 |
 |
 |
 |
Opacity is set to 0 at the birth point |
Opacity is given a positive value at the birth point |
Tips and Tricks |
Home |
Avoiding grid artifacts on the source surface
Rendering with an external volumetric shader.
Creating the illusion for bigger resolution
Max. transp. levels of V-Ray and pure geometry mode
Complex rendering and "Circular reference" message
Fluid simulations in hollow objects
Connecting two simulators in cascade way
Simulation and rendering of fine smoke without using a huge grid resolution
In some cases the cells of the grid are clearly visible over the surface of the geometry used as a source. To avoid this artifact the transparency diagram should be arranged do give zero opacity at the birth value of the source channel. The example below shows two images created with different transparency diagrams. The birth temperature of the fluid is 2000K.
|
|
|
|
Opacity is set to 0 at the birth point |
Opacity is given a positive value at the birth point |
The light cache can cause clearly visible spots over the surface of dense smokes, so it is not recommended to be used in such cases.
 |
 |
Rendered with Light cache - On |
Rendered with Light cache - Off |
Phoenix FD is able to export its content as texture (see Phoenix texmap) and this makes it possible to use an external volumetric shader for the rendering, for example the V-Ray environment fog.
Simple setup for V-Ray environment fog:
To avoid waste of time for test rendering with different displacement settings, the shader can be switched temporary to solid mode, that is much faster than the usual atmosphere mode.
Even without fine details an image can look as if it has a lot of detail, if it has high sharpness and no pixelisation. This effect can be used to produce high resolution looking simulations with relatively coarse grids, using a properly adjusted transparency diagram.
The images below are made with the same simulation input and different transparency settings. The left one is made with simple smoke option, which gives the natural distribution of the smoke without any tricks. The second image uses a transparency diagram containing one peak near to the zero, that produces sharp smoke appearance.
 |
 |
Simple smoke |
Peak transparency diagram |
Rendering in pure geometry mode causes high transparency level and you have to increase the Max. transp levels of V-Ray to a bigger value, given by the following formula: max transp level=grid size*100/step.
The image to the right illustrates an issue that occurs when the Max transp. levels is not high enough
In many cases when the rendering setup is complicated, the user will get an unexpected message stating "Can't make a circular reference". This message appears when the object depends on itself, mostly when some channel is exported as texture and when the same texture is used in the shading. To avoid this message usually a second Phoenix object is introduced, its input path is redirected to the first one, and the whole render setup is made with it. The first Phoenix object should be disabled for rendering, it is used only for the simulation.
 |
 |
An example of a circular reference |
Render set-up to avoid the circular reference |
Phoenix FD has the ability to perform a 2D simulation, if one of the sizes is set to 1. To keep some features like the embedded gravity and pressure decay, it is recommended to keep the z direction active.
Despite the difference in appearance both of the mentioned artifacts have a common reason - too big value of the advancing step compared to the opacity gradient. Depending on the situation there are two ways to solve the problem: to decrease the advancing step or to decrease the transparency gradient. The first solution will increase the rendering time, the second one will decrease the sharpness of the resulting image.
 |
 |
Noise |
Slices |
 |
 |
Low gradient |
Low advancing step |
When simulating fluids in hollow objects one must make sure that the normals of the geometry are pointing inwards.
The simulation of settled liquids is relatively hard for most liquid simulators as usually a residual rest flickering or persistent waves can be seen. PhoenixFD is able relatively easy to achieve settled liquids with perfectly flat surfaces. The most important parameter is the sharpness - set it to zero. If you need it bigger during the simulation, you can animate it, or better to switch on the "static surface correction" that will automatically decrease the sharpness when the velocity is low enough. It's also recommended that the option "Strong surface mode" is enabled. The second important parameter is the lower SPF limit. The practice shows that is good to be about two times bigger than the natural SPF value with advection step=2.
There are many situations when the simulation area is narrow and its volume is not very large by itself, but for some reason its bounding box is a big one and requires a huge grid to cover it. A perfect example for such a situation is a waterfall - the upper area can be covered by a box, which is wide, but not very tall; the vertical area can be covered by a tall and not very wide box; but together they have to be covered by a box that is both wide and tall. A cascade connection of the simulators is suitable exactly for such a situation, because it allows to cover the simulation area with two simulators, each of them having a relatively small grid. This technique is not a core feature (there is no way to connect the cores), but the ability to put a phoenix simulator as a source for another one is used. When a phoenix object is used as source, the implicit surface determined by the effects channel and the surface level defines the geometry surface and the new born fluid will appear over it. The difference, compared the the usual geometry sources, is that we have the ability to set the velocity of the new born fluid to the same value that the fluid selected as source has in this point. As a result, the new born fluid will gracefully continue its movement and it will appear as if this is the same fluid. To avoid the additional speed added by the discharge parameter, set its value very low, for example 0.001. Do not make it zero, because the system will decide that this is not a source and will skip it. The technique can be used for gaseous simulations too, but with a small modification. You have to put the surface level to a very small value, thus making the source surface a box. To transfer the parameters of the gas from the first simulator to the second one, export the used channels as maps and put them in the map slots of the source helper. Do not start the simulators simultaneously, because the second one will run faster (it's empty) - run the source simulator first, and then the second one. As the simulation will usually be ran multiple times, to avoid backward interferences, add the second simulator in the exclude list of the source one.
A well known issue of grid simulations is the numerical dissipation - an undesired diffusion of the grid values caused by the advection. This makes the simulation of fine layered smoke very hard, despite the lack of turbulences. The usual solution is to simply use a huge grid, however this makes the whole simulation extremely slow. Actually, we do not need that resolution for the whole simulation — the low resolution velocity is pretty good — all that we need is just a dissipationless advection method for the smoke. A good solution for this problem is to use particles to represent the smoke and to drive them with a low resolution simulation. Phoenix FD has the tools making this possible - it can drag particles, and can render them as fog.
Enable particle generation in the source, set the rate to about 10000.
Select the particle type of the source to be "Drag"
Perform the simulation. Note that the Velocity channel should be exported to avoid the slice-like artifacts near to the source, and, of course, for motion blur.
Create a PHXFoam object
Enable the "As fog if below" option, and use some big value, say 100 (by default, the particles are of size 1). The big value is needed to ensure that all the particles will be shaded in fog mode.
Perform a test rendering and adjust the transparency and the size multiplier of the PHXFoam shader.
Try to decrease the voxel size of the fog (fog resolution parameter) until the rendering becomes fine enough. WARNING!!! One halving of the fog resolution results in an eightfold increase of the consumed memory! Save the scene before each attempt!
 |
 |
300k cells simulation rendered directly |
300k cells simulation rendered via 10M particles |
A very common question about liquid simulations is how to start a simulation with a volume filled with liquid. Of course, there is an obvious way, by using a preliminary simulation, with a normal source, simulating until the volume is filled and the liquid is settled. However, this could be very slow - sometimes it can take longer than the "real" simulation itself. There are two faster ways to do this, without the initial simulation, as described below:
Using brush mode source
Create a geometry covering the volume that should be filled with the liquid.
Exclude this geometry from the interaction, by adding it in the simulator's exclude list, or by just hiding it.
Create a liquid source, and select "brush" for the "if not solid" option.
Animate the discharge from 100 to 0 for the first two frames.
Disable the velocity check box of the source helper
Using mapper object and VRay distance texture
Create a geometry covering the volume that should be filled with liquid.
Exclude this geometry from interaction, by adding it in the simulator's exclude list, or by just hiding it.
Select this geometry in a VRay distance texture.
Make the texture white in the geometry and black outside the geometry.
Select this texture in a PHXMapper object in the "map" slot, and select the liquid channel to be affected. Make sure that the "Initializer" check box is on.