Katana 3.5v5 docs:

RenderOutputDefine

Specifies output of an image (color, AOV, shadow map, or similar) to a file. In RIB, this means a Display statement.

Connection Type	Connection Name	Function
Input	input	The place in the node graph where you want to define the output settings for rendering.

Control (UI)	Default Value	Function
outputName	primary	Associates a name with the display. Typically primary by default; often shadow for shadow maps, and similar conventions. This name appears in the Render node, along with (or as) the default primary.
type	color	Specifies the type of output. • color - mostly used to render out rgb beauty files, but also can be used for rendering out z, P(point), N(normals), Ci(final shader color) passes. • raw - allows you to directly specify the values for a Display line. Since the output could be anything, Katana doesn't do any colorspace conversion on this output, and can't support tiling. • script - run a script on another RenderOutputDefine, like txmake. • prescript - run a script before the render is started. • none - clears the output. If the output was previously setup by a different RenderOutputDefine node, this removes the entry.
includedByDefault	Yes	When enabled, this Render Definition is sent to the Render node.
rendererSettings
colorSpace	linear	Sets the output colorspace used.
fileExtension	exr	Sets the output file format.
channel	rgba	Sets the channels to output. You can also set a user-defined channel from a PrmanOutputChannelDefine node.
When fileExtension: exr; convertSettings
exrCompression	Scanline ZIP	Defines the exr compression method to use. All methods are lossless (with the exception of Pixar24, which is lossless but quantizes the pixels to 24-bit float). Wavelet is generally preferable as it offers ~2:1 compression even on grainy data. • None - • RLE - • Scanline ZIP - • Block ZIP - • Wavelet - • Pixar 24 -
exrBitDepth	16	Sets the floating point precision of the rendered exr file: • 16 - half float. This is recommended for all color passes. • 32 - full float. This is recommended for all ncf data arbitrary output variables (AOVs).
exrOptimize	Yes	When enabled, the exr file is written out in an a manner optimized for efficient random tile-access. These optimizations greatly improve memory usage and performance for programs, which process images in tiles.
exrType	Tiled	Sets whether the exr file is written to support: • Tiled - random tile access. • Scanline - random scanline access.
When fileExtension: exr
clampOutput	No	When set to Yes, post-render clamp negative rgb values to 0, and clamp alpha values to 0-1. Note: clampOutput has no effect on NaN and inf values.
colorConvert	Yes	When set to Yes, post-render convert rendered image data from linear to the output colorspace specified in the filename. The default value of Yes is suitable for nearly every situation, since the linear output of the render is converted to the colorspace in the filename. A case where you would want to set this to No is if you know the data being rendered is in a colorspace other than linear, such as the re-projection of a log plate, and you want to name the output file log without a linear to log conversion.
When fileExtension: png; convertSettings
pngBitDepth	16	Sets the bit depth of the rendered file: • 8-bit • 16-bit
When fileExtention: rla; convertSettings
rlaBitDepth	16	Sets the bit depth of the rendered file: • 8-bit • 10-bit • 16-bit • 32-bit
When fileExtention: tif; convertSettings
tifBitDepth	16	The bit depth of the rendered file: • 8-bit • 16-bit • 32-bit
tifCompression	LZW	The tif compression method to use: • None - No compression method is used. • LZW - The LZW compression method is used. This is lossless, so it is usually preferable to use it unless there is an issue with compatibility in the target reader.
When fileExtension: tif
clampOutput	No	When set to Yes, post-render clamp negative rgb values to 0, and clamp alpha values to 0-1. Note: clampOutout has no effect on NaN and inf values.
colorConvert	Yes	When set to Yes, post-render convert rendered image data from linear to the output colorspace specified in the filename. The default value of Yes is suitable for nearly every situation, since the linear output of the render is converted to the colorspace in the filename. A case where you would want to set this to No is if you know the data being rendered is in a colorspace other than linear, such as the re-projection of a log plate, and you want to name the output file log without a linear to log conversion.
When fileExtension: jpg
jpgQuality	100	The quality to use when generating the jpg file. Higher values generate larger file sizes, with 100 representing the best quality image and 0 representing the lowest.
rendererSettings parameters continued
computeStats	None	Allows you to compute image statistics as a post process, appending as exr metadata. Select: • None • Raw • Depth Note: In depth mode, zero values and very large values are ignored. In both modes, only the region within the dataWindow is considered.
tempRenderLocation	N/A
cameraName	N/A	Describes the scene graph location of camera to render from. If empty, render from the camera specified in renderSettings.cameraName at /root. The cameraName parameter options are available by clicking the dropdown menu. For more information, refer to the Scene Graph Location Widget Type in Common Parameter Widgets.
locationType	local
When locationType: file; locationSettings
renderLocation		Specify the render location, or bring up the file browser or your studio's asset management browser to select the location to use. For more information, refer to the Asset and File Path Widget Types in the Common Parameter Widgets.