How to create droplets?

[bobweb]

Hello,
I have come across simulations of single droplets made with phoenix I would like to create in a similar way:
http://vimeo.com/91539364
http://www.youtube.com/watch?v=GziCw...LTPRMxUHqd0YOQ

I have tried the whole day countless variations now of surface tension (0-1), gravity (0,01-1), time scale (0,01-1), discharge amount (0,001-50), system units (mm-m), scene scale (cylinder radius 0,2cm - 5cm), resolution (0,01cm - 0,1cm) etc. and already fail at creating the first drop. How did you manage to make the liquid gather in the middle of the cylinder to form a nice drop before falling? In my tests the outer circle mostly fell first. If something like a drop formed, it got split/flattened/stretched on the way down. After impact I got a small bump coming up at most, no pillar so far. Also the initial fillup very often did not stay still until the drop reached the surface, from boiling to building up "stairs" at the edges. I tried it mostly with Smooth 20-80/Classic 2-6, also some with symmetric and forward transfer, but that tends to create geometric patterns like flowing in parallel lines or quadratic splashes.
All with version 3ds Max 2014 / Phoenix 2.2.0.

So I have to admit I am lost. How would you approach such a simulation? Where should I begin with the "right" way?

Thanks,
Bob

[Ivaylo Katev]

ok, here is a simple demo scene, there is some time multiplier (small scale simulations are hard), small viscosity to oppose the instability introduced by the surface tension, and perhaps the most important point - low resolution. the surface tension does not work good for big droplets, and the meaning of "big" is related to the droplet size measured in cells. when you increase the resolution the droplet becomes bigger, related to the cell size, and this leads to different result. this is a bit problem because the rendering needs good smooth surface. if you try to increase the resolution the droplets leave a sharp tail that do not shrink back to the source, as in the real case (which is well reproduced in low res). i suppose this can be overcame increasing the spf and increasing the surface tension, but it would require bigger simulation times and i can test it fast. i suggest to beat it in the renderign, not in the simulation. you can add particles in the liquid (just enable the particle option of the source) and to render them using some particle mesher, for example the vray metaballs or frost. if you use vray metaballs the setup with the official phoenix build is a bit complicated, get nightly build where this is fixed.

[bobweb]

Thanks a lot for your quick reply. Interestingly there was nothing happening when I started simulating your scene - until I installed the latest nightly. So maybe something changed in the meantime which hindered my tests with the "old" version, which is good.
I am still with Vray 2.4, so no metaballs for me at the moment, but I get your point. I will have to do some more tests with higher SPF (so far it keeps completely sticking to the cylinder with SPF 6), also for the initial fillup, which is wildly moving right from the beginning when added to your scene. It makes me really curious how the people behind the vimeo video did it.

By the way I would like to ask you about your general approach with system units. When looking at the sample files from Phoenix there are several different settings, 1, 3, 5, 20 cm etc. Your sample now has 2 cm. What is this choice based on? Is there a rule of thumb I have missed in the documentation? Also your cylinder is quite big for a water tap with about 20cm diameter, is this intentionally oversized to make it work?

Thanks,
Bob

[Ivaylo Katev]

when it's possible i'm using original size of the objects, in this case it is not possible, because the small scale simulations are very hard. the 2 cm choice is actually not a choice, when i started 3ds max, it was set to 2 cm, and knowing that there is no way to simulate in real size, i just adjusted the time scale to make the simulation possible.

[squintnic]

This arbitrary scale issue seems to be a real show-stopper. Is there any intention to make phoenix work in real world scale?

[Ivaylo Katev]

you mean the small scale problem? we recommend to use the real world scale always when possible, but for low resolution this is impossible (rather unacceptable slow) and the really working solution is to violate the scale. you can see this in many movies, it's not a big problem because in the most cases the watchers can't recognize it.

[bobweb]

After a few more tests I dare say that this is close to impossible, at least completely simulated. You could animate the drop manually and treat it as a solid object in the simulation, and then have lots of trial and error simulations with different size and speed of the drop and of course several different Phoenix settings, until you get the behaviour you want. Unfortunately my PC is not fast enough to do all these tests in a reasonable time, so I gave up for now. Frankly, I also got lost to name the problem with each sim asking which scene parameter's fault it could be. The main problem I could not solve in any simulation is a resting initial fillup surface. Whenever the surface tension gets higher, it moves by itself, the highest SPF I tried was 16 and still got a stormy ocean or at least some "waves" building up when the surface tension was lower.

[nicsquint]

yes, also had no luck whatsoever. The drops either dance around weirdly or flatten out and look like blood splats.

The scale issue is a big problem, I don't care about the watcher I'm more interested in the ten other people in the pipeline that have to deal with rescaling everything around a fluid sim - I haven't got far enough along to try rescaling the sim to match existing elements.

[Ivaylo Katev]

no need to rescale anything, just use scale multiplier in the dynamics panel.

[nicsquint]

Do you mean time scale? There is no scale attribute in the dynamics rollout, or is this elsewhere?

[Ivaylo Katev]

ah you are right, there is no scale in the official release. well, in this case use the gravity multiplier, actually this is the only parameter related to the real world and only it is affected by the scale. use gravity multiplier 0.1 is equal to 10 times bigger scale.