I have compared the performance of the different color spaces available in HDMI using the Spears and Munsil Zone Plate and Multiburst tests : RGB video levels, YCbCr 4:2:2 and YCbCr 4:4:4.
The Oppo BDP-83, firmware 905, is connected in an Eizo HD2442W monitor through HDMI (1)
The differences are small. I don't know if this could have any visible effect in a movie. So it's mostly an academic exercise.
YCbCr 4:2:2 performs best, followed by RGB, followed by YCbCr 4:4:4, which darkens the most the edges of the pattern.
I've taken some pictures of the screen (2). The pictures are intentionnally blurred from the lens in order to avoid extra moire and bandwidth issues in the camera.
On the left, the chroma zone plate test. On the right, the luma zone plate test, for reference.
The multiburst test gives more information :
The fact that the last square of both sequences of Multiburst luma tests are brighter than the others can be explained by the fact that they are the only patches that feature only black and white levels. All other patches have grey pixels in order to make the transition from white to black. These grey pixels must have been interpolated from black and white with a lineal scale intead of a 2.2 gamma scale. That's why they are too dark. That's a problem with the test conception.
We can see that the horizontal chroma multiburst test is darkened, but not the vertical one, in YCbCr 4:2:2 color space.
In YCbCr 4:4:4, they are both darkened.
In RGB, they are both moderately darkened.
In RGB and YCbCr 4:4:4, the Oppo performs the chroma interpolation. But in YCbCr 4:2:2, the Oppo only performs vertical choma interpolation (result visible on the horizontal pattern), while the Eizo monitor performs the horizontal interpolation (visible on the vertical pattern).
Thus we can say that the Oppo darkens the high frequency chroma pattern while interpolating, while the Eizo doesn't.
Since darkening is a matter of bandwidth, we should expect more moire and un-smooth edges in YCbCr 4:2:2, because there is a trade-off between aliasing and bandwidth. The deal is with horizontal bandwidth, therefore the left and right side of the zone plate chroma test should look bad.
Here are close-ups of the left bottom side of the pattern, in YCbCr 4:4:4, YCbCr 4:2:2, and RGB respectively (expand the pictures to see the full resolution) :http://3141592.pio2001.online.fr/pic...cs_zpz_444.jpghttp://3141592.pio2001.online.fr/pic...cs_zpz_422.jpghttp://3141592.pio2001.online.fr/pic...cs_zpz_rgb.jpg
On the top left corner, moire seems to be the same. And I fail to see a difference in smoothness on the right hand side.
Therefore it seems that the chroma interpolation could be improved in the Oppo, and that meanwhile, YCbCr 4:2:2 is the best setting for this screen.
Note that no setting presented any problem big enough for the test to be considered as "failed", as shown in the Spears and Munsil BD manual.
So, as I said in the beginning, this seems to be a very minor problem, and it is unlikely that any movie could present a visible difference.
(1)Primary output = HDMI
TV aspect ratio : 16/9 wide/auto (the Eizo maps the 1080 signal pixel-wise, overscan test OK, black bars on top and bottom because it's a 16/10 1920x1200 monitor).
TV system = multisystem
Output resolution = 1080p
1080p24 output = auto
Deinterlacing mode = auto (same result with 2:2 odd and film bias for YCbCr 4:4:4)
CUE correction = auto (same result with off for YCbCr 4:4:4)
HDMI deep color = off
(2)Taken with a 90mm fixed focal lens in order to avoid edge derkening in the lens, or vision angles too wide from the panel, fixed manual exposure and ISO (different for the overall pictures and for the three macro zooms), bright regulator Off for the HDMI input on the Eizo.