trying to understand x.v.Color

john-beale wrote on 6/20/2010, 7:00 PM

I have two AVCHD camcorders than have an "x.v.Color" mode, in additional to the normal Rec.709 color mode. In an attempt to see what is going on in this new mode, I shot a standard color-checker chart using the Panasonic TM700 in both normal and x.v.Color mode, and did the same with a Sony XR500V camera. I also shot stills with a Canon 20D DSLR. Lighting was 9:30 am direct sunlight this morning, in a clear blue sky in Mtn.View CA. Color temperature 5025k (white balance based on a 20D RAW exposure of a calibrated grey card). The TM700 was manually white-balanced to that card; the XR500V has no manual white balance that I know of, only "white balance shift" which I left unshifted.

Below are screenshots of each of the clips loaded into Vegas 9e. The color looks a bit different in x.v.Color mode vs. not, but to tell the truth, I expected more difference than this. I'm aware that Vegas does not yet support x.v.Color but what is really happening here? Looks to me like with this target at least, most of the difference between xvColor and Normal, could be achieved by just a slight tweak with a color balance tool. I assume Vegas' meters are based on the RGB data after conversion from YCbCr. Is there any accessible editing tool that has a vector scope & waveform monitor based on the the native YCbCr data?

http://www.bealecorner.com/TM700/GMB-ColorCompare/index.htm

Comments

GlennChan wrote on 6/21/2010, 7:04 PM

1- Your color chart doesn't look like it contains any colors which fall outside normal color gamut.

What if you shoot the reflections off any CD/optical disk. The surface of the disk acts as a diffraction grating and will generate pure wavelengths. If the reflection in the CD is that of a black surface, the colors will be even more pure.

2- It looks like the camera is doing some sort of color enhancement in normal mode... because the colors shouldn't be more saturated.

john-beale wrote on 6/21/2010, 11:39 PM

Hey, thanks for checking out my experiment. Good idea about using a CD as a grating, but I have a "real" rectangular transmission grating and used it to make a grating spectroscope in a long cardboard box. On the slit I had a CFL lamp (for calibration via mercury emission lines) plus a nice bright 200W clear filament lamp as a continuous source. Both shining through some spun diffusion to help illuminate the grating more or less evenly. I added a Vegas screenshot, and the Y,Cb,Cr planes extracted direct from the .m2ts file here:

http://www.bealecorner.com/TM700/index.html

It looks like we are getting some decent excursions in the Cb, Cr planes. I have to do some more thinking about what that vector scope plot is saying, though. I guess when you feed pure (nearly) single wavelengths into the sensor, you get response from only one CCD at a time (this is a 3-CCD camera with dichroic beamsplitter). So that's why most of the vectorscope is mostly showing me just three directions, I guess those are the primaries of the sensor as seen in vectorscope space. It does seem to exceed 100% saturation especially in the green (lower-left).

GlennChan wrote on 6/22/2010, 11:07 PM

I guess when you feed pure (nearly) single wavelengths into the sensor, you get response from only one CCD at a time (this is a 3-CCD camera with dichroic beamsplitter)
A pure wavelength may cause 2 of the CCDs to respond, in some cases... because the wavelength isn't split perfectly by the beamsplitter. For most wavelengths the beamsplitter will isolate that perfectly, but there are transition areas in between. As I understand it.

2- I don't think Vegas' vectorscopes are necessarily designed to be helpful for wide gamut.

What would make sense is some form of display that shows you:
A- What is legal
B- What is wide gamut / outside normal gamut
C- What is illegal even in a wide gamut system

3- The way wide gamut is encoded is that when you go below black, the transfer function changes so that more colors can be packed in. Charles Poynton's book has some information on this, though there are 2 competing formats for wide gamut and I don't know the exact details and I don't even know which one is being used by Sony now.
I doubt that vegas's scopes handle the change in the transfer function. (Nor should they.)

Sorry, I don't have equipment that handles xvColor so I can't tell you much more than that.

apit34356 wrote on 6/22/2010, 11:42 PM

"A pure wavelength may cause 2 of the CCDs to respond, in some cases... because the wavelength isn't split perfectly by the beam-splitter." this should never happen unless the the wavelength is close to the cut-off. Dust in the lens assembly, oil on the len surface, moisture in/on the lens,scratches, micro air bubbles in the glass of the beam-splitter, can cause issues. A interesting event occurs sometimes when light is reflected from the surface of the CCD( areas around pixels) back into the beam-splitter ( usually excessive photon density). Another interest event that is cause within the CCD surface plane, as the scan starts, some moving electrons will discharge photon energy, affecting a number of close pixels and generating the effect above ( sometimes call noise). Most of the CCDs problems are related to moving the electrons from pt to another pt; more energy required to move them, more discharges....... Sony's new designs cmos/ccds lower this type of issues I believe but there many interesting solutions.

john-beale wrote on 6/23/2010, 7:55 AM

Looking at http://www.bealecorner.com/TM700/TM700-CinemaColorOn.PNG it seems the full spectrum really is chopped into almost pure R,G,B in this camera (except at the brightest point of the R-G border, where things are really saturating). That agrees with other 3-CCD cameras I've looked at.

I calibrated the YCbCr image based on the CFL emission lines and generated a plot of the Cb, Cr response vs. wavelength, for pure colors, as you see here:
http://www.bealecorner.com/TM700/

The measured (min,max) range for this frame, chosen to be at an exposure giving me maximum saturation values, is Cb: (10,232) and Cr: (28,242) over the full visible spectrum. The normal limit in YUV video encoding is (16, 240), so we aren't exceeding the normal range by a whole lot.

I rather suspect Panasonic simply removed the [16,240] chroma range clamp but left the sensor chip and analog electronics precisely the same as in the older, non-"x.v.Color" capable unit. Just a hunch.

GlennChan wrote on 6/23/2010, 11:37 AM

this should never happen unless the the wavelength is close to the cut-off.
That's what I'm saying.

The measured (min,max) range for this frame, chosen to be at an exposure giving me maximum saturation values, is Cb: (10,232) and Cr: (28,242) over the full visible spectrum. The normal limit in YUV video encoding is (16, 240), so we aren't exceeding the normal range by a whole lot.
You should try looking at the RGB values. You are looking for values that go significantly below legal black level. (In normal cameras, you have values that go below black level due to noise from the sensor, compression noise, and/or sharpening.)

john-beale wrote on 6/23/2010, 9:08 PM

> You should try looking at the RGB values.

I'll try that, but the camera doesn't generate RGB (except I guess via HDMI). I wanted to look at Y'CbCr because that is the actual camera output, what is stored in the .MTS or .m2ts file. The RGB values are generated from that, in my case by the software on my PC that is displaying the MTS file (presumably following REC.709).

None of my software speaks the new xyYCC standard, so I presume it is going to clamp R,G,B values to zero and not display negative ones. Sony talks about how the x.v.Color standard can represent negative R,G,B although I suspect no existing camera actually does that.

john-beale wrote on 7/8/2010, 10:53 PM

In case of interest, I made a video consisting of one frame for each Y' value showing on an xy graph each (Cb, Cr) value. So this is each Y'=constant slice through the 3D Y'CbCr gamut volume. Of course this is in arbitrary "Panasonic TM700" units, not necessarily any well defined standard colorspace. But the input was defined, being all single-wavelength colors over the visible spectrum 400..700 nm.

page: http://www.bealecorner.com/TM700/
video: http://www.bealecorner.com/TM700/TM700-YUV-hist1.mp4

You can see that the camera is recording Cb values right out to 255 for Y' in the range (36...80), and we get Cr = 255 at Y' = 77. Cr also reaches down to zero around Y' = (148..190). It would be interesting to compare this (with the same full-spectrum input) to other cameras- haven't had a chance yet.