Author Topic: ** Corona Animation ** -- relationship setting quality/time  (Read 165491 times)

2013-10-10, 01:34:50
Reply #165

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
understood!

therefore the method Bucket approaches reality!?

2013-10-10, 01:36:44
Reply #166

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
Eian, but from where in the world you write?

2013-10-10, 01:55:44
Reply #167

Eian

  • Active Users
  • **
  • Posts: 42
    • View Profile
Hmm not exactly, Bucket mode just adds adaptation so you can have better control over speed/quality while still maintaining unbiased approach which is PT+PT.
It also helps to evenly distribute noise in an image where some areas need more time to clear noise than others..

As of now I'm located in Greece by the way. How about you?
« Last Edit: 2013-10-10, 02:00:33 by Eian »

2013-10-10, 02:03:04
Reply #168

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
I live in Milan.

Eian , I applied to 80 Max ray ahahaha...the refractive surface is much better in different ..

tomorrow night ... place the image

Meanwhile, it is a test to see only the difference

2013-10-10, 02:30:17
Reply #169

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
I go to sleep and leave the pc to work
tomorrow morning before leaving the house, forwarding the image

good night
and thanks Eian

Greats
Raf

2013-10-10, 02:42:12
Reply #170

Eian

  • Active Users
  • **
  • Posts: 42
    • View Profile
Ok good, welcome mate, anytime,

Goodnight..

2013-10-10, 07:35:18
Reply #171

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
good morning

PT+HD
Max ray depht = 60

2013-10-10, 08:10:05
Reply #172

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
Hello Eian

well this weekend we have for Max ray "values"

Comparison between PT+HD is BUCKET
and relations with the rendering time and image size...

2013-10-10, 15:26:34
Reply #173

Eian

  • Active Users
  • **
  • Posts: 42
    • View Profile
Hey,

I don't think you have to go above 32 Max ray depth for the above scene because the objects that refract light are not that complex..
As for the fireflies you have to tweak the settings under bucket mode..

If you want you can send me the scene and I'll try to figure what settings work out best..

Cheers..

2013-10-10, 18:26:34
Reply #174

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
Hello Eian
Yes

I'll send you the scene this evening.
Right now I'm 'back from work

Which version uses 3dstudio?

2013-10-10, 19:18:39
Reply #175

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile

I gave it to stop VFB with | Time 4h:05min |
and the processor has finished at the | Time 4h:32min |

After the stop, the CPU has worked almost 30min

2013-10-10, 20:57:22
Reply #176

maru

  • Corona Team
  • Active Users
  • ****
  • Posts: 12709
  • Marcin
    • View Profile
Maybe some problem with RAM or HDD?
Marcin Miodek | chaos-corona.com
3D Support Team Lead - Corona | contact us

2013-10-10, 21:31:23
Reply #177

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
Hi maru

Very likely.
( I need to format my pc, it's a year and a half  )

This has happened when I increased the max ray 60.  On this occasion I noticed this aspect

I'm linking the scene .max vers.2014
https://mega.co.nz/#!V9diXDQS!Yo9bFwKKpz0Cow-5nUvE6HRGzHof9yeuff3rGYHDB0U

In the meantime is working in the bucket method with one of the parameters tested and with Max depht ray = 32

very fast ....and is also decisive to two three passes


Thanks
Raf

2013-10-11, 19:08:01
Reply #178

rafpug

  • Active Users
  • **
  • Posts: 722
    • View Profile
Hello

What allows you to do the falloff map on refraction?

Thanks
Raf

2013-10-11, 19:26:41
Reply #179

Eian

  • Active Users
  • **
  • Posts: 42
    • View Profile
Hey rafpug,

I got your scene and went through some optimization..

First of all I merged your scene from scratch in order to make sure there were no "wrong" values in settings to start with.
In render settings I used PT+PT, which is faster from HD since its an exterior scene, and which yields more physically accurate results as well.
I set "Path tracing samples" to "24" and used "Bucket" mode with a setting of "3" for both "subdiv AA" and "adaptive steps".
"Pass Limit" was set to "1" since I define the level of quality under "Bucket" settings.
I also set "Max ray depth" to "64" even though increasing it seems to introduce some kind of a visual glitch in this specific scene caused by some light trapping on the left chair.
General rule though concerning "Max ray depth" is that in order to get physically accurate results you need to increase this value until a point is reached where increasing it further gives the same result as before.

Examining a bit further the above I noticed that the IOR (Index of Refraction) of the material used for the chairs was way too high ("5.00").
There are no materials in real life with such high IOR apart from metamaterials which are artificial and subject of a whole different story..
This  setting is also what caused such great impact in rendering time when you increased "Max ray depth".
Seems to me like there was some kind of endless looping that were caused by some trapped rays due to such high IOR.
I set it to "3.00" which I believe is still too high for the material you're trying to achieve but I didn't want to deviate too much from your settings.
Furthermore there is no actual need for the "Falloff" maps you connected to reflection and refraction to simulate "Fresnel" since the IOR value does about the same through a more physically accurate approach.

I also made the actual sun disk invisible in reflections and refractions to get rid of the fireflies..

Apart from the above changes, which seemed a bit obvious, I didn't go into a deeper optimization
however these changes alone had an enormous impact in rendering time and a better refined image..

Since though this thread is about animation, there are ways from which faster results can be achieved,
for example by reverting all values under render settings to default and using Bucket mode with "Subdiv AA: 2" and "Adaptive steps: 3" gives an acceptable result in under 5 mins..

Oh and talking about speed, I'm using an overclocked Core-i7 @ 3.5Ghz (the render stamp can be deceiving).

I'm uploading the previous result as well for the sake of comparison as well as an image with a different material for the chairs,
I believe it's closer to what you might be trying to achieve..

Hope all the above will prove helpful in any way..

Cheers..
« Last Edit: 2013-10-12, 17:55:11 by Eian »