Random wrong rendering of splashes/foam

[ilmelovfx]

I have a problem when rendering PhoenixfD liquid sim.
I made a simple sim based on a famous tutorial, the grid size is about 18mil cells, no fancy stuff, just a boat mesh going straight line.
Every 10 or so frames there's a visible difference in volume and light of splashes/foam. First noticeable occurs at frame 16, then at frame 26, then at frame 33, etc... I did resim two or three times just to check if I did something wrong or if I changed some of the parameters, but the issue remained.
I use BF+LC with 2500 subdivs, retrace 8. Bucket render is set to 1-16, 0.05 (image) and 1-16, 0.05 (dmc). I use 4 pcs for DR rendering (two i7+two xeons), all pcs have Vray 4.30.01 and PhoenixFD 3.14.00, inside 3ds max 2020.3.
I've attached an animated gif so you can see the differences in frames, and sim + render settings.
Any clue why this is happening?

[georgi.zhekov]

Hey,

Does this always happen on the same frames? Does the issue reproduce if you render only on one machine?

[ilmelovfx]

Hi Georgi,

I re-rendered from frame 10 to frame 40 on one single PC, the main workstation, the one that holds the .aur cache files, and yes the issue is present, in the same frames as before (but I didn't resim this time).

[georgi.zhekov]

Can you send the scene over along with a few cache files where the problem is present to the support mail with a link to this thread so we can investigate what is going on?

Thanks!

[ilmelovfx]

File sent! Thank you

[georgi.zhekov]

Hey,

Thanks a lot for the scene. There seems to be a problem when you're rendering in Point mode with disabled scattering and using Physical Camera. I have logged this as bug in our system, under issue number PHI-5431
We will update the thread when we have something on this. In the mean time as a workaround you can use the Approximate scattering or use a regular Max camera and you should be good.

Thank you,

[ilmelovfx]

Hi Georgi, thank you to you and all Chaos support team. I'll use the workaround.
Btw, today I upgraded to PhoenixFD 4.10.00 and installed Vray 4.30.02 and I'm testing the same scene right now. I will post additional info here if there's some improvement.

EDIT: no, it behaves in the exact same way (difference in volume and light on frame 26); also the sim calculated with PFD 4 is way different from PFD 3.

In the meantime, do I have to add "PHI-5431" to the thread title? I think this could be useful

[georgi.zhekov]

What did you change?

[georgi.zhekov]

Ah - if you meant that the issue is reproducible with Phoenix 4.10 - indeed it does reproduce with it as well. But the workaround should do the job until we fix it.

Can you show us how the different sim looks like?

Thanks!

[ilmelovfx]

Nothing changed, i resimmed all the timeline with new versions of software, with the same PFD parameters that I sent you.
Yes, the issue is reproducible with PFD 4.10. I tested the workaround and it works well, though the splashes/foam become brighter but this is not a big problem.

In the following animated gif you can see the different sim (PFD3 that I sent you vs. PFD4 I did today: there are noticeable differences in the details of splashes/foam, you can see them especially near the rear (stern) of the boat). This is still the rendering without the scattering workaround. Click on image to see it non-blurry.

In my opinion it looks better in version 3.x but ok...

[georgi.zhekov]

I'm glad the workaround is working - you can make the color of the foam a bit darker and that should compensate a bit for the scattering.

As for the differences in Phoenix 4 we have changed the FLIP solver so that the Foam Falling Speed varies depending on the bubble sizes and the Initial Vel. Rand. option is applied not just to newborn Splash, but to splitting splash particles as well. You can find a list with all the changes here - https://docs.chaosgroup.com/display/PHX4MAX/4.00.00

[ilmelovfx]

Thank you Georgi, I will take a look to the docs. Now I'm testing workaround + new software versions in DR rendering just to make sure everything works and is set-up correctly.