In this exercise, you'll render Anthony Honn's vehicle model in a studio environment and in an outdoor setting. Furthermore, the car is rendered using a different shader on the body for each layer. The first step is to create a new render layer for the scene.

An object's visibility can be on for one render layer and off for another. Likewise, if an object is on a display layer and a render layer, the display layer's visibility affects whether or not the object is visible in the render layer. This is easy to forget, and you may find yourself unable to figure out why an object that has been added to a render layer is not visible. Remember to double-check the settings in the Display Layer Editor if you can't see a particular object.

You can use the Relationship Editor to see which layers an object belongs to. Choose Window

To create a different lighting and shading setup for a second layer, you'll use Render Layer Overrides. An override changes an attribute for a specific layer. So, for example, if you wanted Final Gathering to calculate on one layer but not another, you would create an override in the Render Settings window from the Final Gathering attribute. To create an override, right-click next to an attribute and choose Create Layer Override. As long as you are working in a particular layer that has an override enabled for an attribute, you'll see the label of the attribute highlighted in orange. Settings created in the master layer apply to all other layers unless there is an override.

This next section shows how to use overrides as you create a new layer for the outdoor lighting of the car.

Notice that you do not need to add cameras to render layers when you add them to a scene. You can if you want, but it makes no difference. The cameras that render the scene are listed in the Common tab of the Render Settings window.

If you're rendering an animated sequence using two cameras with different settings as in the carComposite example, you'll want to use overrides so you don't render more images than you need.

After creating the overrides for the cameras, it is still possible to render with either camera in the render view. The overrides ensure the correct camera is used for each layer during a batch render.

Figure 12.8. Create a layer override for the rendering camera on the studioLighting layer.

The flexibility of render layers becomes even more apparent when you apply different shaders to the same object on different layers. This allows you to render alternate versions of the same animation.

A material override applies a material to all of the objects within a particular layer. To create a material override, right-click on one of the layers in the Render Layer Editor and choose Overrides

Render layers can use blend modes, which combine the results of the render to form a composite. You can preview the composite in the Render View window. Typically you render each layer separately, import the render sequences into compositing software (such as Adobe After Effects or Autodesk Toxic), and then apply the blend modes using the controls in the compositing software. Maya gives you the option of creating a very simple composite using render layers, which you can view in the Render View window.

Blend modes use simple algorithms to combine the numeric color values of each pixel to create a composite. A composite is created by layering two or more images on top of each other. The image on top is blended with the image below. If both images are rendered as Normal, then the top image covers the bottom image completely. If the blend mode is set to Multiply, then the light pixels in the top image are transparent and the darker pixels of the top image darken the pixels in the bottom image. This technique is often used to add shadowing to a composite. If the blend mode of the top image is set to Screen, then the darker pixels are transparent and the lighter pixels brighten the pixels of the lower image. You can use this to composite glowing effects.

The blend modes available in Maya are Lighten, Darken, Multiply, Screen, and Overlay:

In this exercise, you'll use blending modes to create soft shadows for the render of the car in the studio lighting scenario.

This is a good way to preview basic composites; however, in practice you will most likely want more control over how the layers are composited. To do this you should use more advanced compositing software such as Adobe Photoshop (for still images) or Adobe After Effects or Autodesk Toxic (for animations). Autodesk is working toward automating the Maya to Toxic workflow so render layers can be imported into Toxic while maintaining their respective blend mode settings.

As an example you'll start by creating an ambient occlusion render pass for the space helmet scene. In previous versions of Maya you could create a render layer and use the Ambient Occlusion preset. The preset would automatically shade all of the objects in the layer with a custom ambient occlusion shader network. In Maya 2009 the mental ray renderer has a built-in ambient occlusion pass, which creates ambient occlusion shadowing in a render pass without the use of a shader network. This gives you another option for creating ambient occlusion passes. Ambient occlusion passes are frequently used in compositing, so you'll start with this type of pass.

The previous example demonstrated how to create a single render pass. In this example, you'll create multiple passes for reflection, specular, depth, and shadow.

Depth passes are particularly helpful in compositing. They can be used to add, among other things, camera depth of field in the composite. For example, using Adobe After Effects's Lens Blur filter, you can apply a depth pass to control the focal region of the render based on the luminance of the depth pass. There are also plug-ins available for After Effects, such as Lens Care by Frischluft, which create realistic lens effects far superior to the standard After Effects blurs. Using a depth pass for depth of field in After Effects dramatically reduces Maya render times, because mental ray's depth of field can take a long time to calculate. Furthermore any changes you make to the depth of field, such as the focal region, are done in After Effects and do not require re-rendering the entire scene. The same is true for motion blur. You can create a 2D or 3D motion vector pass and then use a plug-in such as Reel Smart Motion Blur to add motion blur in the composite rather than in the initial render. This is a huge time-saver.

You can add render passes to a render layer in the Render Layer Editor. To do this, right-click on the layer and choose Add New Render Pass. You can choose the type of pass from the pop-up list (Figure 12.25). The pass then automatically appears in the Associated Passes section in the Passes tab of the Render Settings window.

Figure 12.24. Clockwise from the upper left are the reflection, depth, specular, and shadow passes for the helmet scene. The elements of the passes appear dark because they have been separated and rendered against black.

Figure 12.25. You can add render passes to a render layer by right-clicking on the layer in the Render Layer Editor.

Render Pass Contribution Maps can be used to further customize the render passes for each layer. A contribution map specifies which objects and lights are included in a render pass. By default when you create a render pass and associate it with a particular render layer, all the objects on the layer are included in the pass. Using a contribution map you can add only certain lights and objects to the render pass. The whole point is to give you even more flexibility when rendering for compositing. This exercise demonstrates how to set up contribution maps.

Lights can also be included in contribution maps. If no lights are specified, all of the scene lights are added. In the minigun scene the directional light is the only light that casts shadows; the other two lights have shadow casting turned off. You can use pass contribution maps to create a shadow pass just for this light.

Render pass sets are simply a way to organize large lists of render passes in the Passes tab of the Render Settings window. You can create different groupings of the passes listed in the Scene Passes section, give them descriptive names, and then associate the set with the render layer or the associated contribution maps. If you have a complex scene that has a large number of passes, you'll find it's easier to work with the pass sets rather than all of the individual passes.

You can create a render pass set as you create the render passes or add them to the set later. In the Create Render Passes window, check the box that says Create Pass Set, and give the pass set a descriptive name (Figure 12.32). The new pass set appears in the Scene Passes section in the Render Settings window along with all of the newly created passes (Figure 12.33). To associate a render layer with the new pass set, you only have to move the pass set to the Associated Passes section. All of the passes included with the set will be associated with the layer even though they do not appear in the Associated Passes section.

Figure 12.32. You can create a pass set in the Create Render Passes window using the options at the bottom.

Figure 12.33. Associating the set with the current render layer associates all of its contained passes with the layer. You can use the Relationship Editor to add and remove passes from the set.

To verify which passes are included in the set, open the Relationship Editor for Render Passes (Window

You can add a new set in the Passes tab of the Render Settings window by clicking the Create New Render Pass Set icon. You can then use the Relationship Editor to add the passes to the set. A render pass can be a member of more than one set.

File tokens are a way to automate the organization of your renders. If your scene has a lot of layers, cameras, and passes, you can use tokens to specify where all the image sequences will be placed on your computer's hard drive, as well as how they are named.

The image sequences created with a batch render are placed in the Images folder of the current project or whichever folder is specified in the Project Settings window (see Chapter 1 for information regarding project settings). Tokens are placed in the File Name Prefix field found in the Common tab of the Render Settings window. If this field is left blank, the scene name is used to label the rendered images (Figure 12.34).

Figure 12.34. The File Name Prefix field in the Common tab of the Render Settings window is where you specify the image name and tokens.

By default if the scene has more than one render layer, Maya creates a subfolder for each layer. If the scene has more than one camera, a subfolder is created for each camera. For scenes with multiple render layers and multiple cameras, Maya creates a subfolder for each camera within the subfolder for each layer.

You can specify any directory you want by typing the folder names into the File Name Prefix field. For example if you want your image sequences to be named marshmallow and placed in a folder named chocolateSauce, you can type chocolateSauce/marshmallow into the File Name Prefix field. However, explicitly naming a file sequence lacks the flexibility of using tokens and runs the risk of allowing you to overwrite file sequences by mistake when rendering. You can see a preview of how the images will be named in the upper portion of the Render Settings window (Figure 12.35).

The whole point of tokens is to allow you to change the default behavior and specify how subfolders will be created dynamically for a scene. To use a token to specify a directory, place a slash after the token name. For example, to create a subfolder named after each camera, type <camera>/ into the File Name Prefix field. To use a token to name the images, omit the slash. For example <scene>/<camera> results in a folder named after the scene containing a sequence of images named camera.iff.

Figure 12.35. A preview of the image name appears at the top of the Common tab of the Render Settings window.

Here are some of the more common tokens:

Note that the capitalization of the token name does matter. If you had a scene named chocolateSauce that has a render layer named banana that uses a specular and diffuse pass with two cameras named shot1 and shot2, and you want to add the version label v05, the following tokens specified in the File Name Prefix field

<Scene>/<RenderLayer>/<Camera>/<RenderPass>/<RenderPass>_<Version>

will create a file structure that looks like this:

chocolateSauce/banana/shot1/specular/specular_v05.#.ext
chocolateSauce/banana/shot1/diffuse/diffuse_v05.#.ext
chocolateSauce/banana/shot2/specular/specular_v05.#.ext
chocolateSauce/banana/shot2/diffuse/diffuse_v05.#.ext

Use underscores or hyphens when combining tokens in the directory or image name. Avoid using periods.

Tokens for OpenEXR files

The OpenEXR format can create multiple additional channels within a single image. Each channel can contain an image created from a render pass. If a scene has one or more render passes and you choose the OpenEXR image format, you can use the Frame Buffer Naming field to specify the name of each pass. This feature is available only when OpenEXR is chosen as the file format and the scene has one or more render passes. You can use the automatic naming setting or enable the Custom option in the Frame Buffer Naming drop-down menu. You can then use the Custom Naming String field to choose the token you want to use.

For multiframe animations, you have a number of options for specifying the frame range and the syntax for the filenames in the sequence. These settings are found in the Common tab of the Render Settings window. To enable multiframe rendering, choose one of the presets from the Frame/Animation Ext menu. When rendering animation sequences the safest choice is usually the name.#.ext option. This names the images in the sequence by placing a dot between the image name and the image number and another dot between the image number and the file extension. The Frame Padding option allows you to specify a number of zeros between the image number and the extension. So a sequence named marshmallow using the Maya IFF format with a Frame Padding of 4 would look like marshmallow.0001.iff.

The Frame Range settings specify which frames in the animation will be rendered. The By Frame setting allows you to render each frame (using a setting of 1), skip frames (using a setting higher than 1), or render twice as many frames (using a setting of 0.5, which renders essentially at half speed).

It is possible to render backwards by specifying a higher frame number for the Start Frame value than the End Frame value and using a negative umber for By Frame. You would then want to use the Renumber Frames option so the frame numbers move upward incrementally.

The Renumber Frames option allows you to customize the labeling of the image sequence numbers.

The rendering cameras are specified in the Renderable Camera list. To add a camera, expand the Renderable Camera menu and choose Add Renderable Camera. To remove a rendering camera, click the trash can icon next to the renderable camera. As noted earlier in the chapter, you can use a layer override to include a specific camera with a render layer.

Each camera has the option of rendering alpha and Z depth channels. The Z depth channel stores information about the depth in the scene. This is included as an extra channel in the image (only a few formats such as Maya IFF support this extra channel). Not all compositing software supports Maya's Z depth channel. You may find it easier to create a camera depth pass using the custom passes (passes are described earlier in this chapter). The render depth pass can be imported into your compositing software and used with a filter to create depth of field effects.

Figure 12.38. You can add and remove renderable cameras using the Renderable Camera menu.

When Maya renders a scene, the data stored in the Frame Buffer is converted into the native IFF format and then translated to the file type specified in the Image Format menu. So if you specify the SGI format, for example, Maya translates the SGI image from the native IFF format.

Most popular compositing packages (such as Adobe After Effects and Autodesk Toxic) support the IFF format, so it's usually safe to render to this file format. The IFF format uses four 8-bit channels by default, which is usually adequate for most viewing purposes. If you need to change the file to a different bit depth or a different number of channels, you can choose one of the options from the Data Type menu in the Framebuffer section in the Quality tab. This is where you will also find the output options, such as Premultiply (see Figure 12.39).

Render passes uses the secondary Frame Buffer to store the image data. You can specify the bit depth of this secondary buffer in the Attribute Editor for each render pass.

A complete list of supported image formats is available in the Maya documentation. Note that Maya Software and mental ray may support different file formats.

Figure 12.39. Specify bit depth and other output options using the Framebuffer settings in the Quality tab of the Render Settings window.

When you are satisfied that your animation is ready to render, and all the settings have been specified in the Render Settings window, you're ready to start a batch render. To start a batch render, set the main Maya menu set to Rendering and choose Render

One of the most useful options is the Verbosity Level. This refers to the level of detail of the messages displayed in the Maya Output window as the render takes place. On the Mac, progress messages should be displayed in the Console Window (Window

Figure 12.40. Detailed progress messages for each frame can be displayed in the Output Window.

To start the render, click the Batch Render (or the Batch Render And Close) button. As the batch render takes place, you'll see the Script Editor update (Figure 12.41). For detailed information on the progress of each frame, you can monitor the progress in the Output Window.

To stop a batch render, choose Render

When the render is complete, you'll see a message in the Script Editor that says Rendering Completed. You can then use F Check to view the sequence (File

Figure 12.41. The Script Editor shows the progress of the batch render.

A batch render can be initiated using your operating system's command prompt or terminal window. This is known as a command-line render. A command-line render takes the form of a series of commands typed into the command prompt. These commands include information about the location of the Maya scene to be rendered, the location of the rendered image sequence, the rendering cameras, image size, frame range, and many other options similar to the settings found in the Render Settings window.

Command-line renders tend to be more stable than batch renders initiated from the Maya interface. This is because when the Maya application is closed, more of your computer's RAM is available for the render. You can start a command-line render regardless of whether Maya is running or not. In fact, to maximize system resources, it's best to close Maya when starting a command-line render. In this example you can keep Maya open.

In this exercise you'll see how you can start a batch render on both a PC and a Mac. You'll use the solarSystem_v01.ma scene, which is a very simple animation showing two planets orbiting a glowing sun.

This scene has a masterLayer render layer, which should not be rendered, and two additional layers. The solarSystem layer contains the sun and planets, which have been shaded. It uses the mental ray renderer. If you open the Render Settings window to the Passes tab, you'll see this scene uses two render passes: diffuse and incandescence. The second layer is named orbitPaths. It contains Paint Effects strokes that illustrate the orbit paths of the two planets (Figure 12.42).

In the Common tab of the Render Settings window, no filename prefix has been specified and a frame range is not set. Maya will use the default file structure when rendering the scene, and the frame range will be set in the options for the command line.

Figure 12.42. The solarSystem _v01.ma scene has been prepared for rendering.

Mac Command-Line Render

For a Mac, the Maya command-line render workflow is similar except that instead of the command prompt you use a special Terminal window that is included when you install Maya. This is an application called Maya Terminal.term and it's found in the Applications

Figure 12.45. The Maya Terminal window is installed with Maya in the Maya 2009 folder.

You need to navigate in the terminal to the scenes directory that contains the scene you want to render.

1. Copy the solarSystem_v01.ma scene from the DVD to the scenes directory of your current project on you computer's hard drive.

2. In the Finder, open the current project directory.

3. Start the Maya Terminal application, and type cd at the prompt.

4. In the Finder, drag the scenes folder from the current project on top of the Maya Terminal. This places the path to the scenes directory in the Terminal.

5. Press the Enter button. The Terminal window is now set to the project's scenes folder, which contains the solarSytem_v01.ma scene.

The commands for rendering on a Mac are the same as they are for Windows. You can continue starting with step 6 in the previous exercise.

It's possible to create a text file that can initiate a series of batch renders for a number of different scenes. This can be useful when you need a machine to perform several renders overnight or over a long weekend. This can save you the trouble of starting every batch render manually. This section describes how to create a batch script for Windows and Mac.

Windows Batch Render Script

1. Move the scenes you want to render into the renderScenes directory of the current project. Give them a name that distinguishes them from the original scenes in the scenes directory just to avoid confusion.

2. Create a new plain-text file using Notepad.

3. In the text file, type out the render commands exactly the same way you would initiate a batch render. Use a new line for each render. For example:

   render -r file -s 20 -e 120 -cam renderCam1 myScene.mb
   render -r file -s 121 -e 150 -cam renderCam2 myScene.mb
   render -r file -s 1-e 120 -cam renderCam1 myScene_part2.mb

4. Save the scene as a .bat file, and save it in the same directory as the scenes you wish to render, usually the renderScenes directory. The file can be named anything, but it should end in .bat. Make sure the format is plain text, for example, weekendRender.bat.

5. When you are ready to render, double-click on the batch script (for example, weekendRender.bat) (see Figure 12.46).

You'll probably want to close Maya to maximize system resources for the render. Maya will render each scene in the order it is listed in the batch file. Be very careful in naming the files and the image sequences so one render does not overwrite a previous render. For example, if you render overlapping frame sequences from the same file, use the -im flag in each batch render line to give the image sequences different names.

Figure 12.46. An example of a batch script file

At render time, all the geometry in the scene, regardless of whether it is NURBS, polygons, or subdivision surfaces, is converted to polygon triangles by the renderer. Tessellation refers to the number and placement of the triangles on the surface when the scene is rendered. Objects that have a low tessellation will look blocky compared to those with a high tessellation. However, lower-tessellation objects take less time to render than high-tessellation objects (Figure 12.47). Tessellation settings can be found in the shape nodes of surfaces. In Chapter 3, the settings for NURBS surface tessellation are discussed. The easiest way to set tessellation for NURBS surfaces is to use the Tessellation controls in the shape node of the surface. You can also set tessellation for multiple surfaces at the same time by opening the Attribute Spreadsheet (Window

You can also create an approximation node that can set the tessellation for various types of surfaces. To create an approximation node, select the surface and choose Window

Figure 12.47. In the top image, the background structure has a low tessellation setting. In the bottom image, the same geometry is rendered with a high tessellation setting.

The editor allows you to create approximation nodes for NURBS surfaces, displacements (when using a texture for geometry displacement), and subdivision surfaces.

To create a node, click the Create button. To assign the node to a surface, select the surface and select the node from the drop-down menu in the Approximation Editor; then click the Assign button. The Unassign button allows you to break the connection between the node and the surface. The Edit button allows you to edit the node's settings in the Attribute Editor, and the Delete button removes the node from the scene (see Figure 12.48).

You can assign a subdivision surface approximation node to a polygon object so the polygons are rendered as subdivision surfaces, giving them a smooth appearance similar to a smooth mesh or subdivision surface. In Figure 12.49 a polygon cube has been duplicated twice. The cube on the far left has a subdivision approximation node assigned to it. The center cube is a smooth mesh. (The cube is converted to a smooth mesh by pressing the 3 key. Smooth mesh polygon surfaces are covered in Chapter 4.) The cube on the far right has been converted to a subdivision surface (Modify

Figure 12.48. The Approximation Editor allows you to create and assign approximation nodes.

Figure 12.49. Three duplicate cubes are rendered as smooth surfaces using an approximation node, a smooth mesh, and a subdivision surface.

When editing the settings for the subdivision approximation node, the Parametric Method option is the simplest to use. You can use the N Subdivisions setting to set the smoothness of the render. Each time you increase the number of subdivisions, the polygons are multiplied by a factor of four. A setting of 3 means that each polygon face on the original object is divided 12 times.

The anti-aliasing settings in the Quality tab of the Render Settings window are used to control and reduce flickering and artifacts that can appear along the edges of 3D objects or within the details of high-contrast or detailed textures applied to surfaces. Flickering in high-contrast textures is not noticeable in still frames, only when a sequence of frames is rendered and played back. This is why it's often important to test short sequences of your animations when adjusting the quality.

To minimize flickering and artifacts, you can adjust the settings found in the Anti-Aliasing Quality section of the Render Settings window (see Figure 12.50).

Figure 12.50. You can control the anti-aliasing quality in the Anti-Aliasing Quality section of the Quality tab in the Render Settings window.

The two main controls are the Sample Levels and Anti-Aliasing Contrast. When adjusting the sampling level, you can choose among Fixed, Adaptive, and Custom sampling modes.

Sampling refers to how mental ray determines the color of a pixel in a rendered image of a 3D scene. A camera emits a number of rays into the scene (known as Primary Eye Rays) to determine the initial color value that will be created for the individual pixels in the final image. After sampling occurs, filters are applied to average the color values for each pixel in the final image. Secondary rays may be cast into the scene to determine raytracing effects such as reflections and refractions.

The type of Primary Eye Ray is determined by the Primary Renderer setting in the Features tab of the Render Settings window. Usually this is set to Scanline for most renders. Secondary rays are specified by the options in the Secondary Effects section of the Features tab (see Figure 12.51).

Figure 12.51. The settings in the Features tab determine the type of Primary Eye Ray cast by the rendering cameras to sample a 3D scene.

mental ray evaluates the render using tiles. A tile is a square section of what will become the rendered image. As mental ray renders the image, it stores the samples for each tile in the sampling Frame Buffer. Once the samples have been collected for a tile, filters are applied to average the colors. The filtered samples are translated into color values and then stored in the image buffer. A tile defines a region of samples that will become pixels in the final image.

The Min and Max Sample Levels determine the sampling range. mental ray uses the Min Sample Level as a starting point and then evaluates the scene and determines whether any parts of the scene require additional samples. If so, more samples are taken. The number of additional samples is determined by the Max Sample Level.

Sample levels increase by a factor of 4. A Sample Level of 0 means that each pixel in the image receives 1 sample. A sample level of 1 means 4 samples per pixel. A sample level of 2 means 16 samples per pixel and so on.

Negative values mean that each sample contains a number of pixels. A sample level of −2 means that one sample is used for an area of 16 pixels in the rendered image. Again, the number of pixels in negative sampling increases by a factor of 4. As you change the sampling, the number of samples that will be used is displayed in the Render Settings window under Number Of Samples (see Figure 12.52).

As the sampling rate is increased, the render time increases significantly. The Sampling Mode option can be used to increase the efficiency of the sampling process. When Sampling Mode is set to Fixed, mental ray uses the sampling level specified by Max Sample Level. This is often useful for rendering with Motion Blur. When Sampling Mode is set to Adaptive Sampling, mental ray uses one level below the Max Sampling Level as the minimum and one level above the Max Sampling Level as the maximum. In other words, when Max Sampling Level is set to 2, the minimum becomes 1 sample per pixel and the maximum becomes 16 samples per pixel. When Sampling Mode is set to Custom Sampling, you can enter your own values for Min Sample Level and Max Sample Level.

Figure 12.52. As you adjust the Min Sample and Max Sample Levels, the range of samples is displayed next to Number Of Samples.

So how does mental ray know when a pixel requires more sampling? Mental ray uses the Anti-Aliasing Contrast value to determine if a pixel requires more sampling. Lower Anti-Aliasing Contrast values increase the likelihood that addition sampling will be required.

The best approach to take when setting the sampling level is to adjust Anti-Aliasing Contrast gradually before increasing the Min and Max Sampling Levels. You'll notice that the Production Quality preset uses Adaptive Sampling with a Max Sampling Level of 2 and an Anti-Aliasing Contrast setting of 0.1. If you find this is not sufficient, try lowering the Anti-Aliasing Contrast setting. Try a setting of 0.05. If your image still has artifacts, you may then consider raising the Max Sample Level.

If you image has large blank areas and only small portions of detail, try setting a Min Sampling Level of 0 or −1 and a Max Sampling Level of 2. This way sampling is not wasted in areas where it is not needed.

Other options include adjusting the textures, applying filtering to the textures, or using the mental ray overrides on individual surface shape nodes to adjust the sampling for whichever surface is causing problems.

You can get a sense of how the sampling is being applied in a render by activating the Diagnose Samples option (found in the Raytrace/Scanline Quality section in the Quality tab of the Render Settings window) and performing a test render. The rendered image will show areas of higher sampling in white and lower sampled areas in black (Figure 12.53).

Figure 12.53. You can visualize areas of high sampling in a test render when Diagnose Samples is activated.

Filtering occurs after sampling as the image is translated in the Frame Buffer. There are a number of different filters you can apply, found in the menu in the Multi-Pixel Filtering section of the Render Settings Quality tab. Some filters blur the image while others sharpen the image. The Filter Size fields determine the height and width of the filter as it expands from the center of the sample across neighboring pixels. A setting of 1 × 1 covers a single pixel, and a setting of 2 × 2 covers four pixels. Most of the time the default setting for each filter type is the best one to use.

The Box filter applies the filter evenly across the image's height and width. Triangle and Gauss filters both add a small amount of blurring to the pixels, while Mitchel and Lanczos both sharpen the image.

The Jitter option reduces flickering or banding artifacts. Jittering offsets the sample location for a given sample block in a random fashion.

The Sample Lock option locks the sampling pattern. This reduces flickering by forcing mental ray to sample the scene the same way for each frame. It can be useful in animated sequences and when using Motion Blur. If Sample Lock does not reduce flickering, try Jitter as an alternative.

The Rasterizer is an alternative to the Scanline and Raytrace renderers that uses a different sampling algorithm. You can choose Rasterizer as your Primary Renderer in the Features tab of the Render Settings window. Rasterizer is most often used for scenes with motion blur or scenes with hair. It handles large amounts of overlapping thin surfaces (such as hair) quite well.

When Rasterizer is chosen as the Primary Renderer, the options for Raytrace/Scanline Quality sampling in the Quality tab become unavailable. Instead you can use the Visibility Samples and Shading Quality settings in the Rasterizer Quality section of the Quality tab (see Figure 12.54).

Visibility Samples works much like the Max Sample Level setting in the Raytrace/Scanline Quality options; however, the Rasterizer is not adaptive. The default setting of 0 means that one sample is used for anti-aliasing. Shading Quality sets the number of shading samples per pixel. The Rasterizer caches the samples with the geometry in the scene, so sampling actually travels with moving geometry. This is why it is an ideal algorithm for use with motion blur. Motion blur is discussed in Chapter 2.

Figure 12.54. When you choose Rasterizer as the Primary Renderer, the Rasterizer Quality options become available.

Raytrace Acceleration settings are found in the Raytrace settings in the Render Settings Quality tab. The purpose of these settings is to improve the speed of your renders when raytracing is enabled. Changing the settings does not affect how the image looks, only how fast it renders.

Raytracing is a process where rays are shot into a 3D scene from the rendering camera. When Raytracing is used as the Primary Renderer (this is set in the Features tab), each ray shot from the camera is compared with every triangle face of every piece of geometry in the scene to determine if additional rays are required for reflection or refraction (recall that all geometry is tessellated into polygon triangles when the scene is rendered).

To reduce the number of comparisons made between each ray and each triangle, mental ray offers several versions of the Binary Space Partition (BSP) algorithm (see Figure 12.55), which is meant to accelerate the process. The BSP algorithm divides the scene into sections; each section is a 3D container known as a voxel. The number of calculations required for each ray is limited to the number of triangles within a section of the scene as defined by the algorithm. In the Acceleration Method menu, you'll find Regular BSP, Large BSP, and BSP2. The Large BSP method is meant for use with large scenes; it does a better job of managing memory through disk swapping than Regular BSP. The BSP2 method does not allow you to set the Size and Depth options. BSP2 is also useful for large scenes and whenever a large number of instances are present.

The voxels created by the BSP method are adaptive. If needed, a voxel can be subdivided into smaller voxels. The lowest level of division is a voxel that contains only triangles; such voxels are known as leaf nodes.

You can set Bsp Size and Bsp Depth. Bsp Size refers to the maximum number of triangles a voxel can contain before it is subdivided. Bsp Depth sets a limit on the number of times a voxel can be subdivided. When tuning the performance of raytrace acceleration, you want to balance the size of the voxels against the voxel depth. A large voxel size means that each ray must make a large number of comparisons as it enters a voxel. A large depth value requires more memory to store the calculations for each voxel as well as more time for determining the placement and division of the voxels. Voxel depth has a larger impact on render time than voxel size.

You can use a separate BSP tree (a BSP tree is the hierarchy of voxels created in the scene) to calculate raytraced shadows. This is activated by turning on Separate Shadow Bsp.

Figure 12.55. mental ray offers three versions of the BSP Acceleration Method for improving the speed of raytrace calculations.

Tuning BSP settings often takes a fair amount of testing, trial, and error. You can create a visual representation of how BSP is calculated by activating the Diagnose Bsp setting. This allows you to see a color-coded version of the scene based on the BSP size or depth.

For the most part the scene appears in blue, indicating that the voxel size is more than adequate. You can try lowering this value, but for this particular scene a size of 10 is fine. If you prefer to have a more detailed report of the impact of the BSP Acceleration settings, set the Verbosity of the output messages to Progress Messages (you can do this in Render Current Frame

Figure 12.56. The Diagnose Bsp option color codes the render based on the voxel depth set in the Acceleration options.