Scene transformations and rendering volumes

[Ankit Pruthi]

Now i've noticed this on a couple of projects. If we take a camera exported from other softwares such as maya into houdini, and scale it down by a factor of 100, to account for cms to metres conversion, volumes tend to render very very dense, almost as if the density itself has been multiplied by a 100 times.

What is the reason behind this behavior?
I've noticed the same happen if there are no transforms anywhere, but a transform is done inside of a sop on the volume.

Below are the attached images showing the difference.

 

 

 

 

[pezetko]

The problem is scaling.
Good habit is always reseting the scale on the camera (and all objects to the 1,1,1), otherwise you get some (unwanted) surprises (different render time, sampling noise, displacement, depth culling, density.. etc.)
You didn't scale lights in your examples, so you got different results. Scale matters!

[koen]

If you interpret the density field as a concentration, scaling a volume up or down will increase or reduce the amount of "stuff" a ray travels through. if you look at beer's law: ( http://en.wikipedia.org/wiki/Beer%E2%80%93Lambert_law ) the absorbtion relates linear to the  path length. When you scale the volume, you change the path lenght.

 

Cheers,

koen

Edited August 13, 2014 by koen

[Ankit Pruthi]

If you interpret the density field as a concentration, scaling a volume up or down will increase or reduce the amount of "stuff" a ray travels through. if you look at beer's law: ( http://en.wikipedia.org/wiki/Beer%E2%80%93Lambert_law ) the absorbtion relates linear to the  path length. When you scale the volume, you change the path lenght.

 

Cheers,

koen

 

Yes that makes complete sense. It also explains why a volume which is about a 1000 metres in size needs to have a density of the order of 0.001 or so to start looking different. So, if i understand it correctly, it depends on how many units the ray travels according to camera space coordinates.

 

Now, what would be the appropriate way to convert a camera imported using alembic from cm units into metres?? Something that works only on the x,y and z transforms and does not effect the scale? I'm not even sure if it makes sense to do that.

As of now the workaround is scaling the density down or up by the same factor used on the camera.

[hall]

I used to re-build the camera from Alembic using just the point orientation to the new camera, it's a bit work, but I looking for a better solution as well.

[eetu]

I used to re-build the camera from Alembic using just the point orientation to the new camera, it's a bit work, but I looking for a better solution as well.

 

You could check Jordi's setup from page 118 of his General Concepts guide at http://www.sidefx.com/index.php?option=com_content&task=view&id=2711&Itemid=132

 

edit: fixed page number

Edited May 10, 2015 by eetu

[tfreitag]

haa! thanks eetu (and of course jordi) for the link ..after spending serveral hours try to rendering volume with max imported camera, this is nice trick, with blend node.

...and its page 118.

Edited May 10, 2015 by tfreitag

[Ryan]

I am assuming you are using a Null to scale your camera.

 

You could also unhide the scale parameter in the camera and put in scale x, y, and z:

 

origin(opinputpath(".",0), "", "SX")

origin(opinputpath(".",0), "", "SY")

origin(opinputpath(".",0), "", "SZ")

[Jebbel]

Thanks guys...this information solved my problems I was having with camera rays. The bigger your scene is...the more apparent the issues become (weird noise, flickering and clipping). We're scaling our maya scenes to 0.01 before alembic export and I rebuild the camera in houdini using the method described in page 118 ->  http://www.sidefx.com/index.php?option=com_content&task=view&id=2711&Itemid=132

 

cheers