by_Hafunui.jpg

Does this even use absorbtion, there is no colour shifting at all. Also, not seeing any strong reflections in that example either, just refraction.



or this:
http://www.google.hr/imgres?imgurl=http%3A%2F%2Fimg1.etsystatic.com%2F0 00%2F0%2F6164032%2Fil_fullxfull.330127729.jpg&imgrefurl=http%3A%2F%2Fwww.etsy.com%2Flisting%2F97 950511%2Fvintage-orange-glass-bowl-glass-dish&docid=8Ko7aXZfG0Gn3M&tbnid=bl3Oxq4nLdMibM&w=1360&h=1154&ei=8VWqUYOtFMn24QSLxYDYBg&ved=0CAcQxiAwBQ&iact=ricl

Do you see the strong red parts in this example? It's shifting colour, just like it should.

Matt
06-03-2013, 11:45 AM
Also, Absorbtion has no units.

From Wikipedia on Absorbance:

"Although absorbance is properly unitless, it is often reported in "Absorbance Units" or AU. However, many people, including scientific researchers, wrongly state the results from absorbance measurement experiments in terms of arbitrary units"

lardbros
06-03-2013, 12:10 PM
Maybe the shader could be altered so that the colour you actually choose for absorption is the colour you get though? Otherwise, having a high absorption when trying to get orange and getting red, is pretty frustrating creatively.

http://retroartglass.com/store/image/1wiib/Glass_Archives_Vintage_Retro_Orange_Glass_Kitchen_ or_Bar_Canister_Set.jpg
Yes, orange glass does appear to go red with the higher absorption, but is there any way it could be easier to control?

I'm sure there must be some fairly simple maths that could determine the absorption and hue, and make sure the colour at the least absorped part could be the chosen colour in the colour picker?! This would make it much easier for me to use, and others I'm sure.
All about getting the results you expect, not JUST physical accuracy.

:D

Not having a moan, just little things like this add up during production and can make the work tedious.

Lewis
06-03-2013, 12:12 PM
Do you see the strong red parts in this example? It's shifting colour, just like it should.

Nope not same thing, try to mimic that in LW with dielectric node and it'll either be too weak in tint/color or orange will go completely RED. We need sommethign in middle. IMHO Dielectric should get few more inputs like CarPaint shader has RealFresnel, paint/spec etc. etc. If NT is going to be stubborn and not try to adjust/update it to users wishes that's fine (it's your software you can do whatever you want with it :)) but please then make "DielectricPRO", "DielectricFake" , "Dielectric2" or whatever name it shold be and give us wanted options :p. Thanks.

P.S. I'll never understand why do you guys sometime have to "fight" / "defend to the death" some features what clearly are at least in grey area (Instancing comes to my mind instantly) if so many users complain. Even if we all are wrong what's the problem with makind/adding more Artis friendly updates i.e. add some advanced feature if users so badly want it ? What's the harm there? I bet if would be less work to add/program it and say "her eit is as you users wanted :p" then spend days and days "defending" something that's not defendable from users POV :D. Aren't we all on same side here wantign to have better features/options in LW ?

Matt
06-03-2013, 12:17 PM
Nope not same thing, try to mimic that in LW with dielectric node and it'll either be too weak in tint/color or orange will go completely RED. We need sommethign in middle. IMHO Dielectric should get few more inputs like CarPaint shader has RealFresnel, paint/spec etc. etc. If NT is going to be stubborn and not try to adjust/update it to users wishes that's fine (it's your software you can do whatever you want with it :)) but please then make "DielectricPRO", "DielectricFake" , "Dielectric2" or whatever name it shold be and give us wanted options :p. Thanks

It's not about being stubborn at all Lewis. Just trying to make people aware of how a feature works. If you want fake, you have the standard layer editor.

Tobian
06-03-2013, 02:38 PM
Yes but Matt, I've also explained several scenarios where it could use more functions, such as an in-depended reflection and refraction blur. Thin film coatings, wax / polish coatings, scattering and sss, dye layers, metal coatings. All of those would add to the realism, be correct in physics and are REALLY hard to do (read impossible) Using 'Standard layers' editor.

Matt
06-03-2013, 03:12 PM
Yes but Matt, I've also explained several scenarios where it could use more functions, such as an in-depended reflection and refraction blur. Thin film coatings, wax / polish coatings, scattering and sss, dye layers, metal coatings. All of those would add to the realism, be correct in physics and are REALLY hard to do (read impossible) Using 'Standard layers' editor.

Well, that's a different request than "Dielectric has issues / isn't working properly" which this thread was eluding to.

Tobian
06-03-2013, 03:34 PM
Well it is and it isn't. While some of those features are quite exotic, some are basic.. Like the ability to have different reflection and refraction blur. More complete Fresnel options, (because not all materials have the same shape Fresnel curve for the same refractive index), or a material library would also make it more complete for 'simple' glass. As I also showed earlier, Spec is weird in general for materials, but I think that's just as much of an issue of Lights as it is materials.

Matt
06-03-2013, 03:55 PM
Had a crack at the orange bowl. Without matching the lighting exactly, clearly it will be somewhat different than the photo. But I have to get on. Hope it helps a little.

114697

Scene file:
114696

MSherak
06-04-2013, 03:50 AM
Had a crack at the orange bowl. Without matching the lighting exactly, clearly it will be somewhat different than the photo. But I have to get on. Hope it helps a little.

114697

Scene file:
114696

Excellent! Thank you..

allabulle
06-04-2013, 04:04 AM
Thanks Matt. I'll give it a go too. I'm having trouble sometimes with absorption and colours too. Maybe some good brand new presets for a future LightWave release would help this kind of situation. Along a brief (or not so brief) explanation on it's proper use, perhaps. For now, thanks to chime in.

lardbros
06-04-2013, 04:52 AM
Cheers Matt... will check this out when I get home.

Very nice example though :D

djwaterman
06-04-2013, 05:00 AM
Well just to re-direct this thread a little, I know it's not the issue under discussion, but what Tobian said made me think that yes, we really need a node that can give us double reflection layers.

Awesome orange bowl Matt.

Lewis
06-04-2013, 12:46 PM
Good one Matt, thanks for scene.

For your result well I'm pleased with what you get except that black edge like i said to you on skype. Everythign looks good to me i just don't like black edge. I don't have orange bowl at home to test in reality but i do have Bleu and it's not having so black-ish edge :).

I rendered your scene last nigh as small/short animation and somethign got me thinking that this black edge has osmethign to do with glancing edge/reflection, Check this and see how when camera moves more up of bowl those edges suddenly change it's darknes very much (in just 2-3 frames) and become more realistic i.e. blue instead black.
https://www.dropbox.com/s/96v97kk6doqcqgt/Orange_Bowl.rar

Saying all that i still thing we shodul get much more Adjustable "Glass node" or whatever name it could have as I said in my very first post.

XswampyX
06-04-2013, 01:22 PM
I have to agree with Lewis. The edges are just too dark.
Here's the scene rendered with another renderer. You can clearly see the difference on the edge of the bowl.

http://i465.photobucket.com/albums/rr16/xXswampyXx/edge_Render_zpsfd7bdf67.jpg (http://s465.photobucket.com/user/xXswampyXx/media/edge_Render_zpsfd7bdf67.jpg.html)

Lewis
06-04-2013, 01:30 PM
I have to agree with Lewis. The edges are just too dark.
Here's the scene rendered with another renderer. You can clearly see the difference on the edge of the bowl.

http://i465.photobucket.com/albums/rr16/xXswampyXx/edge_Render_zpsfd7bdf67.jpg (http://s465.photobucket.com/user/xXswampyXx/media/edge_Render_zpsfd7bdf67.jpg.html)

Thanks man, finally somene who understand my "issue" with this dielectric color shift and problem i find unrealistic regardless of light conditions/environment as Matt suggested it has to be 100% match to photo to be realsitic.

I wonder what's next now i.e. why does Octane look better/more realsitic if LW Dielctric is 100% accurate scientifically/phisically.

Matt do we have slight bug in Dielectric afterall or ?

XswampyX
06-04-2013, 01:40 PM
I think it's too dark because it doesn't take into account any light rays coming in from the sides, or top of the bowl. It seems to trace the thickness of the material as though it was in a solid block of the same material. Mathematically correct maybe, but only in one aspect?

Just a thought.

Tobian
06-04-2013, 02:10 PM
Hmm I am not sure it is too dark.. the LW version just looks like it has a more strong absorption effect, so it's more orange and darker in the deep ray areas. Increase the Octane absorption so the colour matches the LW one, then we can be sure.

Lewis
06-04-2013, 02:24 PM
Hmm I am not sure it is too dark.. the LW version just looks like it has a more strong absorption effect, so it's more orange and darker in the deep ray areas. Increase the Octane absorption so the colour matches the LW one, then we can be sure.

That's the thnig if you decrease absorption to somethign low (like 10 or 20) then bowl is not even orange anymore and if you want it to be this deep orange (which Matt got it right) and set Absorption like Matt did to 150 then you get black edge. Seems that somehow we can't have both in same time (deep orange bowl color and brighter edges).

In Matt'' scene i like edges with Absorption of 100 but i like bowl with absorption 150 and don't like edge darkenss then, seems weird, almost like i'd have to make two separate materials, one for edges and one for bowl itself , which I often did in past like workaround for that specific problem :).

Pavlov
06-04-2013, 02:27 PM
dark edges are same issue i get in my picture. If you look, chairs have quite dark edges while the rest of the chair is barely visible, and i also boosted reflection with DB-W surface booster.
If i wanted to make chair more visible, i get black borders. Since i have that chair and i know plexyglass IOR, i'm sure it doesnt behave that way.
Maybe dielectric works nicely on organic stuff, but for flat, thick surfaces it can show limits.

Paolo

Iain
06-04-2013, 02:53 PM
I think you could argue either way here. I don't see much wrong with Matt's render or XSwampyX's. They both have really strong points and slight flaws.
Using different renderers might just complicate things.
If you search for images of coloured glass bowls you will see a huge variation in edge darkening. It's affected by all the physical characteristics of the material as well as direct and ambient light.
I have had issues in the past with Dielectric and its lack of controls but it is a fantastic short cut tool in a renderer that does not claim to be 'physically correct'.

Iain
06-04-2013, 02:57 PM
Absorption of 100 but i like bowl with absorption 150 and don't like edge darkenss then, seems weird, almost like i'd have to make two separate materials, one for edges and one for bowl itself , which I often did in past like workaround for that specific problem :).

I dont think that is weird in cgi. Maybe renderers are going more in the direction of cameras but we're a long way from that yet.

Tobian
06-04-2013, 03:08 PM
I think that the darkening effect occurs in nature. Personally I have seen it so I don't think it's so wrong. Xswampy's render does not have as high an absorbtion to match the intense dark orange in Matt's - if he cranks it up, I suspect that the edge will get just as dark. Observationally I think it will.

The problem is that in plastics you often have multiple layers and other optical effects going on, such as scattering, which gives transmissive effects, or causes the edges to appear to glow. This is also caused by UV transforming pigments, which can cause them to appear to be non-energy-conserving (they absorb UV light and give out visible light). Coloured glass is often just a stuck on cellophane coloured film, so it gives a constant thickness effect, which absorption won't do (hence dye layers). So really we could use a dedicated plastics material, especially because that's the type of work many LW artists do (plastic bottle production shots).

XswampyX
06-04-2013, 03:54 PM
I've tried to match it Tobian, and if you only use absorption to get your colour then yes it looks approximately the same. :)

Maybe I'm just not used to seeing quality glass.

So how do we get cheap glass with the dielectric node?

http://i465.photobucket.com/albums/rr16/xXswampyXx/LW_Glass_Vs_Octane1_zps3a80de36.jpg (http://s465.photobucket.com/user/xXswampyXx/media/LW_Glass_Vs_Octane1_zps3a80de36.jpg.html)

Tobian
06-04-2013, 04:09 PM
Yeah much as I thought. LW's model is correct, it's just really really tricky to control and make right, because there's no scale (which as matt points out is an arbitrary number but many other apps have one anyway). I agree XswampyX, what we need is a plastics shader really.

bobakabob
06-04-2013, 04:21 PM
I think you could argue either way here. I don't see much wrong with Matt's render or XSwampyX's. They both have really strong points and slight flaws.
Using different renderers might just complicate things.
If you search for images of coloured glass bowls you will see a huge variation in edge darkening. It's affected by all the physical characteristics of the material as well as direct and ambient light.
I have had issues in the past with Dielectric and its lack of controls but it is a fantastic short cut tool in a renderer that does not claim to be 'physically correct'.

Agreed, a simple Google search for images of glass bowls reveals extreme variations from intensely black edges to subtle transparency. There are so many parameters involved. CG is just an illusion after all and at times even the real world can look fake :)

Dave Jerrard
06-04-2013, 05:56 PM
A few things:

First, sure nodes, and in particular, the Material nodes, have some problems, but of those, Dielectric is actually pretty error-free. The problems I've seen here all stem from user error.



hi all,
anyone here gets nice results with dielectric ? on thick items i always get a poor result, with weak reflections and no fresnel.
Im mean, glass is nearly not transparent and very reflective at glancing angles, dielectrinc node alone seems to produce less dynamic results... kinda an air bubble. Absorption is nice but it affects color filtering, but how do you control fresnel or empower reflections ?

I always get what I want, or very close to it.

Fresnel is controlled, by the IOR of the surface (or more accurately, the differences between the two IORs at the interface between two materials). The higher the difference, the more reflective the surface. Dielectric does this correctly. It's the one thing it does right and it does it well.

The main problem that I've seen here that everyone is running into is they're not thinking physically. When you look at a real glass object, you're looking at it in a high dynamic range environment. If you have a window in the room you're sitting in, and it's daylight outside, compare the brightness of the light coming through that window with the brightness of your monitor. Chances are, it's a lot brighter. So is a typical lightbulb. Here's a photo of the room that I'm in with the current lighting conditions, and it's cloudy outside:
114706

Notice the window is brighter than the monitor. It's actually being clipped at 255 because it's so bright, and that's all bounced light, even along the closet doors on the far right. I'm getting a lot of highlights on everything in the room from that window, but barely anything from the monitor. Did you notice the reflection of the monitor on that coke glass? Me neither, and I'm looking directly at it.

Glass, water and other transparent objects need an environment to reflect & refract, or they'll just look dull. To get real-world matching results from Dielectric (or any refractive surface), you need to match the real-world lighting. That means lighting that's much brighter than the standard 0-255 range. A HDRI environment works best, but you can simulate this with super luminous objects as well. This leads to one of a few problems I have with the Material nodes: They do NOT have appropriately bright specular reflections of LightWave's lights. Why they call it Specular in the other Materials is beyond me, but apparently it's just too hard for them to change the labels to what they really should be - REFLECTION - since that's what they do. As someone else has shown, they well reflect objects perfectly, but lights are way too dim. I actually asked Antti about this years ago, and asked for a specular boost but he said it would require an additional ray. Since at the time, the only way to do this in LW was to mix the Material with a specular shader (set to something like 100,000% and blended at 0.001%), I was already using an additional ray. Then Mike Wolf showed that no additional ray was needed...




For stronger reflection use Material Booster from db&w tools.

This is now my preferred method of boosting the specular reflections. Generally I boost them to about 2000%. I don't need to touch the reflection channel since this is actually accurate. Though, this plugin does let you adjust things here as well, so feel free to use it for artistic purposes.



Jim - ok but... isnt dielectric supposed to provide realistic glass and transparent mediums ?
so why should i workaround it... imho it's up to NT to code something which works without too many fixes :)

Paolo

It does. The only reason you feel the need to workaround is because you're not using a correctly lit environment, and probably trying to just use the standard LW lights, which do not reflect correctly. They only reflect at about 5% as bright as they should, or less.




I do NOT like the way absorbtion works with Dielectric (an absorbtion of 2.... 2 Ducks? 2 taj Mahals? 2 metres?, nope not metres...) but that's a separate issue. I do wish Dielectric had more control and more features, such as adding thin film from the car paint shader, and a dye and forward scatter feature :)

Yeah, the absorption values are pretty arbitrary. They are also scale dependent. Scale the object up and it will get more opaque. Scale it down and the absorption will disappear, just like you'll get with real glass. The thinner the glass, the less it will absorb. Stick a spoon in some coffee and you can see the same thing happen. Near the surface, the spoon is perfectly visible, but the deeper it goes below the surface, the darker it becomes. This is a setting that you just have to muck around with to get what you want. Just another reason to build to scale. Look at this glass spider. The legs are almost color-free while the body is much redder. It's a single surface.

114726

Mike Wolf also has another node called Material Mixer. Ken Nign offers a Material Blixer, and there's others. These let you modify the output of materials to your liking. I've used these, and the Material Mixer, to do things like add Kappa or Theta to Dielectric for back-scattered dust on glass and iridescent oil stains on metal.
114724 114725
Here's Dielectric with Kappa and an iridescent reflection:
114710 114716



Oh yes , definitely I'd like Dielectric to get more options for boosting reflections, spec and possible coloring without adding too much of tinting which absorption do and you can't get some colors with absorpition 'coz if you type in high values like 500 or so of something (whatevert that number is ;)) you get some colorshifts that totally change actual color you might be wanting. It might be techically correct (I don't argue on that) but it's too limiting so if you can't change this dielectric then make "Dielectric Pro" with all the tweaks/options :).

I'd rather have MORE materials than have a few over-featured ones. Carpaint went so insane with all those new features that's it's a chore to use for more simple things. I miss the simpler version. Yet we have two Sigmas and three skin materials. I wouldn't mind seeing a new Dielectric for these special cases, like maybe a Coated Dielectric, or maybe even a dichroic (http://en.wikipedia.org/wiki/Dichroic_glass) one. One thing I would like to see done with Dielectric is to fix the useless Dispersion it has. If you turn this up, expecting a rainbow of colors, you're going to be sadly disappointed. You will only get three different images in the refraction, each with a different primary color. No blending. No yellows or purples or oranges, just a red image, a green one and a blue one. At small Dispersion values, it's not too bad, but I mean small. And this is another part that is needlessly arbitrary. What the hell is that amount a percentage of? How does it relate to the actual dispersion values, like you might find for gemstones (http://www.rockhounds.com/rockshop/gem_designs/properties.html)?




Real glass is probably less reflective than might be thought, that's for sure, and it may be that Dielectric is 100% accurate (though I need convincing), however for those situations where accuracy is less important than the effect, more control would be a great help not a hindrance. :D

As mentioned before, the reflection is determined entirely by the refraction index. At glancing angles, it can get to basically 100%, but it drops pretty low at facing angles, but not quite to 0%. But again, if you don't have the proper environment set up, you're not going to get the results you expect. Something that's a 255 level white can only reflect as white in areas where the reflection level is actually high enough, and that's those glancing angles. With an IOR of 1.5, that drops to about 4% when the surface is facing directly at the camera. 4% of white is pretty low. With an HDR environment, you can have colors that are a lot brighter. The Kitchen HDR that comes with the Benchmark Marbles scene has an area that's over 30 times brighter than 255, 255, 255. 4% of that is still over 100%, so even at that facing angle, you can get a full white highlight.

Again, to fake this, just run the Material through Mike Wolf's Material Booster. Another trick I use to make material a little more reflective is to use the Fresnel node, but run it through a BoxStep first. In the BoxStep, I lower the Begin value to -0.1 or -0.2. This pushes the lower input values up a bit, without risking any overloading of the glancing angles. To see how this works, try it on the BoxStep node in the Functions. That has a graph that shows what's going on.



Yes, I think Dielectric is largely correct, other than a couple of issues, but it's a lack of advanced features which is the problem (and not even just cheating!)

Well, it IS just a Dielectric node. It's not anything more than that.




The Materials have also got a number of tools which aren't replecated elsewhere (like correct thin-film in Car paint) which would be pretty in any of the other shaders too. In all honestly I would rather they have basic energy conserving materials that perform correctly and more tweak-able hand-cooked ones with energy conservation built in, and more power features to do boosts.

Delta is an energy conserving material. I use this one a lot. Add a Fresnel node, run through a BoxStep like I mentioned above and it works great. Run that through the Material Booster (something that LightWave should have as a standard node, but I'm not going to hold my breath) and you can get pretty much anything you want. IFW2 has an Iridescent node.




guys, it's not that is looks correct or incorrect... it doesnt have any fresnel effect :)
Reflection and transparency are pratically the same at any angle. It just CANT be correct, imho.

Paolo

Oh yes it does. Here's a little trick I've used for the past few years now. I've had to do a lot of airplane cockpits and windows. Obviously, dielectric was made for this, so I used it. But I don't always have the time to use Ray Trace Refraction, so I turn that off. Dielectric would still calculate the Fresnel for the reflections, as you can see in this image:

114711

That's a Dielectric teapot, with an IOR of 1.5 and refractions turned off.

Unfortunately, this feature is busted in 11.5. But the important thing is they didn't rename Specular to Reflection and break anything.




I seem to have missed that one!! :ohmy:

I thought the Volume Stack only affected the Dielectric node :(

Volume stack affects refraction everywhere, whether it's materials, shaders or even plugins like NatureFX. In fact, it's better to NOT have air polys anymore.




It gets to a point when Absorption is at 16, and in order to make it Orange the Hue has to be on 041, which is actually pretty yellow... if I punch the absorption up even more, it ends up red again. Very odd indeed... and most definitely wonky!!
Those absorption values are only valid for that specific object. The larger the object, the lower the Absorption amount should be if you want to still be able to see through it. Likewise, the thinner the object, the higher it needs to be to have it tinted. Scaling the object will affect this as well; it's a distance based effect. Here's a teapot again, that's only 15cm tall, which makes its side pretty thin. The Absorption level here is 300:
114712



Oh Matt, nice photos of glass... also very nice caustics!!! ;) Hint, hint!

You can get that with radiosity:
114714


I noticed some strangeness, precisely:
- flat glass sheet - i.e. stair parapet. Glass border in most cases looks darker, Dielectrc show it bright in some places, even if there's nothing to refract/reflect.
- I tried to simulate plexiglass chair, properly modelled , i.e. http://www.100matrimoni.it/uploads/kartellluisghost.jpg and results are not 100% convincing, you get too little reflection with correct parameters (plexi has IOR 1.25).

Where did you get that from? Plexiglass is 1.49. Water is 1.33. Plastics are denser than water, though some can come close. None of the plastics listed on this site (http://refractiveindex.info/?group=PLASTICS&material=PMMA) are lower than 1.4.



Also, getting dark edges where plastic is thicker was tricky at best. At last i got something i like, but i had to use material multiplier to boost reflection otherwise it looked like a soap bubble... lighting is correctl sunsky only and a pair of inv2 falloff ara into lights.
114675

Paolo

Sunsky isn't going to do the trick. You'd have to crank the sun in that to a severely high value to get it to have a bright enough reflection. Make sure the model is made correctly. All polygons should be outward facing, and no air polygons are necessary. Make sure Exclude From V-Stack is OFF as well as Double Sided, or it won't refract correctly. Also try an HDR image, like one found here (http://www.cgadvertising.com/pages/free-materials/hdr-maps.php). I like the Studio G environment.


Yes but Matt, I've also explained several scenarios where it could use more functions, such as an in-depended reflection and refraction blur.

Dielectric has this already, though they're both linked to the same Roughness value. But again, this is the Dielectric material, which does exactly what is promises - a dielectric simulation. It's not a "Something Like Dielectric" material. We already have that with other nodes. But more artistic materials would be welcome.



Thin film coatings, wax / polish coatings, scattering and sss, dye layers, metal coatings. All of those would add to the realism, be correct in physics and are REALLY hard to do (read impossible) Using 'Standard layers' editor.

You mean something like these:
114710 114715 114716 114727


He Who Needs A New Back.

Again, mixing and blending materials will get you that now. I did most of these images years ago.

Dave Jerrard
06-04-2013, 06:01 PM
dark edges are same issue i get in my picture. If you look, chairs have quite dark edges while the rest of the chair is barely visible, and i also boosted reflection with DB-W surface booster.
If i wanted to make chair more visible, i get black borders. Since i have that chair and i know plexyglass IOR, i'm sure it doesnt behave that way.
Maybe dielectric works nicely on organic stuff, but for flat, thick surfaces it can show limits.

Paolo
What's your Ray Recursion set to? Is Ray Trace Occlusion turned on?


He Who Really Needs A New Back.

allabulle
06-05-2013, 12:33 AM
Thanks for that very informtive post, Mr. Jerrard (whose chair might need some adjusting).

lardbros
06-05-2013, 04:43 AM
As usual... very nice post there Dave!

With regard to my specific little bit... I realise my settings were specific to that object, at that scale etc.

All I was wondering was whether the shader could somehow fudge the chosen colour of the absorption, so in fact that chosen hue was the most prevalent colour on the object. I believe this would be tricky, as it would require some kind of intelligence within the shader, so it knows the scale/thickness/etc of the object. Maybe this is possible?! It would help for artistic control...



As a side note... in 11.5 the Dielectric shader doesn't show the partial internal reflections as it should. Well, it's different to the F9 anyway. Think this is now fixed for the next release though...

Lightwolf
06-05-2013, 06:05 AM
Well you can split the dieletric and multiply the reflection by any modifier, so you could use anther fresel or an inverse fresnerl or whatever you required.
But that increases computation time if the dielectric is computed more than once as a consequence. Thus the booster.

Cheers,
Mike

Tobian
06-05-2013, 08:44 AM
A few things:

First, sure nodes, and in particular, the Material nodes, have some problems, but of those, Dielectric is actually pretty error-free. The problems I've seen here all stem from user error.


Agreed, to a point :)




I always get what I want, or very close to it.

(snipped for brevity)

Glass, water and other transparent objects need an environment to reflect & refract, or they'll just look dull. To get real-world matching results from Dielectric (or any refractive surface), you need to match the real-world lighting. That means lighting that's much brighter than the standard 0-255 range. A HDRI environment works best, but you can simulate this with super luminous objects as well. This leads to one of a few problems I have with the Material nodes: They do NOT have appropriately bright specular reflections of LightWave's lights. Why they call it Specular in the other Materials is beyond me, but apparently it's just too hard for them to change the labels to what they really should be - REFLECTION - since that's what they do. As someone else has shown, they well reflect objects perfectly, but lights are way too dim. I actually asked Antti about this years ago, and asked for a specular boost but he said it would require an additional ray. Since at the time, the only way to do this in LW was to mix the Material with a specular shader (set to something like 100,000% and blended at 0.001%), I was already using an additional ray. Then Mike Wolf showed that no additional ray was needed...



I used to think of it as 2 different materials, but a dielectric is just one material, with a refractive and reflective component. Some materials are various layers of material (for example Car paint shaders) but those things do get tricky to simulate!

Agreed, most people don't add nearly enough dynamic range in their scenes for realistic shading. Part of this is because high dynamic range causes other issues with rendering, such as blow outs, clipping, poor AA and especially fireflies.





This is now my preferred method of boosting the specular reflections. Generally I boost them to about 2000%. I don't need to touch the reflection channel since this is actually accurate. Though, this plugin does let you adjust things here as well, so feel free to use it for artistic purposes.



Oh hmm yes this could be a good way to solve that. I shall play. Of course it's just a fudge and it's most likely not physically correct, just gets tiring having to hack materials which are sold as such.




(snip)

Yeah, the absorption values are pretty arbitrary. They are also scale dependent. Scale the object up and it will get more opaque. Scale it down and the absorption will disappear, just like you'll get with real glass. The thinner the glass, the less it will absorb. Stick a spoon in some coffee and you can see the same thing happen. Near the surface, the spoon is perfectly visible, but the deeper it goes below the surface, the darker it becomes. This is a setting that you just have to muck around with to get what you want. Just another reason to build to scale. Look at this glass spider. The legs are almost color-free while the body is much redder. It's a single surface.



Yep the units are pretty arbitrary, but if they are based off Beer's Law, then they have a scale.
114736

According to this it's based on Centimetres, so it would be handy to have a way to alter the inherent scale. While I agree on 'building to scale' It would be nice to be able to adjust this. (thanks to Luis for the link)





114726

Mike Wolf also has another node called Material Mixer. Ken Nign offers a Material Blixer, and there's others. These let you modify the output of materials to your liking. I've used these, and the Material Mixer, to do things like add Kappa or Theta to Dielectric for back-scattered dust on glass and iridescent oil stains on metal.
114724 114725
Here's Dielectric with Kappa and an iridescent reflection:
114710 114716

I'd rather have MORE materials than have a few over-featured ones. Carpaint went so insane with all those new features that's it's a chore to use for more simple things. I miss the simpler version. Yet we have two Sigmas and three skin materials. I wouldn't mind seeing a new Dielectric for these special cases, like maybe a Coated Dielectric, or maybe even a dichroic (http://en.wikipedia.org/wiki/Dichroic_glass) one. One thing I would like to see done with Dielectric is to fix the useless Dispersion it has. If you turn this up, expecting a rainbow of colors, you're going to be sadly disappointed. You will only get three different images in the refraction, each with a different primary color. No blending. No yellows or purples or oranges, just a red image, a green one and a blue one. At small Dispersion values, it's not too bad, but I mean small. And this is another part that is needlessly arbitrary. What the hell is that amount a percentage of? How does it relate to the actual dispersion values, like you might find for gemstones (http://www.rockhounds.com/rockshop/gem_designs/properties.html)?




Yeah and this is the problem....

For one it requires a lot of technical skill and messing on to get these types of materials, there's several gaps in shaders, and there's the issue of BSDF integration.

Keeping a fully energy conserving workflow with materials and or shaders is heinously complex to do, and you often have to fight against what they were trying to do (get the material to only output one type of shader is often tricky).

Now to get technical here, whenever you mix up multiple shaders, you stack up on render times, because each shader and or material has to be calculated. For the most part the Materials have internal BSDF integration: All of the shaders are added together and computed as a unit, so render faster - Bi-direction Scattering Distribution Function. Core had this as a shader feature, which helped speed things up a lot! Car paint for example, has a diffuse shader, and 2 reflection shaders. Sure, you can make your own, but it will render a lot slower.

There is a niggling omission in LW as well. You can't combine sss and transparency without a pre-process (for the SSS). So yes, you can use shaders like Kappa or SSS2, but not 'skin' which yields better results. I've tried it and while it sometimes works, it most often results in hideous render noise and errors. It's why I gave up on my scattering glass material tutorial, because I hit issues and got very frustrated. (and you can't play as easy with pre-process shaders, as they don't work in VPR).

With regards to the Dispersion, yeah it works fine to a point, but you start seeing the splitting. It's also VERY slow, because they calculate 3 seperate reflection shaders, with different offsets and add them together. What you want is a chromatic dispersion, which currently isn't possible in LW.

I agree though about more special situations shaders, not monster huge ones, and better native tools for mixing them together, with BSDF integrators.




As mentioned before, the reflection is determined entirely by the refraction index. At glancing angles, it can get to basically 100%, but it drops pretty low at facing angles, but not quite to 0%. But again, if you don't have the proper environment set up, you're not going to get the results you expect. Something that's a 255 level white can only reflect as white in areas where the reflection level is actually high enough, and that's those glancing angles. With an IOR of 1.5, that drops to about 4% when the surface is facing directly at the camera. 4% of white is pretty low. With an HDR environment, you can have colors that are a lot brighter. The Kitchen HDR that comes with the Benchmark Marbles scene has an area that's over 30 times brighter than 255, 255, 255. 4% of that is still over 100%, so even at that facing angle, you can get a full white highlight.

Again, to fake this, just run the Material through Mike Wolf's Material Booster. Another trick I use to make material a little more reflective is to use the Fresnel node, but run it through a BoxStep first. In the BoxStep, I lower the Begin value to -0.1 or -0.2. This pushes the lower input values up a bit, without risking any overloading of the glancing angles. To see how this works, try it on the BoxStep node in the Functions. That has a graph that shows what's going on.

Well, it IS just a Dielectric node. It's not anything more than that.



Yup the Dynamic range of the image thing again. Linear Colour Space also plays a big factor here too, as 1% becomes 5% and 5% becomes 25% luminance, at the lower end of the luminance.

The only thing I would disagree with you here is 'only Dielectric' - Technically both Delta and Dielectric are Dielectric shaders. Technically all the Skin shaders are too. A Dielectric material has a low Fresnel number reflectance, which reflects mostly white. It also has transmission of light. Some materials transmit light, absorbing some of the colour (glass) some have a denser structure and scatter the light (Skin, Wax) and if the distance of the scattering effect is lower than the wavelength of the light, then basically you have Lambertian diffuse. A Physically correct 'Dielectric' shader should have (forward scatter + Backward scatter + transmission x absorbtion + Reflection = < %100) Which is a huge simplicifation, but is what 'Dielectric' materials consist of.







Delta is an energy conserving material. I use this one a lot. Add a Fresnel node, run through a BoxStep like I mentioned above and it works great. Run that through the Material Booster (something that LightWave should have as a standard node, but I'm not going to hold my breath) and you can get pretty much anything you want. IFW2 has an Iridescent node.




Yup, and it's fast to use, faster than plugging a Fresnel into Reflection and Diffuse, because of the integration.





Oh yes it does. Here's a little trick I've used for the past few years now. I've had to do a lot of airplane cockpits and windows. Obviously, dielectric was made for this, so I used it. But I don't always have the time to use Ray Trace Refraction, so I turn that off. Dielectric would still calculate the Fresnel for the reflections, as you can see in this image:

114711

(snip)

Volume stack affects refraction everywhere, whether it's materials, shaders or even plugins like NatureFX. In fact, it's better to NOT have air polys anymore.



Yup and the even more annoying thing about Dielectric is that it does it better than Reflection+refraction mediated by Fresnel materials. Just.. everything else :)





Those absorption values are only valid for that specific object. The larger the object, the lower the Absorption amount should be if you want to still be able to see through it. Likewise, the thinner the object, the higher it needs to be to have it tinted. Scaling the object will affect this as well; it's a distance based effect. Here's a teapot again, that's only 15cm tall, which makes its side pretty thin. The Absorption level here is 300:
114712

You can get that with radiosity:
114714

Where did you get that from? Plexiglass is 1.49. Water is 1.33. Plastics are denser than water, though some can come close. None of the plastics listed on this site (http://refractiveindex.info/?group=PLASTICS&material=PMMA) are lower than 1.4.



Yes, a frustrating part of the absorbtion is it's per object, making preset materials... failure :) it is possible to get pretty caustics using MC radiosity, set to Directional rays, but my god it's slow, and it's monochrome. Tinted materials don't work properly, and there's also no dispersion, so it's less pretty than it could be.




(Snip)

Dielectric has this already, though they're both linked to the same Roughness value. But again, this is the Dielectric material, which does exactly what is promises - a dielectric simulation. It's not a "Something Like Dielectric" material. We already have that with other nodes. But more artistic materials would be welcome.



No, it's not possible, with Dielectric, to have a different roughness value for Refraction and Reflection. It's perfectly plausible for them to be different, in nature. Frosted glass is not 'rough' all the way through, just sanded on the top.




You mean something like these:
114710 114715 114716 114727


He Who Needs A New Back.

Again, mixing and blending materials will get you that now. I did most of these images years ago.

yup, but mixed materials are really slow to render, and very hard to make correctly energy conserving, or make, period, using native tools. Thank goodness for the third party, but it doesn't fill all the gaps. No offense, but showcasing how you can do quite exotic things on teapots doesn't solve the issue of rendering speeds and practical useability, much like having to use a material booster on 50 materials is a major ballache!

Thanks for going into detail though Dave!

Celshader
06-05-2013, 09:23 AM
No offense, but showcasing how you can do quite exotic things on teapots doesn't solve the issue of rendering speeds and practical useability, much like having to use a material booster on 50 materials is a major ballache!

If Dave went into his node workflows for production, we'd be here all night. ;)

114745 114746

Tobian
06-05-2013, 09:35 AM
Yup... So yeah Compounds please! :) I am also quite famous for my node madness, though those do probably win :) The annoying part there is a lot of the complexity will be because of workarounds...

Dave Jerrard
06-05-2013, 10:27 AM
Yep the units are pretty arbitrary, but if they are based off Beer's Law, then they have a scale.
114736
I've seen earlier version that used much smaller values, so I doubt they are.




According to this it's based on Centimetres, so it would be handy to have a way to alter the inherent scale. While I agree on 'building to scale' It would be nice to be able to adjust this. (thanks to Luis for the link)

You can always modify the value by the object's scale. That would make it automatic.







Now to get technical here, whenever you mix up multiple shaders, you stack up on render times, because each shader and or material has to be calculated. For the most part the Materials have internal BSDF integration: All of the shaders are added together and computed as a unit, so render faster - Bi-direction Scattering Distribution Function. Core had this as a shader feature, which helped speed things up a lot! Car paint for example, has a diffuse shader, and 2 reflection shaders. Sure, you can make your own, but it will render a lot slower.One of the reasons I don't like to use Carpaint too much. It was better when you could assign different numbers of samples to the different aspects. My original marble scene used 4 samples on the wood part but only 1 on the clearcoat, which gave really good results. I guess someone found that to be too useful so it was taken away.




There is a niggling omission in LW as well. You can't combine sss and transparency without a pre-process (for the SSS). So yes, you can use shaders like Kappa or SSS2, but not 'skin' which yields better results. I've tried it and while it sometimes works, it most often results in hideous render noise and errors. It's why I gave up on my scattering glass material tutorial, because I hit issues and got very frustrated. (and you can't play as easy with pre-process shaders, as they don't work in VPR).Well, the SSS shaders are slower than my bowels after a week long pizza/mac & Cheese fest (not to mention, completely ***-backwards wrong). I don't even bother with the preprocessing materials. I find they take longer to get good results.




With regards to the Dispersion, yeah it works fine to a point, but you start seeing the splitting. It's also VERY slow, because they calculate 3 seperate reflection shaders, with different offsets and add them together. What you want is a chromatic dispersion, which currently isn't possible in LW.It would be faster to use a stochastic method that's filtered. One sample that varies the IOR and filters the result. I've done this myself by animating the IOR for frame. I've done it with the Advanced Camera too, creating a shromatic abberation effect by animating the field of view and animating its color filter to match, and it renders pretty quick. In the case of dielectric, the color gradient was a bit trickier to set up. Unfortunately, this doesn't work with photoreal blur since for some reason, textures can't be animated like this. But with the Classic Camera, or just multiple Motion Blur Passes, it works fine.




The only thing I would disagree with you here is 'only Dielectric' - Technically both Delta and Dielectric are Dielectric shaders. Technically all the Skin shaders are too. A Dielectric material has a low Fresnel number reflectance, which reflects mostly white. It also has transmission of light. Some materials transmit light, absorbing some of the colour (glass) some have a denser structure and scatter the light (Skin, Wax) and if the distance of the scattering effect is lower than the wavelength of the light, then basically you have Lambertian diffuse. A Physically correct 'Dielectric' shader should have (forward scatter + Backward scatter + transmission x absorbtion + Reflection = < %100) Which is a huge simplicifation, but is what 'Dielectric' materials consist of.
I didn't want to get that technical. As it is, I had to trim my post to fit the 15000 character limit a few times.




Yes, a frustrating part of the absorbtion is it's per object, making preset materials... failure :) it is possible to get pretty caustics using MC radiosity, set to Directional rays, but my god it's slow, and it's monochrome. Tinted materials don't work properly, and there's also no dispersion, so it's less pretty than it could be.
They're not monochromatic. You can see that in the Benchmark Marbles scene, and here:
114747

You are right about the dispersion not affecting them though. And it doesn't work with LW lights. HDRI environments look great though.

As for the presets, I never use them for surfacing. There's always so much tweaking I need to do on them that it's just faster to start from scratch, or copy from another surface that's similar and already part of the same object or scene and make the adjustments.




No, it's not possible, with Dielectric, to have a different roughness value for Refraction and Reflection. It's perfectly plausible for them to be different, in nature. Frosted glass is not 'rough' all the way through, just sanded on the top.But that same roughness affects both the reflection and the refraction the same since those both occur at that surface. It would still be nice to be able to control them separately though.




yup, but mixed materials are really slow to render, and very hard to make correctly energy conserving, or make, period, using native tools. Thank goodness for the third party, but it doesn't fill all the gaps. No offense, but showcasing how you can do quite exotic things on teapots doesn't solve the issue of rendering speeds and practical useability, much like having to use a material booster on 50 materials is a major ballache!I wasn't trying to. Complaining about adding a node like Material Booster to adjust something is kinda silly though. Might as well complain about having to add any node then. What? I have to add a color node? What the...

As for the teapots, they were on hand and quick to use and showed that much of the stuff people were talking about was already possible. If it can be done on these, it can be done in production. Actually, production tends to do things simpler than this and then comp the hell out of it later. I tend to so things up front, in camera as much as possible. And these teapots were a lot simpler than stuff like this:
114749 114750 114751 114748
And all of those are production surfaces.

He Who Doesn't Think He's Done Anything Larger Than Those Bomber Wings.

lardbros
06-05-2013, 11:00 AM
Wish there was a way to pick certain posts... and add them to a pdf to be saved for later! :D Including the large full size pics!

Nice asteroid... the second one is luuuuuurvly!!

Tobian
06-05-2013, 12:08 PM
Ugh, the multipart replies take ages to do. I'm just numbering my responses, and yes, I got word count errors too.

1) yeah it's supposed to be Beer's law, I have no idea if it is.

2) Yup, it's just really fiddly to make it look right, lots of tests, instead of something more predictable

3) Yeah that's a separate issue. they do need to have shading sample multipliers, and importance sampling, as a lot of samples can be thrown at materials which need a lot less. Facing surfaces should have a lot less samples fired at them, as the Fresnel is a lower % of the visible surface.

4) Skin is a really nice material, it's the best one so far. The older one's are horribly inaccurate and produce weird unpredictable results. It's not perfect, it's just better. But for the above reasons I detailed, can't be mixed in with transparent materials. Shame.

5) Yes I agree, it would be better. Hopefully in a future version they do that. Sadly I suspect it's never going to be fast to calculate, but they could try to optimise it would be great. If you have samples of your method to pass onto Antti it would be great, as that sounds very cool.

6) Well I had to read through many technical papers to get that. it's kinda obvious if you think about it, it's just hard hard maths :D

7) Hmm sometimes I have just gotten monochrome results. Possibly something to do with whatever materials I have used. Very nice results. Mega-uber-slow to render in my experience though, and little use practically. Would be better if they had a separate, dedicated caustics engine, which likewise also had stochastic dispersion.

8) Agreed. On surfaces like say Jade, or thin wax though you have very high gloss finishes, and very perturbed interiors. It's probably better handled by SSS but thin SSS is kinda transparent.

9) Hmm but since you already pointed out that Spec was wrong, it should still be complained about. There's a difference between 'flexible' and 'workaround fix' The latter is just a pain to have to deal with. So many times my 'complex' materials have just been compensating for a lack of features, bugs or issues, not just to ADD stuff, like your material mixing examples do (which are very cool btw!). I know they are just cool examples, but the point I wanted to make is there are inefficiencies in the way the nodal material system works, and there could be more optimisations, such as BSDF integration and better tools for material mixing, some of which have to be compensated for by third party tools.

Matt
06-05-2013, 01:10 PM
Yep the units are pretty arbitrary, but if they are based off Beer's Law, then they have a scale.
114736

According to this it's based on Centimetres

It clearly says, that absorbance has _no_ units. You can't say it's based on centimeters because part of the equation has those values.

That's like saying speed is measured in minutes because the equation speed = distance / time has time in it.

Matt
06-05-2013, 01:19 PM
Thanks man, finally somene who understand my "issue" with this dielectric color shift and problem

Hope you're not trying to assert that I don't understand what you were saying there Lewis! ;)



why does Octane look better/more realsitic if LW Dielctric is 100% accurate scientifically/phisically.

It's a different renderer Lewis, that's why, and the settings are different.



Matt do we have slight bug in Dielectric afterall or ?

No, as you can see here, Octane renders the same when the settings are matched more accurately.


http://i465.photobucket.com/albums/rr16/xXswampyXx/LW_Glass_Vs_Octane1_zps3a80de36.jpg (http://s465.photobucket.com/user/xXswampyXx/media/LW_Glass_Vs_Octane1_zps3a80de36.jpg.html)

MSherak
06-05-2013, 01:20 PM
If Dave went into his node workflows for production, we'd be here all night. ;)

114745 114746

Looks like some of my real-time CGFX shaders. Shows all the great control you have with nodes in LW. Love it.

Lightwolf
06-05-2013, 04:37 PM
Nitpicking...

It clearly says, that absorbance has _no_ units.
Yup. It is just a coefficient, which is essentially a multiplication factor and thus has no unit (and requires none).

That's like saying speed is measured in minutes because the equation speed = distance / time has time in it.
Well, yes and no, the unit of speed is m/s.

On a lighter note, I'm very happy to see the booster node brought to it's intended use, yay.

Cheers,
Mike

Dave Jerrard
06-06-2013, 07:20 AM
3) Yeah that's a separate issue. they do need to have shading sample multipliers, and importance sampling, as a lot of samples can be thrown at materials which need a lot less. Facing surfaces should have a lot less samples fired at them, as the Fresnel is a lower % of the visible surface.
Well, in 10.1 and earlier we had that control on sampling, and I used it a LOT. Every one of those surfaces is irreparably busted in 11. And every one of those scenes takes much longer to render now too.




4) Skin is a really nice material, it's the best one so far. The older one's are horribly inaccurate and produce weird unpredictable results. It's not perfect, it's just better. But for the above reasons I detailed, can't be mixed in with transparent materials. Shame.Never had much luck with it. I'm doing stuff now that use Simple Skin. I tried Skin, but it was just rendering too noisy. Simple Skin was cleaner, and since I was trying to keep render times down, I went with that.




7) Hmm sometimes I have just gotten monochrome results. Possibly something to do with whatever materials I have used. Very nice results. Mega-uber-slow to render in my experience though, and little use practically. Would be better if they had a separate, dedicated caustics engine, which likewise also had stochastic dispersion.You radiosity settings are probably too high. I rarely ever use more than 4 RPE and almost never use the interpolated modes since they always look like crap, or they just lose details. The few times I have used them, which was on about three scenes, were scene that had lots of foil in them which cause a lot of sparklies. On hindsight, I could have limited the dynamic range to solve that issue.



8) Agreed. On surfaces like say Jade, or thin wax though you have very high gloss finishes, and very perturbed interiors. It's probably better handled by SSS but thin SSS is kinda transparent.
I've had good results with Sigma, but that's very version specific. It seems to be changed on every damned release. BUt at least they're not breaking old scenes by fixing labels... :screwy:


9) Hmm but since you already pointed out that Spec was wrong, it should still be complained about. There's a difference between 'flexible' and 'workaround fix' The latter is just a pain to have to deal with. So many times my 'complex' materials have just been compensating for a lack of features, bugs or issues, not just to ADD stuff, like your material mixing examples do (which are very cool btw!). I know they are just cool examples, but the point I wanted to make is there are inefficiencies in the way the nodal material system works, and there could be more optimisations, such as BSDF integration and better tools for material mixing, some of which have to be compensated for by third party tools.
Well that does carry over to pretty much every other part of LW. It's not something exclusive to nodes.


He Who Has A Long List Of Gripes That Aren't Welcome.

erikals
06-06-2013, 08:22 AM
If Dave went into his node workflows for production, we'd be here all night. ;)

114745 114746

where is the scream smiley... ?

Dave Jerrard
06-06-2013, 09:11 AM
where is the scream smiley... ?

Those are the simple ones.


He Who Hasn't Posted Any Volumetric Nodes Flows Yet.

allabulle
06-07-2013, 04:25 AM
I just realized you are the ace pilot on the plane, Mr. Jerrard. :)

Nice materials, by the way.

erikals
06-07-2013, 09:10 AM
didn't notice at first.. was that in the movie?... http://erikalstad.com/backup/misc.php_files/smile.gif

Dave Jerrard
06-07-2013, 09:17 AM
didn't notice at first.. was that in the movie?... http://erikalstad.com/backup/misc.php_files/smile.gif

I actually don't remember. That render was actually just a test with an HDR environment, but I can't remember if I put the names on before or after the models were used.

And yes, that is Dielectric on the canopies.


He Who Did Get Caught Changing The OH To JM On Those.

erikals
06-07-2013, 09:41 AM
any artist with respect for himself/herself should do similar... http://erikalstad.com/backup/misc.php_files/wink.gif

to few artists do these days...

Nicolas Jordan
06-07-2013, 04:08 PM
If Dave went into his node workflows for production, we'd be here all night. ;)

114745 114746

Just when I thought I had seen it all. That looks crazy complicated! :bowdown:

Tobian
06-07-2013, 04:28 PM
1) To be fair to the new sampling, the way that LW samples in LW11 is different, in that the same value nets you a LOT more blur, which is always going to be hard on your render. You also get more smearing on glancing angles, which is both correct, but problematic, as it creates a problem for the AS to clean up, and if you don't have enough min samples, it doesn't always clean up right (3 fireflies will never result in a clean disk!)

2) Skin is a curious beast, getting the sss/diffuse ballance is a little like fighting in molasses, it's kind of weird and I never seem to know what I am doing. With Simple skin you can always easily build an energy conserving network anyway. In some ways I wish, again, we had all of the tools Antti built into the materials as shader, so we could more easily roll our own materials (a transparency shader, with absorption, a reflection shader with thin film, an sss shader, with no preprocess) and with BSDF integration. Kind of like in....... CORE :p

3) for interior work like I do, I just can't get away with that trick, it results in hideous hideous noise which never cleans. It's more feasible for exteriors as you get a cleaner more balanced sample from skies. For tes scenarios like that, yes I would use that technique, but for full scenes it's prohibitive.

4) yes on occasion I have got good results from sigma, but you can't mess on with it with VPR, because it has a preprocess, and preprocesses cause their own issues for animation (not that I do much hehe). SSS shaders can't mix with transparency unless they have a preprocess.

5) yep it's the spec hit in general. The Spec seems to match the shape and size of the spot better with materials, over the original shaders, just the luminance is well out of whack.

Oh and also from a tip from Lightwolf, I spotted that materials don't calculate for material mixer, if they aren't called (if the alpha value is 0 or 100%) so that saves some faffing on for the future.

Nicolas Jordan
06-07-2013, 04:48 PM
I'm not a fan of the "unified sampling" in Lightwave 11.x mainly because it prevents the user from being able to control light samples on a per light basis and shading samples on a per shader basis. Ideally users should have the option of being able to control these either globally or locally as needed but if I had to pick one way it would be local control of sampling.

Tobian
06-07-2013, 04:58 PM
I am a fan of it, only it is simply not complete. Needs sampling multipliers, and importance sampling...

Dave Jerrard
06-07-2013, 05:06 PM
3) for interior work like I do, I just can't get away with that trick, it results in hideous hideous noise which never cleans. It's more feasible for exteriors as you get a cleaner more balanced sample from skies. For tes scenarios like that, yes I would use that technique, but for full scenes it's prohibitive.I'm actually working on an interior setup right now using Monte Carlo, and it's cleaning up nicely with 3 min and 96 Max, Threshold of 0.01 (what I'd get in 9.6 with AA of 3 and a Threshold of 0.02). Lots of white walls with a few brightly colors objects. It's subtle, but nice.


4) yes on occasion I have got good results from sigma, but you can't mess on with it with VPR, because it has a preprocess, and preprocesses cause their own issues for animation (not that I do much hehe). SSS shaders can't mix with transparency unless they have a preprocess.


Actually Sigma II does the preprocess. Sigma (I) doesn't. Had some great results using this for ocean water too.





Oh and also from a tip from Lightwolf, I spotted that materials don't calculate for material mixer, if they aren't called (if the alpha value is 0 or 100%) so that saves some faffing on for the future.

I noticed. That's one of the reasons I usually keep the transitions between materials pretty sharp. Those mustangs were using five materials, all blended. I also use Switch a lot since that's only going to render one material at a time.


He Who Is Having A Rough Time Rendering Lines Right Now.

Dave Jerrard
06-07-2013, 05:12 PM
I'm not a fan of the "unified sampling" in Lightwave 11.x mainly because it prevents the user from being able to control light samples on a per light basis and shading samples on a per shader basis. Ideally users should have the option of being able to control these either globally or locally as needed but if I had to pick one way it would be local control of sampling.

Totally agree. It's also about 30% slower than 9.6, ray per ray. And now if we run into a situation where a surface is rendering noisy, but everything else is ok, you can't just up the samples on that surface. No, you have to do it on EVERYTHING. 11 busted a LOT of surfaces I did in 9.6, where I was doing things like dropping samples on reflected rays which made things quite a bit faster. 11 just gave me the big middle finger on all those.

Same with lights. A big area light might need a higher quality, but the smaller ones aren't usually in need of that hit. But it's not like we need to get stuff rendered quickly now, is it?


He Who Has Never Had To Explain The AA Settings So Often Before 11.

Dave Jerrard
06-07-2013, 05:25 PM
I am a fan of it, only it is simply not complete. Needs sampling multipliers, and importance sampling...

It just royally pisses me off every time I load an old scene, or even old objects that were done in 9.6 or 10. Stuff that was set up the way I wanted is now ****** and I have to set those two settings every damned time, usually after I started a render and noticed it's taking too long or looking like crap. I tend to just kill 11 and render in 9.6 again since it does the job right, and faster.


He Who Is Glad None Of His Surfaces Were Broken By Renaming A Label Or Anything Else That He Could Fix Easily.

Tobian
06-07-2013, 05:51 PM
Right, but Sigma 1 doesn't have any transparency...

and yeah Dave if you're using mostly diffuse, but my scenes use heavy reflection blur, and with no importance sampling, I'm stuffed too!
http://www.andrewcomb.com/wp-content/uploads/2013/05/test192.jpg http://www.andrewcomb.com/wp-content/uploads/2012/11/test1861.jpg
For reflection blur the rules work differently than for diffuse, and those scenes would take a million billion years without interpolated.

and yes very fine lines are a pain for me too :D

For specific scenario yes, 11 can be slower, depends what you're trying to do. For reflection blur however the results are just better, and my materials are so much more simple now I don't have to use DP's reflection, which was pretty slow to use.

Dave Jerrard
06-07-2013, 06:16 PM
Right, but Sigma 1 doesn't have any transparency...Yes it does. That's why I was using it for an ocean.




and yeah Dave if you're using mostly diffuse, but my scenes use heavy reflection blur, and with no importance sampling, I'm stuffed too!
Mostly materials, with blurred reflections. The trick is to keep samples slow on everything and let the AS clean things up. Worked better in 9.6 since you could fine tune every surface, light and node.



and yes very fine lines are a pain for me too :DIt's not the fineness of them, I have no trouble there. I want to animate their thickness. Can't do that. Looks like I'm back to using tubes again.



For specific scenario yes, 11 can be slower, depends what you're trying to do. EVERY SCENE I've tested with from a simple luminous ball to full radiosity scenes. EVERY ******* SCENE.




For reflection blur however the results are just better, and my materials are so much more simple now I don't have to use DP's reflection, which was pretty slow to use.[/QUOTE]

That's a matter of opinion. I like the new reflections for some thing, but not others, and they came at the cost of surfaces that were set up the way I wanted them. We lost the old reflection blurring, which just ****** me over again on some scenes. Yet another thing that the user can't get back no matter what, short of having to go back a version or two. Yep, there's that big middle finger again.


He Who Hates Losing Features He Uses All The Time.

Tobian
06-07-2013, 06:37 PM
I can't see an input labelled Transparency. There's translucency though?

yeah I am aware of that trick Dave, I wrote a tutorial about it hehe :D

I really hated the old reflection blurring, so no crying from me, but I understand your pain! :) Hmm, yes, I have seen some people complaining that LW 11 was slower at rendering, but then I, Like yourself was optimising the **** out of my scenes for some time. That said I do prefer the new AA, it just needs the features I mentioned, to give more control to further optimise stuff, while allowing for a generic low maintenance mode for less mental users :D

Tobian
06-07-2013, 06:39 PM
Sigma 2 has transparency.. sigma2 has a prepass phase...

Dave Jerrard
06-08-2013, 09:58 AM
Sigma 2 has transparency.. sigma2 has a prepass phase...

It's not labeled in Sigma, but it does transparency. It has to in order to work. Crank the Distance up and you'll see it...

Here's a few images that show this, THe first four use an Absorption value of 0.01:

Sigma with a 100mm Distance:
114808

A distance of 1 meter:
114809

A Distance of 10 meters:
114810

And a distance of 100 meters:
114811


Here's 100 meters with the Absorption increased to 2:
114812

These were all rendered with Sigma, not Sigma2.


For comparison, here's Sigma2 with similar settings, as close as I could get to the above. Distance is 100m (but looks identical to 100mm), and Transparency set to 100%.
114813


He Who Doesn't Like Nodes That Preprocess Because They Take So Long To See.

Tobian
06-10-2013, 02:43 PM
Hmm it's odd and transparency-like but it is a bit weird how it looks. The last one looks about right though, with more absorption. Sigma 2 also looks slightly odd as it's casting too much shadow. Hmm.

Pavlov
06-12-2013, 07:37 PM
Hi,
Dave - my scene is modeled correctly, faces, materials etc. That was just the light i wanted. Notice that sunsky's sun is NOT hitting the chairs and there's no hole behind the camera. Sun just hits the stairs. As you see my chairs are very transparent and little reflective, while edges are too dark... i saw this chair in same light conditions and it reflects much more, while edges are never so black. So we can say Dielectric works in some conditions, but here it should behave a little different - or we shuld be able to make it behave different.

Beside this, reading other posts above, what i miss most is a common BSDF node which handles diffusion, reflection, transparency and SSS all in one, and all driven by IOR (not gradients please). This would really be a step ahead in material creation. I'm used to BSDF in Maxwell and it's a whole different world.

Again: does Conductor node have any know issue with Unified sampling ? It stays unnormally noisy even wiith high AA..

bye
Paolo

Dave Jerrard
06-12-2013, 08:52 PM
Hi,
Dave - my scene is modeled correctly, faces, materials etc. That was just the light i wanted. Notice that sunsky's sun is NOT hitting the chairs and there's no hole behind the camera. Sun just hits the stairs.
Point the camera at the sun in Sunsky in that scene and tell me how bright it is. Also tell me what your sun light's intensity is. No go to that location you saw the real chair and take light measurements of the area, including of the sun.


As you see my chairs are very transparent and little reflective, while edges are too dark... i saw this chair in same light conditions and it reflects much more, while edges are never so black.

Your IOR is too low. Plexiglass has an IOR of 1.48, NOT 1.25. That means you're not getting the amount of refraction you should be getting, and your reflections are also too low. You can't expect to match a real world refraction if you're not matching the refractive index of it. Dielectric is doing exactly what you told it to do; in this case, a chair made of something less dense than water. In fact, the closest thing I could find to having a 1.25 IOR is baby wash, with an IOR of 1.26 (http://forums.cgsociety.org/archive/index.php/t-513458.html). That was after looking through several pages of refraction lists:
http://refractiveindex.info/
http://hyperphysics.phy-astr.gsu.edu/hbase/tables/indrf.html
http://www.robinwood.com/Catalog/Technical/Gen3DTuts/Gen3DPages/RefractionIndexList.html
http://macro.lsu.edu/HowTo/solvents/Refractive%20Index.htm
http://forums.cgsociety.org/archive/index.php/t-513458.html

[/QUOTE] So we can say Dielectric works in some conditions, but here it should behave a little different - or we shuld be able to make it behave different.[/QUOTE]

Dielectric is a model of real world physics. As such, it can't just selectively work for one object and not another. If you're not getting the results you want, you're doing something wrong. Again, I point to your refraction index being too low.



He Who Tries To Get His Refraction Index Right Before Going On To Other Parts Of His Refractive Surfaces.

Danner
06-13-2013, 05:12 AM
One time I coulnd't get dielectric to give me a deep green color I wanted, I changed my color space settings and I saw it right away.

Pavlov
06-13-2013, 05:26 AM
hi,


Point the camera at the sun in Sunsky in that scene and tell me how bright it is. Also tell me what your sun light's intensity is. No go to that location you saw the real chair and take light measurements of the area, including of the sun.Your IOR is too low. Plexiglass has an IOR of 1.48, NOT 1.25

BTW i cant do it, but let's imagine there's no sun. In a pure overcast day, this chair will reflect more.. yes IOR is clearly wrong, but obviously you can guess first thing i did was ranging from 1 to 2 :) and results were not improving, reflection wise. At this point i should imagine there's kinda a thin coating on the chair itself, made just to improve reflection. So the onyl way to get this material is mixing two materials, Dielectric node and another which provides coating (i dunno which one tough).

thanks,
Paolo

Dave Jerrard
06-13-2013, 11:26 AM
BTW i cant do it, but let's imagine there's no sun. In a pure overcast day, this chair will reflect more..

No it wont. The chair will still have the same reflectance. What you're changing is the amount of light for it to reflect. Again, you need to think in HDR terms, and look at the following page again:
http://refractiveindex.info/?group=PLASTICS&material=PMMA

Thats the IOR for Plexiglass, and it includes a graph showing the reflectivity from 0 to 90 degrees from camera. Here's that graph, just showing the non-polarized light curve (since that's all LightWave currently deals with):
114908
That graph is actually interactive on the page, and you can see there that the surface is less than 4% reflective up to 28 degrees, when it reaches 4.001%. It doesn't even get more than 10% reflective until that angle passes 62 degrees, so most of the surface of those chairs is not going to be very reflective. So if your environments, in this case your white walls, are showing any shading, which they do, then that means they're rendering at a brightness that's less than 255 per channel, or less than 100% intensity. Multiply that by 4% for angles less than 28 degrees, and you'd get a very dim reflection, just like the real thing. The error isn't in Dielectric; it's in either your expectations or your lighting.


yes IOR is clearly wrong, but obviously you can guess first thing i did was ranging from 1 to 2 :) and results were not improving, reflection wise.
Here's the reflectance for Diamond, which has a refraction index of 2.30804:
http://refractiveindex.info/?group=CRYSTALS&material=C
114910
Here you can see that Diamond is less than 25% reflective up until just under 40 degrees.

Here's Silicon Nitride, which about as close to a 2.0 IOR as I could find - 2.00848.
http://refractiveindex.info/?group=CRYSTALS&material=Si3N4

And its reflectance graph:
114911
This is less than 12% reflective up to just over 40 degrees.

Refractive surfaces aren't as reflective as people generally think, but that's because they're used to seeing them under conditions where there's light sources that would be overexposed if shot under the same exposure settings that the refractive object looks good at. 4% of a 100% diffuse white wall will give a reflection that's only 4% as intense, but if that wall was lit so it was 4 times brighter, its reflection would also be 4 times brighter. The wall would just be getting clipped at 255 under normal 8-bits per channel output. You can try rendering in HDR and then adjust the exposure in Photoshop or HDRShop.



At this point i should imagine there's kinda a thin coating on the chair itself, made just to improve reflection. So the onyl way to get this material is mixing two materials, Dielectric node and another which provides coating (i dunno which one tough).
That's not the only way. My preferred method is to use Mike Wolf's Material Booster. This doesn't require an extra material. It takes the output of a single material, internally splits it into the four shading modes and then multiplies those by several user defined values before recombining them into a Material output. Just remember, if you boost the reflection, you should also decrease the transparency by the same amount to keep within the energy conservation rules.


He Who Still Hasn't Figured Out The Curve Inputs On That Material Booster Yet.

Pavlov
06-13-2013, 12:11 PM
hi,


No it wont. The chair will still have the same reflectance. What you're changing is the amount of light for it to reflect. Again, you need to think in HDR terms

well, i forgot to say that i made some cuts in the wall behind camera. From one of these, comes the light on the tissue on the right. I guessed that, with a 800% susnky, these cuts would have been visible, but it were not the case. In fact...


My preferred method is to use Mike Wolf's Material Booster.

.. that's what i did :) i raised a bit reflection (which would have been even lower without it).

thanks,
Paolo

Dave Jerrard
06-13-2013, 01:28 PM
Can you send me that chair? I'll give it a shot here.



He Who Needs A Break From The Stuff He's Been Working On For The Past Few Weeks.

Pavlov
06-13-2013, 02:34 PM
hi Dave,
i sent you the chair in PVT. BTW, thanks for the effort !
Paolo

Dave Jerrard
06-19-2013, 08:30 PM
Had a go at the chair this weekend.

Before I get into that, I want to show the source material for it. You've already posted these two images:
http://www.100matrimoni.it/uploads/kartellluisghost.jpg

http://forums.newtek.com/attachment.php?attachmentid=114675&d=1370220936

There's really nothing wrong with either one, except that the refraction in the second one is a bit on the low side. The interior of the room in that is actually fairly dim. I wouldn't expect any bright reflections on those chairs. If you compare the reflections on those with the reflections on the piano and the floor, you'll notice they're about the same. Now compare that with the reflections in the glass and the chair in this photograph:
http://www.kilkennypembrokehotel.com/files/img/rooms/bathroom_disabled.jpg

Notice that even at a pretty acute, glancing angle, the reflections in the glass on the left are not that strong. You can barely make out the tiles in it, since they are almost lost in the bluish-green color of the glass itself.

Likewise with the chair, where the only real reflections that you see are from the very bright light sources in the ceiling.


Now, on to your chair. You mentioned that you were using an IOR of 1.25, but the chair you sent was 1.3. I used what you had in the model. For a basis, I rendered the chair, cloned several times across a tile floor, under an HDR environment. I also added a half-meter sphere that has the same surface, to better illustrate the effects of the Dielectric, and to show the reflection of the environment. I turned off the Absorption on both, and fed the Dielectric directly to the Surface node, bypassing the booster, so we can see the refraction better. These images use the environment for the lighting; no LightWave lights are used here.

115044 115045

The first image is without Partial Internal Reflections, the second image has those turned on. You can see the amount of reflection does increase a bit with those turned on. In this case, the render hit was about 10% longer, if that. These are not instances. Instances don't work well with refraction (random chairs seem to get a recursion of 1 and it's not limited to Dielectric, but refraction in general), so I had to use actual clones of the chair.

There's not a lot of dark edges here, but then again, there's no absorption going on to create them. Any dark edges are the result of a reflection of the dark part of the environment. There are a fair number of reflections here. Let's add your Absorption of 100:

115046 115047

Again, the first image is without internal reflections, and the second has them on.

That looks better, and this is back to what your model had. At this Absorption, the ball became practically opaque, so all we see here is the reflection on the surface. Chairs now have a few more dark edges. They now look a bit more like this:
http://www.100matrimoni.it/uploads/kartellluisghost.jpg

Now let's apply Material Booster, using the 120% Reflection boost you had set up:

115048 115049

Again, the first image is with Partial Internal Reflection turned off and the second has them on. All further images will have them on.

It's a subtle difference. If you look at the first image here and compare it with the second image immediately above it, you'll see that the earlier image is actually more reflective. Partial Internal Reflections has a greater effect than a 120% boost.

Here's the chairs with your colors now. I've taken the boost off for this:
115050

They look fine and have some strong reflections in the back, and there's some nice rich colored luminous edging going on. The only thing wrong is the refraction. You mentioned these were supposed to be Plexiglass, so I rendered them with that refraction index, which is 1.4813:
115051

Everything just little more reflective and the refractions are notably stronger as well. Yes, there are differences, but that's to be expected since the critical angles have changed as well as the internal reflection angles (due to refraction), so we're seeing a different part of the backdrop reflected & refracted than before. Overall, this is much more reflective than the boosted 1.3 IOR.

However, I checked the website for this chair, and they make it out of injected polycarbonate (http://kartellstorela.com/shop/louis-ghost/), NOT plexi. That has an even higher IOR, of 1.5645. There's a reason it's used in optics, like eyeglasses.

So, I applied that to the chairs and got this:
115052

Now the surface reflections are getting even stronger.

Here's the same scene under three other HDR environments:
115053115054115055

Here they are under a Sunsky:
115056115057

In this case, I boosted the Sun's image in the Sunsky plugin to 1000% and used the Material Booster to boost the Specular value by 10,000%.

And here's the scene again, but under a few small bright light sources, made by using the Crust texture in Textured Environment, cranked to 10,000% opacity:
115058

I took the texture off the floor to better match this photo:
http://kartellstorela.com/shop/images/2087/


I didn't do anything to the geometry of the chairs, though they do have a few bits of internal geometry that should be removed. I cleaned them of for later testing, making them completely air-tight, and a single piece (except for the feet). I'll cover those in the next post.


He Who Took A Few Days Between Shots On This.

Dave Jerrard
06-19-2013, 08:43 PM
Ok, here's my version of that same chair. I cleaned the geometry up a bit, fixing internal polygons that shouldn't be there and a few long triangles, ending up with a single water-tight model that's a single piece (aside from the feet). I surfaced it from scratch, using a color gradient for instances, but I used the one that Pavlov had set up in his surfacing. I didn't bother with any bumps here since the original photos (http://kartellstorela.com/shop/images/2087/) looked pretty smooth.

I did the same progression as before, starting with no Absorption. Here's the chairs without Partial Internal Reflections, and with it.

115029 115030


Not a lot of difference, but I wouldn't expect to see much. The geometry is about all that's changed here, aside from the lack of bumps. Reflections & refractions just look a bit smoother.

On my model I actually dialed in an Absorption of 120 that I thought looked right, which is pretty close to what Pavlov had. Out of curiosity, here's the chairs rendered with half of that - and Absorption of 60 - just to see how that looked. From now on, I'm only showing renders that used Partial Internal Reflections.

115031

At 60, the half meter ball is still a bit transparent. You can see some refractions of the bright windows through it. I also left the color applied to the chairs here. The color is a straight feed from the original color gradient. At this absorption, the colors are a bit muted.

Now for the 120 I found to look good:

115032

This is close to Pavlov's again. The saturation range on this is about the same as on the renders in the previous post. Some nice strong reflections again. And this is still with an IOR of 1.3.


Now I bump that up to the 1.49 that Plexiglass has:

115033

Naturally, this shifts the reflections around on the interiors. The bright reflections in the legs of the pink chair in the upper left are gone, since they were actually on the inside edge, but we see much brighter reflections in the surfaces of the yellow chair in the upper right.


And now for polycarbonate, which has a higher refractive index of 1.58, which is higher than some types of glass and higher than quartz.

115034

Again, reflections shift around, but they get stronger.

Now I started to play a bit with the colors. I ran the original gradient into a ColorTool, and boosted the saturation to 200%. This is a useful trick to get a more acrylic look.

115035

Here's a few variations on the environment. I'm still using the more saturated colors here.

115036 115037 115038 115039 115040 115041

The first three are HDR photos, the fourth is the Crust texture as an environment, and the last two are CGI HDR environments.

Since I was already playing wit the color a bit, I decided to bump the brightness to 200% too. The result looks a lot like fluorescent plastic.

115043

Since fluorescence is a result of UV light causing a material to re-emit light in another wavelength, in addition to what it's already reflecting in visible light, and LightWave doesn't calculate UV light, this isn't really breaking the law of the conservation of energy. Fluorescent materials look so bright because they're both reflecting visible light and emitting their own light from their reaction to invisible UV light.


He Who Spent A Bit More Time On This Than Planned.

Dave Jerrard
06-19-2013, 08:49 PM
Here's some closeups of the main geometry that I cleaned up.

In these images, you can see the edges aren't actually connected, and there's a long triangle down the side of the leg on the left.

115059115060115061115062


Here's my version of the same area.

115063115064115065115066


He Who Likes Rendering Refractions For Some Reason.

allabulle
06-19-2013, 09:17 PM
Wow.

Tobian
06-21-2013, 04:58 PM
Beautiful examples Dave, thanks for doing this. The only real crits is the blow-outs on the render (but they can be handled by tonemapping, rather than clipping, as LW does by default)

Part of the problem is that the specular is a little wrong too, and this is true in most CG: have a look at this sigraph paper from Disney last year...

https://disney-animation.s3.amazonaws.com/uploads/production/publication_asset/48/asset/s2012_pbs_disney_brdf_slides_v2.pdf

Page 23, they call it 'tails' - Spec (and reflection) don't really get this effect, where you get a subtle smear of reflection round the main hotspot. That's really hard to fix! It would require maybe some function-like falloff, but reflections and materials don't have inputs.

I hadn't really understood what the 'partial internal reflections' tickbox before now.. neat, thanks!

Pavlov
06-21-2013, 05:02 PM
Dave, awesome in-depth test !
I'm in a terrible deadline now (and i keep meeting an issue, i'll write a post in apposite section), but i'll read carefully and ask some questions whch remain not clear asap.

Thanks a lot,
Paolo

lardbros
07-03-2013, 06:12 AM
Think these pages need to be ripped and put onto the LightWiki page... with Dave's permission of course! :D

Pavlov
07-03-2013, 07:18 AM
still have to find some time to answer, cant just write a "ooh, nice" post. Quite interesting one, and some questions to make, mainly on the back boxes at the base of each leg.
Yes, reflection seems to work well and this proves lighting conditions are more important than i thought, but yet i find something odd in absorption, but i want to make some in depth tests before writing. Maybe they depend just on geometry not being totally watertight.

bye
Paolo

Tobian
07-03-2013, 08:13 PM
Well... I had a little idea, based on this thread, about how to solve some of the issues mentioned here, about the trickiness of absorbtion and manipulating 'fixed' materials...

So I give you material Tweaker! :)

Michael Wolf whipped me up a version 2.0 of the Material booster, which works now with colour and more advanced modes. Want to tint a Dielectric with a colour? Multiply the Refraction with the colour.. Simples! :) Want to fake a thin film.. make a rainbow incident gradient, and multiply the colour with the reflection and Spec output.

Currently in Beta, but I am sure there'll be a version for general release soon! :)

djwaterman
07-03-2013, 09:01 PM
Thanks for the combined efforts, and looking forward to it.

Andy Webb
07-04-2013, 03:33 AM
:thumbsup:

lardbros
07-04-2013, 06:49 AM
Ooooooohhhh!! :D

Thanks Tobian and Michael! :D

Matt
07-04-2013, 08:41 PM
Great in-depth tests Dave.

spherical
08-01-2013, 09:51 PM
While searching for data on recent, if any, changes in IFW2, one topic popped up that is dear to my heart--this one. As many of you may remember, back in the days of the 9.x and 10.x Open Betas, I was interested in, and instrumental in assisting development of, the Dielectric material we have today. Here's the test that I put every iteration of Dielectric as it was developed through. I think you'll remember it:

116056

I was particularly interested in not only correct Fresnel and Total Internal Reflections but also Partial Internal Reflections. I work with glass IRL, so have many examples with which to compare results. We have our own hotshop and use LW to fully flesh out a blown glass design before we fire up the furnaces to make it for real. Accuracy is important.

What eventually came from Antti, with a couple of nice bug fixes recently from Mark, is a fine Real-World material. It even has a non-Real-World component that I suggested and called Fast Dielectric; where you can now turn off Partial Internal Reflections for non-critical renders or objects in the distance where you wouldn't see them, and get a speed boost.

When this thread presented itself, naturally I was interested to see what was going on. Because most people just don't fully understand it, Dielectric can be thought to be broken when it really isn't. Jen, Matt and Dave have done their usual superb in explaining/demonstrating why it is what it is and that it is not broken. However, there are a couple of further points that were not addressed that may help clear the air, if not the glass.

First: Total Internal Reflections

In the test render, you can see that TIRs work, as does Fresnel. The right hand wall of the cube does not show the green and gold rods that are past it, but rather the blue rod with the black sphere (blue rod's shadow is turned off for clarity, as it was getting too complex to sort out with all shadows on). Don't be confused with the upper red rod appearing to be seen through the TIR at the right and through the correct Fresnel First-Surface reflection to the left. That rod is at camera level. You can see that this is true by the highlight on it in the TIR on the right wall.

Partial Internal Reflections are on, as can be seen by the ghost images of the upper white ball, the three center spheres and the far reflection of the yellow rod appearing just under them on the back and side wall. It is correct.

Where the TIR principle comes into play with the orange bowl is that the glass is essentially a fiber optic. It was speculated that light entering from the top and bottom of the bowl would light up the edge. Not really. TIR prevents it from making the "turn" to shine out the edge. A ray may be heading for the edge when it enters but it is refracted; changing the angle. Once the internal ray passes the Critical Angle, it instantly flips internally and goes out the opposite side or bounces around until it encounters an incident angle that is under the Critical Angle and then exits. There is no getting around this. Short of internal scattering from impurities, that we are not dealing with in this material, no rays that enter the sides of a higher IOR medium from a lower IOR medium will exit through an adjacent side.

This is how fiber optic filaments work. The TIR of the low incident angle rays keep the rays inside the filament until it reaches the other end. They just keep bouncing off the side walls until released.

This brings into play: Ray Limit

I opened Matt's orange bowl scene and rendered it, untouched. The Ray Limit was set at 6. I upped the Ray Limit to 64, the current max in LW (why?) and rendered again, not touching anything else. The results are below.

116057 116058

Flipping back and forth between these two bowls, you will see not only brighter orange appear in the edge at and near its center, but through the bowl walls as well. With the exception of secondary TIR bounces that create bright spots to the far left and right, the edge really won't get much brighter farther from center, as the Fiber Optic principle, if you will, sends those rays off to the left and right. Rotate the bowl around and the brightest portion of the edge will always be at the center.

I would wager, however, that there would be more orange farther out and brighter in the center if the Ray Limit could have been increased past this arbitrary cap of 64. There is no point in having a cap for this, IMO. The limit is there to decrease render times from needless recursions. If you want them, though, that is another story. Reminds me of the 640K that no one will ever exceed memory statement.

Additionally, when working on a glass top for an orrery, I noticed that the Ray Limit would influence the brightness of the edge and what could be seen in it, depending upon not only the limit itself but, once past the limit to begin to show brightness, whether it was odd or even. If the rays from the light were limited at a point when they did not reflect in the lower surface, the edge would go dark. No matter how much farther up I set the limit, the odd/even would control the brightness.

This brings into play: Absorption

It was speculated that Dielectric treats the absorption as if the object were a block and assumed that this is incorrect. Well, it's a bit more complicated than that but is correct; but in another way. The Fiber Optic principle is directing the rays along the entire cross section contour of the flared bowl. A 12" dia. bowl can easily have a 24"+ effective dia. by measuring the length of the path that the rays actually follow. This will make the color more saturated and darker, purely because of the "size" of the disc it is passing through.

This brings us to: More Control

While there is the existence of the PIR switch, this is as far as I think that the material should progress, featurewise -- as a Real-World material. Not having control is what it's all about. It is glass and similar. It should stay that. If you want additional coatings to alter the appearance, then apply them in the same way that is done IRL; by adding another material to the surface through the addition of a node. Making it a one-size-fits-all material just becomes confusing and definitely fosters what could be otherwise considered perversion away from the accurate nature of it. It took a lot to get it to where it is. Learn what it needs. Relish in the results. Leave it alone.

And finally: Painting What the Mind Expects to See

This is a principle that I came up with that makes some of my paintings successful. Determine what the viewer thinks should be and give it to them. A lot of the complaint posts about Dielectric stem from the above principle. People "think" that the reflections should be brighter but, upon measurement of same, aren't. It is the ability of the eye to constantly change "exposure" that fools the mind into thinking that things are "better" than they really are. The reflections may appear brighter because the surrounding environment is darker and the iris opened to compensate; providing a better "exposure" for the retina. The reflections stayed the same. IF that is the aim, then go for it. Come up with a MyDielectric node if you like, where accuracy is in the back seat, or running along outside, and you can tweak the heck out of it all to your heart's content. Like has been said, CG 3D is all fake.

aquinde
08-09-2013, 09:52 AM
Is anybody else having problems with spectral dispersion, as in Prisms, in dielectric with 11.5.1 (Mac)?

After reading through this excellent thread I feel that my little prism test scene should work just fine. I get nice refraction but the white light shooting into the prism is shooting out white, no matter how high I crank up the Refraction Dispersion %.

Anybody have any ideas? Scene files are attached if anyone cares to have a look.

Thanks

116250116251

Tobian
08-10-2013, 05:40 AM
Ok to get any kind of dispersion, you need to have *some* roughness value, otherwise the shaders can't split the rays up, to spectrally shift them. While LW dispersion is not physically accurate, I wouldn't recommend putting it over 100%, and indeed, the higher you put it up the more obvious the fact it's just 3 channel splits. It's nice to see that the caustics effect is picking up on the colour shift, though it's still a bit meh :) I also reccomend chamfering your prisms just a tiny bit, so their edges pick up on some of the light... You can get some caustic-like effects, if you use luminous geometry (not lights) and directional rays enabled in radiosity, but it won't do the colour shifts.

aquinde
08-10-2013, 07:23 AM
Ok to get any kind of dispersion, you need to have *some* roughness value, otherwise the shaders can't split the rays up, to spectrally shift them. ...

Thanks Tobian. How ironic that you need roughness for optics in LW.
Anyhow, as little as 1% and my prism test starts to look like something.
This bit of info, plus that very handy Material Boost node from DB&W Tools plugin, and I'm good to go.

116276

Tobian
08-10-2013, 08:46 AM
Not really ironic, to make realistic properties, you need some roughness, to disperse the min and max values. In physics there's no such thing as zero perturbance on a surface.

aquinde
08-10-2013, 08:57 AM
Not really ironic, to make realistic properties, you need some roughness, to disperse the min and max values. In physics there's no such thing as zero perturbance on a surface.

True. I realized that once I thought it through a bit more.

S0nny
10-03-2013, 01:30 PM
Really informative thread, good job!

Anyway, I have a question, maybe not strictly related to node proprieties but, here we go: how do you guys manage to save png32 with transparency without having 'holes' in the object using dielectric node?

Here an example using the bowl, try open it in photoshop and you can see the alpha literally creates holes at certain angle of material.

Lewis
10-03-2013, 02:42 PM
Try saving TGA32b, Photoshop removes alpha forom PNGs wihtout asking :(.

Tobian
10-03-2013, 02:49 PM
Yeah that's a bug with dielectric. I usually disable the alpha for dielectric materials, because of that. If you go into advanced options for the surface, change the alpha value to 'constant value' and set it to 255 (pure white). That makes it opaque, which causes other issues, but I am not sure what to suggest there.

S0nny
10-03-2013, 05:34 PM
@Lewis
Just tried, but the holes are still there. The only difference I can see is that PNG32 has embedded alpha/transprency and TGA32b has a separate channel. You can see the alpha channel in attach.

@Tobian
Your trick seems work... but, if I need to put a backdrop image in pp behind a dielectric object I completely lose any transparency.
example: imagine a dielectric in front of a window, if I need to replace the backdrop the dielectric doesn't overlap correct because is not transparent anymore.

Compositing in this way can be a pain, I had many problems because of this bug, in Photoshop and AfterFx.

My workaround with static renders is, if the holes are tiny and not too much noticeable, just put a backdrop that match the light/ambient in the scene.
In other cases, I overlap as many layer of the same render (or just the portion I need), to get transparency 'less transparent', because the holes are not 100% value, it sums for every layer on top. I hope I explained myself well...

Tobian
10-03-2013, 07:04 PM
Yes, like I said it's not a great solution if you need to composite things. The best solution is to simply try and do your compositing in Lightwave, which I know is far from ideal if you need to make a lot of adjustments.

Jim M
10-04-2013, 02:16 AM
"Photoshop removes alpha forom PNGs wihtout asking"

Not in my world.

...

Joining the party late... Have you tried putting a small amount of absorbtion on it? Also you could play with setting the alpha to srgb etc to change the alpha value space.

Lewis
10-04-2013, 02:20 AM
"Photoshop removes alpha forom PNGs wihtout asking"

Not in my world.

It think it started to happen around CS3 or CS4. Before that it was fine. Now whenever I drop 32bit PNG (with alpha ofc) it show/loads it without background while in 32bit TGA it puts alpha channel normally and shows all with background so user can remove it at wish. What photoshop version you use?

Jim M
10-04-2013, 02:39 AM
Ah your phrasing is confusing..

You mean it embeds the alpha into the image rather than creating a seperate alpha channel.
i.e. it gets rid of the pixel data... not the alpha channel ;)

Lewis
10-04-2013, 02:44 AM
Ah your phrasing is confusing..

You mean it embeds the alpha into the image rather than creating a seperate alpha channel.
i.e. it gets rid of the pixel data... not the alpha channel ;)

Well yeah sorry for confusion but yes it happens in Pshop while in LW it's looking normal.

Try to load this render (in attachment) in to Photoshop and then in LW image editor and you'll see there is no glass/background in Photoshop (and dielectric glass on table edges has same problem as S0nny have) while it's fine in LW. If you save it from LW into TGA32b then it'll be good in Photoshop also and you'll have Alpha channel layer for transparent surfaces as it should be.

cheers

Jim M
10-04-2013, 02:57 AM
Yep, Photoshop provides a different functionality for PNGs with embedded alphas.
Note that Photoshop has the functionality to have colour data stored in the transparency though ...Not that lightwave is accessing that.

S0nny
10-04-2013, 10:02 AM
Photoshop has the functionality to have colour data stored in the transparency though ...Not that lightwave is accessing that.

Exactly. I think that the problem could be solved if we could set the value of the alpha transparency directly from lightwave (maybe from the dielectric node?), because lightwave actually saves the colour datas in the png.
As you can see in these three images the value of the transparency is way off to be usable, but retains colour informations.


First image is the default png, you can see the gray part as bottom polygon, the white is transparency
Second image is the same png, with a bottom layer as background added in pp
Third image is a stack of 24 png one on top of the other, with the bottom layer as background added in pp



It's tricky but somewhat does the job, of course it's far from perfect, there are no deformation etc... For animations? I don't know, any suggestion?

Mr_Q
04-07-2014, 10:12 AM
Dave is a wonderful technical artist. Always impressive. If their ever was a "Shader's TD", it's him.

alexoctagon2009
04-28-2014, 10:25 AM
I've read through all this post, really interesting stuff. I pretty much agree with everyone, that the dielectric node is really good and seems pretty realistic, it's really helped my renders of bottles, but at the same time In my view it really does need a lot of extra bits of functionality to help get the result right to the eye.

Like many said, the way the colour darkens and changes colour with absoption needs a control adjustment, I come up against this all the time, as does splitting the reflection and refraction which I often want to be slightly different. Realistic or not, we're aiming for what the client wants to see above all.

The thing I want that I don't think has been mentioned on the thread is a transparency setting. I would like to create something semi transparent like an orange juice or limoncello type liquid, or a coloured plastic bottle on many occasions and just end up doing it the manual way. I think Tobian or Dave J said to combine it with an SSS shader that would be perfect and get the best of all worlds.

If someone at Newtek could add all these things into a new version of the dielectric node that would be great :)

Pavlov
04-28-2014, 11:05 AM
I've read through all this post, really interesting stuff. I pretty much agree with everyone, that the dielectric node is really good and seems pretty realistic, it's really helped my renders of bottles, but at the same time In my view it really does need a lot of extra bits of functionality to help get the result right to the eye.
)

Recently i solved using DB&W's Materials Booster, which can boost reflections independently, and give result which fits representation needs when realistic behavior doesnt seem... real :)


Paolo

bobakabob
04-28-2014, 11:27 AM
Yes, although Dielectric is as physically real as Lightwave can get it would be great to see a variant where you can mess with the parameters of 'realistic' glass. That's the fun of working in CGI - breaking the rules. I keep resorting to the ancient LW glass presets as they're far more flexible.

spherical
04-28-2014, 03:17 PM
Use Material Tweaker, found in db&w Tools (http://wiki.db-w.com/tools).

abdelkarim
11-09-2014, 01:25 PM
guys we know this dielectric has bug . in the exmple of chair easy . why not have 2 surfaces name . one all edge of the chair . and 2nd its rest of it . and u can make edge hard and other as u want . easy control .