remove splash on grid boundaries

[kimgar]

easy to get carried away, but at least i learn a lot in the process

basically, i wanted to remove the splashing at the grid boundaries where the simulation meets the infinite sea :


which turned me to the nightlies that doesn't splash at all


which led me back to the official build and mask out the borders in post, problem solved, only not within phoenix

[Ivaylo Katev]

aaah i see
this is usually made by adjusting the birth threshold, increase it in order to remove the splashing of the areas where the splash birth is low.
one day ago there was a change in the splash birth algo, in order to improve the places where the splash is born. i would recommend you to try how it works, it may help you

[kimgar]

hmm, no luck, tried it with 5/10 release, no splashes...

[Ivaylo Katev]

what is your threshold and birth rate? i'm pretty satisfied by the results now, finally we achieved the long pursued appearance of the wave peaks.

[kimgar]

in the official build i had birthrate=0,01 and threshold=5; this worked fine
in the nightly i have birthrate=100 and threshold=0,01; no splashes, especially concerning fluid/geometry interaction. the turbulence field is low in strength but upping it to much will create gaps between grid and infinite ocean. this is a static object so there is not much action to create the splashes except the wind turbulence spacewarp

[Ivaylo Katev]

what is this circle in the center of the simulation?

[kimgar]

that would be my collision geometry, the lower part of the rig from this shot



the film is already finished and the client was happy. it is now just a matter of research to improve the workflow for the next project

[Ivaylo Katev]

now i understand your problem
actually, what produces the activity? don't forget that the displacement waves are fake, they can't produce splash. i suspect you have no activity in the simulation.
in the old builds you have splash, because we had a bug, the thin layer of liquid produced splash, even without motion.
however, this bug is already fixed. in this thread i suggested you to use the ocean texture to produce real waves. did you try this? it's not so easy, but you goal is not easy at all.

[kimgar]

in this situation the thin layer bug was perfect the only activity in the simulation is a wind spacewarp which worked nicely (with only 13 mill gridsize) except for the boundary splashes which was fixed in post.

i did try your ocean texture mapper scene, but couldn't produce any splashes, i suspect it was because of some nightly installed, i will have another go at it with latest build and official

[TSR_trix]

Very nice mood kimgar in your shot. Could we see the whole sequence?

[kimgar]

thanks trix, the client hasn't gone official with it yet, but i'll make an upload when he does. shouldn't be too long

[kimgar]

here is the animation, it is never perfect but the thin layer thing worked nicely in this case. wish there would be an option for this in the next patch instead of removed altogether, but i guess you can't have everything

http://www.indok-as.no/customerfiles....rev4.720p.mp4

[Ivaylo Katev]

very nice, only the particle count seems too low, and it was possible to turn on the mb for the splash, it is fast with the point shader.

[kimgar]

the splash particles is very low so i ran a resim with 10x splash particles but ran out of disk space during the night, got files 200mb+ each now

one question; is the velocity channel discarded in the resim files? i can't resim the files already resimmed

[Ivaylo Katev]

by default yes, this was fixed because many users got the same problem. 200MB is even small cache size for production involving foam and splash, in the nightlies we have added detailed control of the particle channels export, because by default they export everything and the files become huge. we also changed the restore concept, now by default every 15th frame can restore the simulation (the backup interval option), and the internal state is not exported if you select to export the velocity, all these changes are directed to decrease the cache size. this night i ran a simulation with 50M particles and 60M grid, the cache files were 700M for the non restore frames and 2G for the restore frames