Hacker News Reader

Comment by kmacleod | Hacker News Reader

I worked for a video titling company in the 80s during their transition from Analog titling to Digital titling. One key thing I learned there is that CRTs have about three times the horizontal resolution folks give them credit for. Most folks measure horizontal resolution in how many clear pixels you can go from light to dark and back. What they don't account for is the pixel-shift, you can position those start and end transitions at three or more times that resolution.

Nowadays we call that sub-pixel rendering or antialiasing. But in the 80s so many people were convinced TVs could only do 640x480. Our titling systems were typically doing 2400x480 to get good quality character aliasing and image shading on CRTs. This is still somewhat true today analog-wise but common sub-pixel rendering or antialiasing does achieve the same effect on LCDs.

View on HN

Strictly speaking, because the signal is continuous, analog video doesn't really have horizontal resolution, does it?

Exactly. Black and white TVs worked very much like oscilloscopes and had almost no definition of horizontal resolution. Very few color broadcast TVs were made with three separate beams and three separate color filters (like rear-view projectors). Your typical color TV effectively created "pixels" by their use of shadow masks and three separate beams. The analog signal, however, could still be in transition along the entire horizontal sweep. The biggest difference between high-end studio monitors and your average TV was how packed the shadow mask was.

I don't think that's quite how it worked.

The colour was created by different phosphors on the inside of the screen, there was nothing different about the electron beams. The number and pitch, the resolution, of these different phosphor dots determined the resolution of the screen.

The shadow masking was to prevent the beam for, say, the red phosphors sweeping blue/green subpixels when moving from one pixel to another, since it wasn't practical to turn the beam off and then back on once the steering coils had been changed. Steering was continuous, so without shadowmasks, you would illuminate on the neighbour subpixels you pass on the way.

You could have done it all with a single beam, if you really wanted, but it's not very practical - you'd need to sweep slower since you could only illuminate 1 subpixel at a time, it'd take much longer to to steer, illuminate at the right level, and move to the next with the right beam power selected.

> I don't think that's quite how it worked.

I think he described exactly what you did (in fewer words) in the second half of his post. ("Your typical color TV ..."

The first half discusses that some projection screens had 3 different tubes. ("Very few broadcast TVs ...")

Either that or he edited his post.

My objection is:

> Your typical color TV effectively created "pixels" by their use of shadow masks and three separate beams.

It's true that almost all colour TVs are pixelated and have shadow masks and three separate beams, but it's got the cause and consequence the wrong way around.

The pixels are created by the phosphor dots, not the mask/tube design. It's possible to use this design without three tubes or the screen, strictly, but a consequence of this design suggests the optional addition of a shadow mask and additional tubes as a natural, later progression for improved performance.

> It's possible to use this design without three tubes or the screen, strictly,

Without shadow masks or an aperture grille seems pretty dang hard. Even ignoring the alignment concerns, it requires a much, much higher bandwidth.

The shadow mask and aperture grille approaches were the only approaches that ever really worked for color CRTs.

Phosphors and some kind of masking is how you make different colors on a single tube. Indeed, the shadow mask or aperture grille was offset 3 times and used to etch for the phosphor coatings in most manufacturing processes.

> l addition of a shadow mask and additional tubes as a natural, later progression for improved performance.

There's just one tube-- multiple guns/beams, though.

That really opened my eyes. I played GTAIII a long time ago on a portable black and white TV, and I vividly remember it looking so much better than our main TV (which was still a tube at the time)

AFAIK, you're still limited by the bandwidth of your analog signal. If you tried to treat it as if it had unlimited horizontal resolution, you'd use so much bandwidth that you'd end up stomping over the color subcarrier (and probably the audio subcarrier too).

No, I think it's the other way. The horizontal lines are digital, the meridians are continuous.

Think about it this way: lines are discrete, because they are separated by an HSYNC pulse that is used to fly back the electron beam from right to left. But the 'pixels' within a horizontal line are continuous: there is nothing in the video signal that delineates them.

To the number of horizontal pixels is essentially limitless (if you ignore the bandwidth limit and the CRT scan raster) while the number of vertical pixels/lines is always discrete.

there are a fixed number of horizontal lines. The number of implied vertical lines is determined by Nyquist's theorem.

To realign your decoder to interpret the earlier poster: their "horizontal resolution" is the frequency along the horizontal axis, i.e. the density of signal changes within one scan line.

No, resolution is not a digital concept. Pixels are a digital concept[0], but in an analog signal there is still a limit on how much frequency content you can resolve.

CRTs additionally impose their own limit on how sharp the picture can get horizontally. Confusingly, these are called "television lines", even though you'd think that refers to the horizontal scanlines that do discretely divide up the picture vertically. Color CRTs further limit the amount of color resolution independent of luminance resolution, and how they do that depends on what specific phosphor pattern they use, how they separate electrons from each color gun, etc.

[0] And one that's just as misleading as sample-and-hold diagrams of digital audio that make people go out and buy really expensive DACs for reproducing ultrasonics they can't hear

It has the NTSC Subcarrier of ~3.58Mhz, and that limits color resolution.

Let's say you're displaying a 640x480 screen on a TV. Your chroma information is sampled from an area about 2.1 pixels wide. So even if the signal is continuous, your color information is the result of sampling a boxed area centered on that pixel.

And because you don't want the wave pattern to be displayed as luma, you also take the average luma of that area. Now you have a blurred fat color pixel, but it can be positioned anywhere continuously.

NTSC is just one specific implementation of analog color video transmission, has nothing to do with CRT functionality outside of North America/Japan/a few other places and generally when speaking about CRT monitors for computers, arcades and the like they operate on discrete RGB signals with a wide variety of timing: https://en.wikipedia.org/wiki/Coordinated_Video_Timings, NTSC isn't a thing.

Furthermore, even when using NTSC video timings, the 3.58 MHz subcarrier is specific to transmission of composite video (carrying brightness and color combined in the same signal). A DVD player for instance with a component or SVideo output produces NTSC timing but there is no subcarrier because the color signal is discrete.

Reply on news.ycombinator.com