[GM5]

I have to agree with Lars - great thread and tests. Hope some good comes out of these tests

[MetinSeven_com]

Surely the only way forward for this is for NL to commission someone to re-calibrate Maxwell. I would suggest someone with an artistic eye and who can argue for visual artistic qualities rather than linear mathematical correctness. Ideally a modern day Leonardo DaVini type of person with grounding both in mathematics science and art.

Well, okay then, I'll have a look at the Maxwell code.

[PA3K]

So here it is:

i made little labyrinth, and placed one 100W 30x30cm emitter inside. All walls, floor and emitter are just one polygon on each side. Top views from maxwell camera.

Result from alpha:

beta:

and V1:

Since with the same setting V1 looks different i changed settings to match alpha and beta result (in maxwell-studio i changed emitter efficiency from ~12 to maximum 683, burn from 0,8 to 0,3 and ISO from 100 to 50):

Then i changed brightness and contrast of all images to +40, +40 in photoshop to see how images responds to b/c changes:
alpha:

beta:

V1:

and changed V1:

Difference can be clearly seen. Since alpha and beta are almost identical (beta better respond to b/c changes-oposite wall can be seen), V1 after b/c changes shows burned-out light and shadows worse then alpha (? V1 burn is set to 0,3 ?). It was rendered on the same computer P4 Prescott without HT 2,4GHz overcl. at 3,1GHz SL ~15. Alpha was fastest 50min, beta 1h 18min, V1 more than 3h. At the same SL V1 look more clear (but like with antialiasing and noise filter, beta looks sharper). I also rendered that scene with alpha maximum bounces changed from 8 to 16, but no visible difference. Seems it has nothing to do with bounces. I just didn`t changed scale (next time). But for sure beta has wider dynamic range. There was no way for me to change parameters in V1 to have similar tonal and dynamic range to beta. Like maxwell went digital.

I know that this little test is not complex at all and also not scientific like yours SJ. I just want to contribute to find some problem solution and way how to get render images better. Hope it helps somehow someone.

Here is that little labyrinth with Physical sky and sun.

beta:

changed V1:


Patrik

[Mihai]

Well the original renders from alpha/beta don't look identical at all. V1 looks closer to beta in that regard, except in V1 the light is stronger further away from the emitter, so naturally you get a more blown out render when raising the contrast so high.

[Frances]

Beta:
http://img230.imageshack.us/my.php?imag ... eta6dw.jpg

V1:
http://img230.imageshack.us/my.php?imag ... 1up4uz.jpg

If you scrub between these two, you will see that the light distribution isn't the same, the efficiency setting isn't comparable. Patrik, I came across a post by tom that stated that a beta D65 equivalent in V1 would be 6500K with 125 efficiency. So, I wonder what the V1 version would look like with 125?

My test isn't very scientific. The emitters are boxes with 100w D65's for the beta and the equivalent emitter in V1.

Beta (tried to match noise level and it rendered about 20 minutes longer):


V1:


V1 with 1000W emitters and 12.5 efficiency:

Notice there is more noise in the 2nd V1 render after 60 minutes. I think jacking up the intensity and using a low efficiency may cause noise to resolve more slowly?

Also, to anyone who knows, efficiency is not a percentage, is it? What is the scale for it?

[Mihai]

It's how many lumens are emitted per watt. More efficient bulbs put out more lumens for the same amount of watts consumed.

[Mattia Sullini]

I know many of you already told me about the fact that ISOs have no influence on the light distribution and that playing with gamma and burn i would obtain the same results. In this thread is coming forward, at least this is what i have understood of it, that gamma and burn actually have influence on light distribution. And in fact, for example, auto exposure gives me different results depending on the ISO i set. This means that using an ISO not properly set for that particular scene would lead to use a "wrong" combination of burn and gamma...or not??? I need to feel an idiot at least once a week and then i would really appreciate if someone would explain me the reason why i am wrong again!

[Mihai]

iso/fstop/shutterspeed, don't affect the amount of energy that is in the scene, from the emitter. You are simply making the "film" more or less sensitive to light.

[PA3K]

Mihai, you wrote "V1 the light is stronger further away from the emitter, so naturally you get a more blown out render when raising the contrast so high". Yes you are right, but that is the problem. If you look at second beta picture (b/c+40,+40), you can see illuminated wall in second cell (top one). In V1 picture burn was set to 0,3 (less contrast) so naturally if light is stronger further away from emitter, why is illuminated area in second cell smaller than in alpha (no illuminated wall in second cell after b/c changes)? I wanted to know, how far can the same emitter illuminate labyrinth by indirect light. By other words SJ tests showed same result-more light near and less light far. It looks like scale is different (but not in scene).
Frances i can`t see those images, because i`am not registered on cgarchitects.

[iandavis]

everyone is comparing maxwell to the perceptual innaccuracies of film. The problem with this of course, is that light, materials and our expectations of them are fixed, and photography doesn't change any of that. Different films and cameras only selectively record portion of the light spectrum in various biases. Light propegation and how much light is reflected/absorbed from objects is an 'earth-constant'.

My perception of V1 is that light is too quickly absorbed. Now, is this due to incorrect material settings, as mahai suggests, or the calculations themselves are in error. The fact of the matter is, if we need to adjust or allow for V1's new rendering 'style' then the default settings should reflect this (no pun intended). I have seen VERY good results with V1, but like myself and my own experience I'm betting they have pumped up the materials to FIX what is wrong. So, it SEEM like the midline, absorbtion curve, or whatever you wish to call it is set a tad low. If the rendering engine is unbalanced so that we need to artificially inflate either the surface reflectivity or the lighting power I suggest the short term solution would be tutorials from next limit on how to get realistic results with an engine that is a little on the contrasty side.

If anyone remembers the RCs where slammed for exactly the same thing.

2c

[SJ]

Hi Patrik,

nice test-work, although not consistent enought from it’s setup to draw conclusions from it. I thought about setting up some kind of labyrinth to compare light propagation too
What has immediately caught my eye is the different quality of sunlight in your renderings. Beta looks much more realistic. V1’s sun is way to dimished …and some more is off too. I think illumination through physical sky with sun is a thing that has to be tested. I know I was fascinated everytime by sunlight renderings made with Beta. V1 disappoints me in this regard again and again.

[samsam]

I know I was fascinated everytime by sunlight renderings made with Beta. V1 disappoints me in this regard again and again

word

[Matthias Rudolph]

i rebuild the cornelbox of the v1.0 announcement and rendered it with v1.0 and beta1.2.2a. maybe PA3K could analyse those to renders regarding the light propagation etc.

all the values are the same at both renders, the emitter strength, the color-values, shutterspeed, gamma, fstop, etc etc etc.

BETA:


V1.0:



every image cooked 12 hours, the beta gained up to sl18.8, v1 to sl 15

/cya

[Mihai]

Are you sure all the values are the same, because the illumination would not be the same at all if you put a 100W emitter in beta, and do the same in V1.

[Matthias Rudolph]

all the values are the same, that means i used 9watt emitters(the lamp contains 3 equal emitters, each 9watt). the rest was not changed. i´m aware of that emitters in beta are not the same like in v1, but i´ve not changed any value to receive an isoline comparison