www.BinaryAlchemy.de :: View topic - compression type artifacts visible when volume fades out
 SearchSearch   RegisterRegister  ProfileProfile   UsergroupsUsergroups   Log inLog in 
If you create a new post, please use a topic that describes your problem
Documento sin título
 
compression type artifacts visible when volume fades out

 
Post new topic   Reply to topic    www.BinaryAlchemy.de Forum Index -> Shader Problems / How do I...?
View previous topic :: View next topic  
Author Message

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Mon Jan 18, 2010 8:58 pm    Post subject: compression type artifacts visible when volume fades out Reply with quote

I am doing renders in Maya, going for some dissipating ocean spray, and have been getting some strange blocky artifacts as the volume particles fade out - looks kind of like bad jpg compression. It is visible in the alpha as well so I don't think it is a cell size issue - also the 'blocks' appear to be larger than the lookup cell anyway. I haven't ever noticed this before, so I was wondering if anyone else has this problem. I am trying to upload a rendered sequence but the files are quite large.
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Mon Jan 18, 2010 9:44 pm    Post subject: Reply with quote

I assume that cells are ignored because the first cell test does not hit any density.
Do you have very fine smoke?
Please try to set the marching detail to "fine".
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Mon Jan 18, 2010 10:28 pm    Post subject: Reply with quote

it is very fine smoke, and my march settings are set to 'high'. I was wondering if there is any other setting to tweak, either in non-simple mode or in the extra attrs. I am also using a density limit of 5. Also, is there anything to tweak in the render settings? I don't think mental ray has any volume-specific settings unfortunately.

edit: does particle count make a difference with this?
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Mon Jan 18, 2010 11:44 pm    Post subject: Reply with quote

All settings beside the lookup table and the marching tab settings do not affect it.
And Mental Ray does not have any settings that influence the volume renderer.

If you have only one picture, then I could probably say if it is the error that the cell thinks it is emtpy.
Happens with cigarette-like smoke if the smoke stream/line happens to be in the middle of the cell.
And to check if the cell is empty the shader checks the edges of the lookup cell only.
But then the lookup table cells have to be 3-5 times larger than this smoke line.

Please enable preview mode, reduce = --- and enable "show cells detail level".
Then you should see the cell size in the render view.
Just to check.

About the problem:
Marching tab, disable simple mode.
Set Light samples per cell to 7. (also increases the samples for the "empty"-check)

Then the other parameters (they are for the quality, not because of the problem ):
max adaptive subdivision: 2
Density threshold: 0.16
Use table if alpha: 0.9
UseTexture Min: -0.003f
UseTexture Max: 1.7f
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Tue Jan 19, 2010 1:58 am    Post subject: Reply with quote

Holger,
Still no luck with those settings. I was able to upload a small sample of images. Upon further testing and re-reading this posting, I would say that my smoke isn't really 'fine', but rather low density with above average detail.
Thanks
Josh

http://www.youshare.com/Guest/555a1647156bd906.zip.html
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Tue Jan 19, 2010 2:47 am    Post subject: Reply with quote

Definitely not what I thought. Strange artifacts.

Does Moving/rotating the camera a bit help?

There is only this one geometry shader in the scene, nothing else? No particle cloud, no second geoemtry shader?

You have used the maya surface shader as material?

MRay Ray depth is set to a high value?

Your MRay tile size is not that small that it could be different tiles?

BA_fractal is the only texture shader applied? What happens if you remove the BA_Fractal connection and set a constant value instead. One that gives you a similar over-all look. Still artifacts?

Any shader parameter driven by something?



Please set the lookup interpolation to none and render frame 85, then I can see what the size of the cells are.
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Tue Jan 19, 2010 3:23 am    Post subject: Reply with quote

I have tried most of these suggestions but still no luck. There was a geometry shader in the original file but I still got the artifacts after removing it. I have uploaded a test file that should give you the same artifacts, hopefully you could troubleshoot it directly. There will need to be a particle cache created of course.

http://www.youshare.com/goshone/6274e3cf318cd42d.mb.html
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Tue Jan 19, 2010 4:25 pm    Post subject: Reply with quote

Downloaded the file, I think I can take a look in a few hours.
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Tue Jan 19, 2010 8:13 pm    Post subject: Reply with quote

Thanks for taking the time to help me out. I have been struggling with this for a week now. I was thinking... could it have anything to do with scene size? I am currently testing a scene at 10x scale. I will follow up with any results.
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Tue Jan 19, 2010 11:23 pm    Post subject: Reply with quote

I tried to cache the scene, but the cache files seem to be empty.
Can you send me one frame. Perhaps 85?
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Tue Jan 19, 2010 11:27 pm    Post subject: Reply with quote

make sure you cache all the particles, as there are particles emitted from other particles. I will try to upload some cache files in the meantime.

edit: I had disabled the dynamics on the base_particle. Set 'is dynamic' to on, and it should cache fine.
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Tue Jan 19, 2010 11:32 pm    Post subject: Reply with quote

It was the dynamic flag.
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Tue Jan 19, 2010 11:41 pm    Post subject: Reply with quote

PS: As the cache files are getting large, perhaps you should create a new "cache particles" mel script.
Maya-Cache to local temp, then use the BA_particle_converter to copy-convert the files to the network.

See http://www.binaryalchemy.de/develop/shd_vol/cache_research.htm
Scroll down for Maya
You half the time you wait for caching.
And you use 1/10 of disk space.


Or you can ask your R&D guys if they can write a completely custom exporter.
SDK for the ba cache files is included.
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Wed Jan 20, 2010 12:56 am    Post subject: Reply with quote

Ok, I am rendering the mist_partcles.
But I noticed it is slow for such simple smoke.
Do you really need a fractal recursion level of 5?
Max 3, usally 2 should be enough for a particle.

Then you can directly pipe the gradient into the V coord of the ramp.
No need for a shader in between.

Then the place3DTexture:
What effect do you want to get with these external coordinates?
And I do not know which effect this kind of texture projections can have.
As volume UVW coordinates are different than default object UV coordinates.


This is the image I got on frame 85, layer mist_lowcon:




It looks a little bit more dense, but no artifacts.

Which version of the shaders are you using?


.
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Wed Jan 20, 2010 2:31 am    Post subject: Reply with quote

schoenberger wrote:
Ok, I am rendering the mist_partcles.
But I noticed it is slow for such simple smoke.
Do you really need a fractal recursion level of 5?
Max 3, usally 2 should be enough for a particle.

OK, I thought it was abnormally slow as well. Doing a test to duplicate your results and waiting on a render. Typically they like really detailed volumetrics here, so I have to push the noise detail and usually a value of 4 isn't even enough. Is the recursion level what is slowing it down so much? It really slows down when the volume gets transparent. I understand it has to march further in that case, but I wish there could be some optimization there.


schoenberger wrote:

Then the place3DTexture:
What effect do you want to get with these external coordinates?
And I do not know which effect this kind of texture projections can have.
As volume UVW coordinates are different than default object UV coordinates.

On of the pipeline TDs here is responsible for that. Supposedly that is the correct node setup that mental ray expects. After looking at one of the sample files, I noticed that you don't use them in your networks, so perhaps this is unnecessary.

schoenberger wrote:

Which version of the shaders are you using?

We are currently using BA_Volume_Cloud v3.6.20


I tried to duplicate your work, by recaching the particles and rendering frame 85. I did see some artifacts still. I am re-rendering frame 65 but when it is done, I will post both. But did you change the settings as described above, by lowering the noise recursion, because my image looks a bit different, including the artifacts.
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Wed Jan 20, 2010 2:46 am    Post subject: Reply with quote

goshone wrote:

Typically they like really detailed volumetrics here, so I have to push the noise detail and usually a value of 4 isn't even enough. Is the recursion level what is slowing it down so much?

It depends on the particle size. If you have huge particles, then you perhaps need to increase the level.
Just do a side-by-side rendering with level 2.
Sometimes you could need more levels even for smaller particles if you have scaled the fractal to large.
Imagine you have a huge noise pattern but you scale in and use only a tiny part of it. Then you need of course more detail.

goshone wrote:

On of the pipeline TDs here is responsible for that. Supposedly that is the correct node setup that mental ray expects. After looking at one of the sample files, I noticed that you don't use them in your networks, so perhaps this is unnecessary.

It is probably not the right way for the volume. The BA shaders have a texture projection build in.
And if you want to use other MRay shaders for particles, then use the BA-texcoord shader to get your UVW coordinates. It's faster in the volume and animated scale/rotation is applied right.

Note: The particle texture is what slows down the rendering most (perhaps with the shadow calculation, if you have a lot of lights or use soft shadows/multiple shadow rays for a light).
Each shader you add to the particle texture slows down it more and more.

goshone wrote:

We are currently using BA_Volume_Cloud v3.6.20

Ok, I was testing with 3.7.
Many scenes with volumes at the moment or can you update?

goshone wrote:

But did you change the settings as described above, by lowering the noise recursion, because my image looks a bit different, including the artifacts.

No, I just rendered.
(I disabled the light shading and made it white as the alpha shows the artifacts too)
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Wed Jan 20, 2010 3:37 am    Post subject: Reply with quote

So I will render a sequence of images tonight. I was unable to see any clear artifacts but they are much more noticeable in motion.

I am also testing with a larger scene. So far it seems to be performing better but a full rendered seq will provide more insight.

I did notice a serious speed increase when disconnecting the particle gradient that was connected to the 'density shape -> intensity' parameter. I wanted to have little or no noise at birth to quickly introduce noise once the particle was a few frames old. But I can probably live without this. I also disabled the particle gradient density input, just to simplify things and I assume speed up renders.

there are only a few options now as the cause of this:
1. corrupted cache - could this possibly be the cause?
2. density limit and high noise recursion
3. different BA version - I think we can update this relatively easy
4. could many overlapping particles create problems, fading out at different rates?

anyway, I will continue with testing and let you know what I find
Thanks for the help
Josh
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Wed Jan 20, 2010 3:45 am    Post subject: Reply with quote

Reasons for the Issue:
1. No. Then the issues were particle shaped.
2. No.
3. Probably, perhaps in conjunction with your texture coordinate setup.
4. No (Only if you have more than 500 overlapping in the same area)


PS: I have not checked if you use distortion in your fractal. That is also slower.
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Wed Jan 20, 2010 4:09 am    Post subject: Reply with quote

here are some of my rendered images with the artifact still visible.
Only changes I made were:

cell size = 1.0
noise level = 3

http://www.youshare.com/goshone/6b7ad2db207bbd70.zip.html

ps. from now on I will not use the external coordinates. I disconnected them and there was no difference.

Also, in some of my earlier scenes, I may have had that many overlapping (>500).
And no distortion on the noise.
Back to top
View user's profile Send private message

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Wed Jan 20, 2010 7:48 pm    Post subject: Reply with quote

This is what was rendered last night:

http://www.youshare.com/goshone/36cda706c93f39ec.zip.html

Very obvious artifacts there. This was from the scene I rebuilt at a larger scale - not from scratch but just adjusting fields and particle behavior.
I am starting to think it is due to many overlapping particles. Notice that the artifacts are mainly happening at the base of the volume where many particles are accumulating.
Back to top
View user's profile Send private message

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Wed Jan 20, 2010 8:49 pm    Post subject: Reply with quote

Yes, it could probably the particle count in one space location.
The artifacts are then from a MRay function that finds the surrounding particles in space.
Try to use 1/3 of the particles and increase the density/alpha of the particles*3.
(should be also 2x times faster at render time)

Or try that the particles do not get that big. Then there are less overlapping.
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

schoenberger
Site Admin


Joined: 02 Mar 2005
Posts: 3786

PostPosted: Mon Jan 25, 2010 3:23 pm    Post subject: Reply with quote

PS: it could be that your version does only support 250 particles at the exact same location in space.
(instead of 500 of the current version)
_________________
Holger Schönberger
Binary Alchemy - digital materialization
Back to top
View user's profile Send private message Send e-mail

goshone



Joined: 11 Oct 2006
Posts: 68
Location/Company/Country: hydraulx, Santa Monica

PostPosted: Tue Jan 26, 2010 1:01 am    Post subject: Reply with quote

ok, thanks for the tip...
We will upgrade to the latest version.
I was able to get better results by modifying my sim to have less particles in the bulk and more where they count.
Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic   Reply to topic    www.BinaryAlchemy.de Forum Index -> Shader Problems / How do I...? All times are GMT + 1 Hour
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
 
Documento sin título
 



Powered by phpBB © 2001, 2002 phpBB Group



Number of shameful bots caught by Anti-Spam ACP: 1667