[breberka]

HSV=[0,0,0] is only color ... i mean ... you should have absolutelly black body ... means ... absorbtion coeficient of this body should be >> 1. is it correct?

[Thomas An.]

HSV=[0,0,0] is only color ... i mean ... you should have absolutelly black body ... means ... absorbtion coeficient of this body should be >> 1. is it correct?

Right ! well sort of. Absorption and scattering are all turned off. In Maxwell "turning-off" means set to Zero.

So, its Abs=0 and Scatter=0

[JDHill]

In RC2, the bending phenomenon still seems to be present. The output at the left seems to be straight, while more near the center it is not.

s.l. 13.00 | severely over-exposed | area-light off


I hope I'm doing this right.

I'll update the image as the render proceeds.


~JD

[Thomas An.]

I have a bad feeling about this one

Did you let Rhinoll convert the dielectric material or did you make the material from scratch in the studio ? Also, I think you need to enable-dispersion (from within studio) before we can see colors and stuff...

[JDHill]

Hi Thomas,

You are correct - this is 100% Rhinoll. The model requires changes to run in Studio, though. I've posted a bug about it - when rendering from Rhinoll, single plane mesh emitters only emit in the direction of the normals, while in Studio, they emit in both directions.

In the shot above, I was primarily interested in noting that the light pattern seemed to follow the same pattern as your test. I aborted the above test, and had been letting another one go for about 9 hours with the prisms made entirely of glass:

s.l. 20.81 | ISO 400 | shutter 750 | area-light on


This test doesn't seem as conclusive about much of anything, as far as I can tell.

I have not been clear though, in reading through all of the tests, on one point. I see that in your example, the middle grouping gives a very curved output, but the others, while straighter - are also curved. Was the question of curved output directed toward all groupings, or just the center one?

Next, I'm going to change the model and let it run from Studio.

By the way, although I'm pretty comfortable with computing in general, I am by no means knowledgeable in the 'World of Maxwell'. [read: you can't possibly offend me with any correction or clarification .]


~JD

[Thomas An.]

Hi JD,

I tried a similar test yesterday after seeing your effort. Even with dispersion turned on the result (after 7 hours) was entirely incorrect (unless I grossly misunderstand how to use materials or unless Rhinoll translates dielectrics in entirely incorrect ways)

I think the RC1 core (we are still using the RC1 core event though we downloaded RC2) is not ready for this at all. It appears to be very crude in all tests from all users that I have seen. Lets not waste energy. I say wait for a core update.

Was the question of curved output directed toward all groupings, or just the center one?

Yes, all of them had a bend (some more than others). Well, if the developers were to fix things so that the middle grouping was properly straight, I suspect, by induction, this would autocorrect the slight bend of the other two groupings as well.

[JDHill]

Hi Thomas,

Thanks for the clarification. I have a suspicion that it is likely a problem of one of the core algorithms.

I once wrote a 2D surface for vector-based drawing. One feature of this surface was the ability to simulate the mousewheel zoom behavior of the major 3D applications - zooming always towards the cursor. Being 2D, it used matrix scaling and translation to simulate this.

In the beginning, objects would seem to follow a hyperbolic trajectory as you zoomed in and out. I could fine-tune different variables to get it very close to working, but as the zoom factor became more extreme, invariably the target would translate out of view.

Eventually, I found a part of the function which was producing a non-linear relationship between the scale and the translation, and once this was corrected, the surface worked as intended.

It may be possible we are seeing a similar problem in this scenario.


~JD

[Thomas An.]

Eventually, I found a part of the function which was producing a non-linear relationship between the scale and the translation, and once this was corrected, the surface worked as intended.

It may be possible we are seeing a similar problem in this scenario.

True. This very plausible. I remember for example the alpha had a problem with dispersion where the order of the colors in the spectrum was incorrect, but this was later fixed for the beta. So the dispersion algorith has had issues before.
http://www.maxwellrender.com/forum/viewtopic.php?t=1922


In your latest test I see a vertical grouping of dots; which is somewhat encouraging.

[max3d]

Hi Thomas,

I was rereading this thread as I'm still puzzled by the curved dispersions.
Did you retry this scene in RC3 and if so could you reproduce it?

[Thomas An.]

Hi Thomas,

I was rereading this thread as I'm still puzzled by the curved dispersions.
Did you retry this scene in RC3 and if so could you reproduce it?

Hi max3d,

As mentioned earlier tried RC2, but couln't produce any dispersion at all and the refraction was way off... meaning that the beams were going through the first prism at the wrong angles.

There was no dispersion at all other than very strong fireflies.

Haven't tried RC3 per se, but I doubt it would be any different.

[hdesbois]

Dispersion still unreliable in RC3. Better let it alone for now.
HD

[max3d]

Ok. I'll trust you both on it and don't spoil time on it. I had a simple experiment in mind, but that only makes sense if the renderer has the basic functionality again and it still shows the strange curving.

[Nakonieczny]

Beautiful test.
My 3 cents - as for a bug, shape of bending is too similar to rainbow bending. Maybe its a... some thing to check. In nature I many times saw this kind of dispersion, on the sky . Its not like bug in color order - this one its too close to life. Of course it can be coincidence. But even the direction of bending is the same - in most causes youll see only part of the rainbow, an its always (in my opinion) bend this way.

Adam

[JDHill]

Adam :: I could be wrong (as always ), but I believe the rainbow takes a curved shape because it is created by a cloud of (spherical) water droplets. From one single vantage point, you see a spherical shape form, as a result of the angle formed between the Sun, your eye, and any individual droplet. In each instance, the color reflected by that single droplet will be determined by this angle.

~JD

[Micha]

... I have not read anything here, only an observation: from the window of an airplane I have seen a rainbow on the clouds under the airplane - a closed circle.