20

It is well known that when emulating classic games on modern displays, you need to be careful not to distort the aspect ratio. Generally speaking, CRT screens were 4:3, and specific pixel aspect ratios of many classic arcade games, consoles and computers are given here.

Of course, title safe area was a concern, particularly for home computers that had to work on many different models of TV sets. It's hard to say exactly what was the title safe area for NTSC TV, but the Commodore 64 is generally accepted as having come pretty close to it. From the above table, 320 pixels / 8.19 MHz = 39 microseconds, which sounds in the right ballpark.

NES: 256 pixels / 5.37 MHz = 47.6 microseconds. That is significantly past the title safe area and into overscan, but that's fine, that was a feature for a dedicated console; it matches one's memory of playing Super Mario Brothers, where objects would scroll onto the right of the screen without any visible border.

Missile Command: 256 pixels / 5 MHz = 51.2 microseconds. Hmm. That is even considerably further into overscan. Resources are being wasted generating pixels the player will never see...

... unless. That's an arcade machine. It came with its own screen. Sure, TV picture tube minus the tuner, 4:3 overall aspect ratio. But maybe they could have tweaked the display electronics to shrink the horizontal scan, so more of that 51.2 microsecond active scan line is visible.

But in that case, the calculated pixel aspect ratio of 27:22 would be no longer valid. And in general, it would be harder to figure out exactly what was the pixel aspect ratio of arcade machines. Calculating it from the pixel clock in the same way as for home computers and consoles would no longer give quite the correct results.

Is the above conjecture correct? Or did arcade machines actually use the same parameters as TV sets, and did Missile Command just accept some extra wasted pixels?

  • 2
    Key quotation from cited article: "Some coin-operated machines use a different active scanline period, and the arcade operator is expected to adjust the monitor's picture size to match a test pattern displayed in a particular game's service mode. " – Damian Yerrick Mar 24 at 19:15
35

CRTs don't have pixels, they don't work that way. Also, arcade monitors expose all the picture controls at the back so it is possible to adjust them quite extensively. Operators would have made sure that the picture was the right shape and in focus near the edges.

Since the controls are all analogue and no-one bothered to measure the display geometry the aspect ratio was just close to 4:3, basically whatever looked good to whoever set it up.

The picture controls adjust the width of the beam scanning area, for example, so the pixel clock is irrelevant. The owner just turns the knob until the picture covers most of the screen. Games were designed with this in mind and the horizontal resolution was just that - resolution, not an attempt to create a particular aspect ratio.

with thanks to Tommy, here is an image from an actual CRT. In the upper right there is a zoomed in area from Sonic's foot, which shows some of the vertical bars that people often assume are pixels are in fact not evenly lit. The tops and bottoms of some of them are at different brightness levels.

enter image description here

The bars you can see are actually the shadow mask, and are unrelated to pixels. They define the maximum horizontal resolution that the CRT is capable of resolving, but even then the relationship is not 1:1.

  • 3
    Drats, I should have waited instead of writing many words. Good short version here. Except, of course are games created with a certain ratio in mind - the ship of defender must be a sleak fighter, not a bumble bee. But you are right, that these ratios are more of a guideline than the strict laws a CAD application will prefer. – Raffzahn Mar 22 at 13:27
  • 7
    "CRTs don't have pixels, they don't work that way. " What does that mean? The screen, at least on color CRTs, is separated into discrete regions. – Acccumulation Mar 22 at 14:44
  • 5
    @Acccumulation see youtu.be/Ea6tw-gulnQ but basically what you see on the TV is the result how it produces colour via different coloured phosphors and masking, not pixels. – user Mar 22 at 14:55
  • 6
    @Acccumulation The red, green, blue phosphors, the discrete regions, on a colour CRT don't have any correspondence to the pixels being generated by the arcade machine's video circuitry, and it's these pixels that the original poster is talking about. – Ross Ridge Mar 22 at 17:56
  • 1
    @Acccumulation the "regions" are just holes in the shadow mask. That video explains it better than I can here, but they are not pixels and don't have to be uniformly lit. The beam intensity is analogue and the limit on resolution is more to do with the limitations of the electronics than the "regions", which are just holes in the shadow mask. There is no correlation at all between them and pixels though, unlike an LCD monitor. – user Mar 24 at 12:39
14

Did arcade monitors have same pixel aspect ratio as TV sets?

Short answer: No, not necessarily.


Long Answer:

  1. To start with, 'Title Safe Area' is an idea to define the parts of one transmission to be displayed even if any of the many receivers is maladjusted. It's nothing inherent to the TV signal or its definition, it's a safeguard against less than correct receivers.

  2. Next, Pixel ratio is, even on a TV, complete arbitrary. TV sets a frame timing, but not a pixel one. The number frames per second and number of lines within a frame is defined by the respective standard, but that's it.

  3. The number of pixels per line is, even on TV, infinite. Of course, as they get defined 'tighter', as more restrictions due transmission and electronics appear. It's up to the application what restrictions are acceptable or not. In general, colour resolution on TV is less than luminescence resolution.

  4. When it comes to the tube, none of these standards set any numbers. While they are made to fit certain standards, a CRT offers complete freedom, not only for pixels per line (as in TV) but also for number of lines per frame and frames per second - as long as it's within the CRTs abilities.

  5. Parameters are set here mostly by screen mask (for colour games) resolution, phosphor used and electronics used for deflection.

As a result, even by using standard CRTs and average electronics resolution and pixel ratio can diverge quite from what a TV supports - and in both cases Pixel resolution is arbitrary due the fact that signals are analogue and not bound to any standard.

Pixel ratio in video games is thus defined by the generating computer system video circuit (assuming a correct alignment of the installed game/CRT). If a game (hardware) developer chooses a format of 512 lines by 1024, then that it is. Depending on the arcade cabinet setup (screen covers, etc), this may end up in any arbitrary ratio, not just 2:1 or 8:3. Being freed from TV specs is a great thing, isn't it :))

  • 2
    I wouldn't say that the pixels per line is infinite - more that while there is a limit of the amount of detail that can be discerned on any TV, framing it in terms of pixels is meaningless for analogue TV standards. While you can sometimes make approximate comparisons with digital resolutions in terms of how accurately you can describe detail, these are just that - approximate comparisons. Exactly how accurate they are will depend on the thing you're displaying. – Muzer Mar 22 at 17:43
  • 1
    @Muzer Well, that's the 'restrictions apply. If a TV can do like 200 lines (thats 400 pixel) wit the full dynamic of black to white, it can do 800 pixel with the restriction that each neighboring pixel may not differ by more than 50% luminosity and so on - and if that reminds anyone of the Amiga, yes, it's much like the restriction of its (in)famous HAM mode regarding colour change. – Raffzahn Mar 22 at 19:12
  • Per Nyquist's sampling theorem, the number of "pixels" in a fragment of a baseband analog signal is its duration times twice its highest frequency. For NTSC composite video, usable luma bandwidth is 3 MHz (with chroma at 3.0-4.2), but 240p/480i component (such as arcade RGB) can hit 8 MHz. – Damian Yerrick Mar 24 at 19:26
  • 1
    @DamianYerrick Nop, that's the minimum sampling frequency needed to sample such a signal. Not related to pixel or alike. – Raffzahn Mar 25 at 1:54
  • "luminance" is the normal word for the intensity/brightness part of the signal (as opposed to chrominance), not "luminescence". The dictionary definitions are "the intensity of light emitted..." vs. "the emission of light by a substance that has not been heated". So luminescence just describes the fact that light is being emitted at all because electrons are hitting phosphors (en.wikipedia.org/wiki/Cathodoluminescence), not specifically the intensity. – Peter Cordes Mar 25 at 6:38
7

Asteroids is an example of an arcade cabinet that didn’t even use raster graphics, but vector graphics. Battlezone and Lunar Lander were others.

They used similar technology to the Tektronix 4000-series terminals of the ’70s, or the IBM 2250: A cathode ray fired into the back of a glass screen coated with phosphors, like in an old-fashioned black-and-white TV, but instead of moving across each row of the stream in a regular pattern and switching on and off, it instead followed a more complex path to draw geometric line art.

  • 1
    Interesting - but you should expand this to explain what the effective resolution in lp/mm was - that's about as close to "pixel ratio" as you can get with an analog display – Carl Witthoft Mar 22 at 18:00
  • 1
    @CarlWitthoft You might enjoy: youtube.com/watch?v=FkHjG759ABY – Davislor Mar 22 at 18:08
  • 1
    Battlezone and Lunar Lander were also vector graphics games. The IBM 2250 and the CDC 6600 console were vector graphics terminals. – rcgldr Mar 24 at 18:24
  • @Davislor Cool -- but I was distracted by the 1-second clip, badly out of focus, of a pingame. Now I'll waste hours on ipdb.org to figure out what game that was. – Carl Witthoft Mar 25 at 12:38
  • @CarlWitthoft - there is no "effective resolution" of such a display, as the monitor itself is analog in both axis, much as a raster CRT is analog in the horizontal one. The calculations and graphics driver hardware would have a resolution, but not the monitor. – Chris Stratton Mar 26 at 22:43

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.