13

Another question about color computers in the days of CRT televisions!

There is a consensus that pixels should be square, and so they typically are on purpose-designed monitors. However, looking at a close-up of pixels from a CRT TV:

Phosphor pattern

... they definitely do not look square; they look to me about as rectangular as the screen itself, maybe 4:3, which suggests horizontal resolution should be about the same as vertical resolution?

The other thing that jumps out at me is that the columns of pixels are offset half a pixel height, which is the last thing I was expecting on a system that scans by horizontal lines. What's going on with that? Does the electron beam half-light two pixels every other pixel? Does it have anything to do with interlacing?

  • 6
    That's just one of many phosphor patterns used by electron-beam televisions. – Mark Jan 29 '17 at 22:01
  • 3
    BTW, there's no kind of consensus that pixels should be square. Especially on older homecomputers, pixels were shaped by whatever accident the necessary timing constraints produced. Non-square pixels where common. – dirkt Oct 6 '17 at 6:02
  • 3
    @dirkt there's a reason why newer systems and formats have all converged on square pixels, though — they're far easier to work with. Who wants a system where you can't even rotate an image 90 degrees without changing its size? – hobbs Oct 11 '17 at 6:47
  • 1
    @traal My 1280x1024 monitor had square pixels, and it was great, the squarest monitor I ever owned. It's a pity it died; you can't buy monitors like that anymore. – rwallace Nov 11 '17 at 11:10
  • 2
    @traal 1280x1024 monitors do, in fact have square pixels. They're 5:4 aspect ratio displays, not 4:3 (at least for LCDs, CRTs that supported 1280x1024 were typically 4:3 displays). Probably the most recent common PC resolution with non-square pixels would be 16:10 1200x800 displays, which have pixels that are slightly squished in the horizontal (1200x750 would be square). – mnem Nov 14 '17 at 22:38
32

CRT TVs are analog devices, there is no "pixel", but the size of the spot limits the resolution of the image, as the size of silver nitrate crystals limits the resolution of photo films.

Several arrangements of the colour stripes on the CRT tubes have been designed, they are a compromise between resolution, luminosity and the precise alignment of the electron guns and shadow mask.

Trinitron CRTs, with vertical stripes, made very good computer monitors. For interlaced analog TV signals, the pattern above with staggered pixels is probably better and more mechanically stable than a rectangular array.

At the age of CRT displays, there was no exact alignment of pixels to the phosphors patterns, like it is common nowadays with LCD screens. The analog bandwidth and dispersion of electrons made every pixel a Gaussian circle. The coloured mask is overlaid over this circle, phosphors are unevenly lit.

(Of course, you can ask Wikipedia about "Aperture Grille" vs. "Shadow Mask")


(I did not expect so many upvotes, this answer deserves an upgrade !)

In a way, I did not answer the reason why a computer could not align the dots in memory to a precise coloured spot on a cathodic tube.

It is because is it is an open-loop system! There is no feedback between the displayed image and all the electronic stuff between the very low voltage VGA signals and the very high voltages driving the CRT.

A CRT is mechanically stable, and the manufacturers painfully optimized the construction for a precise alignment of phosphors to the shadow masks and electron guns, with thick hardened glass and temperature stable alloys. So the "blue" gun only lights blue phosphors.

There is not such precision for the electronic part. Even minimal variations of the temperature and the wear of components can slightly change the electron acceleration voltage and the voltage put into deviation coils. Even Earth magnetic field or a nearby loudspeaker (and there are loudspeakers in TVs and many computer displays !) can move the position where electrons fall.

Without a very complex and precise regulation mechanism, sensors in the CRT able to measure electric charges or where light is emitted, there is no way to align pixels to phosphors.

Electrostatic deviation,which was used in oscilloscopes, is a bit more precise but need very deep tubes. An average size TV would be one meter deep and weight one hundred kilos. BTW, there were also special tubes for oscilloscopes and RADAR able to keep charges for some time, giving persistent images...

And here, after this long digresssion we wander back to paleo-computosaurs.

Williams-Kilburn CRTs (https://en.wikipedia.org/wiki/Williams_tube) were used as computer memory, measuring the electric charge of each pixel. Even the wikipedia article mentions the issues with having to constantly recalibrate the fixture. Here, the pixels on the CRT and the video RAM are actually the same thing! Too bad the hardware for probing pixels is hiding them !

  • 5
    Relevant: A Pixel is Not A Little Square (PDF). – user4379 Jan 29 '17 at 21:20
  • 1
    Indeed, the shape of the electron gun's beam is a Gaussian distribution – Nayuki Jan 30 '17 at 0:12
  • 5
    I don't know whether it's worth pointing out: original black and white screens don't have any sort of pattern mask — they're a completely open canvas for the electron gun. Colour masks just act to combine the output of three electron guns on one surface. A hypothetical ideal colour mask would have individual elements so tiny that no human can see them without a microscope. The size of the projected beam(s) is completely unrelated, indeed the exact same tubes were often used for PAL and NTSC, despite different line heights. – Tommy Oct 5 '17 at 13:20
  • @user4379 that paper is totally irrelevant to both the question and answer. It has its place, but it's not here. – Mark Ransom Oct 17 '17 at 23:01
5

Besides TEMLIB's excellent answer relating to the physical characteristics of the TV, it's also worth pointing out the effects of the video input itself as well.

For analogue video input, the picture is encoded as a continuously varying value, in the form of a wave. The higher the horizontal resolution you want, the more detailed the variations in this value need to be, so the higher bandwidth you need to use, for various mathematical reasons. So the horizontal resolution is based on the chosen bandwidth - since there's only so much useful electromagnetic spectrum, this of course is a compromise.

But for vertical resolution, the story is different. An analogue TV signal is encoded with a specified number of lines (there are generally only a small number of choices that a TV will support; but a CRT designed as a computer monitor might well support more), with a special signal at the start of each line so the TV knows to switch the beam off and move it back to the start of the next line. This is further complicated by some of these scanlines being invisible, due to them being in parts of the screen where the distortion from the convex CRT would be too great, or due to the TV needing time to move the beam back to the top left to start the next frame.

But besides these issues, the vertical resolution is well-defined, whereas the horizontal resolution is not.

Of course, if you connect a digital computer to an analogue TV, there will certainly be a defined resolution at the stage when the digital picture is converted into an analogue signal for output. And many old computers do indeed use pixels that end up being rectangular when displayed on the screen.

  • 3
    Such as it may be relevant: industry practice when digitising analogue composite video is to sample at four times the frequency of the colour subcarrier. So, no, there aren't actually discrete pixels, but the industry has decided that 910 samples/line NTSC and ~1,135 samples/line for PAL is appropriate for preservation. That's for composite before separating colour; if it's a colour image then actual fidelity will be less. – Tommy Oct 17 '17 at 15:08
  • 2
    @Tommy absolutely. The frequency of the colour subcarrier will effectively limit the horizontal resolution of colour TV signals; and for black and white, it will be constrained by a standardised bandwidth limit. So there will absolutely be a limit beyond which no more horizontal resolution can usefully be discerned. The whole thing about analogue signals being better than digital is a silly myth. But to try to demarcate well-defined pixels in the horizontal will not be possible, whereas it would be in the vertical, was simply the point I was trying to make. – Muzer Oct 17 '17 at 15:24
  • 1
    Oh, yeah, sorry I was just trying to back up your point that they pick a particular bandwidth, which creates an effective limit, and for colour composite those are the particular discrete approximations they much later picked based on the bandwidth limits they had earlier imposed. So, some concrete numbers to back you up. – Tommy Oct 17 '17 at 15:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.