Basic Guide to Color Calibration using a CMS (updated and enhanced) - Page 3

I agree ! It's easy to spot the the differences the CMS/CRS system made when displaying 75% colorbars (I used 75% windows when setting the colors in the AE1000's CMS), but not with other color intensity levels.

anbjornk

Did you measure to see if the Panny is changing all intensity levels or just the one sampled - or are you just agreeing it easier to see the changes at the brighter levels? The Sanyo changes it for all intensity levels - but only for a narrow hue range around the sampled color.

FWIW

This is what HomeTheater mag gets for the Sharp 12KmkII

Red Color Point: x=0.655, y=0.333 Green Color Point: x=0.325, y=0.633 Blue Color Point: x=0.147, y=0.054



The text points out the green is tinged yellow compared to the mkI - so make sure you are comparing to the proper version in the reviews to see if you are tracking. But theirs is not as far yellow green as some of you. FGMs before looks closest to this.

I just agreed with you I do not know what the AE1000 system do exactly. but I know it's not good enough. Are there testpatterns I can use to check these things?

Reason I'm asking is that if I use your methodology I may nail the 30%/90% combination (or similarly 30%/80%) with 0.313,0.329 but still find that the middle ranges (40%-70%) are showing 0.310, 0.321 or similarly. I figure that I need to alter the Blue cuts/gains to "lift" these up. I'm taking measurements pointing at the lens (emissive) at about 5 feet.

As you can tell, I'm "missing" some education here on the mechanical aspects of how the controls alter the grayscale. I think I'm trying to make the grayscale alignment be as smooth and close to D65 as possible (lowest dE) throughout the range. I have found that I run out of blue at the top-end of the Ruby. Guess that's the nature of the Xenon lamp.

What I've found, with my particular projector, is that to bring the "middle" in line with the suggested IRE points, I need to:

a)really reduce Blue bias while increasing R/G bias, and
b) reducing Red gains while doing some slight adjustments to G/B gains.

Any help in understanding these mechanisms would be appreciated. What's interesting is that the Ruby is trickier to work with than the Sanyo Z5 I've also calibrated.

Here is a comparison of the two methods for adjusting color/tint on my Benq PB6200 FP. First, I adjusted the color and tint using the blue filters and Getgray(see file: blue filter method.jpg). Next, I used the 75% method described at the beginning of this thread (see file: 75% method.jpg).

As you can see, there is quite a difference between results of each method. I am not sure which one to go with. The blue filter method result has smaller delta E values so I would guess that I should go with that one. However, with what I have read on this forum, the 75% method is more accurate. Would it be better to just adjust the tint in the blue filter method to bring cyan closer to the target? The third image (compromise.jpg) is a color and tint both set to default (0). Maybe I should just adjust the tint (bring cyan up to the target) on those settings - let me know what you think. - Thanks.

Here is what Home Theater Mag measured:

It may not be the decoder, I think he is using a magenta filter on the output.

dE is your measure of accuracy, but but do know what flavor of dE you are using (they are not all equal). If you are having to make tradeoffs between having the correct quantity of light vs. the correct quality of light, then that can be a hard choice as Tom indicates.

Bill

angryht

aligning the secondaries by coordinate may not work as well if you have not first done the greyscale. It is obvious when your cyan and magenta are too blue/purple and your whites are too blue/purple that this is a greyscale issue - not a tint issue.

This is indicated by the alignment that you got with the filters - it is "correct" for that incorrect greyscale - the secondary lines are nearly crossing at the white point. The cyan will not be on target even if you have aligned with filters simply because the white point is not on the D65 target.

Now if you have aligned the greyscale best you can - then trying to hit the secondary coordinates per the specs rather than alignment with filters or incorrect white points - is a reasonable compromise - if you can. But you will never accomplish that without color/tint per CMY - as the global tint can only rotate around the incorrect white point you will only get one secondary correct.

And even if you do get the white point at D65 - the secondary lines will not perfectly cross because the red/green primaries are off.

I am not using any filters on the output. I am taking readings off of the screen.

Attached is the grayscale achieved prior to doing any adjustment of the color and tint in the user menu. When I was done with the adjustments I did a quick check of the grayscale using the 10 points and it was consistent with the values prior to the adjustments. In other words I think my grayscale is pretty good.

I think I will give this a shot.


Maybe one or two clicks to split the difference?

Thanks again for the responses.

Agreed. A slightly different take on Tom's procedure that may work on some people's units:

1) Adjust the Red, Green and Blue Primaries in the CMS. The secondaries should move as the primaries are adjusted, but don't worry about their accuracy just yet.

2) Calibrate the grayscale. As Kraz mentions, the secondaries will move around as the balance between the primaries is adjusted to hit white.

3) Adjust the location of the secondaries in the CMS. The secondaries ought to be fairly spot-on with most "normal" color systems after #2, but some may not be. This will especially be true if there is a dedicated "scondary primary", like the DLP color wheels that have Yellow filters in them (I'm not sure of any CMS-enabled displays that have odd color wheels, but I'm sure they are coming; this is true for using an external box with color controls, though).

4) Adjust the color decoder (color/tint). This ought to be changed internally as the CMS is adjusted, but it may not be.

You may want to swap #3 and #4 in my procedure above, depending upon how your display is designed/responds. Also, always re-check the grayscale after #3 and #4. A well-designed unit should not have the grayscale altered by changing secondaries, but reality trumps theory so you have to check.

Bill

If your grayscale was correct - then you have one of those displays with the color decoder assuming a 9300K greyscale - something that has happened before (an Optoma that had one scan rate but not the others do this). If that is the case then try adjusting the 9300K preset to be D65 if you can.

I personally find incorrect yellows most offensive so that is the compromise I would lean to.

Krasmuzik: I think you found the issue!!! Brilliant!!! I am pretty sure have a color temperature that is set to 9300K, at least that is what I think, based on the service manual and some independant testing. The color temperature in the user menu has values from -50 to +50. It appears that the -50 is the default for 6500K, the 0 is for 9300K and the +50 is for the 11500K. When I did the gray scale adjustment before, I had set the projector to 0 to adjust grayscale not knowing that the color decoder may be assuming a color temp based on the user menu setting.


I will set the color temp to -50, adjust grayscale and then try the color/tint adjustments.

I'll let you know how it turns out.

Thanks again, I really appreciate your thoughts.

I spent several hours re-calibrating my Epson TW700 projector with this method. I have some questions:

The Epson Tw700 projector only allows you to finely adjust using either RGB adjustments (grayscale) OR RGBCMY adjustments (colourspace) and not both at the same time. (i.e. when adjusting grayscale using RGB, the TW700 locks you into a predefined colourspace that within the CIE is accurate for red, blue and yellow, but off for green, cyan and magenta. Furthermore, this predefined colourspace when red is 21% of white, only shows blue as 4% of white and green as 95% of white).

Thus, which is better:
To have a well calibrated colourspace and poorer grayscale? (using other adjustments in the menu for setting better grayscale while adjusting colourspace, I can't get dE below 4 from 40-100 and below 10 for 0-40)?

Or is it better to have a well calibrated grayscale (dE below 3 for 20-100) but having an inaccurate colourspace?

Sorry for the convoluted question, but I really must know. Overall I think grayscale hits more accuracy points between grayscale and colourspace.

It is blue that is off the most. Red and Green actually track pretty close together to ideal. Blue is 8-15% off in general.
Here are my dE values by IRE:
20=17
30=12
40=11
50=8
60=7
70=7
80=6
90=6
100=3
Overall, the colour temperature histogram shows of course a cooler temp.

The above calibration was accomplished by setting Colour temp to 6000K, and adjusting a setting called Skin Tone on my projector, whereas the RGB calibration with proper grayscale tracking allowed me to use a Colour temp value of 6500K.

I'm just used to have a dE below 3 across the board when using the RGB controls.
Just to clarify, when grayscale is calibrated with RGB, and thus tracks very well, the CIE chart in turn shows Red, Blue, and Yellow are right on the reference points for 709. Green is the most off (lying above the reference), and cyan and magenta are slightly off.
BUT, when I try to make red, green and blue balance as percentages of white (as per your calibration tips), that's where I see them the most off from the 21%, 70% and 8% they should be at (as previously mentioned).

How can the CIE look pretty good for blue, but be so off on the percentages when adjusting colour overall?

Spent several hours last night calibrating TB1 mode for both my HTPC (DVI/HDMI cable feeding 720p) and my Xbox 360 (component cable feeding 720p, but 480p for the GetGray as I don't have HD test patterns and that's how the 360 works).
Also worth mentioning is I use a 106" Carada BW screen, and the projector is 15' away.

I used the Pantone Eye-One LT probe with HCFR software & GetGray patterns. I used the 709 colour reference setting as most of what I watch is HD material, but this does cause a problem (see section "What's left to do").

I used the RGBCMY calibration, rather than the RGB calibration. I had previously been running with a calibrated RGB only for both components, but as mentioned in the Cine4Home review, the colours were oversaturated, so I decided to go further with the RGBCMY calibration.

Calibration Method:
1. Gamma: Multiple Grayscale readings to adjust gamma curve (achieved Gamma of 2.26 for HTPC & 2.23 for 360).
2. Overall Colour: Read 75% white window and adjusted overall colour setting until Red showed as 21% of the white reading.
3. Grayscale: Adjusted Colour Temp and Skin Tone settings while doing continuous readings and looking at grayscale until dE was the lowest from 40-100 IRE.
4. Colourspace: Adjusted each RGBCMY Hue and Saturation setting while taking continuous readings and looking at the CIE chart. Dial-in each colour to the best reference triangle areas. This is tricky, for as each colour was dialed in I also looked at the continuous data readings and readjusted Saturation onthe currently calibrated colour so that I achieved the best compromise for each colour's % of reference white at 75%, which is the following:

Red = 21% of reference white
Green = 70% of reference white
Blue = 8% of reference white

It's important to note that I never compromised a good placement on the CIE chart for more accurate colour balance on the data readings, since for green these two were in conflict.
5. Final adjustments: Performed a complete grayscale, primary, and secondary set of measurements and made minor adjustments where needed.

Results:
The result is a colour space pretty much dead on D65 and a grayscale with a dE of about 4 from 40-100 IRE, so I'm ecstatic about the results. I'll post my chc files later to this post.
However, I really do have to get used to D65. The colours are severely unsaturated compared to what I have been watching up until now.
The other thing I now notice much more with the properly calibrated projector is a substantial increase of noise/grain in the picture. This must be a result of the largely reduced saturation in colours (my colour is at -5, with most individual RGBCMY colour saturations averaging about -90). Any explanation here would be appreciated.

Projector Settings:
I'll post my complete settings later by editing the post, but here is what I can remember off the top of my head:

HTPC:
White Level = 13
Black Level = -5
Colour = -5
Tint = 0
Colour Temp = 5500K
Skin Tone = 3
DVI Level = Expanded

Xbox 360:
White Level = -4
Black Level = -3
Colour = -5
Tint = 0
Colour Temp = 6000K
Skin Tone = 3
Setup Level = 0%

Eye-One vs Spyder2Express:
The Eye-One is more accurate through the IRE, and especially in the 20-40 range.
For me it's well worth the extra expense, mainly because I'm obsessed with properly calibrated display devices
It also takes much faster readings than the Spyder for the 40-100 IRE range, and slightly slower for the 0 to 40 IREs when you choose "Average many readings for dark" setting in HCFR.

Whats left to do:
1. Enjoy the beautiful picture of course, but also I need to find some good settings from these 2 component calibrations for my Sonicview 8000HD FTA receiver, which of course I cannot play test patterns on. Best I can do is record the HDNET patterns and adjust white, black and colour and tint.
2. Rec 709 vs 609: On the 360, the GetGray patterns are displayed on in 480p rather than 720p, as the 360 sends it this way. Since I don't have HD patterns like DVE HD (just came out), I cannot calibrate properly for 709, even though that's the setting I used in HCFR even though the 360 outputs the patterns in 480p. I think I will create two settings, one calibrated at 709 and 609 to see if it matters for HDDVD material. Any advice on this would be appreciated.

Thanks. That helps a lot.

I have tried to calibrate the Optoma HD7100 pj which has the following adjustments:
-In the user menu: contrast, brightness, sharpness, colour, tint, gamma, color temperature in 500 steps plus x,y adjustment, and RGBYCM gains and offsets.
-In the SM: Color wheel adjustment, lightness, colour and tint for RGBYCM, VGA auto calibration, MST 9888 component YPbPr, AD 9888 VGA RGB and SAA 7118.
Since I am using the DVI input, colour, tint, VGA autocal, MST 98088, AD 9888 and SAA 7118 are grayed out and NOT available. Therefore, I have used the RGBYCM lightness, colour and tint adjustments instead of overall colour and tint to do the calibration described by Tom in this thread.
Attached is the before file, done with a DTP-94 colorimeter reading off a white eggshell screen placed about 2-3 feet from the lens surface. The DTP-94 was just above the pjs lens looking towards the screen. Doing the reading in this way I have improved the low end range/accuracy substantially as compared to reading from the screen proper at 10 ft from the projector. The patterns are those of the HCFR dvd disc which is in PAL and the pj plays it in PAL if in auto mode. Is this OK for calibration purposes? Should I have the projector to output a NTSC signal from the PAL dvd input signal? Or should I have the dvd player to output a NTSC signal to the projector?

Although the signal from the Bravo D1 dvd player has been upscaled to 720p I am using the rec. 601 colour space for the measurements by HCFR because I am unsure as to what the dvd player and the pj actually do with that signal; any comments on this? Is there a way to know if the dvd player/pj are using rec 601 vs. rec 709?

I have used the DVI input, set the gamma at 2.2, colour temperature at 6500K and sharpness to softest setting. All other settings are default.
I first tried to adjust the gamut as best as possible, then I tried to adjust the lightness value of the RGBYCM as a percentage of white lightness. Then I refined the intersection at the D65 point by slightly adjusting the Y tint and M tint and then run and adjust the gray scale until it looks OK. I had to do a couple of small adjustments to the gamut and the D65 point position which had been affected by the gains and offsets adjustments for gray scale. Then I had to re-adjust again the lightness values as percentage of white lightness until they are all within +/- 2% of the respective theoretical value. Attached is the after file.
I feel that I could improve a bit the dE of B in the gamut and improve the gray tracking from 20-80 IRE with a little tweaking of the gains and offsets.
However, I have not touched the contrast and brightness. From past experience, I know that to achieve the blackest blacks and the whitest whites I could use something like B=-13 and C=+9 but this will make a bell shaped gamma log curve and add colour to the grays at 80 IRE or so.
Any comments/observations welcome.
I would also like to report that when adjusting the lightness of the primary colours, I was expecting HCFR to report an equivalent change to the adjacent secondary colour lightness but that did not happened. It is like the concept of the lightness of say yellow = Glightness+Rlightness did not apply. In other words, if I increased the value of green lightness the values of yellow and cyan were not affected (or very little affected).
Can anybody explain??
Thanks for looking.
Fermin

 

start.zip 1.0615234375k . file

 

after.zip 1.1806640625k . file

Brightness?

Oh, sorry. Mine is about 11.2