easy way to drop a volume of fluid?

[Tony_Morev]

or i'm missing some very important reason why they should be untimed?

you mean besides user convenience and disk (plus potentially time) savings?
you're a funny guy, Ivaylo!!!

BTW speaking of initial states and loading cache and such
in RF if a simulation is stopped and the program closed, next time you open the file and click simulate it automatically continues from where it stopped before - VERY CONVENIENT for lengthy simulations

also at any point in the simulation (with anywhere from 1 to 3 clicks) what you see can be turned into initial state and time reset to 0

i haven't tried PH Restore function yet, well i kinda have only to see it error out, but my point is the UI and defaults need to be made more convenient

RF doesn't have Pause function, instead it has Reset, so any simulation that's stopped automatically continues from disk or RAM unless the Reset is pressed (or "Reset to Initial state")

it's easier to use than Pause that only works for RAM, Stop that puts an end to the simulation without an ability to continue and Restore that i have yet to see do anything other than tell me i have an "Invalid backup file"

seriously it just seems so much more rational and this is what ultimately reflects on peoples ability to quickly learn Phoenix as well as your ability to sell it

[Tony_Morev]

good news! restore worked for me for the first time, may be it was a 1.2 problem

[Tony_Morev]

play with the surface balance parameter.

Now, i can see how "play" can apply to something more or less interactive
but if it takes several hours between changing the setting and seeing the result it's important to understand what the parameter does

so what do those numbers mean exactly? and in what units if aplicable?

thank you

[Tony_Morev]

odd ...
so i have two sources in one simulator one drop-fill the other a regular flow

if both are active the simulation runs fine
if only the flow is active the siimulation runs fine
if the flow is off (geometry removed from the source helper) the drop-fill first falls and then exlodes all over the place
this is in beta
i haven't tried it in 1.2

[Ivaylo Katev]

good news! restore worked for me for the first time, may be it was a 1.2 problem

yes it's possible. are you sure that you had the velocity exported? phoenix searches the cache files in backward order for velocity and restores the simulation from the first that has is.

but if it takes several hours between changing the setting and seeing the result it's important to understand what the parameter does

well, i'm doing this restoring the simulation in the point where the liquid is almost settled and after 10 frames the tendency is clearly visible.
the exact meaning of this parameter is too technical, in few words it introduces an "error" in the conservation calculations, that should compensate the error introduced by the classic advection. in the most cases (but not always) the classic advection produces positive error, i.e. the liquid after the transfer is a bit more than the liquid before the transfer, and to compensate this you need negative value of this parameter. keep in mind that this parameter is global, it affects the entire grid and if you have two different glasses with different levels of the water, it's very probably that you will be unable to achieve balanced level for both of them. btw i feel that the current name of the parameter is unappropriate, perhaps i have to change it.

[Tony_Morev]

Ivaylo, is there any way to stabilize the forward trasfer calculated fluid?

i picke it over classic since i have 30M cell grid and finding the right surface balance wasn't as easy you made it sound

the problem with forward transfer is that the fluid just won't settle! more so it's not going in circles like some describe, it's plain and simple BOILING!
that's what made me believe in the first place that the meshing is wrong and there's tearing in it
no tearing, it was just big ass bubbles raising of the bottom

it doesn't matter if i do drop of script cell fill
for the first few frames everything appears to be acting fine and then it comes to a boil
viscosity slows it down temporarily but not enough to make it go away, at least not in my lifetime

here's my first try, i stopped the sim after realizing the water above is moving to fast and the water below needs to settle first
and ever since i've been trying to make it settle
if you only look at what the basin is doing that stuff just goes on and on ....
http://vimeo.com/27053310

i have better bigger examples if you need to see them
for now i went back to classic, if filled with a script there is no turbulence at all, but growing is hard to manage

[Ivaylo Katev]

i wish there was a way
the ft advection has many advantages compared to the classic, however it just can't be used for settled liquids, due to the damn boiling! actually these two issues, the boiling and the growing, are the two sides of one coin, they are caused by the residual velocity after the conservation process.
what you mean with "right surface balance wasn't as easy you made it sound "? if the limits of the UI do not allow to achieve the value needed to stop the growing, you can set the needed value by the listener. just to be sure that you know this - the settled liquids are possible only in constrained spf mode, and because of this if you just copy the settings from another working scene, the chance of success is not big.
btw, just asking - why you need prefilled pool with settled water for the waterfall? the water will be disturbed from the waterfall anyway and you just can start with empty pool and wait until is filled by the waterfall.

[Tony_Morev]

i'm going to start answering from the bottom up ...

i need the settled water because the flow is going to be very slow and the turbulence is going to create be some sort of subtle ripple pattern
the physical size is about 1.5 human heights across the store threshold, so it's not a big watrfall, but it's not a tiny one either
as a matter of fact the actual basin is bigger than what i'm simulating and i will only use anywhere from 50 to 100 frames of the simulation
so i will have to crop the animation before the ripples start noticeably bouncing back from the rocks

by "wasn't as easy as you made it sound" i meant your reply when you said how you can preview the effect relatively quick
with the 30M grid nothing is quick so, no - it's not the UI limitation
after all the whining you've heard from me about superficial stuff like UI i finally have something very serious to complain about
and that's unfortunately the core of the program ...
i think you're overlooking something, there has to be a way to make this work
RF's hybrido is a grid simulator - the fluid doesn't boil and doesn't grow
Blender has a grid simulator also - their fluid doesn't act weird either
i haven't played much with houdini, but i believe they also use grids at least to some extent
so clearly there is a solution out there
and if it's not a bug that's causing all the problems and if you can't find a way to implement the surface balance automatically, then you need to consider something as fundamental as creating a new core for fluids only
at the moment the ease of fluid simulations with phoenix ranges from hard to IMPOSSIBLE
mainly based on the size and complexity
if you leave things the way they are you'll end up having a gas simulator that occasionally can fill a cup of water, provided the user has no access to RF where he can do it a lot easier

this is very serious Ivaylo, you ability to sell the software hinges on this
if all you're selling is a sibling of FumeXF that has a dysfunctional fluid sim built in, then even the marketing exposure that V-ray provides for you won't help much

the reason why i bought a copy is because i has hoping to get an all-in-one simulator
and so far the only thing that i liked about Phoenixe's fluid sim was the ability to interactively change settings while the simulation is running
BTW, seeing how that's not the case with other fluid simulators, if you have to take it away in order to make the core more stable, do so
i'd rather have a predictable simulation than the one i have to constantly steer as i sit and watch it run

i'm far form giving up on Phoenix in case you think i sound too negative
you have a lot going on, and i have no doubt there is a solution to the issues you have with fluids
it just needs a fresh critical look, instead of coming up with patches like "surface balance"

[Tony_Morev]

Ivaylo, is there a way for you to "close" the system in classic advection mode for fludis?
a filled cell preservation mode of some kind - what goes in will either come out or stay, irnoring any compression, as it's negligeble in fluids
or may be use it as a check to calculate surface balance
you've probably though of it already, but just in case ...

[Ivaylo Katev]

i think you're overlooking something, there has to be a way to make this work
RF's hybrido is a grid simulator - the fluid doesn't boil and doesn't grow
Blender has a grid simulator also - their fluid doesn't act weird either

well, if you can easy produce settled liquid with rf or blender, we really have a problem. but is it really like that? all the movies i have seen do not reach settled state.

[Tony_Morev]

this is a still shot of almost the same geometry done in RF
it's a test and is rough in many ways, but there is no boiling
yes there's some residual movement in the ware surface as id was also a drop fill
but was settling very noticeably i just didn't want to wait longer
plus it dropped from a higher point than the phoenix simulation



and here's the sim preview

http://vimeo.com/27096815

i didn't do this particular setup in blender as i need a flow also and i'm not sure it's currently supported outside of particle realm
it only allows a volume drop in the simple grid simulation, but in other tests i've done with it blender proved to be the easiest simulator to use
with almost no parameters to tweak and very predictable results, again no noticeable boiling or growing

if you can't make a fluid settle 100% it may be fine
after all the process has somewhat of an exponential nature, i'm guessing, and it just takes long
a slight movement is most of the time acceptable

now, i'm NOT saying that the math you use is comparable with RF or blender in any way and that you're just doing it wrong, all i'm saying is there are simulators out there that at least from a dummy user stand point appear to have "similar" type solvers and they work fine

of course i could be way off, and what you're doing is much more advanced, but if that's the case, more reasons to break off the fluids and simplify the solver
you have to parts that work to an extent: 1) forward transfer creates amazing looking flows, 2) classic has an ability to settle
so marry them! either find a way to combine them or allow different sources to be solved via different algorithms within one grid or allow a change of solver during one simulation (i haven't tried it actually, may be it already works ) all of this will allow you to stick with what you already have

there's always another way - start from scratch!
focus on fluids only, discard any gas influence (which right now is still part of the sim, as changing the preview options clearly shows - there is a whole bunch of stuff floating up the air above the fluid surface) and see what you can come up with

[Tony_Morev]

here's another longer run of forward transfer
you can see when i've upped the viscousity it slows the boiling
but then goes right back to it

http://vimeo.com/27100945

BTW another feature i feel is missing is per-source control of fluid parameters

and here's my second attempt with classic (i don't know why vimeo interlases the videos, may have something to do with stepping in the mesh)

http://vimeo.com/27101743

this one had surface balance at -0.1
apparently it wasn't enough the basin water level still grows
but being a scripted fill this time it has no turbulence at all

speaking of surface balance, is it grid resolution dependant?

[Ivaylo Katev]

BTW another feature i feel is missing is per-source control of fluid parameters

unfortunately here we can't beat the particle simulators.

speaking of surface balance, is it grid resolution dependant?

yes, you can't just increase the resolution and resim with the same value expecting to have the same result.
well, seems like you have serious experience with the liquid sims, so if you consider this a problem then very probably it is, i can't underrate your oppinion.

[Tony_Morev]

yes, you can't just increase the resolution and resim with the same value expecting to have the same result.

is there a way to predict how it will react to the grid size change?
like your SPF "rule"
and in general it's safe assume that 99% of simulations even the very small (physically) ones will first be started at a larger cell size for testing
which means, if you can't automatically adjust every setting that needs to reflect the cell size change, there at least should be some sort of "LOCK" feature that will change multiple settings simultaneously as the cell size changes to deliver more predictable results at different resolutions, provided you can come up up with a dependency rule

and truly that rule doesn't even have to be very accurate
think about it, your first advice in the help file is to stick with real world sizes as this is real world physics, right?
so in real world sizes haw large or small can a sane user set the grid cells, well i would say 1mm-5cm depending on the setup
anything below that is probably unnecessary and anything above is too chunky
so that could be you working range to establish any correlation of the settings

this could be a major step forward because it will make things quicker to test

[Tony_Morev]

does McCormack sound familiar?
that guy's stuff is used as one of the advection methods in houdini