[bignose]

I have the same problem with physical sky. It works fine as long as I don't change any setting for it. But if I do (whatever, time, city, date, anything) it renders black.

Already had the same problem with the previous version.


Bignose

[Thomas An.]

As it says in the subject, I get pretty good and even reasonably fast results with the skydome, but physical sky just produces blackness.

What settings do you use for the physical sky ? (city, hour, GMT ?)

For every city you choose make sure you insert the correct GMT (Rhinoll will not do it automatically, but it should hopefully in the near future).

So it is possible you are ending up with a midnight situation.

[Micha]

Hi,

I had some problems with "Berlin" too, but after I have choose GMT 6, it works approx. like I like.

If this not work, try 12 hour? This was my start point for my environment tests.

[Thomas An.]

- Once the GMT/location thing is actually corrected, where is supposed to be north in the model. Convention would suggest up in topview is north, but somehow this doesn't seem to be the case ... either that or I haven't found the right correction for the GMT/location problem yet.

Yes, the Y-axis is supposed to be north. If not then its a bug.
Oscar mentioned at one time that in future releases there will be an indicator pointing the location of the sun.

Could anyone explain what actually is rendered differently with the "low quality" switch on.

Use it only as a rough draft to set up the scene illumination. It gives the full intensity of the illumination almost imediately (just by drawing a few dots here and in their full intensity), but the caustics are calculated very sparsely and you will always have pesky dots here and there. I am not sure how it is really handled this internaly.

[Micha]

Sometimes I have got sharper images with more details and less noise if I render an image with "low quality" in double size. After denoise and scale down, the images was nice clear (double size helps the denoiser to clean the image without to loose to much details).

Alsos it seem to be in the moment that big scale images are only possible in "low qualit" mode.

[Thomas An.]

Sometimes I have got sharper images with more details and less noise if I render an image with "low quality" in double size. After denoise and scale down, the images was nice clear (double size helps the denoiser to clean the image without to loose to much details).

Alsos it seem to be in the moment that big scale images are only possible in "low qualit" mode.

Very interesting Micha, but you need to tell Oscar because you seem to know something that the developer doesn't.
http://www.maxwellrender.com/forum/view ... ality#1474

[Micha]

Here I have down a test some days befor. I think, the "low quality" looks sharper and show a finer noise.

[Thomas An.]

This test is not representative. In this test you are using only difuse materials. This is known, but not too useful.

You have:
-No dielectrics
-No caustics (metals, or through glass)
-No speculars
-No indirect conditions

This image is so basic that Maxwell can do it even at low quality.
Flamingo can do this image too (with a lightdome) in one hour.
AIR can do it in less than 5min.

The image is just too basic.

[Micha]

Maybe it is to basic, this was a simple first test.

My test based on an idea from some where in the forum. Somebody has post some realy big images with bottles and has swear on "low quality" and "scale down". My extra thesis, big scale images make it easier for the denoiser to find the noise without to catch the small details.

In two weeks I hope to find more time to make more test in this direction (with caustic ...).

It's a pitty, that I never get an answer to my post http://www.maxwellrender.com/forum/view ... ow+quality. I have seen the question for the "low quality" often, but nobody has got an deeper answer from the NL team.

Do you remember you to the test scene with this cube walls in the Rhinoforum? A basic scene with diffuse materials only. The Maxwell image was the best at the high quality GI level in the same time like AIR. AIR can make a good approximation in a very short time, but it need a very long for high detailed GI and with much GI bounces. Flamingo Lightdomes are simple fakes. Good for a nice lightning, but the expert eye can see, what a lightdome is.

[Thomas An.]

It's a pitty, that I never get an answer to my post http://www.maxwellrender.com/forum/view ... ow+quality. I have seen the question for the "low quality" often, but nobody has got an deeper answer from the NL team.

I don't think anyone really knows. The developers do not comment / disclose about the engine internals (maybe for patent/confidentiality reasons due to competition). So we can only speculate.
Personally, I go by what Oscar suggested.

Do you remember you to the test scene with this cube walls in the Rhinoforum? A basic scene with diffuse materials only. The Maxwell image was the best at the high quality GI level in the same time like AIR. AIR can make a good approximation in a very short time, but it need a very long for high detailed GI and with much GI bounces. Flamingo Lightdomes are simple fakes. Good for a nice lightning, but the expert eye can see, what a lightdome is.

True. I agree. AIR and Flamingo will give an approximation. But for something very simple (and given enough effort) they can still fool us to believe its real.
Now, I did not try the cube walls scene with AIR, but a long time ago I tried a cube room in AIR for 4min:

[Micha]

Was this scene in AIR GI with color bleeding or without (occlusion mode)?

[Thomas An.]

Was this scene in AIR GI with color bleeding or without (occlusion mode)?

I rememeber in all my AIR tests I used color-bleeding.
(Did not like to use occlusion alone)