Chaos Corona Forum

General Category => Gallery => Topic started by: peterguthrie on 2013-11-18, 15:47:07

Title: 1st test with corona
Post by: peterguthrie on 2013-11-18, 15:47:07
Hi everyone,

This is an old project of mine, and seeing as it is very simple I thought it would be a nice easy start for trying out corona. Each one got to about 200 samples i think or passes, or whatever its called ;) and took about 2 hrs.

Original set done with vray here: http://www.peterguthrie.net/archive/#/plainspace/

So... how do I get access to nightlies then!

Peter
Title: Re: 1st test with corona
Post by: michaltimko on 2013-11-18, 16:06:05
I was waiting on your "first" post here hehe! As usual, very nice starters. Hope to see more images from you soon.
How do you feel working with corona comparing to Vray ? 

Title: Re: 1st test with corona
Post by: Ludvik Koutny on 2013-11-18, 16:11:06
They look nice, but i think such a simple scenario should render a bit faster. Do you use PT+PT or PT+HD? As for the daily builds, once you reach 5 posts, you will see daily sub-section section in alpha builds section. There, you will just post a link to this thread, and you will get daily build key ;)
Title: Re: 1st test with corona
Post by: peterguthrie on 2013-11-18, 16:25:37
I'm quite sure it can be faster, indeed I hope it can! I used PT + HD.

I've only had 1 day playing about with corona so far, so most settings still a bit of a mystery which is partly why I've asked Juraj to do a tips & tricks type post on my blog so that I and others can get up to speed.

will check out the alpha builds section, thanks!
Title: Re: 1st test with corona
Post by: Juraj on 2013-11-18, 16:29:47
They look nice, but i think such a simple scenario should render a bit faster. Do you use PT+PT or PT+HD? As for the daily builds, once you reach 5 posts, you will see daily sub-section section in alpha builds section. There, you will just post a link to this thread, and you will get daily build key ;)

Why would you classify this as "simple scenario" ? Seems to be pretty lot of GI work. And I think all materials are specular, even the matte paint, as it does look so from last image.
It could be pretty high RGB value for white, though I can't know.
Title: Re: 1st test with corona
Post by: Ludvik Koutny on 2013-11-18, 16:37:03
Yes, materials with too high albedo could cause some problems as well... or coronalights with directionality... Some of these vulnerabilities will hopefully get removed in future builds. The one with directional light is already gone, but there is no stable daily with that yet. As for albedo, it will always pay off performance-wise to keep diffuse and reflection ranges in reasonable boundaries. My top for diffuse is RGB 220, which i consider to be a white sheet of paper.
Title: Re: 1st test with corona
Post by: michaltimko on 2013-11-18, 16:40:30
Regarding settings, there is no big magic behind imo. Playing with PT samples or with MSI/HD does the trick but, for the still images, changing those settings doesnt have drastical impact on speed/quality ratio.
For me, MSI parameter is most important for interior renderings.

Also, you can cover openings with planes with CoronaPortalMLT.

Here is my pretty simple test, how portals can speed up the rendering : http://forum.corona-renderer.com/index.php/topic,1551.msg11805.html#msg11805
Title: Re: 1st test with corona
Post by: RolandB on 2013-11-18, 16:48:01
Here is Peter... waoooo, what some great artists here. I stay !
Title: Re: 1st test with corona
Post by: peterguthrie on 2013-11-18, 17:13:09
Yeah probably every material has glossiness, and GI wise i dont think its the easiest scene to light.

Let me check if I had any directionality on the IES lights.. and what my white walls were set to.

re: portals, I thought I read somewhere that they werent always an advantage... ? I am using a HDRi sky.

happy to try one render with and one without to see if it speeds it up.
Title: Re: 1st test with corona
Post by: Ludvik Koutny on 2013-11-18, 17:14:52
Yeah probably every material has glossiness, and GI wise i dont think its the easiest scene to light.

Let me check if I had any directionality on the IES lights.. and what my white walls were set to.

re: portals, I thought I read somewhere that they werent always an advantage... ? I am using a HDRi sky.

happy to try one render with and one without to see if it speeds it up.

In this particular case, portals will probably make difference as the skylight opening is not that big. They are very simple to use, just put a large quad over there to cap the opening and apply portalMTL on it :)
Title: Re: 1st test with corona
Post by: michaltimko on 2013-11-18, 17:17:54
Portals should help with scenarios like you have here, where you need to "push all light thru that hole".
I have heard that they can slow down renderings in scenes with big openings, but i have never noticed any slowdown and im using portals almost in all my scenes.
Title: Re: 1st test with corona
Post by: peterguthrie on 2013-11-18, 17:31:35
Yeah probably every material has glossiness, and GI wise i dont think its the easiest scene to light.

Let me check if I had any directionality on the IES lights.. and what my white walls were set to.

re: portals, I thought I read somewhere that they werent always an advantage... ? I am using a HDRi sky.

happy to try one render with and one without to see if it speeds it up.

In this particular case, portals will probably make difference as the skylight opening is not that big. They are very simple to use, just put a large quad over there to cap the opening and apply portalMTL on it :)

does the normal of the quad matter? ie. facing in or out?
Title: Re: 1st test with corona
Post by: michaltimko on 2013-11-18, 17:32:51

does the normal of the quad matter? ie. facing in or out?

Keymaster (dev) told me that no, but im always face-in.
Title: Re: 1st test with corona
Post by: Ludvik Koutny on 2013-11-18, 18:15:01
Yeah probably every material has glossiness, and GI wise i dont think its the easiest scene to light.

Let me check if I had any directionality on the IES lights.. and what my white walls were set to.

re: portals, I thought I read somewhere that they werent always an advantage... ? I am using a HDRi sky.

happy to try one render with and one without to see if it speeds it up.

In this particular case, portals will probably make difference as the skylight opening is not that big. They are very simple to use, just put a large quad over there to cap the opening and apply portalMTL on it :)

does the normal of the quad matter? ie. facing in or out?

It does not, but i adopted a habit of always having normals facing inside of an interior, in case functionality of portal changed somewhen in future. :)
Title: Re: 1st test with corona
Post by: Ondra on 2013-11-18, 23:00:48
does the normal of the quad matter? ie. facing in or out?
No. Rules for portals currently are:
1) no double portals (2 layers of portals in the same window)
2) keep them as low-poly as possible

Restriction #2 was already removed in the unstable builds.
Title: Re: 1st test with corona
Post by: Paul Jones on 2013-11-18, 23:22:52
Nice to have you on board Peter :)
Title: Re: 1st test with corona
Post by: modern_babylon on 2013-11-19, 04:20:22
Looks nice Peter,

I remember the originals of this set well. Always liked the contrast and balance you managed with the warm/cool lighting. Looking forward to your future corona posts.

Cheers
Title: Re: 1st test with corona
Post by: peterguthrie on 2013-11-19, 13:52:13
I did a whole load of tests, but nothing really conclusive.

- I think the main problem may be the IES lights, has there been progress in this area in the daily builds?
- changing light samples multiplier from 2 to 5 made the most noticeable improvement for the same rendertime.
- adding a portal made no noticeable difference.
- in the original set I had the MSI at 400. Changing to 100 was an OK trade-off. Dropping to 20 made the image noticeably darker.
Title: Re: 1st test with corona
Post by: peterguthrie on 2013-11-19, 13:52:36
p.s. thanks for the warm welcome!
Title: Re: 1st test with corona
Post by: Ludvik Koutny on 2013-11-19, 13:55:37
I did a whole load of tests, but nothing really conclusive.

- I think the main problem may be the IES lights, has there been progress in this area in the daily builds?
- changing light samples multiplier from 2 to 5 made the most noticeable improvement for the same rendertime.
- adding a portal made no noticeable difference.
- in the original set I had the MSI at 400. Changing to 100 was an OK trade-off. Dropping to 20 made the image noticeably darker.

400 is way wayyyy too much...  Even 100 is too much...  but of course default (20) will cause some bias. MSI is really a trade-off value, it let's you to adjust ratio between bias and rendering speed, as for IES, i think there are some improvements made in very recent builds, but those are unstable at the moment. If you have access to dailies, you could try 2013-09-17 build ;)
Title: Re: 1st test with corona
Post by: Ondra on 2013-11-19, 14:31:09
The daily builds have support for the standard 3dsmax lights, so you could use a photometric light with IES profile, that generates zero noise. This is currently faster than proper area lights with IES, because I haven't gotten to sampling directional light yet. And IES is in its core nothing else, than directional light.
Title: Re: 1st test with corona
Post by: peterguthrie on 2013-11-19, 17:01:32
thanks Keymaster and Rawalanche, will try daily build as soon as I build up the courage.

Just a quick question re: HD Cache, are there settings there I could be tweaking or should I leave them as default?
Title: Re: 1st test with corona
Post by: Ludvik Koutny on 2013-11-19, 17:10:24
thanks Keymaster and Rawalanche, will try daily build as soon as I build up the courage.

Just a quick question re: HD Cache, are there settings there I could be tweaking or should I leave them as default?

Lowering precomp density to 0.1 is good for very quick previews, if you want to iterate quickly on something, just be sure to put it back to 1.0 for finals.

Then PT samples in HDcache increase sample accuracy, so you may try 1024 or even 2048 if you have any scenario that is very complex for indirect lighting.

And positional sensitivity increases sample density...  increasing it makes it more stable in animations, and could also improve accuracy for stills.

Non the less, all around, defaults work mostly fine. I usually raise HDcache PT samples to 1024 for finals, and lower precomp density to 0.1 for quick previews.
Title: Re: 1st test with corona
Post by: peterguthrie on 2013-11-19, 17:14:49
thanks Keymaster and Rawalanche, will try daily build as soon as I build up the courage.

Just a quick question re: HD Cache, are there settings there I could be tweaking or should I leave them as default?

Lowering precomp density to 0.1 is good for very quick previews, if you want to iterate quickly on something, just be sure to put it back to 1.0 for finals.

Then PT samples in HDcache increase sample accuracy, so you may try 1024 or even 2048 if you have any scenario that is very complex for indirect lighting.

And positional sensitivity increases sample density...  increasing it makes it more stable in animations, and could also improve accuracy for stills.

Non the less, all around, defaults work mostly fine. I usually raise HDcache PT samples to 1024 for finals, and lower precomp density to 0.1 for quick previews.

exactly what I needed, thanks!
Title: Re: 1st test with corona
Post by: blackhaus on 2013-11-20, 16:47:58
Welcome aboard Peter!

pretty interesting tests, the lighting is amazing, thanks corona for make this
so easy, with a good understanding of lighting/camera/photography we can reach
very good results.
As I always say, at the end of the day what really matter is the image.

Hope see new stuffs from you my friend,
Cheers
Fernando