I’m trying to create custom output transforms in DCTL, using the parametric ACES transforms in DaVinci Resolve.
Should the mid-point luminance (Y_MID) be set to 15 nit for a 600 nits P3D65/PQ output transform like it is for 1000, 2000 and 4000 nits?
I’ve read it should be 4.8 nits for a 10000 nits output transform in this post:
TLDR; I would recommend that you just set your peak Y_MAX to 600nits and Y_MID to 10nits in your custom Output Transform. Then you can add an exposure shift to your image if you want it overall brighter.
This depends on your use case. There are 2 basic philosophies on how one should map HDR in relation to SDR. #1 is that you can keep all mid-tones and normal values exactly the same and just extend the curve above 100nits so that it doesn’t roll off until 1000. This effectively keeps mid-gray at 10nits (SDR video) and just uses the additional luminance range for headroom and to extend highlights. This preserves the greatest “match” between SDR and HDR (because most of the values are the same) but generally most people also want their HDR image overall brighter.
Option #2 does #1 but also includes an overall exposure shift so that mid-gray and the overall image is a bit brighter. This is what the HDR transforms do in ACES 1.1. We ended up at 15nits for mid-gray at 1000nits, but then didn’t go any brighter for 2000 or 4000 nits peak luminance. (The reasons why not are a long story, but technically the only difference in the 2000 and 4000 nit OTs is that the highlights extend up a bit farther - if there is scene data that reaches those values…)
The reason the value is a parameter is to allow for any of these scenarios, but unfortunately we never published any guidance on how to derive a value for “somewhere in between” or any recommendations for where mid-gray should be for a 600-nit luminance.
In ACES 1.1, the mid-gray and peak luminance are controlled independently. So holding to the design theory, the mid-gray should probably be set somewhere between 10nits and 15nits. But technically, the difference you will see if you change that mid-gray value will be an exposure shift, so the answer to your question, I would recommend that you just set your peak Y_MAX to 600nits and Y_MID to 10nits. Then you can add an exposure shift to your image if you want it overall brighter.
(By the way, the 4.8nit value you cite is the mid-gray for dark-surround theatrical where the peak luminance is only 48nits. This is ~10nits in “SDR video” land.)
Rest assured, in the ACES 2.0 development, that mid-gray value does vary automatically based solely on the peak luminance value (but the option to make it “match” exactly also remains available, for those use cases where it is needed). And we will publish guidance on how exactly to use the transforms to achieve the needed use case.
Thanks a lot for the recommendation and explanation.
The use case for the 600 nits ODT would be to do a first pass grade on a 600ish nits OLED monitor when our reference monitors are not available.
I’ve tried both 10 and 15 nits and 15 nits feels better because the overall exposure is closer to the 1000 nits ODT on a 1000 nits monitor. I don’t suppose there would be any problem sticking to 15 nits for this specific scenario?
In case we use a lower contrast ratio monitor, I understand Y_MIN should be set to the minimum black level of the display. How about the STRETCH_BLACK field? Should it always be on?
In that use case, yes, I would agree then to keep it at 15 so that it will match what you see on the 1000=nit.
Well…“stretch black” is sort of a hack to force the OT to make CV=0. It’s not the ideal way to fix the actual problem but I had limited options at the time the transforms were introduced. Yes, I would leave it set “on”.
(Unless you are trying to simulate some other display’s black level in PQ - i.e. if you wanted to flare up the blacks to say 0.01nits or something to simulate what SDR black might look like, but I’m assuming you don’t want that…)