As I really wanted to fix the white balance to D65, and still keep the best colors possible, I decided to try to achieve this through the Service Menu.
One should be very careful, of course, about entering the Service Menu. This can void your warranty and cause you to pay for any resulting service calls. The choice of navigation buttons on the remote control is not obvious, there are some menu choices that immediately invoke an action without further confirmation in the submenu, and there are no cancel buttons. I only went down this path after reading zerog6's contributions on service menu calibration in the A750 thread.
Before using the key sequence to enter the service menu, (Off, Menu, 1, 8, 2, On), or the sequence 0000 to unlock the Advanced submenu in the SM, you need to know that the left/right buttons on the remote choose new values, enter the corresponding submenu or initiate the action. Use only the up/down buttons to navigate in the menus; use Return to back your way out. Be very wary of any choice that says "Calibration", especially ADC Calibration. This can cause the set to reprogram its internal values using the presumed incoming test pattern, and is at best an awkward problem to solve. Also, be aware that entering the service menu causes the set to undo any user menu calibration you have set, even if you change nothing. So be sure you have written down any values you have set. And be sure to write down or photograph any service menu screens before you change them, so you can put them back.
My goal was to try to get the best calibration for movie viewing, as close as possible to what is achievable on the A550/A650 sets. On a first pass, I was able to get much closer, but it's not yet the best possible. In particular, I have not yet found any way to move the green primary closer to the Rec709 point.
Given what I had learned with the User Menu calibration, I decided that the right process was to follow three steps:
(1) Set the basic backlight/contrast/brightness/color controls.
(2) Adjust the white balance controls.
(3) Adjust the color points for best xyY values. This fixes the secondaries for any effects of changes in the white balance.
Iterate as necessary between (2) and (3), until it is as good as possible or exhaustion sets in.
(1) Having already calibrated once using the User Menu, I set the basic controls to the same values.
Based on my experience with the User Menu, I opted to set the Color level for an average error in the Y values of the colors. In particular, I set it a little low, as the Red Y was too high, even though set the initial Y value of green and blue a little low. As you will see below in (3), this technique which works on the A550/A650 was a mistake here, as the factory values for B550 color controls are set near their highest Y value. There is not much room to push the Y values higher. So it would be better to err a little higher on the Color setting, then use the color controls to reduce the Y values.
There is a wider choice in Gamma selections in the Service Menu, with the MGamma option. I looked at this, but returned it to Off, so I have not measured the effects of other choices, e.g., M4, M3, etc.
(2) The white balance controls in the Service Menu have different ranges than the User Menu. More importantly, there are only two colors here, red and blue. As green is the largest component and determines most of the overall white level, it seems to be controlled by the gamma setting. Here you can adjust only the red and blue to achieve the best balance relative to green. This is fine, as it matches some of the professional guidance available on this site, to leave the green alone. As is typical, it took three iterations to get a reasonably even grayscale before going on to the color controls.
(3) As I mentioned above, I tried using the primary hue controls to move the green and red color points. I did not measure any significant changes in the green primary. On the red primary, I was trying to tweak it a little closer to blue. After doing a full color calibration, I measured the intermediate saturation points and found that there was now a big curve toward magenta in the red line from 0 to 100% saturation. Apparently hue had a larger effect on the intermediate saturation points than the endpoint. So I put it back near its initial setting and remeasured to verify that the red points were back to nearly colinear.
The Sat controls on the primary colors adjusts primarily the Y values for each color. Given that the RGB primaries are close but not on the Rec709 targets, I used a spreadsheet I published in another thread
to calculate the optimal Y values to minimize the error for a given set of primaries.
For the secondaries, it was reasonably straightforward to use Hue to move these close to the Rec709 points, then use Sat to adjust the Y values. The only confusion is that the Hue controls are not a simple scale. 0x3F can yield a point far away from 0x40. I've not yet decided if they are doing signed arithmetic or sign-and-magnitude arithmetic mod 128, but using the continuous measure feature of HCFR it is easy to play near the limits of the values and do a binary search to approach the best point.
In hindsight, I wish I had set the Color control higher, as in some cases I maxed out the Sat control for a color trying to raise its Y value, without quite reaching the target.
As you can see from the attached CIE diagram, the secondaries are close, while now maintaining a D65 white. The RGBYCM dE values using CIE2000 are now 1.56, 2.60, 1.46, 1.79, 2.13, 0.47, so they are not terrible but not perfect. The grayscale, though, is the best I have ever achieved. PS3 grayscale curves are generally very flat, but the Service Menu grayscale controls do seem to give a finer control on the white balance than the User Menu controls. The end result is not quite as accurate as I might have obtained on an A550, but with another pass I'm sure I could reduce the errors on the secondary colors. The final colors look very accurate, and the B550 certainly gives a deeper black level, and better overall contrast than the A550.
After calibrating for the PS3 on the HDMI4 port, I thought I would need to copy these values to the other sources, in particular the COMP1 port for the cable box. When I changed sources in the Service Menu, the new values were already present. I did not have to enter them again. So it appears that the color calibration is global for the entire set, i.e., there is only one. And the white balance for each color tone in the service menu is also global.
Possible areas for future research, for the next person to go down this path:
(1) What is the real effect of Hue on the primaries? Is there some way to move Green, in particular?
(2) I did not measure the differences on the B550 between Auto and Native colors. I set Auto before I entered the Service Menu, and I assumed that my Service Menu measures match what I would have seen had I re-measured outside the Service Menu when I was done.
(3) As the color changes seem to be global, what is the impact on game (Standard/Normal or Dynamic/Cool) of changing the Movie Warm2 colors? Does enabling movie mode processing the service menu affect lag during games?
(4) What is the meaning of the other Gamma choices in the service menu? Do they change the overall shape of the gamma curve?
For reference, these were the control settings I entered in the Advanced Service Menu, with the original value in parentheses. One should assume that each set is different, and I know that different source devices are different, so one cannot predict that these same values would be appropriate for anyone else.
Movie Contrast 88 (95)
Movie Brightness 47 (45)
Movie Color 48 (50)
Movie Sharp 10 (20)
Movie Tint 50 (50)
Movie BkLight 4 (5)
MSubGamma -1 (0)
Red Gain 157 (134)
Blue Gain 48 (52)
Red Offset 131 (129)
Blue Offset 138 (139)
Red Sat 0x13 (0x17)
Red Hue 0x00 (0x40)
Green Sat 0x1E (0x1C)
Green Hue 0x40 (0x40)
Blue Sat 0x1B (0x18)
Blue Hue 0x42 (0x40)
Cyan Sat 0x1E (0x18)
Cyan Hue 0x08 (0x0f)
Magenta Sat 0x15 (0x16)
Magenta Hue 0x10 (0x40)
Yellow Sat 0x1E (0x13)
Yellow Hue 0x50 (0x40)
PS3_090625.chc.zip 11.3271484375k . file