Technical question on TV colo(u)r space

[fluggo99]

This one's been nagging me ever since I started using MSP. It has color filters in every program that limit the colors you can use to the NTSC or PAL colorspace. I hate the color choices once you've done that; they're all washed out.

But that's not my question.

My question is, why are these colors unusable in NTSC? Is it just not a good idea, or will the feds track me down for using illegal colors? :P

Get technical; I like it.

[Patrick]

Well, this ain't technical (sorry ), but won't the non NTSC or PAL colours simply bleed when viewed on a television (as opposed to a computer monitor)?

[DrMordrid]

Think of a 24 bit palette in terms of how many levels of each primary color, red, green and blue, are possible. This comes out to 255 levels.

Due to bandwidth limitations that date back to the 1950's TV's can only display approximately 204 levels of each primary color without them bleeding across the screen.

Sometimes the intensities can't even go that high without problems, depending on the individual TV's weirdnesses. Most often this happens with red, but green and blue can bleed as well.

Even full intensity white, 255, 255, 255, can cause horrible problems in the form of the video signal crossing over to the sound band. You can get buzzing and all kinds of video artifacting. Ever hear that from an older TV when a B&W checkerboard or striped pattern was being shown? Filters prevent most occurances now, but in the old days it was interesting....

What the output filters do is limit the intensity level of the video data to those legal and capable of being displayed on a TV screen without these distortions.

Examples: if you set up a 255,0,0 red it will change it to no higher than a 204,0,0. A 255,255,255 white would be similarly affected. Some filters will make the max intensity even lower just for insurance.

This will make the colors seem somewhat washed out or pastel-ish on the computer display, but also will prevent more problems than you can imagine when the video is displayed on a consumer TV screen.

In practice it's a good idea to keep the levels even lower than 204 in case of the occaisional out-of-spec TV getting a shot at your project. I try to keep red at or below 185 & green or blue at or below 195 when generating any video-destined bitmaps just so I don't get any suprises.

Sorry, but no bright crimsons until we get progressive scan digital TV.

Dr. Mordrid




[This message has been edited by DrMordrid (edited 08 April 2000).]

[fluggo99]

Coolness! But that raises more questions...

Okay, so there's some bright colors TV's shouldn't display. Couldn't we just lower the brightness coming out of the computer? Why a software solution?

It would seem to me that limiting colors to below 204 would be wasting the upper 51 levels. (which, since it's spread across red, green, and blue, is actually wasting 132,651 colors) Software and hardware designers had that little extra room for more detail; why didn't they use it? And make 255 representative of the brightest color you could use?

Then again, the palettes in MSP look like they have diamond-shaped sections cut from them, meaning that the color restriction is limited only to certain hues. (Dang, this is confusing!) What light can you shed on these?

[Patrick]

Fluggo, have you been spending too much time in your room again?...

[JoakimC]

Isn't this because YCBCR can't represent all colors that RGB24 can.

[fluggo99]

What? I'm sure-- positive, in fact!-- that staring at a monitor all day doesn't do a thing to your eyes. And just because there's no ventilation doesn't mean...

[Patrick]

[DrMordrid]

Yes fluggo99, it may be a waste of the upper 51 intensity levels but the bandwidth is the real limiting factor. Full intensities to the existing 24 bit RGB levels would require re-doing the whole TV broadcast spectrum.

Why bother when progressive scan is just around the corner?

Dr. Mordrid

[salacious]

When DVD files are encoded it actually reduced the spectrum range of the YCbCr range to between 15 and 235 (or is it 240). The lower level is sometimes called 7.5IRE. So even with DVD outputting progressive scan you are losing spectrum range. I would convert the YCbCr range back into RGB space and see what the RGB colour range is (since this must be in in your TV) but I don't remember all the colour space conversions in my head (only Y=.59R+0.3G+.11B).

Salacious

[Grigory]

The limitation is imposed NOT because of bandwidth, but to escape simple signal overloading condition. The levels below 7.5 IRE are used for syncronization, and regular video signal must not go below this level.
The levels over 235 are restricted because you have to leave some space for color subcarrier and escape the situation when the composite signal amplitude goes out of range.
This is TV transmitter modulator limitation, but not a bandwidth (frequency range) limitation.

Grigory
P.S. All conclusions above remain unaltered, however.

[jeff b]

...of course, then there's the sloppy cable companies that blow the crap out of all white backgrounds and slop over into the audio!

[jrr]

As nobody's mentioned it I thought I'd also point out that a lot of the limitations in displaying colour on a CRT screen (slightly different for flat screens & projectors) is the type of phosphors used.

The NTSC standard is designed to use a specific red green and blue phosphor set by the SMPTE, similarly the EBU sets the standard in the PAL world.

The choice of phosphors is a comprimise between selecting a material with certain colours to mach the human red, green and blue cones in our eyes and the efficiency and decay time of the electron->photon production in order to get a clear bright image.

There is a relatively large range of pure (highly saturated) colours that cannot be produced by the phosphors on the screen and so even if the colours can be generated in a computer, they can never be accurately displayed.

In most cameras a matrix is used to calibrate the red green, and blue CCD responses, so what is recorded onto tape is already within the colour gamut of CRT's. So, to your eye, what you see on screen looks similar to what would see it in the natural world (although to any other animal it would look nothing like the original).

Incidentally in the broadcast industry (eg in serial digital interfaces (SDI) and the DigiBeta tape format) they are increasingly using 10bit 4:2:2 sampling and so the question of reducing the dynamic range is pretty well overcome.

James

PS see these references for more details on colour: http://www.inforamp.net/~poynton/PDFs/Guided_tour.pdf http://www.color.org/wpaper1.html



[This message has been edited by jrr (edited 10 April 2000).]

[Patrick]

Welcome back Grigory. I haven't seen you post for awhile. And if Kris sees this, welcome back to you also. Where have you guys been hiding?

[Grigory]

Hi Patric,

I was hiding behind DV Raptor DV format editing .
I also did some job and had little free time.

Now, I have changed MB to Abit BP6 and dual Celeron 510.
This setup made unworkable my old good BT848 capture card. It was capable to capture in YUV colorspace from the very beginning and I used it for my VHS captures successfully for two years.
I decided to come back to RR_G + G400 DH combination, but abandon Matrox MJPEG format.
That is why I came back to this forum.
Nice to meet you again.

Grigory