Maya 2016 Ext2 SP1 + VRay 3.4 problems with cached particles and motion blur

[spacematt]

Hi all, I'm having some serious problems with cached particles and motion blur. First of all I know that to get motion blur with cached particles I need to cache enough subframes to match the shutter interval and motion samples. With that knowledge everything (kinda) works as expected in Maya 2016.

However there seems to be a problem with Extension 2 causing Maya to sometimes load the wrong cache frame, and to cause VRay to sometimes not load any cache data at all. Does anyone else see the same problems? Simple steps to reproduce...

Any particle system (preferably a simple one because the render times can be astronomical)
VRay Motion blur - duration 1, interval centre 0.5
Create an nCache with evaluations every 0.5 frames (should be enough right?).
Render a few frames and you should see one of:

• no particles
• non-blurred particles all concentrated at the origin
• motion blur streaks all ending at the origin

Works fine in Maya 2016/VRay3.3. Increasing the temporal resolution of the cache improves individual frames but it doesn't seem to be consistent across the animation, causing some flickering. I think this is a Maya problem, because around frame 50 or so I can see a frame doesn't load correctly into the viewport (this happens in any scene), but in Maya 2016 VRay3.3 seemed to just load the nearest cache sample, instead of just loading nothing, so maybe a problem with both?

Just one last thing, no matter what Maya/Vray version the attribute 'use interpolated cache' doesn't seem to to anything, but maybe that's just me. It's been a complicated day...

All this on everybody's favourite... OS X (10.10.5)

[spacematt]

UPDATE:

As far as I can tell VRay 3.4 in Maya 2016.5 cannot batch render cached particle animation at all, regardless of motion blur or nCache settings.

Also Maya is definitely having a problem reading the correct cache sample. Intercepting the time input to the cache node, and adding a small amount (0.0045) seems to fix it in the viewport, so I'm guessing they have a float rounding error. But that doesn't fix VRay which seemingly used to fall back to an earlier cache sample if it didn't get what it needed, but now just seems to find a bunch of zeros.

Would love to know if anybody else can reproduce this, especially on Linux which is a possible workaround for us.

[yolov]

From the tests that I've run it seems to be an Autodesk issue, the particles don't render when cached at all in batch. You can cache them to vrscene and them port back to maya with a VRayScene node, but I'm not sure how this would help your case.

[spacematt]

Thanks Alex, I have a case open with Autodesk, but I fear they're going to come back to me with the workaround I posted above - adding a small value to time. That masks the problem entirely in Maya, and I fear they're going close the case and bounce me back here. I've no idea what data Maya is feeding to VRay and whether it's entirely bad (all zeros) or just not quite where VRay is expecting it to be. Perhaps your tests have revealed more information which would help me keep my Autodesk case open?

[nfowler]

Has anyone successfully cached particles to a vrscene. I just get an empty transform. No child emitter, nParticle or VOPs.
Also, animation cache doesn't go above 999.

Maya 2016.5 Vray 3.40.02

[Morbid Angel]

I cached nParticles to its ncache format just fine from 2016.5 ext2 sp1.

However I rendered the particles from maya 2015 (I don't support vray for 2016.5 yet)

[yolov]

I still can't render cached nParticles with any renderer in 2016.5, not only V-Ray. That's why I think it's a Maya issue. Caching to alembic works, though, but the abc cache has to be created *before* any other particle cache is made.

[spacematt]

Definitely a Maya problem. Autodesk are aware, they have promised a fix asap, and they know the cause, but it persists in Maya 2017 for now. It's to do with how the time input is interpreted in the cache node. The particles are even wrong in the viewport, although the problem is subtle there. Look closely and you'll find a single frame jumps/holds where the wrong cache frame is loaded. The first is usually around frame 55-65.

[yolov]

Yes, 2017 has the same issue.