Page 1 of 1

sky portal

Posted: Wed Feb 27, 2013 8:12 pm
by photomg1
would love to see some form of sky portal in maxwell , my only experience of these in another render engine would be modo's version of a portal where they gather rays and focus them into a specific area .Similar to an emitter but they emit no light only gather the light from behind them and focus them into an area .

I did this quick test (sorry about the ugly image I just wanted to illustrate my point).The originals were at 1500 x 1000

enclosed room with emitter at 30 mins sl 12 bench 208
Image
no emitter just lit by skydome 30 mins sl 16.26 bench 1172
Image
no emitter just lit by hdri 30 mins sl 16.24 bench 1165
Image

So what I'm requesting is something which behaves as the emitter in the first image but is taking its colour from what is behind it.As it renders far cleaner in the same amount of time.Which would be fantastic for getting cleaner interiors with a hdri/sky for example.

Re: sky portal

Posted: Fri Mar 01, 2013 10:57 am
by jespi
I would also like to see portals implemented within maxwell. Any kind of performance improvement is very welcome in maxwell.

Re: sky portal

Posted: Tue Mar 05, 2013 10:01 pm
by Ha_Loe
Sky portals in a biased render are a way to speed up things, because you can ignore light/geometry that is not "visible" through the portal or precalculate the light distribution across the portal. Can't see where this fits into an unbiased concept.

If you want to have natural light bouncing around exterior geometry lighting an exterior scene, you will have to deal with extreme contrast, which in turn takes longer to clear the low light interior. If you don't need the exterior simulated, replacing the windows with an emitter is basically the same as a sky portal...

Re: sky portal

Posted: Wed Mar 06, 2013 1:36 am
by photomg1
Ha Loe,
Its not the same as an emitter.Why else do you think they exist in other render engines if its just an emitter. If it was offered in the software , I'm sure it would
be an option as to whether you could use it or not.Hence those wishing to have classical unbiased render could choose to use it or not, just like you have the option
with caustics on/off and dispersion at present.These options are there to save on render time and noise.
If you have an emitter in the window and want to have another light i.e sun shining through as well you would have to turn the original emitter in the window invisible
to global illumination .You would then be back to the original problem of noise.The emitter in the window visible to global illumination traps the rays in the room hence
the faster clearing of noise.But that is hardly unbiased either is it.
Cant really understand why anyone would be against them other than for scientific reasons , even then just dont use them if they become an option and those are your needs.

Re: sky portal

Posted: Sun Mar 10, 2013 10:04 pm
by Ha_Loe
photomg1 wrote:Ha Loe,
Its not the same as an emitter.Why else do you think they exist in other render engines if its just an emitter. If it was offered in the software , I'm sure it would
be an option as to whether you could use it or not.Hence those wishing to have classical unbiased render could choose to use it or not, just like you have the option
with caustics on/off and dispersion at present.These options are there to save on render time and noise.
If you have an emitter in the window and want to have another light i.e sun shining through as well you would have to turn the original emitter in the window invisible
to global illumination .You would then be back to the original problem of noise.The emitter in the window visible to global illumination traps the rays in the room hence
the faster clearing of noise.But that is hardly unbiased either is it.
Cant really understand why anyone would be against them other than for scientific reasons , even then just dont use them if they become an option and those are your needs.
I never coded my own render engine but did some research a while ago and have some experience in game graphics (which in essence is very biased rendering), so please take the following more as an educated guess:

Biased engines calculate each light effect (specular, diffuse, ambient occlusion, global illumination,...) in independent and separate passes. So each effect can be optimized. They generally only calculate light and geometry that is visible to the camera. In games, portals are used to partition geometry into rooms and check for visibility. If the portal is not visible, you don't need to bother about any geometry inside the connected room... Likewise for specular light: if the portal isn't "visible" to the object, you don't need to check any of the rooms lights. Since only few polygons will be actually lit by the portal, this can speed up the tracing.

GI in biased engines is simulated by sampling the specular light that hits the surfaces and reemitting this light a set number of times. Again, a portal can be used to check, whether a portion of the room is actually lit from outside. Portals typically cut this bouncing off and separate GI in the rooms, speeding calculations up.

Unbiased engines cannot separate geometry into rooms, because the whole concept is about emitters casting light into the scene and collecting it with a camera. If you would want to separate rooms, you would need to calculate luminous flux across the portal, emitt the light into the next room, ignore or handle light, that emerges from this room or bounces back... essentially making the whole thing either highly inefficient or a biased engine. You are right, with an emitter window you just manually bias the process and an internal solution might spare the time. But for maxwell to calculate the portal would mean a whole render for each portal. ...I hardly can see the potential speedup here.

If you have two largely different emitters like a candle and the sun, a biased engine needs to balance the number of photons each source emitts into the scene. With only few photons passing through the windows and few emitted by the candle, while thousands harmlessly hit the exterior walls, the camera will take forever to collect a sufficient amount of photons, with or without a portal. Emitter only interiors clear faster, because the emitters are better balanced. If you make one emitter bright as the sun and add a single candle, you will be back to square one with a noisy interior. The only kind of biased thing, Maxwell does is to check for fully closed rooms (i.e. windows sealed with visible emitters) and ignore all sources outside that enclosure.
In contrast, the biased engine just checks for the visible lights for the current pixel and adds up their intensity. So the fact that biased engines have fewer problems with sun+candle situations is probably not because of sky portals, but because of the biased concept of rendering.

...so this is why I think that asking for sky portals is like asking for fire to be wet.

On a side note: dispersion and caustics are effects that add to the realism of the light calculation. They are calculated as each photon interacts with the scene. Because they can have little effect to the overall look of the scene, you can skip them and save the time. Do not confuse them with the "caustics" simulation of a biased engine.

Re: sky portal

Posted: Mon Mar 11, 2013 6:54 pm
by photomg1
Ha Loe ,
that is a very imformative reply .Thank you , but your explanation does differ from my understanding of portals in an unbiased engine .I also have a license of octane render which also has portals which work in that software using the path tracing option(unbiased) . I'd post the diagram here and the explanation but I don't want to lose my license (just I incase I want to use it in the future :)). Its free to join their forums , go to forums>resources and sharing> Look for a sticky post >TUTORIAL/DIAGRAM: How portals work and how to use them .

Its more of a one way material with light coming in but not going out . Also shadow rays in the room are fired at a random points on the portal so that nearly all the shadow
rays in the room find the portal.Giving much improved convergence .So less samples are needed to get a clean render.(4x to 2x speed up )

Re: sky portal

Posted: Mon Mar 11, 2013 7:05 pm
by numerobis
photomg1 wrote: I'd post the diagram here and the explanation but I don't want to lose my license (just I incase I want to use it in the future :)).
:shock: :?:

Re: sky portal

Posted: Mon Mar 11, 2013 7:42 pm
by photomg1
I dont use it at the moment as the octane plugin to modo is very poor ( very out of date ,has smoothing problems which needs workaround).Plus the lack of displacement and texture limitations are annoying(game breaking at times) .But it does have portals !

It didn't seem right to copy and paste the information given to registered users in their forums here.Hence me posting directions to the post : )

Re: sky portal

Posted: Mon Mar 11, 2013 11:52 pm
by jespi
I think Fry Render also had portals, so it's not only a feature for biased renderers.

Re: sky portal

Posted: Tue Mar 12, 2013 12:12 pm
by Ha_Loe
photomg1 wrote:Ha Loe ,
that is a very imformative reply .Thank you , but your explanation does differ from my understanding of portals in an unbiased engine .I also have a license of octane render which also has portals which work in that software using the path tracing option(unbiased) . I'd post the diagram here and the explanation but I don't want to lose my license (just I incase I want to use it in the future :)). Its free to join their forums , go to forums>resources and sharing> Look for a sticky post >TUTORIAL/DIAGRAM: How portals work and how to use them .

Its more of a one way material with light coming in but not going out . Also shadow rays in the room are fired at a random points on the portal so that nearly all the shadow
rays in the room find the portal.Giving much improved convergence .So less samples are needed to get a clean render.(4x to 2x speed up )
I sure will have a look at the forum. Here's from what you describe: If you block outgoing light, that will cancel the unneccessary calculations for light from the inside hitting outside geometry. On the other hand, if you don't have any outside geometry, the rays would travel into infinity and not interact anyway... testing for a hit of the portal or missing it will take about the same time, so I don't see a speedup in this case.

The balancing problem with sun vs. candle will not be solved by the portal. "Shadow rays" are a biased concept. This sure will improve convergence, as this is what biased rendering is all about.

Re: sky portal

Posted: Tue Mar 12, 2013 1:36 pm
by photomg1
Which is exactly what I suggested at the end of my first post on this subject , and already happens in maxwell if I place an emitter in the window visible to gi.

At the end of the day I love the look of maxwell renders but want faster clearing interior renders. I don't care how it does it ,if it looks the same I'm happy with that.I feel
fairly sure I wouldn't be on my own with that philosophy .

Re: sky portal

Posted: Tue Mar 12, 2013 4:49 pm
by Gary Bidwell

Re: sky portal

Posted: Thu Mar 14, 2013 12:27 am
by photomg1
Hi Gary,
do you use iray i just wondered if they ever resolved their issues with it, as I noticed that post is from a couple of years ago. I did have a look and they are in the latest 2013 build release for iray.