Velocity Blur on fluids weirdness. Is it a bug or just me?!

Hi there,
Been playing about with latest SVN (build from Mpan in particular) to render out a test scene. I’m using basic comp with DOF blur and Velocity blur to learn how nodes (and blender tbh) work.

The DOF is no problem but the Velocity blur seems to be not being applied to the smaller particle elements of the fluid object. Is the velocity not working on these (see below close up) or do I need to change a threshold somewhere?
http://homepages.nildram.co.uk/%7Ejamiemur/CG_SPHERE_FLUID_TESTS/flood_still_lores.jpg
Full-res jpg Here.

Another example frame here:
http://homepages.nildram.co.uk/%7Ejamiemur/CG_SPHERE_FLUID_TESTS/Missing_Velocity_Blur_close_up.jpgFull-Res complete.jpg here.

There’s an h264 avi movie here: cg_sphere_fluids.avi
You can see the sharp smaller elements whizzing about un-blurred which kind of breaks the motion blurs desired effect.

Here’s the node graph:
http://homepages.nildram.co.uk/%7Ejamiemur/CG_SPHERE_FLUID_TESTS/Nodes_cg_fluid_test_01_mni.jpgFull size image:Nodes_cg_fluid_test_01.jpg

All suggestions/ideas welcome. I’ll keep playing with this one for a bit longer for now.
Cheers

TxRx

Use MBlur.Report bug.

When an object crosses another’s (or it’s own) path vector blur chokes. It deosn’t look like a bug to me. If you take a good look you’kll see that it’s only elements in close proximity to the camera that fail to blur, probably as a result of either fiddling with the threshold settings or values in the vector pass canceling each other out due to opposing directions of travel. Many of the smaller pieces do appear to have blur applied, just not even close to the level of the main body. You can tell because some have diffused specular highlights while other’s are very sharp. It’s a very linear effect that ill tollerates overlapping motions in opposing directions.
Always put vector blurred objects on individual render layers and also don’t mix them with static objects. That will correct most of the problems you’ll encounter with it and you won’t need to fiddle with the threshold settings.

Thanks for replying folks. I’ll separate the fluid from the rest of the scene and try a couple of short sequences/frames to see if that makes a difference. The smaller particles are generated by the larger fluid sim. They’re all one object (just to clarify that).

If the velocity channel is getting choked by the same objects velocity in the background (which sounds as if it’s right) then there may be be a slight bug. If I take the fluid into a separate render layer and it still presents this problem then I may submit a bug report after I try a few things. Even in commercial apps things don’t always work as intended with the vanilla settings straight out of the box, so I should at least confirm this is a bug first before I post a report.
Cheers

TxRx

It’s looking like it’s those smaller elements of the fluids which aren’t producing and velocity/vector data at all. Rendered as a single layer, even looking at the velocity pass, you can see a complete omission of the data for those secondary particles which the fluid body seems to generate. I’ll compile some screen grabs, blends and so on and prepare my bug report tomorrow.

Cheers

TxRx

I could understand if it thought that the fluid domain was not moving, but I would think that each vertex would have to have its own vector…so even a small droplet that may be five verts, each of them must have some sort of velocity vector, or blender would not know to move them for the next frame. So if they have a velocity vector, the vector blur should know how to blur them. I don’t know why it would “stop” for small ones, but I am sure the dev knows.

This is just a wild guess but could it be that a vector pass is only calculated on a per object and not on a per mesh base? As far as I know the fluid sim only produces one object consisting of all the meshes from the fluid. So the vector pass would only incorporate the broad direction the mesh is flowing and not each droplet by itself.

@Musk, That’s what I thought, perhaps Papasmurfs suggestion that there may not be enough verts to possess/calc velocity has something to it but I’ve seen some larger (let’s call them secondary splashes of the main fluid object) have enough verts but still no velocity data.

I thought maybe it’s because they are coming straight at the camera it would be something which wouldn’t show velocity clearly.I’ll have to try two cameras on the same frame and so forth to double check that isn’t the case.

To my initial (untrained in the source code) eye it seems that the secondary splashes which aren’t in possession of any velocity calculations may be a different piece of code which may just be missing the params/routine to calculate the velocity of those particles. They do move across the scene just the same (if not faster) than the main body. They should present some form of velocity data.

I’m at work just now but tonight I’ll try out some other scenes using the same params and scenarios to see if I can replicate this properly. I’ll upload blend files once done.

Cheers for replying

TxRx

I couldn’t see it appear immediately after I reported it but it has been posted.

I tried another angle, removed the DOF and saved out an example still of the lack of velocity data. There has [historically at least] been a few vector blur bugs so hopefully I’ve not made some schoolboy error and submitted it as a bug! Fingers crossed all goes well and it get’s picked up before RC2! :slight_smile:

http://homepages.nildram.co.uk/%7Ejamiemur/CG_SPHERE_FLUID_TESTS/Camera_02_f50_missing_velocity_smaller.jpg
Full-res still here