[Mihai]

OK the settings are as follows:

Emitter intensity: 10 000 000 Watts
ISO: 100
Shutter: 1/200 s
fStop: 128

rendered from C4D through Plugin 0.6b

Just to make sure that these extreme settings don't play a part in the differences, I suggest rendering in beta with something more "normal". I can't understand why you set fstop at 128. You can leave it savely at 11 or even 8, there won't be any heavy blurring, almost none. Then adjust your emitter intensity so the scene has the brightness you want. Then, in V1 start first by changing your emitter intensity so the brightness matches the beta, and in second place, adjust your ss and iso.

[mverta]

Thomas, to answer your question, if you're standing in the room and looking at the lux meter, that's reality. If you take a picture of the event, that's photographic reality.

Maxwell is reproducing photographed reality, using a simulated camera. The presence of ISO, etc. are there because that's the target. When we look at realistic renders, it's photography we're actually comparing against in our brains. Coming from visual effects, this difference is at the crux of what we do, certainly. We don't put lens flares and glare on images because they look that way "for real," but because they photograph that way. Ditto tons of abberant camera junk which is now regularly part of CGI's definition of photoreal. Ditto lighting CGI space ships with "bounce cards", for that matter. If you light them "for real," with a single parallel light and no ambient fill, they never look right. "Right" is a studio model lit with studio lights, because that's how it's usually photographed. Oddly enough, the exact phrase, "it doesn't have to be real, it has to look right" is a mantra amongst visual effects supervisors, who spend all day and night wrestling with all things photoreal. Hey, the word "photo"'s in there!


_Mike

[seco7]

I'm not a photographer by any means but I have to admit I'm not sure where your going here. Aren't photographed reality and "real" reality both moving targets? Is Maxwell to look like a Nikon D2 or a Hasselblad medium format? That is why I like the comparisons to the beta as a known and (largely) excepted quantity. Besides, that comparison is infinitely easier than comparisons to "reality".

[mverta]

They can be (moving targets), and even Maxwell uses a simplified camera model (a single lens system) but the target - by design - has always been clear. I suppose you could say Maxwell is currently sort of aiming at a "simplified" or "averaged" film response, but ideas like different film stock response would absolutely be following with that. But right now, what we're looking at here, in terms of beta/1.0 comparisons, is that beta has at least a generic/idealized photographic response, replete with biases, whereas 1.0 is primarily characterized by a different response, which I agree appears less photographic in circumstances.

And again, these concerns, while tangentially related to other things about Maxwell, like materials, is really not about "accurate" or "correct" - Maxwell is unquestionably more accurate in its calculations and abilities than beta - it's a question of how it represents those calculations in the render. I agree it was generally more photographic in beta, and thus more satisfying, even photoreal. But not 100% of the time, which is why it's so elusive. Do you follow? It's like 1.0 does the math better, but shows it to you through foggy glasses sometimes. (That's not a great analogy, but it'll do).

_Mike

[chrisvconley]

One thing I notice from SJ's plane test is that one difference between 1.0 and Beta is their respective DYNAMIC RANGE or GAMUT of grays.

V 1.0 has a broader range -- the white is whiter and the dark is darker.

Beta's range is less, and so it DISTRIBUTES its lightness more evenly through the range.

I propose this to be THE ANSWER of why Beta feels better than V 1.0.

[mverta]

Well that clears that up.

_Mike

[SJ]

One thing I notice from SJ's plane test is that one difference between 1.0 and Beta is their respective DYNAMIC RANGE or GAMUT of grays. V 1.0 has a broader range -- the white is whiter and the dark is darker.

Chris, your proposal is incorrect. Beta's dynamic range is wider, because it has the ability to catch more f-stops of light on it's virtual film/sensor. Review dynamic range regarding photographic media ...keeping in mind that the brightness-range of M~R's virtual reality is as infinite as in true reality before having it translated into a viewable image.
The term "Gamut" doesn't apply here, because it's the same in both images, being 8 bit RGB. White is white (255,255,255) and black is black (0,0,0) in both images. Only the translation of light-intensity is different in both engines and here, Beta's image sensor is able to translate a wider range of light-intensity into discernable shades of grey = wider dynamic range.
Gamut doesn't apply to the rendering engine in general, because -as stated- it is infinite in it's internal calculations.

[andretto]

Only the translation of light-intensity is different in both engines and here, Beta's image sensor is able to translate a wider range of light-intensity into discernable shades of grey = wider dynamic range.

that seems correct to me (hey, for what it's worth... )

[chrisvconley]

Yes, right -- I got it backwards. I just want to know: do others see the observation I am trying to describe?

In the images provided in "next test: comparison of grey scale and light-falloff" do we agree that Beta's white is greyer and its black lighter?

So the theory is, said another way, Beta is not "blowing out" at the ends like V1 is.

With this restated theory, I repropose it as THE REASON Beta feels better than V1!

[SJ]

I just want to know: do others see the observation I am trying to describe? [...] So the theory is, said another way, Beta is not "blowing out" at the ends like V1 is.

Agreed
By the way, i used the term "light-falloff" as a synonym for "attenuation". Maybe my choice of words is a bit clumsy from time to time... English is just my 2nd language.
Initially, I did suspect the air in V1.0 to have a stronger attenuation than in Beta. This seems not to be the case. In my test i checked the attenuation with those planes that are positioned in different distances from the camera while the tone curve of the resulting image was compared through the greyscale at the right. So it was possible to cross-check what was wrong through the usage of these methods combined.

[sandykoufax]

Thanks for the great test and discussion.

[PA3K]

I have done some little comparison with the same very simple scene between alpha, beta and r1. How can i place images here to this thread?

[samsam]

http://www.imageshack.us/

post the images on this site and then post one of the links on this forum!

[jonathan löwe]

upload your image somewhere, for example on imageshack and then post it here by using the "img"-button which you'll see when replying


should look like this:


edit: hmmm, samsam's to fast for me

[Xlars]

Impressive test-work .. this is one of the most interesting / exciting threads I have seen for a long time