Volume quality VS Stochastic transparency

[catchyid]

Hi,

I am reading Mantra documentation and I don't understand the difference between Volume Quality and Stochastic Transparency parameters? For me, they sound the same: how many voxels we sample in volume.

https://www.sidefx.com/docs/houdini15.5/props/mantra#vm_volumequality

https://www.sidefx.com/docs/houdini15.5/props/mantra#vm_transparentsamples

Thanks,

[stickman]

Interested to see what the experts say... I understood it as follows:
volume quality addresses the resolution of blocks within the defined area... kind of like subD geo
stochastic transparency addresses camera-facing slices, like algorithmically doped up sprites

[catchyid]

I agree with definition of Volume quality, basically that what the docs says, it's like a percentage 0 means no samples, 1 means sample all voxels in a volume, >1 means for sample each voxel more than once. BUT the Stochastic Transparency is strange because when you sample voxels you already takes transparency into account (i.e. a volume by definition will absorb some light, so the more your travel inside the volume the more you become opaque and less transparent, so what makes this parameter independent of the Volume quality?)

[Atom]

I just think of Stochastic Samples as edge quality. It defaults to 4 for faster renders but you really need to increase it to 9-12 or more to remove the dancing particles around the edge of a volume.

[catchyid]

Thanks Atom ...It seems to me that it's taking really long time to render simple volumes and get rid off that salt and pepper look around the edge of a volume (well compared to FumeFX/Mental Ray). I am still testing because I am sure there is a faster way to get good results around volume edges, maybe there are some tricks I am not aware of. One last resort is to blur the image in compositing but this is not ideal

[catchyid]

I've read the documentation couple times now From my understanding: adjusting volume quality is like adjusting pixels sampling rate (i.e. it affects the volume uniformly, so 1 means sample ALL voxels in the volume, period!). However enabling Stochastic sampling is like asking Houdini: can you guess where the important voxels are and sample them only? So instead of uniformly sampling all voxels (or the percentage set in Voxel quality), Houdini tries to only sample important voxels (there must be some heuristics used internally by the rendering engine to determine how important a voxel is, e.g. maybe "close" to a light source, or maybe the density of the volume in a specific area is uniform hence it's easier to treat that area as one voxel...). Since borders are problematic (thin volume), one need to put high values such as 20 or 40 for the sampling number to get them right...