Issue with DCP white being 2 fl lower than white over HDMI, SDI and internal patches.

[Christian Wiberg]

We have a Barco DP4K-19B with a ICMP server.

We calibrated our projector after a lamp change. And we wanted, for the first time, to verify that we are getting the right light level and colours on all sources. First we measured internal patches and patches fed through HDMI and SDI. White was hitting 14 fl and primaries was correct on all of them. We then did a reading with patches from a peak white test DCP and then we were only hitting 12 fl, primaries was also darker.

Macros for DCP and HDMI/SDI are all the same except for input and .pcf (even tried the same .pcf but didn’t affect the fl)

Have read some posts on this forum regarding DCP white not matching internal patch white but no solutions. Does anyone have an idea what we can do to resolve this?

[Marco Giustini]

Are the colours read by your instrument the same throughout the patches? I suspect HDMI was D65 and the DCP is DCI. That is going to have different amount of RGB to get there and that will affect brightness indeed.
I suspect that darker primaries on DCI is because you get pure RGB while in HDMI mode multiple colours might get mixed together to get your desired primary coordinates. If you inspect a test pattern with pixel-size details, you might be able to spot that if your convergence is not spot on - that is, on a blue test pattern you might see pure blue and a hint of green just mixed together.

In the end you should trust the patches played back through the server: that is going through the whole chain and that is what you are showing to your customers.

Also make sure the meter does not change position between readings, particularly on a high gain screen. Use a tripod possibly.

[Steve Guttag]

Color correction will come off of overall luminance. You correct color by filtering off, not by adding to.

When measuring, for DCI, for example, the "white" should be reading x=.314, y=.351 (within the tolerance of the spec) or the "Y" (luminance) value is not accurate. When using projector test patterns, you need to use RGB 12-bit full white, "Color Verification" as your TCGD and the MCGD that you measured. These should yield the correct color for measurement.

When running a peak white pattern on your server, you would use your normal projector macro that should have a PCF with the DCDM_XYZ_314_351 (or words to that effect with the 185 or 239 or "Auto" in its name. Again, your color should read correctly, if everything is set correctly. You should verify that your source DCP is mastered correctly (the Dolby test patterns that they made for their 3D process have a good peak-white pattern). Even with all of that, you should anticipate that internal test patterns will always read higher. In fact, I'd say that internal patterns read about 10% higher than DCP patterns. This is server independent. So, I'd anticipate that if you are reading 14fL with a projector pattern, you'd get something closer 12.7 on a DCP pattern.

Truth be told, DCP patterns are the standard as they test the entire signal path for the projector. Projector patterns...the calibration ones start right on the formatter. The verification ones are on the ICP (or its modern equivalent).

When checking non DCP levels, the same rules apply. The exception is what TCGD you select for the other color spaces. However, a proper external generator can certainly be used and would be considered a higher quality measurement of the real signal path. But, again, the luminance reading would only be valid if the color's x/y coordinates are also correct (x=.3127/y=.3290). You also have different gammas at play between the standards too...which will not affect a projector generated pattern but could affect a source generated one (but full white should be full white).

[Leo Enticknap]

Like Steve says, use a test pattern generated by the server (i.e. a DCP) rather than one that originates further down the signal chain to take the reading.

The gotcha is the method of entering the values. To shoot the colors, you have to play the three color DCPs and peak white, write down the values from the photoradiometer, and then go to the color calibration section of the projector software or web UI to enter them. Most projectors will not let you play a DCP while entering the color coordinates; only internally generated test patterns. Once you've done this, it's a question of activating the new MCGD file, then playing the peak white DCP, and measuring again. If you're within DCI on the colors, then tweak the CLO file for the light level if necessary.

[Steve Guttag]

...or...use the Ti ICP/Enigma program and you can adjust your MCGD in real time AND see the results on your meter in real time (if you have a meter the perpetually updates, like the PCA-100).

[Marco Giustini]

Is shooting RGB DCPs a necessity? I never had issues getting perfect DCI colours using RGB from the projector. I would just make sure you use a white DCP for final brightness (and do remember that you should really do an average when checking brightness, particularly on high gain screens!)

[Dave Macaulay]

Personally I do the MCGD file using the projector color calibration procedure. Using the TI ICP/Enigma control program works but it's more of a chore. I check colour with a peak white DCP, this gives the same x-y result as color verification in color calibration when the "color verification" TCGD is picked.
LSC files have to be saved for each format and input if you want the image brightness to be accurate.
Alternate content will be using (probably) REC709 colour and thus the screen brightness changes - CLO measures light before the light engine. If you use your DCP flat LSC file for alternative content flat, the image brightness will be different.
But if the DCP LSC file is calibrated with a DCP peak white clip, your brightness with a peak white DCP should be pretty close to what you want. It is important to set the light level at your target using manual lamp power adjustment then saving the calibration there. CLO does not track super well, if you calibrate at 20 ft-L and then ask CLO for 16, it isn't going to go to exactly 16. If you use the white pattern... it is brighter than DCP peak white.
Colour calibration compensates for colour shifts due to the lamp, lens, light engine, port, and screen. You can only correct colour by losing primary intensity, ie a bluish image can only be fixed by turning blue down - you can't turn up green and red... so corrected white is always dimmer than uncorrected white.
And the test patterns use "test pattern" MCGD, not your calibrated MCGD. As far as I know only the white color verification in color calibration with the "color verification" TCGD is picked, or playing a peak white clip, will have corrected colour. And you are stuck in calibration where you can't get to lamp power setup that way. You could get a corrected white pattern in the TI program and then do lamp power in Communicator if you like doing things the hardest way... seriously, just use a peak white DCP.

[Christian Wiberg]

Hi. Thank you for all your replies!

The peak white DCP (hitting DCI P3 x.314 - y.351) is the same for patches through SDI/HDMI and internal patches. I have also imported the peak white DCP into DaVinci Resolve and Filmlight Baselight software and both show the right level out on SDI and HDMI when measured. So its only from our ICMP server the peak white DCP is a little lower on luminance.

We are reading white and primaries with a CR-250 (bit.ly/3W5yT75) connected to Lightillusion Colourspace (bit.ly/3UNpmjS)

When running a peak white pattern on your server, you would use your normal projector macro that should have a PCF with the DCDM_XYZ_314_351 (or words to that effect with the 185 or 239 or "Auto" in its name. Again, your color should read correctly, if everything is set correctly. You should verify that your source DCP is mastered correctly (the Dolby test patterns that they made for their 3D process have a good peak-white pattern). Even with all of that, you should anticipate that internal test patterns will always read higher. In fact, I'd say that internal patterns read about 10% higher than DCP patterns. This is server independent. So, I'd anticipate that if you are reading 14fL with a projector pattern, you'd get something closer 12.7 on a DCP pattern.

I get more or less around 12.5-12.7 fl yes. So maybe Steve Guttag are correct in this?



Or if I understand Dave Macaulay right, and I am showing a proper peak white DCP, the light levels should be very close.

And do I also understand correctly that I can’t use CLO to gain up to the luminance, from 12.5 to 14 in my case? I can only lower it with CLO?

[Leo Enticknap]

With a xenon bulb, especially an older one, increasing the current can also throw the color space off a bit. So sure, you can get your colors right, then tweak the CLO file to say "this light level = 12.27 ft-l" (for example), and then re-execute the macro that this CLO file is applied to. But when you then re-shoot the white, it will likely have shifted a little. If it's no longer within DCI, you'll need to reshoot.

[Christian Wiberg]

With a xenon bulb, especially an older one, increasing the current can also throw the color space off a bit. So sure, you can get your colors right, then tweak the CLO file to say "this light level = 12.27 ft-l" (for example), and then re-execute the macro that this CLO file is applied to. But when you then re-shoot the white, it will likely have shifted a little. If it's no longer within DCI, you'll need to reshoot.

If I understand it correct we can't increase the light level as we need to have this projector showing right luminance level and colours for both DCP and HDMI/SDI. And increasing brightness we will have to much luminance for HDMI and SDI I guess.

And in our cinema we are exactly hitting 14fl / 48 nits on the highest wattage for scope and lowest for flat.

[Steve Guttag]

I think Leo is over emphasizing the color changes with respect to the xenon lamp. In fact, it is the most stable, color wise, of the three light methods used for cinema. HMI lamps are the worst (you really need separate MCGD files on HMI projectors for each lamp level). My guess is that for laser it is going to vary by manufacturer and how they deal with their lasers dimming across the colors (some use three lasers with the primary colors, some use three lasers but two primary colors, some use two lasers with one primary color).

When I shoot colors, I tend to preset the level on white so I'm in the ballpark of 14fL before I begin.