10

Many early computers used TV sets as monitors. With an NTSC TV set, you could really only count on about 200 scan lines of vertical resolution, and for horizontal resolution, maybe 192 color clocks at a most optimistic reckoning, or maybe 300-400 pixels usable horizontal resolution if you're willing to accept some color artifacts, the upshot being a TV set could do 40x25 text but not much better than 40 columns and definitely no more than 25 rows.

I get the impression some early computers could get a sort of graceful degradation, could do better than the above while staying in the NTSC ballpark, taking advantage of a monitor if available and falling back to TV otherwise, by generating composite video, displaying this on a monitor if available, otherwise converting it to RF for a TV display.

Presumably you still have to stay within overall NTSC parameters, but that's nominally 240p (when you don't use interlacing); could you that way get closer to being able to actually use 240 scan lines?

Similarly, how high could the practical horizontal resolution get in that scenario, if you happened to have a composite video monitor plugged in? High enough for 80 column text that you could actually use all day without eyestrain?

  • 1
    I think 160 horizontal color clocks would be more accurate. At least if you wanted each horizontal pixel to be a different color with minimal color bleed (artifacts). – cbmeeks Jul 5 '17 at 17:35
  • 1
    Another reason that early computers didn't generally do interlaced mode is simply storage: if each line is unique then obviously the interlaced version requires twice the footprint. Text modes are an exception if the characters are in ROM and the manufacturer can afford double the storage there but are also provide an interesting exception to the non-interlaced pattern: on the BBC Micro by default all of the pixel modes were non-interlaced but the teletext mode was interlaced. And on most TVs, teletext output was interlaced. Though it's unlikely that built-in receivers would use composite. – Tommy Jul 6 '17 at 0:38
  • 1
    Soviet microcomputers used at least 512x256 black and white and 256x256 color. – Leo B. Jul 6 '17 at 5:41
  • 1
    I remember my friend's ancient tube-type color set with a significantly rounded bezel would slightly cut off one of the corners of the VIC-20 screen (maybe 1/2 of a character). A lot of games included centering adjustments, though I don't know how often they were actually needed. – supercat Jul 11 '17 at 17:21
  • 1
    Additionally, the VIC-20 is able to output either interlaced or non-interlaced video. Interlaced video is 59.94fps; non-interlaced is about 59.8fps. Non-interlaced generally looks better because the chroma phase flips every frame rather than cycling through a 4-frame pattern, but on sets with poor power-supply regulation, non-interlaced video makes the screen strech and warp slightly at a rate of about 11.5 times/minute. Switching to interlace mode reduces the shrink/swell speed to about 3,.5x/minute. – supercat Jul 11 '17 at 17:29
7

NTSC provides 227.5 colour cycles per line; PAL is very close to 283.75. In both cases, the visible area is around 80% of the line, but most home computers had a much bigger border than that — e.g. (of those I know offhand) the Acorn machines paint for 40µs, which is 62.5% of the line; the 48kb Spectrum paints for 128/224ths, which is around 57%; over in NTSC land the Atari 2600 paints for close to 70% of a line.

There's a useful caveat though: colour cycles per line is a corollary of the normal line period and the colour subcarrier frequency. Since TVs apply some tolerance in timing you can do what at least both Atari and Apple did in NTSC: provide a slightly longer than usual line, such that per the colour subcarrier frequency it's exactly 228 colour cycles long. That way you can paint a decent number of individual colour pixels and have the effects of colour banding be uniform down the screen and from frame to frame.

In Atari's case at least, that amounts to populating 160 individual colour cycles†.

If anybody had done the same thing in PAL then they'd probably have stretched to a line long enough to contain 284 colour cycles, and painted between 170 and 200 of them. Though the phase alternating part of PAL means that you wouldn't get the same consistency vertically, which is presumably why nobody did. The PAL Ataris just end up no longer being aligned with the colour subcarrier.

That all being said, if you're asking about horizontal resolution in the abstract then don't discount the option of just doing it in black and white. Colour TVs are supposed to suppress any decoded colours if there was no colour burst — that's going back to the transition period when some channels were broadcasting in colour and some weren't, and you didn't want false colours showing up all over the place on those that weren't and which also weren't pre-filtering their signal. On some TVs that might cut the low-pass filter out of the loop. If so then you can do a lot more pixels and expect them to show up cleanly.

Ditto if displaying on a cheap black and white TV as the colour subcarrier frequencies were picked so that a combination of phosphor persistence, persistence of vision and psychology make the colour part of the signal hard for a human being to detect even if a pre-colour TV, with no concept of distinguishing the colour part of the signal, is tuned to a colour broadcast. Therefore you could send one of those a relatively high-frequency signal and expect it to be rendered without deliberate filtering: as a rule of thumb, you can multiply the horizontal resolution by four††.

The Apple II is a machine that will omit a colour burst in text mode to get that benefit on compatible televisions. Some very old machines, such as the ZX80 and ZX81, are purely monochrome so don't output a colour burst at all. ZX BASIC could well look better on a ZX81 than a ZX Spectrum.

So, the numerical answers: 160–200 colour if you want the nicest possible artefacting, 320–400 if you don't mind some artefacts, more like 640 or so if you're in black and white and are lucky with the TV set.

† actually the Atari outputs either 160 or 152 pixels per line because the way it handles object movement means that it sometimes omits the left-hand eight pixels. You'll have seen it in many titles as short black horizontal lines on the left margin that move around as the sprites do.

†† the hand-waving version: if the colour subcarrier is n Hz then per Nyquist you need to capture 2n samples fully to describe it, but colour in NTSC and PAL is encoded in quadrature so there are actually two parts to the colour signal encoded exactly out of phase, which makes 4n. But more persuasively, sampling at four times the colour subcarrier is the official industry recommendation for digital preservation of archive composite video.

9

It depends on whether the TV is color or black-and-white/monochrome. Older B&W TVs (and ordinary TV electronics and tubes converted into monochrome monitors by some el-cheapo monitor vendors), did not block (filter out) color burst frequencies (with the associated IQ bandwidth), and the cheap analog filtering did gracefully degrade as the bandwidth was pushed.

Most old TVs had vertical and horizontal overscan pots (variable resistors) for adjustments inside the back. The edges of the TV display rectangle defaulted to being hidden behind the TV bezel. By using the overscan adjustments to shrink the image, one could fit more width and height (in pixels) on the visible portion of the display within the TV bezel cut-out (and distortion limits).

So the actual usable resolution could vary quite a bit by how the "TV" was set up or adjusted.

280 x 192 pixels was a safe sub-region for a non-specially-adjusted/off-the-shelf analog color TV. Tweak the pots, and one could display more. Using a B&W TV set or tube with modified electronics, even more pixels were possible. But eventually the contrast would roll-off, with the vertical limit set by the inductances of the yoke on the neck of the tube. (High-resolution CRT monitors use custom magnetic yoke designs, very different from the ones manufactured for consumer TV sets).

IIRC, old electronics magazines (circa 196x to early 197x?) had articles on how to mod a B&W TV's electronic circuit chassis (tube or discrete transistor) for use as a monitor or an XY oscilloscope.

6

I'm not sure if I understand the question correctly.

TV composite is bandwidth limited because the color carrier signal blocks higher frequency. This means one particular type of signal cannot "gracefully degradate": If the horizontal resolution is too high, you'll just get colored jumble on the TV, and not somehow a fuzzy image in lower resolution.

RF modulation has nothing to do with it: It's the composite signal itself that is limited that way, and the fact that many color TVs didn't have a composite input and one needed to modulate the composite signal itself onto a RF carrier like for broadcast TV doesn't matter it this respect.

Would it have been possible to use the NTSC timing, and display the full 240 lines on a monitor? Certainly, but that wouldn't make a lot of sense, because with a different timing one could display even more lines.

Composite monitors could easily do 80x25 (with 9x14 dot matrix, IIRC) at the beginning of the homecomputer era (e.g. Apple II 80-column video card). That signal wouldn't display on a TV, at all, even though it used the same RCA plug.

Monitors for business-oriented computers were capable of doing this or better long before. The Xerox Alto did 606x808 bitmapped graphics in 1973.

  • Right, but computers like the Xerox Alto got high resolution by accepting the need to always use a monitor. I'm talking about computers that can still use a TV if that's all you've got. And you can have graceful degradation up to a point. Something like 500-600 pixels of horizontal resolution will, I think, display on a TV, albeit blurry. – rwallace Jul 5 '17 at 18:06
  • As I said, I didn't understand the question - the Alto is an example for an actual resolution of monitors (which is the title). The main problem isn't the horizontal resolution, it's the vertical resolution. You need a different timing for that, anyway, and you also need the color burst part for the TV, but not for a composite monitor, etc. So hardware that would do both TV and monitors was not frequent, and would use different modes. – dirkt Jul 5 '17 at 18:22
  • Well, the Commodore 64 did both, and seems to have used the same mode. – rwallace Jul 5 '17 at 18:34
  • 1
    "That signal wouldn't display on a TV" is outright false. If you set the timing parameters reasonably close, a single signal pin (R, G, or B) of the VGA connector, tied to the hblank and vblank pins, can be fed directly into the composite video of an NTSC TV (as mono) and the result has at worst minor rainbow artifacts. I've personally done this, as well as hooked up all three channels of the same VGA output with an NTSC color encoder chip. – R.. Jul 5 '17 at 22:39
  • @R..: I wasn't talking about talking about PC-type graphics cards, but the Apple II 80 column card and similar ones, where you couldn't set the timing, which is why it wouldn't display on a TV. I know perfectly well that if you can program modes and match the timing, you get a picture on a TV, but that wasn't the issue. – dirkt Jul 6 '17 at 6:45
3

"...i was using my A500 with a amber composite monitor. Terrible monitor for an amiga, except for interlace mode which had almost no flicker."

I don't know which particular monitor was referenced above (perhaps a Zenith ZVM-122-A), but the Commodore 1080 (link2) and 2080 (link2) color monitors also had long-persistence phosphors:

The disadvantages of interlacing include an annoying flicker when displaying 400 or more interlaced lines... The problem is further compounded by the high contrast in which computers display objects; this contrast makes the slower refresh apparent to the human eye.

The only way around this problem is to use a monitor with long-persistence phosphors. However, these special phosphors in themselves create problems: a displayed image may smear as it moves on the screen, and the slow phosphors usually make the overall image dimmer.

This helps explain the earlier quote about how an amber composite monitor, besides being monochrome, was a "terrible monitor for an Amiga" (a multimedia/games machine).

Interlaced mode on an A500 is 400 (NTSC) or 512 (PAL) horizontal lines of resolution, and 320 (LowRes) or 640 (HiRes) vertical lines. A composite connection limits the color bandwidth, so in summary, I would say that the maximum useful resolution of a composite NTSC video monitor designed for productivity is about 320×400 pixels (color) or 640×400 pixels (monochrome).

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.