How to use CIE graph for checking Rec.709 broadcast spec?

Hello!
I recently had a project where I was trying to observe legal limits for broadcast.
On one particular clip, I thought I would check the CIE graph - this is my first time doing this as I've only just heard about it. After my initial grade pass, here is what I found...

CIE After First Grade Pass

CIE_Original.png (669.76 KiB) Viewed 2395 times

The CIE graph shows that I'm way, way out of gamut on this colorful bottle. A bottle that to my eye appeared to be about right, and was only just barely outside of the saturation marks on the vectorscope. However, on the CIE graph, it's going nuts.

So decided to try and reduce the saturation until it came into gamut...

CIE After Desaturation to reach In-Gamut

CIE_Desat.png (652.91 KiB) Viewed 2395 times

Whoa. The bottle looks awful and extremely undersaturated. On the vectorscope we can see that I'm not even close anymore to our hitting our safe boxes, even with extents. There is no picture, but I also tried shifting hue as well as reducing and increasing luma vs hue. Not only did those break the image really bad, but nothing seemed to get me close to in-gamut while also retaining a reasonable saturation level to the eye or on vectorscope. I'm new to this CIE graph and couldn't find great info online about it's use. What am I doing wrong here? Or what would be the correct "best practice"?

I'm using a Davinci Wide Gamut color space with a timeline gamma of 2.4.
The camera is a Blackmagic Ursa Mini Pro, and the input color space is set to use Film Gen 1.

Can anyone provide me some insight on how to actually use the CIE graph effectively, how important it is to not clip the color space zone, etc, or proper techniques for managing when color does go out of gamut?

Best!
-Daniel

Just wanted to add -
After posting I was reading someone else mention the OFX gamut limiter.
I placed this on my last node and took a screencap with the original grade enabled.

Gamut Limited

CIE_Gamut Limited.png (370 KiB) Viewed 2170 times

Interesting...CIE still showing the gamut is pushing what looks like beyond the Rec.709 color space limits, but the image, while missing detail in some areas, is much much closer to what I would expect by eye, and on the vectorscope.

The plot, for me, thickens...

If your working space is rec709 and output is rec709, you can go out of gamut only if you introduce negative values. Positives will not cross gamut border, whatever their magnitude is. In wide gamut intermediate, if clipping below zero or gamut compression is applied you can’t go out of gamut in output either.

Absolute values are kind of irrelevant on gamut plot. 8bit RGB triplet 1, 0, 0 has maximum saturation in cie graph, because it has biggest achievable color separation due to other channels being zero. So one thing to check is clipping blacks in some channels.

And last, cie gamut plot isn’t telling you much about broadcast legal or valid values, if anything at all.

The CIE graph shows that I'm way, way out of gamut on this colorful bottle

I have to wonder what part of the signal you think is out of gamut there

The CIE graph shows that I'm way, way out of gamut on this colorful bottle

I have to wonder what part of the signal you think is out of gamut there

I do see in the one screen capture, that the red/oranges are crushed up along the right side of the Rec 709 triangle, but they don't go outside of it. So as far as I can tell, the image is Rec.7009 broadcast spec. Resolve has done its job and restricted colors to the gamut you specified.

I am not an expert, but I remember thinking similar thoughts: "Crushed blacks and whites (waveform) are bad. So Crushed colors (chromacity) must also be bad." But looking bad, unusual, or extreme is different from being out of spec. In older days, you could have luma on a waveform that is out of broadcast spec, so the luma display and "broadcast spec" had a relationship. But I think with modern editing software with color management, your output will be within spec if you've set up your color settings properly.

The question, then, is what does it look like? That's more of a judgement call, where graphs can give you clues of where to look to make sure you're okay with it. If you see significant portions of your luma (waveform) or chroma (chromacity) crushed, you need to look at the image as a whole and at the area where it's crushing to determine if it's the look you want.

If the bottle's colors are so crushed that it appears to be made of solid plastic of nearly uniform color, that's probably a bad thing. It looks "wrong". But if it appears as you want it to -- in this case natural, but you could also be trying to have the bottle appear unnaturally colored -- it's cool, and Resolve's color management is keeping it from actually having illegal chroma values.

(Or so I imagine. I'm not an expert, and this is the story I tell myself.)

I think there is a somewhat common mixup going on here, where gamut calculated from RGB data is mixed with luma-chroma model chromaticity and related video signal legal/valid thresholds. These are not the same thing and their relation is not that straightforward afaik. Which is also the reason I wrote that CIE gamut diagram does not probably show you much useful if your task is to fit into video signal specifications. CIE diagram xy values are calculated from XYZ color coordinates, but luma and chroma components are calculated from gamma corrected R'G'B' values using different math, plus there is the question of valid ranges (allowed excursion) that are not visualized on CIE gamut diagram.

I think there is a somewhat common mixup going on here, where gamut calculated from RGB data is mixed with luma-chroma model chromaticity and related video signal legal/valid thresholds. These are not the same thing and their relation is not that straightforward afaik. Which is also the reason I wrote that CIE gamut diagram does not probably show you much useful if your task is to fit into video signal specifications. CIE diagram xy values are calculated from XYZ color coordinates, but luma and chroma components are calculated from gamma corrected R'G'B' values using different math, plus there is the question of valid ranges (allowed excursion) that are not visualized on CIE gamut diagram.

Ah, that makes sense!
Also very interesting and good to hear that Resolve is taking care of a lot of the broadcast spec limits under the hood. As mentioned earlier, I am using a Resolve Wide Gamut color managed workflow, as well as using the broadcast safe option on my projects, I just rarely actually deliver for broadcast and wanted to cross my t's and dot my i's. Thanks to all for clarifying some of these less common graphs!
Best,
-Daniel