PDA

View Full Version : I Hate Mr Area Light ...


ghostlake114
12-07-2006, 01:36 PM
Right, I dont want to raise any quarrel about vray and MR, or any offense, but this is the fact I encouter everyday, and I must say: I hate MR area light.
I know when I say that, many will say: NO, it s great, it could give soft shadow,fast setup with physical light, It could emit photon, it.... etc... Right, I know it s useful, but I still hate it.
I will go straight to the mater
========================

First, you could view this pic of slipknot66 (http://forums.cgsociety.org/member.php?u=90683) vbmenu_register("postmenu_1962277", true); . This is great pic of lighting, shading, rendering
http://www.cgnetworks.com/gallerycrits/90683/90683_1091670128.jpg
But, please look at the ceiling near the door, do you see the noise, the strange noise...
====
Another look at dagon tutorial
http://www.treddi.com/forum/index.php?showtopic=8436
You may have known it, you could follow this tutorial and tell how could you advoid all the noise at a aceptable time.
====
You may say that: No, it s acceptable time. So I will give me simple case, an interior render
============
First, I want to give vray and MR the sample condition and render method. I know it seem craz. But the balance method I think is

method of Vray is



Vray 1.5 core
IRR + Lightcache + expotional color mapping.
AA is -1,2 adaptive with area filter
1 Vraylight + directional light for sun.
Vray grey diffuse only material


And MR is



MR 3.5 core
GI + FG + buffer.api for color mapping
AA is -1.2 with box filter
1 MR area light,-physical light- emit photon + directional light for sun
ctrl_shading grey diffuse only with mib_photon_basic. I think this is fastest shader for diffuse color testing only.

====
And now, the only thing left, I try to compare render time and area light noise eliminating
=======
First, a picture of MR
http://i35.photobucket.com/albums/d152/ghostlake114/MR_32sam.jpg
32 sample in area light for noise reduction :(. Time 2.19 min
============
Second, a picture of Vray
http://i35.photobucket.com/albums/d152/ghostlake114/vr_12sam_2.jpg
12 sample (sub-div) in area light for noise reduction. Time 2.04 min
============
I dont want to discuss about math of 2 renderer, but only look at the way each solve with noise
============

Vray area light distribute noise the overall pic, and sample raise help us eliminate noise
MR area light distribute noise the overall pic but noise is more density at the area light edge. That s way we raise sample, the noise at other is out, but the noise at light edge still remain and a very high sample still not kill all of them...
====
It s just my comparision and at this time, I cud not find a way to elimiate the noise with less time. Or maybe my technique is too bad to use MR :(.
But the fact area light noise distribute and how to reduce them in MR is terrible :buttrock:
================
PS: This may be a long thread, hope anyone cud read throgh and help me find a way to get off this stuk :(. A method such as reduce contrast threshold and raise light sample is okie, but I dont want to trade so many render time like that :(:shrug:

dagon1978
12-07-2006, 03:31 PM
ghost, your comparison is a little bit strange on my eyes:

-if you want to compare mr_arealight with vr_arealight why you introduce so much factors?
a better approach could be to disable the Global Illumination and tonemap, and compare the difference in a simple raytracing render
another approach could be to use just the Irradiance Cache (IRM-vray, FG-mray), with similar settings (the FG in mray3.5 is very similar to the vray IRM) and compare the difference in a quasi-gi mode
i made this kind of tests many times, i could say, the new mray area lights can be faster in some scenes and slower in others.. rendertimes aren't so different compared to the vraylights

-you may use an harder scene to test, or an higher resolution, because 10/15/20 seconds on a 2minutes rendering scenes aren't really important (could be a matter of faster/slower exporter etc), try at least a 15/20 minutes test

fabergambis
12-07-2006, 03:46 PM
Please, can you post your MR Area Light (Hi - Limit - Low) Samples? You'll probably find the solution there...With a 25 -2 -5 setting I usually get good results in not much more rendertime.

Quote [dagon]: -you may use an harder scene to test, or an higher resolution, because 10/15/20 seconds on a 2minutes rendering scenes aren't really important (could be a matter of faster/slower exporter etc), try at least a 15/20 minutes test - END Quote.

You're right dagon, but he's probably pointing at the grainy shadows on the ceiling of his test room with MR, compared to the clean soft shadows of the Vray version; ghostlake spent much rendertime to obtain a grainy result with MR. That's the matter I think...

dagon1978
12-07-2006, 05:49 PM
You're right dagon, but he's probably pointing at the grainy shadows on the ceiling of his test room with MR, compared to the clean soft shadows of the Vray version; ghostlake spent much rendertime to obtain a grainy result with MR. That's the matter I think...

i saw, but the adaptivity of the new mr area light (in maya8) works quite well, so, you can increase the samples without a great cost for rendertimes
comparing 2 minutes with (let say) 3 minutes is not a great idea for a good test, especially if you have so many factors in your scene

ghostlake114
12-07-2006, 06:11 PM
@dagon:
I include too many factor for the real comparision in working. GI+FG is the good solution for interior like IRR+Lightcache. The tonemap is only additional for quality purpose.
AA and filter is in the same condition
====
With all that thing, the area light in each renderer works so different. The way MR light distribute noise and eliminate them is really problem that I want to show.
====
I could say in this case, I even raise high sample of MR area light to 100 and low is 25. The noise still somewhere in the light edge.
------
Ya, light edge noise distribution is so anoying, compare with overall noise distribution of Vray light.
==

dagon1978
12-07-2006, 06:23 PM
i have a completely different opinion, and i'm quite sure of what i'm talking about
if you include GI in your test you can't control the time contribution due to the area light sampling, especially if you use such a low rendertimes!
low samples at 25 :banghead: make me think you dont know what you are trying to do, because in a diffuse scene the low sample is not working (is the samples level for reflective/refractive surface, where you see reflective surfaces in your render??)
and the low sample actually increase a lot the FG map calculation :banghead:

and this is a demonstration that less factors = less mistakes

ghostlake114
12-07-2006, 06:34 PM
Ya, high sample only test (100). But I do not get all what you say. What do you mean with "include GI in your test you can't control the time contribution due to the area light sampling, especially if you use such a low rendertimes". I include GI+FG just want to give a total test in a real work, not an independent test only with Area light.
===
The Vray is for demonstrating only, show how vray deal with light sample in a real work.

dagon1978
12-07-2006, 06:46 PM
I include GI+FG just want to give a total test in a real work, not an independent test only with Area light.

and this is the problem, because if you wanna test how an area light work you have to start with a basic light setup, without GI and without any other factor (like a direct light, for what?)
how do you think the GI+FG setup changes the behaviour of an area light (if you use it correctly!)?

ghostlake114
12-07-2006, 06:56 PM
For this simple scene, I try to keep GI and FG in smooth render without GI artifact. As you could see, the only left is the noise come from area light. I do not think this come out from MR area light. --- SORRY, GI AND FG, not MR area light, MY MISTAKE CUZ THIKING WHEN TYPING
The fact, I also view many interior render (not yours, dagon :) ), and they share the same problem as slipknot66 (http://forums.cgsociety.org/member.php?u=90683) in my first post, the weird noise artifact in area light edge :(

dagon1978
12-07-2006, 07:02 PM
For this simple scene, I try to keep GI and FG in smooth render without GI artifact. As you could see, the only left is the noise come from area light. I do not think this come out from MR area light.
The fact, I also view many interior render (not yours, dagon :) ), and they share the same problem as slipknot66 (http://forums.cgsociety.org/member.php?u=90683) in my first post, the weird noise artifact in area light edge :(

yeah, and this is due to the lack of previous mr area lights (points)
the new adaptive system works very good, i can post an example in an heavy scene

ghostlake114
12-07-2006, 07:08 PM
yeah, and this is due to the lack of previous mr area lights (points)
the new adaptive system works very good, i can post an example in an heavy scene


===
It seems interesting, adaptive system. Could you be more specified, dagon...

dagon1978
12-07-2006, 07:41 PM
ok, this is an example of what i'm talking about

Vray 1.5 RC2 for 3d Max
AA:
Adaptive Subdivision -1 2
Mitchell size 4

vraylight

samples 8 (default) 10:44
http://img245.imageshack.us/img245/4231/vray8s10m44sku1.jpg

samples 25 15:48
http://img245.imageshack.us/img245/1867/vray25s15m48sud3.jpg



Mray 3.5 for Maya
AA:
-1 2
Mitchell size 4

physical area light

samples 8-1-1 (default) 3:13
http://img245.imageshack.us/img245/7871/mray8s3m13szp6.jpg

samples 32-1-1 5:56
http://img245.imageshack.us/img245/6382/mray32s5m56sms0.jpg

samples 256-1-1 9:33
http://img245.imageshack.us/img245/7572/mray256s9m33szt7.jpg

ghostlake114
12-07-2006, 07:48 PM
Sorry dagon, so you raise the sample to 256 for noise eliminating
So what do you mean about adaptive system in new MR 3.5 ??? :love:
===
By the way, the render time in this case is so cute :).

cpan
12-07-2006, 07:59 PM
heya dagon :)
never used vray but i have a feeling vray somehow uses the indirect ilumination map ( "fg" ) to smooth out the noise in the shadows...?:curious:

could you test vray with
> a indirect illl OFF and area shadows ON scene and
> a indirect illl ON and area shadows ON scene
to clear this up?

salut,
calin

dagon1978
12-07-2006, 08:00 PM
i mean, with the new area light you can increase the samples without high-time costs
this was impossible in the past area light (point), i usually used some kind of tricks (AA/contrast threshold/ect)
the real difference between the mr area light and the vray lights is visible in simple scenes (like your), the mr area light dont like a surface too close to the light (like the windows) the vraylight instead works better (seems like a cutoff to the closest object, if they aren't important)

but if you move you exclude the vertical part of the windows to the shadows and you scale a little bit down your area light the mray render can be much faster :D

dagon1978
12-07-2006, 08:05 PM
heya dagon :)
never used vray but i have a feeling vray somehow uses the indirect ilumination map ( "fg" ) to smooth out the noise in the shadows...?:curious:

could you test vray with
> a indirect illl OFF and area shadows ON scene and
> a indirect illl ON and area shadows ON scene
to clear this up?

salut,
calin


mmm not with a default setup
but, vray has an option to render the vray shadows with the IRM, this is easily visible in the vray benchmark, where the shadows are very smooth, but have some kind of artifacts ;)
http://img354.imageshack.us/img354/2066/cornelltestnonover7lv.jpg

but, it's possible that the vray light re-use the path of the indirect illumination for accelerate the render, i dont know if it's possible...

ghostlake114
12-07-2006, 08:15 PM
@dagon: ya, agree I never imagine 256 sample is fast like that. In the past with MR 7, 100 sample mean over killing and I never think I cud raise it too high...
However, what I also want to mention here is the total lighting solution :love: . Yes, agree that the Vray light and MR light go so far when they are alone. But in my test (correct me if I am wrong, dagon :rolleyes: ), with IRR + lightcache turn on, the additional time is only a little. Ya, I mean with some kind of 300 lightcache and -3,-2 (or -1) with 50 subdiv. The time for a total pic is only 15 + 5-10 min (I guess that in your scene)
===
But it s a large different when we turn GI and FG on in MR...

dagon1978
12-08-2006, 01:59 AM
But in my test (correct me if I am wrong, dagon :rolleyes: ), with IRR + lightcache turn on, the additional time is only a little. Ya, I mean with some kind of 300 lightcache and -3,-2 (or -1) with 50 subdiv. The time for a total pic is only 15 + 5-10 min (I guess that in your scene)
===
But it s a large different when we turn GI and FG on in MR...

the vray GI is highly optimized, i mean, the area light sampling has not influence in a GI map generation (LC, IRM, etc)
in mental ray when you increase the Area Light sampling you increase the FG map calculation too (here's why it's important to use a right setup with a right low-samples)
the vray approach is surely helpful and i hope this "philosophy" can be integrated to the netx-gen mental ray (they made some step on the right direction with the 3.5, better area light system, much optimized materials, etc)

on the other hand, raytracing in mental ray is faster, simply
so, mental ray could be as fast as vray, also with less optimizations (and this is true for the GI algorithms too)...

this is for yashu, vray:


8 samples (no GI) 10:44
http://img245.imageshack.us/img245/4231/vray8s10m44sku1.jpg
8 samples + IRM (low) 10:49
http://img80.imageshack.us/img80/1732/vrayirmlow8s10m49sys7.jpg


25 samples (no GI) 15:48
http://img245.imageshack.us/img245/1867/vray25s15m48sud3.jpg
25 samples + IRM (low) 11:57
http://img80.imageshack.us/img80/839/vrayirmlow25s11m57sjf7.jpg


and now mental ray

32 samples (no GI) 5m56s
http://img245.imageshack.us/img245/6382/mray32s5m56sms0.jpg
32 samples + FG 9:52
http://img80.imageshack.us/img80/2866/mrayfg32s9m52sbg1.jpg


256 samples (no GI)
http://img245.imageshack.us/img245/7572/mray256s9m33szt7.jpg
256 samples + FG 14:01
http://img80.imageshack.us/img80/9167/mrayfg256s14m01shf7.jpg

i think that vray uses some technique to re-use the GI paths on the area light calculation, but surely the vraylighs are not smoother with the IRM

MaNdRaK18
12-08-2006, 03:13 AM
If I use mib_cie_d as light shader in mr area light - shadows don't work anymore. Why?

dagon1978
12-08-2006, 03:44 AM
If I use mib_cie_d as light shader in mr area light - shadows don't work anymore. Why?

are you using the mib_cie_d in the color of the area light?

MaNdRaK18
12-08-2006, 03:48 AM
are you using the mib_cie_d in the color of the area light?

Nope, I thought I should use it as a light shader?
So, Dagon, should I map it to light color instead?

dagon1978
12-08-2006, 03:52 AM
Nope, I thought I should use it as a light shader?
So, Dagon, should I map it to light color instead?
yes, this is the right setup ;)

MaNdRaK18
12-08-2006, 03:55 AM
Thx a lot, man.
X-tremely fast reply ;)

ghostlake114
12-08-2006, 07:08 AM
the vray GI is highly optimized, i mean, the area light sampling has not influence in a GI map generation (LC, IRM, etc)
in mental ray when you increase the Area Light sampling you increase the FG map calculation too Ya :thumbsup:
I am very interested of your test, dagon, that explain what I am thinking. So, the impact of lightsample on final render with VRay (IRR (+ lightcache) turn on) is quite not much. But the impact of MR area light sample is quite clear and it raise our render time a lot.
========================
the vray approach is surely helpful and i hope this "philosophy" can be integrated to the netx-gen mental ray (they made some step on the right direction with the 3.5, better area light system, much optimized materials, etc) Totally Agree :D , a sample raise in MR area light but cause lesstime increase when FG,GI turn on will hard rock MR.
Again, we wait for MR 3.6 :sad: :shrug: .
PS: just a personal idea, why MR technique is always "old" in comparision with Vray, we have talked toooo much about lightcache - a super GI canculation,VrayMTL -compo pack, a now is impact of MR area light to GI canculation :rolleyes:

ghostlake114
12-08-2006, 11:10 AM
in my simple example case, it s headache when I must raise sample to 256 for noise eliminating. Time for a 640x480 diffuse white is 20 min now
With Vray, for example
http://www.3dvn.com/download/tutorials/max/tut_vray/lighting/noithat.jpg
1600x1200 of this kind of quality cost only 16min in the same PC
===
I still not understand much about render speed of simple scene in comparsion with complex scene... As Dagon mentioned, Vray s so fast for simple scene as mine, but they are the same with complex scene or MR even faster ? (condition for comparing is as above)

Gal
12-08-2006, 11:40 AM
interesting topic,
i wanted to give you guys another issue to ponder about. although i dont have access to vray i have noticed that vray area shadows act much more like an object casting indirect illumination in mental ray rather than source light cast shadows.
take a look at dagon's example i've marked the 4 images with dots of interest.
(http://www.galroiter.com/temp/shadows.jpg)
http://www.galroiter.com/temp/shadows.jpg

even though you can still see the direct shadow indication on the red/green dots on the vray gi render it is still almost invisible and infact it is almost the same value as the surface around it just without noise. compare that to the mental ray fg render, the direct shadow from the area light is just a cast shadow and had no affect on the fg light bounce calculation. what i found in mental ray is no matter what you do (and i believe here is where vray is different) cast shadows are cast shadows multiplying whatever indirect illumination does and always always creating this unrealistic independant shadow. vray on the other hand(and maybe ever single render i've seen was manipulated and i'm wrong) takes cast shadow calculations with indirect illumination as if the shadow was indirect itself,not a cast shadow. the ONLY way to get the same result with mental ray is to use an object light emitter and ofcourse..to get a smooth render you need an impossible sample/radius. for this same scene if you replaced the area light with a white plane it would take about 7,000 fg samples and a radius of 0.005/0.010 and around 2 hours rendertime and still a lot of shadow detail will get clipped.

1 point to consider, notice that the yellow dot on both engines results in the same value which is a surface at an angle higher than 180 degrees(which i think the renderers dont calculate shadows on it but just clamp it to 0% light intensity of the area light, either cause this is the math or to optimize) and on both engines it's the same value.however since the green/red are different it means this is a direct cast shadow problem in mental ray when combined with FG.

this ofcourse is just a theory of mine but i've seen it so many times and frustated me dozens of times. i still have not found a solution to remove or blend cast shadows from area lights into a nice soft indirect illumination light setup, only way is to use objects as light emitters for FG(works best but too slow be useful by itself).
i'm very curious to see what dagon says about this.

ghostlake114
12-08-2006, 12:10 PM
Amazing analys
===
So it s an unequal in MR when we switch FG combine with MR area shadow.

what i found in mental ray is no matter what you do (and i believe here is where vray is different) cast shadows are cast shadows multiplying whatever indirect illumination does and always always creating this unrealistic independant shadow

==
So, It s a weak point of MR now and we could not combine them in a fast render time???
===
@Gal: Simple add the image path with http, ex: http://www.exam.com/01.jpg

Saturn
12-08-2006, 01:22 PM
what i found in mental ray is no matter what you do (and i believe here is where vray is different) cast shadows are cast shadows multiplying whatever indirect illumination does and always always creating this unrealistic independant shadow



This is more a shader problem than a MR problem. Most of shader compute the shadow on direct illumination and then ADD the irradience ( or anything that simulate irradience eg: Ambient Occlusion ) to the result.

And actually if you look to the dot you pointed if the shadow was multiplied by irradience you will still see it black. The problem here is maybe a missing bounce in FG or something like this.

Chris-TC
12-08-2006, 02:44 PM
i mean, with the new area light you can increase the samples without high-time costs

These new area lights that you're talking about, are you sure they're new to mental ray? Or is it something that was newly introduced to Maya?
I don't remember reading anything about improved area lights in the mental ray 3.5 specifications.

I'm asking this because the integration of area lights in Maya seems to be different from that of XSI (we can only specify two values: Samples U and Samples V).

ghostlake114
12-08-2006, 02:45 PM
This is more a shader problem than a MR problem. Most of shader compute the shadow on direct illumination and then ADD the irradience
So you know any shader that do the "vray" thing? I really curious

Saturn
12-08-2006, 03:20 PM
So you know any shader that do the "vray" thing? I really curious

Actually Vray should do that too. What you are looking for it s something that mimic Vray's indirect illumination. In this exemple more light is coming from an indirect bounce in the shadows part.


By the way you will have to be patient because in the next release of T2S shader you will have access to a new soft shadows features :)

dagon1978
12-08-2006, 03:29 PM
i noticed this strange problem yesterday, i was using DGS for this render (cause the scene was already done with it some times ago), is a long time i dont use dgs, i can try to do a test with a different shader


These new area lights that you're talking about, are you sure they're new to mental ray? Or is it something that was newly introduced to Maya?
I don't remember reading anything about improved area lights in the mental ray 3.5 specifications.

I'm asking this because the integration of area lights in Maya seems to be different from that of XSI (we can only specify two values: Samples U and Samples V).


for what i understood it's a new maya shader, not a new mental ray feature
the old area lights (points) use U-V samples in maya too

dagon1978
12-08-2006, 03:55 PM
Ya :thumbsup:
I am very interested of your test, dagon, that explain what I am thinking. So, the impact of lightsample on final render with VRay (IRR (+ lightcache) turn on) is quite not much. But the impact of MR area light sample is quite clear and it raise our render time a lot.

for an high-res render you can compute your FG with few samples for the area light, freeze the map and then render the raytracing part
for a low-res or mid-res it's less important



Totally Agree :D , a sample raise in MR area light but cause lesstime increase when FG,GI turn on will hard rock MR.
Again, we wait for MR 3.6 :sad: :shrug: .
PS: just a personal idea, why MR technique is always "old" in comparision with Vray, we have talked toooo much about lightcache - a super GI canculation,VrayMTL -compo pack, a now is impact of MR area light to GI canculation :rolleyes:
vray has his problems too, especially with many raytracing rays (high glossy, high area light samples, etc), you can clearly see it on my tests, although his optimizations vray is slower then mental ray in an heavy scene

keep in mind, vray is a GI renderer, optimized for GI calculation, and surely it rocks for this kind of works, mental ray has a completely different target, but it can render GI quite well

ghostlake114
12-08-2006, 05:54 PM
I think I see all clear now with MR and VR in this comparision. Thanks all for your hep and info, dagon.
==
Vray agains MR... This seems an endless discussion :D

hominid
12-08-2006, 08:38 PM
Hi Saturn,

By the way you will have to be patient because in the next release of T2S shader you will have access to a new soft shadows features
Do you know if the new T2S shader will be ported to maya as well and will it be compiled into a 64 bit version?

Cheers,
Pete

Gal
12-08-2006, 09:03 PM
please check my shadow example on page 2, since it is your scene example. what do you think? i cant be the only person to notice this matter.

dagon1978
12-08-2006, 11:35 PM
Hi Saturn,


Do you know if the new T2S shader will be ported to maya as well and will it be compiled into a 64 bit version?

Cheers,
Pete
yes ;)

please check my shadow example on page 2, since it is your scene example. what do you think? i cant be the only person to notice this matter.
yeah, when i said "i noticed this strange problem yesterday" i was referring to this example
i have to investigate a bit more to make a judgment

Gal
12-09-2006, 12:02 AM
my mistake, thanks!

slipknot66
12-09-2006, 01:47 AM
Vray agains MR... This seems an endless discussion :D

Nothing to Discuss... lol

slipknot66
12-09-2006, 01:48 AM
i didnt read all posts, but was looking yur renders.. check your photons

KIDI
12-09-2006, 01:34 PM
hey Great post :thumbsup:

can u please teel me where i can find the buffer.api for color mapping .

This image made with lower the framebuffer direct from maya to 0.455 but it makes problems
where is a texture node.

Thanks

dagon1978
12-09-2006, 02:36 PM
my mistake, thanks!
ok gal, i'm trying the scene with turtle and i'm getting exactly the same results of mental ray
so, are we sure the vray render is the right result? it's surely nicer, but... we are talking about a FG/Irradiance-only render (1 bounce)

http://img95.imageshack.us/img95/6361/turtles25625m54ssf3.jpg
http://img95.imageshack.us/img95/8638/turtlefgs25625m54sxc4.jpg

sorry for the quality, i'm not a great turtle user :p

dagon1978
12-09-2006, 02:38 PM
hey Great post :thumbsup:

can u please teel me where i can find the buffer.api for color mapping .

This image made with lower the framebuffer direct from maya to 0.455 but it makes problems
where is a texture node.

Thanks

here (http://forums.cgsociety.org/showthread.php?t=209991)

ctrl.buffer v2.0 rc5 ;)

KIDI
12-09-2006, 03:12 PM
Thanks, u are the one :) .

Gal
12-11-2006, 09:05 AM
dagon well i am not sure about this turtle example since the shadow on the lower left part of the image, the right side gets illuminated by gi and the left has none(lack of bounce). but anyway i believe the vray way is much more correct.
ok think about these 2 scenarios. 1 is when you direct sunlight hitting an object above a floor. in this case the way area lights and cast shadows work fine.
however, 2nd scenario any diffused light situation where there is no visible direct light source.. which is atleast 50-70% of lighting scenarios. for example a room with a window as the only light source from the sun/skys above the room not shining through the window directly. basically any real life situation where photographers use diffuse lighting(almost all of them?) ..car shots, portaits.. so many more. if you look at a lot of the vray character renders you can see how the cast shadows diminish and blend into indirect shaded areas. ofcourse real life shadow isnt cast shadow just lack of light so the independant multiplying cast shadow from area/normal lights in mr or turtle or any other engine is wrong.
i understand this is 3d math and we cant expect real world light engines right now but how would you light a scene with diffused light but still get sharp shadows(diffused light origins can create sharp shadows they are not always soft shadows)? i said before only by using a plane/objects as emitters for FG but it's impossible to get detailed shadows even with the highest of settings/samples.
i dont know what i;m expecting really, i am just pointing out limitations and if vray does a better result (either it's intentionally or just the way it calculates) i would like to find a way to mimic that in mental ray.

thev
12-11-2006, 03:26 PM
You might find the following examples useful:

http://www.spot3d.com/vray/help/150R1/examples_lights.htm#ex4

As discussed there, area lights in V-Ray produce the same illumination as a rectangle with a self-illuminated material with the same color/intensity; and this is the exact same illumination that you get if you use skylight only with the same color/intensity. This is one of the simple tests that a GI renderer must pass successfully.

Best regards,
Vlado

Chris-TC
12-11-2006, 07:35 PM
You might find the following examples useful:

http://www.spot3d.com/vray/help/150R1/examples_lights.htm#ex4

As discussed there, area lights in V-Ray produce the same illumination as a rectangle with a self-illuminated material with the same color/intensity;

Well, theoretically the same should be true for mental ray. But V-Ray's various ways of calculating GI seem to be quite a bit more effective.

If you wanted to use Final Gathering in only in the example in your link you'd have to crank the settings beyond anything reasonable. So that leaves only area lights as a viable option. I'm kind of let down that mental ray hasn't evolved much in terms of GI for so long.

dagon1978
12-12-2006, 12:02 AM
You might find the following examples useful:

http://www.spot3d.com/vray/help/150R1/examples_lights.htm#ex4

As discussed there, area lights in V-Ray produce the same illumination as a rectangle with a self-illuminated material with the same color/intensity; and this is the exact same illumination that you get if you use skylight only with the same color/intensity. This is one of the simple tests that a GI renderer must pass successfully.

Best regards,
Vlado

ehy vlado! i was hoping on your contribution ;)
i do remember these tests, so, are you saying that the mray and turtle solution isn't correct?

dagon1978
12-12-2006, 12:59 AM
eheh after some tests i'm nearly sure that the shadow in the vray renders are wrong :d probably an irradiance-smooth problem

thev
12-12-2006, 09:50 AM
eheh after some tests i'm nearly sure that the shadow in the vray renders are wrong :d probably an irradiance-smooth problem<shrug> And I know for a fact that they are correct, so...? Area light shadows in V-Ray are correct in the sense that they do what they are designed to do. This is not necessarily what mental ray or turtle do. Further on, in 3dsmax, mental ray renders different things depending on if you use photometric area lights or mental ray area lights, and it does yet a different thing in Maya. So "wrong" would be a rather relative term.

Best regards,
Vlado

dagon1978
12-12-2006, 01:56 PM
Vlado i made some test with illuminate planes in vray and the result it's pretty similar to the mray (and turtle one)
the render is a bit slow (cause i'm using QMC) but i can post it tonight
so, it's wrong for what it's intended to do

thev
12-12-2006, 02:19 PM
Vlado i made some test with illuminate planes in vray and the result it's pretty similar to the mray (and turtle one)
the render is a bit slow (cause i'm using QMC) but i can post it tonight
As you wish, but you will not convince me - I've done countless tests myself with V-Ray already and have taken great care to make sure the results are mathematically correct. I believe I already made my point to other readers of this thread and since I don't have any comments other than the examples in the link above, I won't post anything else here.

Best regards,
Vlado

dagon1978
12-12-2006, 03:11 PM
Vlado, i can't see your point, i just trying to understand wich are the right result
i already said that the vray one look surely better, but seems strange
your examples are very good and explain the behaviour of the vraylight quite well
but i'm trying your test in my scene and it seems incorrect, this could be, or maybe i'm doing something wrong, you are free to correct me (also i'm not surely a skilled vray-users), and i would be glad to be corrected here

ciau

mat

p.s.
this is not surely a matter of irradiance because with vraylight and QMC the result is still different to the illuminated plane

ghostlake114
12-12-2006, 07:51 PM
If you wanted to use Final Gathering in only in the example in your link you'd have to crank the settings beyond anything reasonable. So that leaves only area lights as a viable option. I'm kind of let down that mental ray hasn't evolved much in terms of GI for so long.
Totally agree, while other engine give us more GI canculation way, faster or cheaper, or easier to handle, we MR user still suffer a complex GI :shrug:
====
I know we are talking about the complex of light model... But Vlado, could you explain why turn IRM on, in some scene in Vray, render time reduce, but with MR, FG on mean huge time for canculating FG

dagon1978
12-14-2006, 02:37 AM
i'm sorry for the delay, i had an hardware problem on my pc :cry:
this is a crop of the test i was mentioning before


http://img89.imageshack.us/img89/7851/matad1.jpg
it's a lighting pannel with QMC (128 samples)
there is also some strange problem in the corner, but the shadow under the window isn't soften like the vray light render

i hope to complete this test in the next days, i can share the scene if someone else want to test it

Fireantz
12-14-2006, 06:57 AM
hi, master Dagon. fire again. Glad to see u here.
since last month start to learn max n vray. so i m maya , max, mentalray n vray user now.:scream:
i think i can help u for the vray test.:thumbsup:

i didnt read all thread here. so pls shortly brief me.:)

dagon1978
12-17-2006, 12:56 PM
hi, master Dagon. fire again. Glad to see u here.
since last month start to learn max n vray. so i m maya , max, mentalray n vray user now.:scream:
i think i can help u for the vray test.:thumbsup:

i didnt read all thread here. so pls shortly brief me.:)


this is the scene
http://rapidshare.com/files/7850066/jap_vray.rar.html
in the pannel-QMC i forced the gamma a little bit
there is also a vraylight (off), try delete the pannel to see the difference

i would like also the vlado's opinion

Fireantz
12-19-2006, 07:13 AM
well...Dagon,in your vray setting. It giving me a very long rendertime.

For a vray user. mostly we won't use QMC as primaryGI.
here my vray setting for interior (a mostly vray's user setting).
-Irandiance map (-4/0)as my Primary.
-Lightcache as secondary bounces.
-vraylight with 32subdv sampling.

mmm...Dont know this is your answer u wan or not.

test:
http://img186.imageshack.us/img186/2738/lightcachenp6.jpg

dagon1978
01-14-2007, 01:49 AM
well...Dagon,in your vray setting. It giving me a very long rendertime.

For a vray user. mostly we won't use QMC as primaryGI.
here my vray setting for interior (a mostly vray's user setting).
-Irandiance map (-4/0)as my Primary.
-Lightcache as secondary bounces.
-vraylight with 32subdv sampling.

mmm...Dont know this is your answer u wan or not.


i know, my test was for the area light / pannels illumination, not for reduce the rendertimes

in your render you are using a vray light and i notice exactly the same differences i had in my tests
if you look at my previous render (the QMC one) you can see a different fading shadow on the floor, this is my point....
so, vlado, no answers?
this is not what i wish, this is what anyone can see ;)

dagon1978
01-19-2007, 08:04 PM
ok, i had the time to render the QMC scene with the incandescent panel and now i'm completely sure there was not MY WISH, but a FACT
the panel produces a different shadow and different light behaviors

http://img259.imageshack.us/img259/9850/vrayarealightrv9.jpg


well, i'm sure, vlado dont want to replicate more, but now i could say that i wasn't wrong

ciau :)

mat

Gal
01-20-2007, 07:36 AM
dagon i wasnt following(not a vray user) what are you showing here? visually i'd expect the yellow dot surface area to be as dark as the red and green..but it isnt so that means red and green are wrong shadows that are not getting correct bounce light like the yellow surface does? or vice versa.

dagon1978
01-20-2007, 02:07 PM
dagon i wasnt following(not a vray user) what are you showing here? visually i'd expect the yellow dot surface area to be as dark as the red and green..but it isnt so that means red and green are wrong shadows that are not getting correct bounce light like the yellow surface does? or vice versa.

i'm just showing the result of a QMC (1 bounce brute force) render with a panel light (no vraylights) in vray, this would be the "right" result, in terms of light/shadow correctness.
so, NO, red and green aren't wrong, what you expect it's simply different from what it would be

you can also see that the result it's very different from the vray+irradiance(+vraylight) render (and much similar to the mray/turtle one)

Gal
01-21-2007, 12:40 PM
i dont know what that technique is but to my primitive eyes this render isnt right either.
what bounces light onto the yellow dot and not to the green/red? they should have the same value...infact the yellow should be even darker than the other 2 since the only surface that can bounce light onto it is really far away (the other side of the room). agree?

dagon1978
01-21-2007, 01:23 PM
i dont know what that technique is but to my primitive eyes this render isnt right either.
what bounces light onto the yellow dot and not to the green/red? they should have the same value...infact the yellow should be even darker than the other 2 since the only surface that can bounce light onto it is really far away (the other side of the room). agree?

i would agree if you have more then 1 bounce, but this isn't the case
and, why the should have the same value?
infact in the red dot you wouldn't have any light bounces, if you think at the 1 diffuse bounce of the floor and the wall (and this is the only first diffuse bounce you should have on the room, except for the little tables or the chaise longue)
how the red dot can receive light with just 1 bounce?? this is what i was thinking before to do the test with bruteforce and the render confirm my doubts
the yellow dot should have surely more light (not much more, but more!)
the green dot it's strange for me too, but, remember, an unbiased (brute force) render is the solution where all other (biased) techniques should converge, so, i repeat, the QMC (1 bounce ) i posted should be the right solution, or you think that the QMC algorithm is broken?
i can try with other GI engine for this

Gal
01-21-2007, 05:59 PM
the yellow dot i'm talking about is both sections of the wall above and below the window.red is floor shadow,the red dot which is 90 degrees to the light source(window) while the yellow is 180 degrees and the nearest bounce source is far away on the other side of the room which clearly cant light the yellow dot at that distance/intensity. the green is fine i'm not refering to it... if this is just 1 bounce how can the yellow dot wall be lit at all? where is light coming from? and if there is light coming from somewhere to lit it then the red dot should be as lit as the yellow but it isnt so it's a cg cast shadow? not a real light trace?
if you want a real solution i'd use lightscape (maxwell maybe also?).
i dont know dagon, i dont know vray and the rendering solutions so i cant comment about what you are using all i'm saying visually this looks wrong.

dagon1978
01-21-2007, 06:25 PM
the yellow dot i'm talking about is both sections of the wall above and below the window.red is floor shadow,the red dot which is 90 degrees to the light source(window) while the yellow is 180 degrees and the nearest bounce source is far away on the other side of the room which clearly cant light the yellow dot at that distance/intensity. the green is fine i'm not refering to it... if this is just 1 bounce how can the yellow dot wall be lit at all? where is light coming from? and if there is light coming from somewhere to lit it then the red dot should be as lit as the yellow but it isnt so it's a cg cast shadow? not a real light trace?
if you want a real solution i'd use lightscape (maxwell maybe also?).
i dont know dagon, i dont know vray and the rendering solutions so i cant comment about what you are using all i'm saying visually this looks wrong.

the yellow dot it's receiving the light bounce from the floor, this is the first (and only) bounce you have in this scene, the red dot cant receive it! 'cause is part of the floor (how you think the light should bounce from the floor to another part of the floor?), so you should have light only if you add more then 1 bounce, this is clear for me and it's clear also in the vray render, wich is unbiased
with maxwell you cannot control (maybe in the preview engine or in the past releases of the engine) the bounces of the light, same in lightscape (if i remember right), i'm still thinking you are not trying to understand that's a single bounce render, not a full GI render, there is a big difference

dagon1978
01-22-2007, 12:07 AM
...and this is a mray rendering

http://img219.imageshack.us/img219/9184/mrayarealight4iz.jpg

same conditions (no lights, a light panel on the window, path tracing, just 1 diffuse bounces)
as you can see the result it's quite similar (it's just a render region, cause my pc is a little too old to make this kind of renderings)
so, or all the path tracing algorithms here are broken, or this is the right solution for this render :)

Gal
01-22-2007, 02:05 PM
i was refering to the first few tests you've done when i said visually wrong but lets take a look at this 1 bounce render you did. what i simply dont understand is if that is bounce light and not a cast shadow(like the mray render you first did) since there are no area lights now then what illuminates this area? look at the first image.

http://www.galroiter.com/temp/mr_error.jpg

i cant see any surfaces that would bounce light onto there with just 1 bounce. it should be as dark as the corners.


speaking of those corners what shades them to that dark intensity? i would expect the same bounce light to hit those areas(?) and leave a tiny dark edge in the intersection of the walls not this large shadow area. my logic tells me the (?) areas are actually RIGHT but the (check mark) areas are wrong! and at 1 bounce they should be dark if anything, not the other way around. but i do see how light can bounce from the floor to the ceiling as it does now but i dont see how light hits the wall where the window is at a stronger intensity than the ceiling!?!

http://www.galroiter.com/temp/mr_error2.jpg

i highly doubt the yellow line could represent the bounce light coming from the opposite wall and lit the wall where the window is at that distance and not lit the shadow of the table much stronger since it's a shorter distance.

you see what i mean? this is just 1 bounce so it isnt physically accurate but if you think of the math of just 1 bounce these issues shouldnt be in those images. there are no cast shadows from any area lights right? just FG? so why do these issues happen?
i may be wrong and just stupid but these look like errors in the way these engines work, dont you?

i know maxwell/lightscape cant do just 1 bounce that's the whole point i want to see what happens to these problematic areas since they are supposed to be true light tracers.

^Lele^
01-22-2007, 11:42 PM
I have to second Vlado in this -mind you not hard a task-.
The vray light by default ignores the light normals and emits light emispherically in front of the plane.
Hence the brighter tops, compared to a plane with a vraylightmtl applied to a plane, which follows the plane normal more closely.
Ignore light normals - normally, the surface of the source emits light equally in all directions. When this option is off, more light is emitted in the direction of the source surface normal.
Disabling "ignore light normals" will have the vraylight behave precisely as the geometric light.
The issue sits between keyboard and chair, and not with the software, in this particular case.

Lele

dagon1978
01-23-2007, 12:01 AM
I have to second Vlado in this -mind you not hard a task-.
The vray light by default ignores the light normals and emits light emispherically in front of the plane.
Hence the brighter tops, compared to a plane with a vraylightmtl applied to a plane, which follows the plane normal more closely.

Disabling "ignore light normals" will have the vraylight behave precisely as the geometric light.
The issue sits between keyboard and chair, and not with the software, in this particular case.

Lele

really? where's your render? i'm waiting for it :)
the issue sits between the keybord and the mind, in many case :)

dagon1978
01-23-2007, 12:21 AM
i was refering to the first few tests you've done when i said visually wrong but lets take a look at this 1 bounce render you did. what i simply dont understand is if that is bounce light and not a cast shadow(like the mray render you first did) since there are no area lights now then what illuminates this area? look at the first image.

i cant see any surfaces that would bounce light onto there with just 1 bounce. it should be as dark as the corners.


speaking of those corners what shades them to that dark intensity? i would expect the same bounce light to hit those areas(?) and leave a tiny dark edge in the intersection of the walls not this large shadow area. my logic tells me the (?) areas are actually RIGHT but the (check mark) areas are wrong! and at 1 bounce they should be dark if anything, not the other way around. but i do see how light can bounce from the floor to the ceiling as it does now but i dont see how light hits the wall where the window is at a stronger intensity than the ceiling!?!



i highly doubt the yellow line could represent the bounce light coming from the opposite wall and lit the wall where the window is at that distance and not lit the shadow of the table much stronger since it's a shorter distance.

you see what i mean? this is just 1 bounce so it isnt physically accurate but if you think of the math of just 1 bounce these issues shouldnt be in those images. there are no cast shadows from any area lights right? just FG? so why do these issues happen?
i may be wrong and just stupid but these look like errors in the way these engines work, dont you?

i know maxwell/lightscape cant do just 1 bounce that's the whole point i want to see what happens to these problematic areas since they are supposed to be true light tracers.

gal, maxwell/lightscape are true light tracer like mray/vray/turtle/or everything else, the marketing it's probably blinding you, a raytracer is a raytracer, a brute force algorithm is a brute force algorithms, you can tell it maxwell or how-u-like
but if you think you would have a different result you could try with maxwell, no problem, i'm pretty sure the result it's quite right

in your examples you are thinking about raytracing, not about GI, the additional bounce you have is a GI bounce, so, in a diffuse room your rays are splitted around the point hits and not just traced, this is why your floor is lighting the yellow point too :)

^Lele^
01-23-2007, 12:25 AM
A render to compare the two lighting solutions should have the vraylight with direct lighting only compared to the vraylightmaterial with primary bounces.
A render with light+primary would compare to a plane with primary and one secondary bounce, and so on.
The vraylight has a direct lighting component which is amiss from the vraylightmaterial, by design.
And no, do not wait for any render of mine too soon to demonstrate this.

Lele

dagon1978
01-23-2007, 01:12 AM
A render to compare the two lighting solutions should have the vraylight with direct lighting only compared to the vraylightmaterial with primary bounces.
A render with light+primary would compare to a plane with primary and one secondary bounce, and so on.
The vraylight has a direct lighting component which is amiss from the vraylightmaterial, by design.
And no, do not wait for any render of mine too soon to demonstrate this.

Lele

ok, that's probably the problem, tomorrow night i think i could repost the test, so, this is just a semantic misunderstand, 1 diffuse bounce in vray is intended as 1 traced (diffuse) rays
(vraylight+1bouce = 1 bounce = 1 trace depth;
light pannel+1bounce = 0 bounce = 1 trace depth)

thanx lele :thumbsup:

Gal
01-23-2007, 03:03 AM
i'm an "artist" if you can even say that.. so dont hold me on math or physics of these applications but all i am saying is! if that is a gi bounce then it is WRONG!! maybe the math is fine but visually it is wrong to my eye. the floor would illuminate the wall if light would bounce a round a bunch of times in the room but not that intensity comparing to the other areas in the scene which are dark..(i know i know this is a 1 bounce example but the same shows in the previous renders you did.. so that means even at higher bounces it still is ...wrong visually) so ya to me all these renderers if they do this are wrong. i could be mistaken obviously but that's just how i see how light would react in the real world.

YourDaftPunk
01-23-2007, 07:33 AM
But is it wrong? It is supposed to be - after one bounce you have a lot of missing energy of course - it's a render of something your eye cannot naturally see. Your eye/brain does its own funky tone mapping too.

Here are three renders from Indigo which is unbiased metropolis light transport like Maxwell, but free. They are in order:

1) primary ray without bounce
2) primary ray with one bounce
3) no limit

http://www.shawnlipowski.com/forumFiles/jRoom_1ray.jpg
http://www.shawnlipowski.com/forumFiles/jRoom_2ray.jpg
http://www.shawnlipowski.com/forumFiles/jRoom_3ray.jpg

Sorry 'bout the noise, but I only rendered a couple minutes on the first two.

-shawn

Gal
01-23-2007, 02:54 PM
there you go!
thank you shawn, here first image only 1 bounce the window wall is DARK! as it should be with 1 bounce of light, no light should reach that wall..
the next images make perfect sense too notice the window wall isnt lighter than the floor nor should it be! and there are no black dark areas in the corners.
to me this is how light DOES work in real life and that is a correct lighting solution, thank you.
what do you think dagon? agree disagree? and even if this isnt "right" this is the type of lighting solution i was hoping to get with mr that's all.

MaNdRaK18
01-23-2007, 04:59 PM
there you go!
thank you shawn, here first image only 1 bounce the window wall is DARK! as it should be with 1 bounce of light, no light should reach that wall..

Sorry, but Shawn stated that first image is primary ray and NO BOUNCE.
I think You guys speak in different terms, that's why You don't understand each other.
Speaking about 'bounce' rays, Dagon refers to GI tracing, which ISN'T ray tracing as I think You think.

Gal
01-23-2007, 06:21 PM
ya i cant find the right terminology to explain but i think you know what i mean.
oh first image has no bounce? my mistake but the 2nd and 3rd show the right result just as well.

yassein
01-25-2007, 03:09 PM
hi everybody

i have a problem concerning area light

as you can see at this picture ......the area light zone is tooo shining to the degree it hurts the eye' view so .... is there a way to fix it ?

thanks alot

http://xs411.xs.to/xs411/07044/erea.jpg

MaNdRaK18
01-25-2007, 04:20 PM
... as you can see at this picture ......the area light zone is tooo shining to the degree it hurts the eye' view so .... is there a way to fix it ?


Errrrrr.... yes!
Tonemapping.

Gal
01-26-2007, 01:07 AM
easiest thing is to render to 32bit tif and load it in shake, apply hsv curve to it and adjust V...will tone that out easy. you can kinda use photoshop but i wouldnt.. or any 3d tone mapping additions to maya to do it(which is kinda silly since you have to rerender each time to see how it affects it).

Chris-TC
01-26-2007, 03:52 AM
i have a problem concerning area light

as you can see at this picture ......the area light zone is tooo shining to the degree it hurts the eye' view so .... is there a way to fix it ?

thanks alot

I know that tone mapping is the new buzz word, but in your case there's an easier and better way to solve the problem.
You have put an area light with quadratic falloff close to the windows. This creates a hotspot that's unrealistic in your case.

If you want to use quadratic falloff, you'd have to put the area light very, very far away from the window. After all, the light coming in through a real window is not cast by a bright lightsource a few inches away from the window.
But this type of lighting situation is a case where using linear falloff (which is otherwise unrealistic) makes more sense because it's easier to control and gives practically identical results to a quadratic decay light source that's far away.

This way you'll get more believable results than if you just tonemapped the hotspot.

yassein
01-26-2007, 06:24 PM
:bounce: thanks for all of you
that's it
you know the real problem
actually i found your way is right
i change the light falloff to linear and that gave me better result
but i still need to know more about erea light especailly how to you mental ray node
like physical light
thanks for help and i hope some day i can help the others like you do :)
http://i2.turboimagehost.com/b1/259622/erae.jpg

CGTalk Moderation
01-26-2007, 06:24 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.