Stepping patterns on liquid surfaces

[oliver.sullivan]

I'm currently testing PhoenixFD (for Maya, nightly builds), and really liking it so far, but I cannot seem to get rid of the stepping artifacts as visible in this image. At this point all that's happening is an initial fill-up and some time settling.



I'm using the FLIP solver (have tried others too), 110 million cells, have tried from 1 to 10 steps per frame, both open and closed boundaries etc... but can't make it go away. At the moment everything else is working great, but unless I can stop this happening we may have to use <shudder> Bifrost instead... (please no!)

Thanks for any suggestions

[Ivaylo Katev]

can you share more details, or directly send the scene to the support?
before sending the scene, one question - what is the surface level value of the mesher?
if you are starting from old viscous scene, i suppose it may be 0.1, that is totally incompatible with the flip solver results. flip exports the liquid channel that requires surface value 0.5, and any other value causes grid artifacts

[oliver.sullivan]

Surface level is 0.5, but thanks for checking. Assume I should send the scene to support@chaosgroup.com?

[Ivaylo Katev]

yes, this is the mail
very strange, i was ready to bet that the surface level is to blame, the flip itself does not produce step artifacts, because it's particle based, but the result is converted into grid format to make it renderable by the old pipeline based on a grid mesher,
there are no visible step artifacts in this conversion, except the surface level case that i mentioned.

[oliver.sullivan]

I've sent the scene through. I was originally dealing with the artifacts by using some overall displacement to hide it, but this was starting to make smaller details look wrong. I thought it was perhaps an issue with grid resolution or SPF, but these don't seem to make much difference to the issue?

[Paul Oblomov]

Check grid rotation. I remember had smth like this with rotated grid.

[oliver.sullivan]

Check grid rotation. I remember had smth like this with rotated grid.

Not rotated, but translations are non-zero... I'll try zero-ing them just in case.

Setting the Conservation Method to Symmetric rather than Smooth seems to help a bit in a low cell count sim. I'm trying a high count now...

[Ivaylo Katev]

the flip solver is not using this option (conservation method). are you sure this is a flip? can you post an image with the info panel, i.e. where the content of ta currently loaded cache is shown (particle count, channels etc)

[oliver.sullivan]

the flip solver is not using this option (conservation method). are you sure this is a flip? can you post an image with the info panel, i.e. where the content of ta currently loaded cache is shown (particle count, channels etc)

Hmmn, I really thought it made a difference when I changed the conservation method - unless it really is due to me resetting the simulation container's transforms?! The high cell count sim is still running so bit early to tell if it'll look different.

[Ivaylo Katev]

hm, there is not enough info in maya version, ok can you send me the phoenix log file?

[oliver.sullivan]

hm, there is not enough info in maya version, ok can you send me the phoenix log file?

Sent to support, thanks.

[Ivaylo Katev]

confirmed, this is not the flip core. will say more later

[oliver.sullivan]

confirmed, this is not the flip core. will say more later

Interesting...! Thx for looking into it.

[Ivaylo Katev]

the coworker in the maya field said that the flip is not actually connected , added just in the UI

[oliver.sullivan]

OK... so is this something that can be added soon? Or is Maya development some way behind Max?