[Mihai]

Well if I compare 9 watts in beta to 9 watts in V1, I have to raise the wattage in V1 quite a lot to match the watt setting in beta. But in your case, it is much brighter...something must be off.

[Frances]

Well if I compare 9 watts in beta to 9 watts in V1, I have to raise the wattage in V1 quite a lot to match the watt setting in beta. But in your case, it is much brighter...something must be off.

It would be great if we could have a table that breaks down the beta emitter presets into V1 parameters. There really shouldn't be this disparity regarding emitter output.

[aitraaz]

Yikes another maxwell pie-chart

But its true we should put something like that together as a transition from beta illum values to the new V1 ones, it will be quite helpfull

[Mihai]

It's because beta wasn't very well calibrated to real world values, and on top of that you had only a watt setting, with I assume a fixed efficacy rating. The way of setting the beta emitters was really just try and set a wattage and see what comes out. With 1.0 you can actually use the lumens or efficacy rating of the manufacturers lights.

[Frances]

It's because beta wasn't very well calibrated to real world values, and on top of that you had only a watt setting, with I assume a fixed efficacy rating. The way of setting the beta emitters was really just try and set a wattage and see what comes out. With 1.0 you can actually use the lumens or efficacy rating of the manufacturers lights.

Not that this is your problem or responsibility, ( ) but your comment begs the question, "If Maxwell was sold as physically based, where did they get the parameters for the [beta] emitter presets?"

[Mihai]

Well it's this misconception again that if Maxwell internally computes the correct physical behavior of light, the interface will reflect that and that there is no calibration required. If they only provided one type of 40W bulb, then things would be easy, but not very useful

The old beta interface values (emitters, camera) didn't match very well with what one would expect in the real world using those same values. So it's a matter of calibration.

[Thomas An.]

Hi all,

Just a note here: As Mihai mentions.... (it agrees with my findings as well) The shutterspeed or the emitter intensities of beta1.2.2a were completely wrong (off scale in general ... not just for the above render).

It seems V1.0 has more realistic interpretation of the exposure/illumination values.

[Frances]

Well it's this misconception again that if Maxwell internally computes the correct physical behavior of light, the interface will reflect that and that there is no calibration required. If they only provided one type of 40W bulb, then things would be easy, but not very useful

The old beta interface values (emitters, camera) didn't match very well with what one would expect in the real world using those same values. So it's a matter of calibration.

I am wondering about the conclusions you and Thomas An have reached or the explanations you've put forth. No reason for anyone to flip out or anything. I suppose this was all gone over ad nauseum in the beta stage and I must not have been paying attention. The inaccurate calibration of the lights and camera and all that. Seemed to me that whenever anyone got poor results from beta, it was explained that it was a physically correct result.

It's true that "physically correct" can occur and be unexpected if the user's settings are whacko. No program can save someone from themself. But what you're saying is that, using emitter presets in a reality-based way (in other words, no 10 million watt incandescent bulbs) and standard camera settings (no pinholes in oatmeal boxes), the results from beta were not physically accurate? This is what I'm skeptical about, because it is really a revelation to me.

[mverta]

The emitters in beta (to the best of my knowledge) were not calibrated against any real-world, actual values. We did an extensive -and ongoing- validation of emitter values for 1.0 so that users could put in real-world, everyday data and get predictable results. "Physically accurate" means that the propogation of light and interaction with materials follows predictable rules, but that's not dependent on the amount of light you put into the scene being "reasonable" from a real-world standpoint. If you put a billion-Watt bulb in a lamp, Maxwell will deal with it like it was a real thing. And without a real-world values interface, this is what you might've done in beta. 1.0 just went one step farther so you were putting in familiar values. It just helps if you're recreating a scene and you want to mimic working in reality as closely as possible. Ditto ISO and f-stop.

Nature is blisteringly, blindingly complex, and Maxwell has the formidable task of reducing a billion billion parameters in nature to a few simple ones that users can actually deal with and control. The truth is, there is no way you can reduce all of the materials and substances and phenomena in nature to half a dozen sliders. But you can govern the rules of even a simplified system in a physically accurate way. The fact that Maxwell does as much as it does with so little input is a Herculean accomplishment on its own. I, for one, am always pushing to make the experience working with Maxwell as intuitive, and as much like real-world as possible, but there's always going to be a limit. Otherwise, Maxwell would have to come with 2 lighting gaffers and a union rep to hang emitters in my scenes.


_Mike

[tom]

It's all about lumens and lumens = Power (W) x Efficacy (W/lm) Beta didn't have efficacy setting and it worked like constant efficacy. So this doesn't mean beta was not physically correct. Beta was selling different watts of same type bulbs and the release provided this as something customizable, more flexible.

[Voidmonster]

Lambert material, Benchmark 50.96, Rendering time 4h56m44s



Roughness 99, Benchmark 61.16, Rendering time 4h07m15s



Roughness 90, Benchmark 60.76, Rendering time 4h08m50s



Roughness 50, Benchmark 48.41, Rendering time 5h12m20s

I'll post the analysis in a separate message.



Roughness 10, Benchmark 61.36, Rendering time 4h06m25s

[Voidmonster]

In my test, beyond a certain point I didn't try to match SJ's original Beta image. The results I was getting suggested it would be more fruitful to do a comparison between V1 images with different settings.

Here's what I did and what I found.

I chose 7 points in the stairstep image, in roughly the same areas as SJ. I marked down the pixel coordinates of each of the seven points and then used the 5x5 averaged eyedropper to obtain a color value.

I only performed this test on the Lambert, Roughness 99 and Roughness 90 images. Here are the results:

La | 99 | 90
---------------------
1 - 79 | 105| 95
2 - 57 | 76 | 67
3 - 40 | 55 | 47
4 - 28 | 39 | 32
5 - 20 | 28 | 22
6 - 17 | 19 | 17
7 - 11 | 16 | 12

To make more detailed comparisons to the beta image, I would need the MXI file in order to match lighting levels more closely. However, I think so far this test clearly shows that fairly significant changes to light distribution happen as a result of material differences. Note especially the difference between Lambert and Roughness 99.