RGB Parade question

Hello.

There is a footage from camera 8 bit hevc without any correction in the attachment. Why do I see so many colors bellow 0 on parade? Especially in blue channel?

My guess is a lot of spikes in the lighting. It looks like bad mixed lighting to me with fluorescents and god knows what else.

Use clips as needed to get the image under control and make it legal, while not crushing the black level details.

Do you mean that the top image is the original, and it's clipping the shadows, and with a correction you can get more details in the shadows? That would mean that Resolve is applying wrong input levels to the footage.

On the edit page, right click the clips and choose "Clip Attributes...", and change Data Levels to Full. See if that will help.

The top image is the 8 bit h.265 original from NX500. Saturation on the camera was set to 0. Actually I think I had to set it to -3...-5. The blue channel bellow 128 on the second image is the yellow floor. Reducing the saturation is getting blue bottom up.
Full Data Levels almost doesn't do anything.

It is still don't clear for me what is the reason.

I think the colors in the original footage are so saturated that they are crushing the bottom of the parade. So the "blacks" are not crushed because of deep shadows, but because of the saturated colors.

If the color gamut of the camera is srgb/rec709, saturated colors will clip easily.

If the color gamut of the camera is srgb/rec709, saturated colors will clip easily.

The color gamut of the camera is srgb/rec709. That's why saturated colors had to be clipped. But in fact they are not, and I don't understand why.

Perhaps if you post your Resolve settings such as colour management etc we may be able to offer more assistance. There are many settings that can cause the effect you are seeing. It's quite normal for saturated colours with low brightness to clip on the RGB Parade display. I would be more concerned with the slight clipping on the waveform display.

Perhaps if you post your Resolve settings such as colour management etc we may be able to offer more assistance.

Almost everything sets as a default.

If the color gamut of the camera is srgb/rec709, saturated colors will clip easily.

The color gamut of the camera is srgb/rec709. That's why saturated colors had to be clipped. But in fact they are not, and I don't understand why.

They colors are clipped, but it's not so obvious...
If you lower the saturation, resolve will do a new "mix" of the color channels, and it's gonna mostly use the data from the green channel.

How could they be clipped if I see the information bellow 0 and can get it back by reducing the saturation and moving up the shadows?

The yellow mattress is pure black in the blue channel, there would be detail in the "black" if the yellow is not clipped. If you just lift the blacks, the mattress would be plain dark grey on the blue channel, with no detail.

The yellow mattress is pure black in the blue channel, there would be detail in the "black" if the yellow is not clipped. If you just lift the blacks, the mattress would be plain dark grey on the blue channel, with no detail.

Yes, you are right. But if I reduce the saturation the yellow mattress which is pure black in the blue channel is getting up not clipped.

Yes, you are right. But if I reduce the saturation the yellow mattress which is pure black in the blue channel is getting up not clipped.

Go to Color Management and check off the "Make Broadcast Safe" checkbox and try the "-10 to 110" range, and see what happens. Bear in mind it is technically possible to have something that's illegal and out of gamut but not technically clipped.

Go to Color Management and check off the "Make Broadcast Safe" checkbox and try the "-10 to 110" range, and see what happens. Bear in mind it is technically possible to have something that's illegal and out of gamut but not technically clipped.

I Tried -10 to 110 - nothing changed. I tried check off the box - and again nothing.
I guess that happens because the original file in hevc.

Yes, you are right. But if I reduce the saturation the yellow mattress which is pure black in the blue channel is getting up not clipped.

You're hardly gonna see any color clipping in this case, But that is the reason for the crushed parade.

If you can post a short clip I can download, I'll check and give you a definitive answer. Everyone here are just guessing.

If you can post a short clip I can download, I'll check and give you a definitive answer. Everyone here are just guessing.

https://yadi.sk/d/TwEwHV3y3EHkWT

I looked at the clip, and changed the levels to full. got some more info in the shadows. However I could lift the signal and there was still a lot of signal under 0. I think the "negative" blue values are from YUV-RGB conversion that Resolve is doing?

I looked at the clip.

Thank you.

Actually Premiere Pro shows the same. It's a screenshot of original h.265 opened in Premiere Pro without any conversion to prores or another.

Hi Sergey,

Ok. This is complicated because it requires an understanding at a technical level to comprehend.
Your original clip is ok, no issues. Resolve is ok, no issues.

Firstly, the clip is 'full range' i.e. bits 0-1023. It is NOT video range 64 - 940.
When I set the clip to full (data) range in resolve the waveform shows a perfect exposure range from black to white with no clipping. I have attached an image at the end of this post.

The following is regarding how the video data is created by the camera.

As the original clip is YUV (or Y, Cb, Cr to be technically correct) the clip is being displayed as an RGB signal on the RGB scopes. The yellow floor mat in the image has a certain luminance component (Y) and colour component (U,V). The yellow is very saturated and has a fairly low luminance. To achieve the required saturation at that luminance level, the complimentary colour of yellow (blue) must be reduced, ( therefore increasing the complimentary colour, yellow) otherwise the yellow would be too bright if the red and green (mixed together to make yellow) channels are increased to achieve the correct saturation.

It is perfectly acceptable for the signal to dip below the zero line on the scopes. Using RGB scopes the green channel is mostly representing the Y component, as in a video signal, the electrical levels of the RGB channels are not equal due to human eye colour perception.

A quote from a technical paper:

"The Y, or luminance value, is actually a "weighted average" of the R,G and B values for a given color. The human eye perceives green as being brighter than it really is, and blue much less bright. By constructing a luminance signal that is biased toward certain colors, a brightness signal that approximates human color perception can be created.

The equation that is used when encoding the luminance signal is:

.3R + .59G + .11B

This means that the green component is given the most weight and the blue the least when determining the luminance of a pixel. Thus, green and red objects will appear brighter than blue objects, just as in the human eye."

So, set your clip attributes to full range (0-1023) and grade away to your hearts content, safe in the knowledge that this is how it works, and that there is nothing wrong.

Here ends the lesson. Whew! If you still don't understand, you will need to do a television electrical engineering course of study because it gets even more complicated than my attempt to simplify what is going on behind the scenes.

Looks perfect

Screen Shot 2017-02-21 at 10.33.10 PM.png (627.2 KiB) Viewed 6869 times

So, set your clip attributes to full range (0-1023) and grade away to your hearts content, safe in the knowledge that this is how it works, and that there is nothing wrong.

Thank you very much! This is a lot more sophisticated than I thought.
Does it mean that the camera was set right and there was no need to lower down the saturation of the camera?

Camera settings have nothing to do with the issue in this thread. Set them so the image looks good to you. The important thing is proper exposure.

Ok. Thank you.

Hi Peter, thanks for your post.
I had a very similar question in another thread and read your comment, and hope that I got it right:
I was wondering why I see negative values in the RGB parade in the blue channel even after applying a broadcast safe filter. (the screenshot is from FCPX, a file imported from avid with safe color limiter set to RGB minimum 16)
Until now I thought that the parade shows me the "real" RGB values that exceed a value of 16, so if I see something below the line it must be illegal.
If I get your comment right the RGB parade doesnt show the real values and its weighted, so blue is perceived less bright and therefore shifted down in the parade. This way slightly negative blue values are not an issue but negative green values woud be an issue?

It is perfectly acceptable for the signal to dip below the zero line on the scopes. Using RGB scopes the green channel is mostly representing the Y component, as in a video signal, the electrical levels of the RGB channels are not equal due to human eye colour perception.

A quote from a technical paper:

"The Y, or luminance value, is actually a "weighted average" of the R,G and B values for a given color. The human eye perceives green as being brighter than it really is, and blue much less bright. By constructing a luminance signal that is biased toward certain colors, a brightness signal that approximates human color perception can be created.

The equation that is used when encoding the luminance signal is:

.3R + .59G + .11B

This means that the green component is given the most weight and the blue the least when determining the luminance of a pixel. Thus, green and red objects will appear brighter than blue objects, just as in the human eye.

Properly converted YUV->RGB signal should be normalised to 0-100% based on limited/full levels info stored with YUV file. RGB should not really use limited scale- this rather would be confusing.
Also, any compression will mess your scopes and may cause further confusion.
"Perfect 0-100%" YUV data once compressed won't be so perfect anymore, so you would need to remove this variable. If you want to avoid codec overshoots then use uncompressed format. Best if some YUV444 (just not many tools support them).
Y in Resolve is "fake" calculated signal and not sure if it's exactly the same as Y from YUV signal. It may be.

Imagine B&W chess board with perfect 0/1023 values in Resolve exported to 10bit: RGB, YUV444 and ProRes444.
RGB is 1:1 map so it's all perfect.
YUV444 will go through conversion, but when done at 32bit float then it should revert back to RGB keeping 10bit precision, so still should be perfect 0-1023.
ProRes444 - Resolve RGB data will be converted to YUV444 (16bit or 32bit float), as ProRes is alway YUV internally and then it will be compressed. High contrast edges can create overshoots and your perfect signal won't be so perfect anymore. You may see in FCPX scopes values outside 0-100%.

Thanks a lot!