LUTs o' Luck - BUG or DESIGN?

Perhaps someone can speak with authority on this issue that may be a bug, questionable design, or my not understanding the Resolve Color Space workflow correctly.

My setup is 14.1.1 with a win10 box with plenty of horsepower to make Resolve s snappy 4k+ VFx editing system for output to Digital Cinema and UHD.

I've used Spectracal in Resolve with an i1 Display pro to generate 3D cube 17 point P3 65 Gamma 2.4 LUT for an LG 31MU97-B DCI res monitor driven over HDMI from a Decklink Mini Monitor 4k. (the monitor is capable of displaying this with pretty good DeltaE's). When I apply the LUT as the Video Display LUT it looks like Resolve applies it AFTER the Output Color Space instead of after the Timeline Color Space...which would seem to be the logical place for it in the pipeline.

For example, the Timeline Color Space is P3 65, the Output Color Space is the same and the P3 65 LUT is applied to the Video Display. All looks good. If I then change the Output Color Space to Rec 709 because I want to deliver some files for colleagues to watch on their HDTVs there is a visible shift on the LG monitor. It appears the video signal is going from P3 to 709 and then expanding from 709 to P3 on the monitor. That doesn't seem to make any sense. If the Video Display LUT came after the Timeline Color Space there would be no shift with the change in Output Color Space. So if this is indeed the design and not a bug it would appear that I have to do all my color evaluation/ rough grading work in the expanded gamma P3 space with the P3 Video Display LUT and then switch to the 709 Output Space every time I want to output something in 709. Is that correct?

Or am I missing something?

David
Intel i9 7960x @ 4.4GHz
Thermaltake Floe Riing 360 Water Cooler
Asus x299 Prime Deluxe
64GB 3333 Corsair Dominator ram
2 EVGA 1080ti Cards
Decklink Mini Monitor 4k

For others who might be wondering about this, it's been pointed out to me (the diagram on p73 of the UG) that this IS the design. I was aware of that diagram but I thought it might be in error as from the users perspective this pipeline would not seem optimal. However I've been around NLE development a long time and know that there may be software related issues that govern design. Also, the existing pipeline may well be a holdover from the earlier days of daVinci when the application of a "film print emulation LUT in a log workflow" (p.74) would have made sense. Today, that use would be limited to a very small user base. However there may be other applications that I'm missing. If so, I hope someone will enlighten and point them out as I'm new to Resolve.

Otherwise I'll continue to switch output spaces before outputting to a different color space than the timeline-monitoring space but....

....I'd like to make a Feature Request to have the option of changing the pipeline for the LUT to be applied after the Timeline Color Space.

This is how this should work.

The solution for you is simple. Make a REC709 display LUT and load that to view your REC709 material.

This is how this should work.

The solution for you is simple. Make a REC709 display LUT and load that to view your REC709 material.

I'm not sure why it "should work this way unless one is working exclusively in rec 709 and outputting always is 709. But if the project is with RAW source material and for wider gamut output shouldn't the timeline color space where all color operations are done, be done in a wider gamut?

Are there other reasons to work in Rec 709 that are not obvious? If so, it would be helpful to hear about them before I get too deep in teh project.

It is essential that you understand colour management before you attempt to do something like this. All colour management in Resolve is manual, you have to specify your camera source, timeline, output and monitoring spaces and provide LUTs if necessary. There is nothing "automagic" like with Photoshop.

A clarification: you are monitoring your output, not your timeline. If the output space is the same as the timeline space, then there is no change between output and timeline and in that case you could say you are monitoring your timeline. Think of it like this:
Camera space = what the DOP captured your footage in
Timeline space = what your adjustments are affecting
Output space = what you are looking at.

The workflow is then to go camera space -> conversion -> timeline space -> conversion -> output space. You can do this in regular Davinci YRGB mode by using LUTs or colour space conversion OFX nodes in your grade, or you can use Davinci colour managed mode, which clutters up your node graph less.

Ultimately, your monitor is going to have to accurately display the output colours on your screen. If you want to monitor a Rec709 output (for your friends with HDTVs), you will need to make a Rec709 calibration LUT.

It is correct that if you previously graded your shots for P3-D65, if you simply change your output to Rec709, the colours will change. However, if you correctly convert from P3-D65 to Rec709, by using a conversion OFX node, or by setting the output space in colour managed mode, and you change your monitoring LUT to your Rec709 monitor calibration, then the colours will be mostly identical except for colours that are out-of-gamut for Rec709 compared to P3-D65 (highly saturated colours like neon tints and bright red), which is expected behaviour when mapping from a larger colour space like P3 to Rec709.

If you use P3-DCI and try to map that to Rec709 you're going to experience a white point shift unless you make a LUT to convert from P3 to Rec709 by relative colorimetric conversion. As of right now Resolve's colour management can only do absolute colorimetric conversions.

@Micha Clazing
Thanks for the detailed response. I'm using Davinci colour managed mode. The shots are from BM CDNG green screen source material. A lot of the BG plates are RAW stills mixed with Adobe RGB and sRGB sources. As the primary end output is P3-DCI I thought it best to do all ballpark color in P3 timeline color space. the 709/sRGB outputs are for others to view working assemblies so it's only important that they be an accurate but gamut limited representation of the P3 timeline. I've assumed up to this point that Davinci colour managed mode will handle that conversion faithfully when I switch output space to 709 2.4 but I suppose I should check.

I've assumed up to this point that Davinci colour managed mode will handle that conversion faithfully when I switch output space to 709 2.4 but I suppose I should check.

If you have a Rec709 monitoring LUT that is accurate for your monitor, you should be able to assess if the conversion is correct. My personal experience is that I had to change the input gamma, but that may have been because I am targeting Rec1886 gamma rather than 2.4. In any case; don't assume, but verify.