Originally Posted by ajc9988
So, first, it depends on which software suite you are using for calibration. I primarily use HCFR, but if you use chromapure, spectracal, or light illusion, there are going to be some differences.
Next, are you talking about HDR10 generally and the PQ EOTF thereof, or Dolby Vision. As was warned, although DV and HDR10 use the same PQ EOTF curve, different display manufacturers may implement it in such a way that you cannot easily do DV the same way of doing HDR10. My Vizio P50-C1 is an example of that. That is the reason I'll eventually be purchasing the Spectracal Home product, but since my next sets will be either Vizio or TCL with current offerings and pricing, I'm holding out until they put out an autocal version for one of those manufacturers (the TCL X10, for example, uses a 17x17x17 3D LUT autocal feature, but isn't released until around September).
So the following will describe my HDR-10 workflow, not DV workflow because I still have not discovered a good way to do it with the freeware software available (if possible, support your developers for software, patterns like R. Masciola's, R.Pi generator, or Ted's Lightspace if going with those options, etc.), but this may work for DV depending how your TV mfrs implemented DV.
1) Set the Backlight.
To do this, you first need to do a reading of your black and 100% white to see what you are working with. If you are under a peak white of 465cd/m^2 with a roll-off implemented, then you should calculate what your diffuse white value should be, which roughly translates to where the 50% gray pattern should be set for nits. For HCFR, this adjusts the luminance curve. You may also want to just run the entire gray scale to figure out whether the manufacturer implements a BT.2390 roll-off at the near white. That is important to understand if it clips and where, which can effect the diffuse white value, which really only comes in if a roll-off is employed.
After you know if the roll-off is implemented and have the diffuse white value, which roughly sets the 50% gray value, you then put up the 50% gray pattern and adjust the backlight until the Y value (xyY calibration) matches the calculated 50% value (on many sets with lower nit values, raising the backlight value will not change the peak brightness. If yours does not change that value, it will move the curve inside of that by changing backlight, so lining up the 50% gray on brightness should help get it close to where it needs to be. This can vary by set, so try to verify how your manufacturer has implemented their controls on the set.).
2) use the patterns to set your brightness and contrast values. Then use the flashing boxes patterns, if your TV employs a color only mode (like blue only mode or red only mode, etc.), to set the color value.
Sometimes you cannot set these perfectly with HDR. Just try your best with what you know on SDR. Many choose Blue over red or green only mode for numerous reasons which will not be discussed here. You could even check the values and select the color with the highest value on color. This is setting chroma, which is related to saturation, but distinct from it. Tom Huffman discussed this a bit in his guide in this forum.
3) Set the grayscale.
First, start with the offset and gain values to try to lower the dE as low as possible. I personally use dICtCp error value, but many use the dECIE76 for gray scale and dECIE2000 values for colors. That really is a preference, but Calman is mentioning that dITP/dICtCp 720 is what they are going with on HDR. I prefer the look I achieve with that even on SDR. But pick your poison, so to speak.
I like to run gray scale sweeps rather than just doing 30/80 or 30/100 for this, then monitoring the RGB delta without gamma on the entire sweep. This allows me to analyze how the offset or gain is effecting the entire scale, allowing to get the RGB dE as low as possible before moving on. Many recommend trying to minimize any increase on the offset values as it may change the black value, including raising the number of nits there. Be aware of the limits of your colorimeter on reading near black levels as well to know where it may be less accurate. For gain, playing with the green value can change the contrast on some displays, so be careful there as well, as you can often just change the red and blue values to balance the gain without running that risk.
Once you have the offset and gain set, then you can move on to adjusting the gray scale with the 10/11/20 point adjustments by including gamma in the calculation. Some manufacturers set the adjustments within the scale that they work with, so you may have to figure out what the percentage on the TV adjustment pairs up with which pattern you are displaying. You might have the 50% TV value match with the 55% gray pattern, etc. So figuring out which value matches which pattern is key (or if it spans two patterns, trying to balance the setting versus the dE value on the two patterns). This can be the most infuriating and time consuming at first. Once you know the set, you can jump in relatively quickly.
4) Set the CMS values for the color, often HSV/HSB values.
Depending on how much gamut coverage you have, this will vary. The wider the supported gamut, the easier this is to set, obviously. This is why, for the most part, you can just use the 50%amp/50%sat patterns, which are around 100 cd/m^2, to set roughly where the color would be similarly to how you set up SDR using 100%amp/100%sat, except you won't have the steps in between if you use sat sweeps to set it. You will want to get the 50% patterns and 25% patterns more on point than worrying about the higher values as this is where the accuracy will matter a bit more, but if your set can support the full values for DCI P3 gamut, up to 90% REC 2020 on top end flagships potentially on some sets, then you will want to try to get it accurately set on the sat sweeps across all percentages as much as possible (good practice anyways). As R. Masciola pointed out in one of his manuals, I think the DV manual, getting the xy coordinates correct is more important than matching the Y luminance value, so if you have to give somewhere, give on that.
You may find you have to make some tradeoffs. It happens.
5) Go back and verify the earlier settings are still good. You may need to change some of the 10/11/20 point adjustments for gamma after making changes to the color CMS. If so, then double check the CMS doesn't need some extra tweaks after that if you changed anything.
So, it is roughly similar to that of SDR calibration for HDR, to a degree. More things to watch for, but not so many that it should be insurmountable. DV does not always follow the above. It should be close, but not always and will vary by implementation by the manufacturer, as well as limitations of your TV set. You also need to watch to make sure the HDR you are setting has activated with the pattern being displayed before taking the measurement, which means there may be a delay between selecting the pattern and when the set kicks in for HDR to then measure, which can vary by set.
The brighter your set's top value is (meaning nits or cd/m^2), the easier it should be for getting the values set, same with wider supported gamut values.
This isn't a perfect explanation, but should give you a rough idea of calibrating HDR. HLG is much simpler, as it follows more closely SDR calibration. For my Vizio P50-C1, it seemed the CMS values were shared between HDR-10 and HLG, so I stuck with my settings for HDR-10 calibration, then just worked within that to set the more global settings of backlight, brightness, contrast, color, tint, sharpness to get it right, which the grayscale tracked really well in regards to RGB balance by using the HDR-10 settings, and did decent on the gamma value, so no real complaints. That can vary by manufacturer though and may be separate CMS controls for each depending on your set.
Hope this helps you guys somewhat.
Edit: and if anyone with more experience wants to jump in and correct anything I've described above, please do so. This is speaking from my current understanding of HDR calibration, which may not be representative of the industry at large or may be inaccurate compared to someone who calibrates for a living or is more engrossed in the content of color theory than I am.
Thanks for a great post. Still trying to get my head around it but certainly helps (not withstanding the arguments that followed