[deep__dive]

i'm really disappointed of maxwell. i can't belife that i spent so much time to learning and now i know nothing. here another example of a render:



i know, its no so bad, but it have the same problem, it's plain und dull cause it was lit by emitter planes in the windows, the only method witch gets mi reasonable results. i know the materials arent very good, but this isn't the real problem...

and all the advice i can get, can help mit to make the render faster, but it can't solve the problem that i get horrible noise in all indoor renderings when try somthing ohter than emitterplanes in the windows.

how would you make the light setup for this scene?

i have tried with artificial sun und ib with hdri.

that is the result (sl 15.3):



https://dl.dropboxusercontent.com/u/100 ... itchen.mxs

and a renderfarm si no solution, i think all renderer shoud be able to render simple scenes without a renderfarm. i don't wont chance the renderer either, i don't have the same time twice to get to know another renderer...

[Half Life]

If you post an actual Maxwell Scene (MXS file, via pack and go, zipped) it would be much easier for others to give help.

One thing hardly anybody seems to talk about here, which is often critical with interior noise, is the issue of unevenly powered emitters.

Maxwell is optimized to clear noise from more powerful emitters first -- (in any case using the Physical Sky) the Sun being much more powerful than the interior emitters means that the noise from the emitters will take longer to clear... This is one of the headaches of doing interiors.

Ideally, you would want to ramp up the power of your interior emitters (to match the intensity of the sun) and then reduce them to the desired output using Multilight (after or during rendering). Conversely, you could opt to either use a fake sun emitter and/or IBL so as to have a much weaker "sun/environment", thus bringing the power of all scene emitters more into balance.

But the bottom line is: the larger the disparity (between the most powerful emitter in a scene and the least powerful emitter), the more noise the low powered emitter(s) will produce at low sample levels... ultimately it will resolve at higher SL (22-26 is what I tend to find in most cases) that do not involve dielectric materials (glass,SSS, water, etc.).

However, if it is indeed a material issue, doing a clay render would reveal that immediately -- and at that point it would be fairly easy to find the culprit by process of elimination. Bright and/or reflective surfaces will produce more noise than rough and dull surfaces... the fastest rendering material would therefore be a lambert black, simply because it absorbs most of the light and scatters the rest evenly, producing no reflective caustics.

So the most difficult situation to find and resolve is bad geometry -- which is fairly common depending on how the model was made... but this is in no way Maxwells fault.

In this particular scene I see many possible causes of excess noise -- but it would only be conjecture on my part as to what might be the core issue(s)... regardless you may have to adjust your expectations of about render times/noise and interiors in Maxwell 2.x. Or , at the very least, accept that some fakery (like window emitters) are just the only way to get something clean in the timeframe you will accept (foregoing render farms).

Most of the best renders you will find here are absolutely filled with advanced fakery -- and while that goes against the core principles of what Maxwell is supposed to be, it is simply the reality of where we are right now in the evolution of Maxwell.

Best,
Jason.

[deep__dive]

ok, thanks for answer! i have changed the c4d fil to mxs.

i will try to balance the emitters.

[gadzooks]

Well said Half life!

[deep__dive]

when i set the default exposure to ev 7, iso 100, fstop 8, shutter 2 for interior renders, the sun need to have 20 000 000 watts (1 km away, ø9m) to have sufficient influens to the render, do you mean really i shoud put the other emitter to 20 000 000 watts and regulate them down?

in my test the render looks a little bit better, but it's really confusing mi...

[Half Life]

when i set the default exposure to ev 7, iso 100, fstop 8, shutter 2 for interior renders, the sun need to have 20 000 000 watts to have sufficient influens to the render, do you mean really i shoud put the other emitter to 20 000 000 watts and regulate them down?

in my test the render looks a little bit better, but it's really confusing mi...

It is pretty counter intuitive -- and of course the balance between camera exposure and emitter power will play a large part in the exact settings needed to create balance.

However the answer is yes, bring the power of all the emitters in a scene (remembering to account for surface area the emitter material is applied to) as close to the same value(s) as you can -- then adjust them using Multilight back to the desired relative power after launching the render.

This trick only works if the issue is unevenly powered emitters, but this is a common enough problem that it should be documented somewhere other than these forums.

Best,
Jason.

[deep__dive]

when i have a image emission texture or a hdri lightning setup, how do i specify the power of them? or are they except of this rule?

[Half Life]

Both HDR based emitter materials and IBL have intensity settings... just lower these.

Best,
Jason.

[deep__dive]

but i can't setting watt and in reference to ibl i don't know the surface of the light dome?

[Half Life]

Look at each multilight channel individually -- the ones with the most noise need the emitter values to be raised, or conversely the ones with the least noise the emitter values can be lowered. Whichever approach you take the exact numerical settings do not matter -- what matters is that all multilight channels are showing the same amount of noise (as much as possible).

This is what I mean by balance.

Best,
Jason.

[deep__dive]

thanks for your advise, it was really helpfull, im agree with you, this knowlede should be really in the manual, without that its really hard to get a clean render.

[deep__dive]

in this context it could be interesting to know more about the pyhsical sky/sun and the sky dome. does anybody know how much power they have? it could be possible to use them also in interior renderings and avoid the problem that they increas the rendertime and the noise?

[numerobis]

What i can't understand is why this balancing can't be done internally if it is working so well?

Where is the advantage of prioritizing stronger light sources if the result is a noisier image?

[Half Life]

What i can't understand is why this balancing can't be done internally if it is working so well?

Where is the advantage of prioritizing stronger light sources if the result is a noisier image?

It is not always a problem. It is only the issue when it happens to be the issue -- and most often seen in interior shots with sunlight coming into an enclosed space while also having emitters. Which if you think about it is a pretty extreme example of emitter power disparity.

Or to think of it another way, it is not so different from the issue of dielectric materials and sunlight -- an extreme example that causes problems because we (as humans) have no issue seeing any difference between a glass under artificial light, versus a glass under sunlight... but the calculations are totally different (particularly power of emitter, relative size and distance from emitter), and it should always be noted Maxwell is just a fancy calculator.

Give it an easy problem, it gives a fast/good solution -- give it a hard problem and it will take time (or massive processing power) to arrive at the correct answer. And since we don't have prebuilt/assumed answers as cheats for the hard problems (biased rendering), Maxwell must actually take the full time to deal with the tough problems we give it.

So for now, if you want fast answers, you have to avoid giving it unnecessarily hard problems -- thus the advanced fakery. This would be analogous to what might be called wisdom or skill in Maxwell rendering -- knowing when and where a problem is likely to be and how to resolve it, or if it even can be resolved, is part of why a Maxwell user would even be called skilled/knowledgeable.

Unfortunately, the misconception is often repeated (by Next Limit no less) that Maxwell works best if you set up scenes "just like reality" -- however that mindset only works until it hits one of these quirks. At that point, the only practical solution is to break the tie with reality and start "faking it". This is the reason why totally non-real cop-outs like AGS even exists (which to me is the very antithesis of what Maxwell is meant to be).

That is not to say I don't think these are a big issues -- quite the contrary. I think these things should be A-1 top priorities going forward, long before any new features are added I want to see these problems resolved in some way.

All I am saying is: if it were a simple problem to solve, it would have been solved long ago -- because it is certainly a known issue from their side. I learned the emitter balancing trick from Tom, so obviously not only did he know of the problem, but had a suitable workaround in place before I even knew what the problem was.

I have a tremendous amount of faith in most of the people on the dev team, so I will allow some slack for the quirks that exist now -- and try my best to simply remind them (when the time is right) that any remaining legacy quirks are still top priorities in my mind.

It would be nice to be able to truly say "set it up like reality" and not have to make any allowances for fakery... that is my personal goal at least.

I know you understand this. I'm simply putting it out there for the sake of newer users who come across this. I don't want there to be a misconception that Next Limit is negligent or unaware of these problems... quite the contrary, Next Limit/the Maxwell team may be the best company I've worked with from the standpoint of diligence and caring about "getting it right".

Best,
Jason.

[oz42]

Fantastic insight (about emitter balancing) Jason, thank you for your expert knowledge (and easily digestible explanation) once again!