[Bubbaloo]

Why are the emitters black/gray?

[Peder]

PS. Just for fun I tried applying the 3600W emitter to the file with the individual planes. Again rendering to SL 6 and not touching exposure. This would be the indoor sun version according to Mihai. So that means an artificial lighting of 36x3600W= 129 600W.
http://idisk.me.com/peder.lindbom/Publi ... 232335.jpg

Now this looks unrealistic to me. There must be something wrong with the intensity calculation here. I am a lighting designer by trade That much light would wash out the scene even if the exposure was balanced for daylight.

[Fernando Tella]

That room is a bit misleading. It looks like a house room but it's 9m high; what is the size of each emitter plane?. I would try to replicate a room from your house and simulate what's happening there; turn on multilight and play a bit with the wattage and ISO/SS. In real life, if you turn a light on when sunlight is coming through the window the effect is almost none.

The difference between joined mesh and separated is interesting; what happens if you let it run longer, let's say SL13? It probably has to do with Maxwell optimizations, but the final result should be the same.

[Peder]

I think an apology from NL would be apropriate. I have spent many hours trying to hunt this issue down. And I have not felt any great support from NL during this process. Many times I have wondered if I should pursue it further. You need to foster a culture where people dare to speak out if they find issues not making people feel stupid.

In this case it has created professional problems for me delivering the quality and the timeliness my clients rely on. Also I have been forced to use a fairly expensive render service provider just to save my bacon. I did not expect to be let down in this way by a tool that prides itself of it's realistic lighting representation in general and particularly it's unbiased nature.

I have the greatest respect for the work you guys do and rely on it professionally but this time I think you fell down.

[Peder]

Fernando!
Ok you do have a point. The room is bigger than it looks. But still 129kWatts? Doesn't seem right anyway. But the main point I am trying desperately to get across is that as it stands Maxwell is only unbiased after a specific SL or rather gradually converging on unbiased. When this happens depends on how many emitter objects are in the scene. In my original case with 415 objects this point lies well beyond SL 17. You can call this optimization -I would call it biased or even a bug.

It is easy enough to work around now -when I know to join my emitters.

[Bubbaloo]

Why are the emitters black/gray?

Hello?

Look at this again:


It was my understanding that even if the emitter meshes are not joined, you still need to multiply the wattage by the number of emitters. I have done 3 or 4 interiors with V2 using this method with correct results. even a couple of them had sunlight coming into the room. I will test some more as time permits. The fact that you constructed a test scene with a room with a ceiling 9 meters high makes me question the scale of your original scene. Scale plays a HUGE role in Maxwell's calculations.

[Peder]

Bubbaloo I hear you. I am only trying to follow Mihais instructions to eliminate the intensity from the equation. But as you my gut feeling is that you need to set the emitter strength for the entire group of emitters. But I could be wrong. ISO 100 and a shutter time of 100th of a second is slow. So possibly an indoor emitter of 100W could look that weak.

I did a new test with a more normal ceiling height of 3.75m and increased the ISO to 2000:
http://idisk.me.com/peder.lindbom/Publi ... 005114.jpg

And here with joined emitters:
http://idisk.me.com/peder.lindbom/Publi ... 005459.jpg

But as you see my main point stands - that the sunlight is slower to develop with separate emitter objects.

[Fernando Tella]

It was my understanding that even if the emitter meshes are not joined, you still need to multiply the wattage by the number of emitters.

Both images have the same emitter material applied; in this one each plane is an independent mesh:



Joined meshes:



It's like that since multilight sliders are per material and not per mesh.

All your MW2 renders are wrong Bubba, you'll have to repeat course.

[Mihai]

The sunlight will take longer to develop the more separate emitters you also have in the scene, if your projects require that many emitters, better to join them and multiply the emitter strength with the nr of joined emitters. The noise for the rest of the scene won't change dramatically anyway, you'll have to wait for a high SL for the noise to clear, separate emitters or not. But in your case it will be more efficient to join all those emitters.

About emitter strength, it's simple to check, like Fernando did, compare separate or joined emitter strength:



Noise separate vs joined emitters (both images rendered to SL 16 in the same time and roughly the same benchmark):



I think the extra noise you see in the separate emitter image is coming from the slower converging sunlight, if you check the emitters separately using ML, separate or joined will have about the same amount of noise.

Regarding light levels remember it's also the fstop that matters, in this scene it's set to 8 which with an ISO and shutter speed of 100 looks correct to me for sunlit daylight exposure. It's cloudy here today but I quickly checked with my camera an outside exposure at fstop 8/ ISO 100, gave a shutter speed reading of 60-80.

[Peder]

Thanks, Mihai. Confirmed then. Recommendation - join emitters=Yes!!!
Lighting levels -unintuitive perhaps but realistic nonetheless.

[tom]

Which one would fragment the drive more, multiply the seek/read/write access times and finally slow your harddisk (and so the computer) down?

1000 files made of 1 KB or 1 file made of 1000 KB?

Conclusions:
* It's true in v2, you don't need to merge them for ML.
* You may not like to merge them if you have a few emitters.
* You definitely need to merge them all (or at least as groups) if you have tens/hundreds of them especially together with sun.
* You should find new total lumens for merged emitters or change your emission strategy in terms of units. (e.g. lux or luminance)
* Pink preview is a current preview-engine limitation with Multilight. It has no negative effect on your render in any way.

This is just an optimization handicap but all these methods will give you the same look in final render. The word optimization roughly stands for the importance of samples taken per emitter. Otherwise, Maxwell has no sort of tricks or shortcuts as we clearly mention with patience and passion, over and over. It's your benefit to turn optimizations into advantages. Thanks for understanding...

[Peder]

Thanks Tom I now understand. I and hopefully others will benefit from this thread.

[Furumaru]

I have made a similar test. It seems that the joined lights should have the same polycount.



Furumaru

[Peder]

Now that is interesting. Care to make a version with joined emitters and 1 polygon and 64 polygons respectively? It seems the test with joined emitters is less noisy than the one with separate emitters and similar polycount. But that is maybe just a result of the optimization spending more time where there are more polys.

[Peder]

Also if you add another group of joined emitters with only 1 polygon each to the bottom of the boxes how are they affected by the difference between 1 poly and 64 poly emitters at the top? Maybe the number of polys control how much time is spent between different groups of emitters.