Originally Posted by
bob pariseau
Definitive ycbcr 4:4:4 vs. Ycbcr 4:2:2 vs. Rgb data format sizing info from anthem!
the following information has been confirmed with anthem for the avm-50 and statement d2:
Hdmi ycbcr 4:4:4 input or output is always 8 bits per component (24 bits per pixel).
Hdmi rgb (studio or extended) input or output is always 8 bits per component (24 bits per pixel).
Hdmi ycbcr 4:2:2 input or output can be either 8 bits per component (16 bits per pixel) or 12 bits per component (24 bits per pixel) -- according to the max allowed by the device at the other end of the cable -- but with the horizontal color resolution halved. Like this: Ycb, ycr, ycb, ycr, etc.
Note: Hdmi ycbcr 4:2:2 at 10 bits per component (20 bits per pixel) is not supported. Nor is "deep color" hdmi ycbcr 4:4:4 (i.e., more than 8 bits per component / 24 bits per pixel).
Any of the above input hdmi ycbcr streams is accepted by the gennum vxp video processor chip at the full 8 or 12 bits per component (as described above) but is immediately converted to 10 bits per component rgb (30 bits per pixel) - using high precision color conversion matrices - prior to other internal processing inside the vxp.
"high precision color conversion matrices", here, means that the constants and math used in the conversion are actually higher resolution than the data stream (in or out) so that the color conversion process itself introduces no quality degradation.
An input hdmi rgb stream is accepted by the vxp at 8 bits per component and is extended to 10 bits per component (30 bits per pixel) prior to its internal processing.
The vxp does all the rest of its internal processing (de-interlacing, scaling, gamma correction, etc.) using 10 bits per component (30 bits per pixel) rgb.
If 12-bit ycbcr 4:2:2 is to be used for output, the vxp, at the end of its processing, converts its results to 12 bit per component (24 bits per pixel) using high precision color conversion matrices.
If 8-bit ycbcr 4:2:2 is to be used for output, the vxp, at the end of its processing, converts its results to 8 bit per component (16 bits per pixel) using high precision color conversion matrices.
If ycbcr 4:4:4 output is to be used the vxp converts its results to 8 bits per component (24 bits per pixel) -- again using high precision color conversion matrices.
Finally, if rgb output is to be used the vxp rounds its results to 8 bits per component (24 bits per pixel).
---------------------------------------------------
again, see the links collected in the first post of this thread (technology and terminology / data format section) for a more detailed explanation of the differences between these data formats and why you might want to experiment with 12 bit ycbcr 4:2:2 data format when using source or display devices that allow it.
--bob