Originally Posted by Manni01
So I've done lots of tests today and here are my results:
1) The reference color profile on our models is the same as DCI-P3-F, so there is no need to use one slot with DCI-P3F.
2) Instead of calculating the new BT2020 targets from the rec-709 offsets as per Chad's procedure, it is possible to significantly improve the results in BT2020 by calculating specific offsets for the BT2020 profile using P3 targets (we can't use BT-2020 targets because our PJs can't reach BT2020, they reach 65-70% of BT2020 at best).
So here is the procedure I followed:
- I updated Chad's spreadsheet with DCI-P3 targets instead of Rec-709 target. This means that you can calculate offsets with the filter on and at the targets that matter most for current content, i.e. P3.
- I updated Chad's workflow so that it defaults to DCI-P3 / D65 / Power Gamma 2.2
Once you have followed Chad's procedure (or my slightly modified version) to calculate the rec-709 offsets and created the rec-709 custom profile to correct the meter errors at rec-709, load my DCI-P3 workflow, and use the gamut luminance layout to create the offsets for DCI-P3. To do this, do the following:
- In your HDR user preset, select Reference, 6500K, gamma normal on the projector in high lamp, iris fully open. This is important as this will give you the widest possible P3 gamut. Low lamp reaches a narrower gamut, so even if you use low lamp to calculate the offsets select high lamp, iris open. This will make sure that the errors are mostly meter errors and not too much native gamut limitations.
- Run a gamma 33 points+color autocal
- Open the DCI-P3 workflow and measure the current user mode (still on reference, 6500K, gamma normal, high lamp, iris fully open). This will give you the reference reading at various levels of luminance. Copy the RGBW xy just as you did for rec-709, and copy them in my updated DCI-P3 spreadsheet.
- Once all the RGBW data is copied, the speadsheet will give you your new targets for DCI-P3 and BT2020.
- If you want, create a custom profile for DCI-P3. There is little to no consumer content using this, so it's not necessary to do so. However, it is interesting to create one as you can then import and select the new DCI-P3 profile and compare the readings with the reference (non-custom-corrected) one. You should see significant improvements.
- Create a custom BT2020 profile using the new BT2020 targets. This is the profile you'll use for UHD Bluray HDR etc. It's the same as the one you created following Chad's procedure, except that it should give you improved results in BT2020.
- Open a BT2020 / D65 workflow, select the new BT2020 profile in the PJ, select P3 Sweeps (P3 within BT2020) with a normal gamma and check the results. You should notice a significant improvement.
- Select your custom gamma, set 100% white to D65 and measure your actual BT2020 calibration using the multiplier used in your selected Arve curve. Gamut and gamma/greyscale should be fairly close. I attach my final results.
- If you need to use a different setting than HDR High lamp / iris full open, run an autocal using reference/6500k/normal, then select the custom BT2020 profile, custom gamma and set WB to D65 at 100%. There is no need to recalculate a profile in low lamp, it would most likely lead to gamut shrinking.
I did more tests with my Spyder 5 / Spyder 4 and while the Spyder 5 is more accurate facing the screen, it's less accurate facing the projector, which leads to more (unnecessary) correction and a significant loss of brightness (118nits vs 132nits in low lamp HDR, 170nits vs 186nits in high lamp HDR, once fully calibrated to D65). So my last attempt at replacing my Spyder4 by a Spyder5 is over. The Spyder5 is going back as I would never use it given the brightness drop it causes.
Hope this helps. Not much time to follow up on this, but I'm sure those who are fluent in Autocal will pick up on this and will be able to help.
Following this new procedure, I get a color volume of around 98% of P3 in both low lamp and high lamp, so I might go on using low lamp unless I really need the extra pop. The fan noise and heat in high lamp is unpleasant, especially in summer!
Let us know if this improves your BT2020 results or not!
- Thanks for this great write up. I thought I mastered the autocal process a year ago and hadn't looked back before today. Now that I have, I see there are optimizations available, so it is back to the learning curve for me.
I didn't realize CalMan 2017 added support for the Spyder 5 - that's really convenient. I haven't used CalMAN in a long time. Hopefully my license is still valid. My old laptop with it installed no longer works so I think I'll need to contact them to free up that license so I can reuse it.
Some questions on this and related items:
1. In the post above, you say that it "significantly improves" the results. Can you elaborate on the specific improvements, in both subjective and non-subjective terms, if known? I ask because I'd like to know if this is worth the learning curve in terms of what types of improvements I can expect to come from this.
2. Is there any possible downside(s) / tradeoffs whatsoever to using this updated approach?
3. Regarding this: "Select your custom gamma, set 100% white to D65 and measure your actual BT2020 calibration using the multiplier used in your selected Arve curve. "
3a. When you say set 100% white to D65, you mean using the RGB gains? I don't recall making this tweak at the end of any autocals. So this may be a new step, or I may be misunderstanding, or I've made this tweak normally and don't remember doing so.
I thought the point of this was NOT making tweaks to RGB gains, but rather having the custom color profile adjust for this...
3b. When you say "and measure your actual BT2020 calibration the multiplier used in your selected Arve curve" - I don't recall what multiplier I used - or if I specified one versus having the software do it indirectly by setting other parms. How can I figure out what multiple it is? And how do I tell CalMAN what the multiplier is? I've never used that function in CalMAN.
4a. Recently you told me it may be a good idea to try redoing the autocal with BT2020NF. I said I didn't gain much brightness and you said to make sure I did a gamma+color and then a color autocal, otherwise I'm leaving a lot of brightness on the table. I can't recall, so I'll try that all over again. Just to confirm, you mean gamma+color for the main calibration and then +color for each of the sub-changes like iris, CMD etc yes? Or do you mean gamma+color with Reference and then +color with BT2020NF?
4b. You also mentioned I won’t lose any saturation for 90% of the content, especially in high lamp (which I use exclusively with BT2020 content of course), by using the BT2020NF filter. That sounds good on paper, but I'd like to understand better what I do lose for the remaining 10% of the content.
Depending on that, it may still not be worth the tradeoff. I really enjoy the WCG and in certain scenes when watching certain movies I am in awe - just saw jaw dropping moments. As just a couple examples, there are scenes in Pan where guns shoot out a burst of color, like purples, oranges, deep reds etc (not to mention several scenic views in the movie that have some incredible looking colors which seem to be the WCG in full action). Another such example is in the first Kingsman movie - when the people's heads are exploding into busts of color - the colors are so vibrant and I've always assumed this was the benefit of the WCG. Would I be giving up such moments? If so it probably isn't worth the extra 10-15% (since at best that may put me around 50ish nits, still WAY short of 100 nits).
5. Regarding HLC:
5a. We exchanged a post a week or so ago about HLG. Arve chimed in about it. Based on what he said, can I create a custom HLG curve using his tools? Would I do this the exact same way I created my HDR10 custom curve, except perhaps I am starting with a different baseline curve? I may be switching to DirecTV soon, which has a decent amount of 4K content and several live upcoming sporting events now in 4K HDR.
5b. What happens if we use a HDR10 curve on HLG content? Does it look obviously bad, perhaps too dark or bright? Or is it something much more subtle?
6. Just to confirm, when setting bbo afterward I want 68 to be slightly flashing, yes?