1.2 - How NTSC composite video works
Horizontally speaking, NTSC encodes its various components as signal brightness and two color information streams of differing bandwidth.
The brightness component can change at a rate that is approximately equivalant of 700-ish brightness changes per scan line, with the other 20 or so appearing in the overscan area which is typically hidden by the way television tubes are mounted; your milage may vary a little if you have an LCD, but then again, it may not.
Color changes are a function of combining the brightness change with the two color components. These components can change at an average rate of 100 color changes per complete line. However, because one component is slower than the other, not all color changes can be reproduced at that rate. Notice that I described this as a rate; that's because television, real television, is a pure analog signal and although the rate that the brightness and the colors colors can change is limited, the position that a brightnes or color change can occur at is only limied by how recently one already did... if colors haven't changed within 1/100th of a line, then you can have a color change fairly precisely located... at the cost of not having another for a 1/100th.
Similarly, a brightness change (or a green amplitude change... some of you will see why when I describe the math, for the rest, just trust me) can occur at a rate of about 700, but they can start anywhere and so the precision with which either a brightness change or a color change can be located on a scan line is in effect infinite with an analog system. When displayed on a typical color television tube, most of this capability is lost because the display beam only has a finite number of RGB phospher triads it can illuminate, and the analog detail is re-sampled by the "jail-bars" of the phosphor dots or slots. However, this is still true of a black and white set, which has a continuous display surface. Again roughly, greens change the fastest, reds the next fastest, and blues the slowest of all.
These color change ratios (to one another) were designed to mimic the ratios exhibited by your eye's sensitivity to similar changes. Unfortunately, while the idea is sound as far as it goes, your eye's ability to deal with those changes, ratios aside, is so much higher than the change rate video provides, that I would argue that the designers kind of screwed the pooch in this area, but that's a different discussion.
The math is done like this, again more or less, using the R, G and B (red, green and blue) color components:
Brightness = .59 times G plus .3 times R plus .11 times B. That gets you luma, a black and white signal that offers compatability with how the older BW television sets worked. This is also called "Y".
The first color component is simply (R-Y), although as I mentioned above, it is bandwidth-limited so that the color changes are encoded in a broad, blurry way.
The third component is (B-Y) and it is bandwidth-limited even further... slower and blurrier.
The color image is re-created at the display this way:
R = (R-Y) + Y, B = (B-Y) + Y, and G = Y - (R + B), keeping in mind the R, G, and B .3, .59, and .11 scaling factors, respectively.
As far as vertical resolution goes, this is a bit easier to understand. The display is created in two passes. One the first pass, half the lines are painted. On the second pass, the other lines are painted in between the originals. Next time, another two set of lines that are considered a new pair, and so forth. These sets of lines, half-scans as it were, are referred to as the odd and even fields of a frame.
A frame is considered to be definitive of how many lines you see, and it adds up to 400+ (oddfield=200+, evenfield=200+), with the remaining scan lines again typically hidden as a consequence of how television tubes are mounted. The region where they hide is called the "overscan area."
But there's a monkey wrench thrown in here as well. Because the odd and even fields arrive at regularly spaced, separate intervals in time, they only really get to fully describe an image if it is standing still. If the image is in motion, then by the time the next field comes along, the details in the image are in new positions and the lines for the new positions won't be colinear with the previous field in "that frame", thereby sacrificing perceived spatial resolution for higher timewise (faster motion) resolution. Or in other words, you can see action with a 60 per frame rate, or detail at a 30 per frame rate.
"Actual" television resolution is a conflation of what happens horizontally with color and brightness encoding, combined with how the odd and even field displays interact with the motion of the subject matter being displayed. To sum up, you can't look at it as having a specific horizontal or vertical resolution.
The bad news? The bad news is I simplified a lot of issues and skipped over various warts like artifact production and didn't waffle about PAL, which is a little different, and completely ignored SECAM which is completely different (but that's France... it's always the French, isn't it.) You should be grateful.
With apologies to Tom Lehrer, there'll be a quiz next period. Let's not all see the same hands.
Have a comment for me? Click here.
