35

While early microcomputers used analog video outputs (often to use a television as a display), higher end machines such as the BBC Micro or Commodore 128 supported a digital RGB (or RGBI) video output. In the IBM-compatible world, the CGA and EGA display adapters likewise had digital outputs.

When the VGA standard landed in 1987, it was analog, rather than digital. VGA and analog output held sway for well over a decade or two, but then DVI, DisplayPort and HDMI took video outputs back to digital.

Why did video standards abandon digital in the first place, and then return to it several years later?

  • 3
    The answer to "why RGBI in 1981 instead of jumping straight to RGBHV" is probably going to be something boring like, "because memory was expensive and there wasn't a market for >16 colors that justified adding more wires to the monitor cable". – snips-n-snails Jul 22 at 19:33
  • 8
    In the late 90's the "urban myth" anecdotal reason for this was given was that analog monitor cables gave off RF noise that was easily surreptitiously scooped up and read by surveillance teams, and that the same nonexistant nonpersons not responsible for not scooping up data from telecom providers in the modern era weren't responsible for not pulling strings to ensure this became standard. – Darren Jul 23 at 3:32
  • 8
    I think it's more like video outputs were parallel up until DVI, when they transitioned to serial. Even RF video is parallel in signal terms: the various things are folded into different ranges of the frequency spectrum but are then all occurring at once. – Tommy Jul 23 at 16:20
  • 1
    @R..: I think it does, that might be why it's an urban myth that analog was problematic. rtl-sdr.com/… is a DIY tutorial / article about building a TEMPEST snooping system with software-defined radio that can snoop DVI or HDMI signal leakage. (With an antenna in the same room.) With compressed video (like digital TV broadcasts), you need to capture it near-perfectly to get anything, but monitor signals are uncompressed. – Peter Cordes Jul 25 at 1:24
  • 2
    Just as a side note, some DVI ports supplied both Digital and Analog signals – Smock Jul 25 at 10:56
33

Early digital video outputs, like CGA, were not really akin to the later standards such as DVI and its follow-on's. The reason for using multiple lines to carry the different analog portions of the signal to the monitor was to prevent crosstalk interference of these signals.

You can see this in the very early computers like the Commodore 64 and Atari 800 separating the luminance and chrominance signals to produce a cleaner display. Similary, CGA (RGBI) separated each color component from the sync components in order to deliver a cleaner version of all those signals to the CRT, minimizing crosstalk. VGA, which was not limited to only 4 bits for representing color information, continued this approach of separate signal lines for color and sync.

Remember, within the CRT, there are only 3 color signals possible for controlling the 3 electron guns that will produce RGB on the display phosphors. So when VGA took the technology from CGA's 2^4 possible colors to VGA's 2^18 possible colors, it was still only necessary to separate the R from the G and the B signals. There would be no point to using 18 separate lines to carry the 18 bits of digital color data to only 3 electron guns in the CRT.

So, the more essential technology change was the move from CRT's, which are analog peripherals, to LCD screens, which incorporate more digital processing in controlling their individual pixel intensities. With LCD's, obviously there are no purely analog electron guns assigned to the primary colors. This fact, coupled with excellent advances in the bandwidth of serial communications channels, allows modern displays to carry the full digital display signal in its computer-native form to the display's processor.

  • 20
    @Pascal I think you are mistaken. The most common form of color CRT used a shadow mask and three electron guns to achieve color. electronics.stackexchange.com/questions/350071/…. – James_pic Jul 23 at 13:36
  • 4
    @James_pic I think I mistook a schematic drawing in one of my old technical video-books. For pratical reasons it makes sense to use three of them. My bad. Leaving the original comment to not break the flow. :-) – Pascal Jul 24 at 6:51
  • 2
    Many LCD/LED monitors still have VGA input for backwards compatibility, and presumably convert the analog signal back to digital inside the monitor. My first LCD monitor had only VGA input, negating the digital benefit altogether... – Jonathan Jul 24 at 11:33
  • 3
    I can't imagine any way CRTs of the time could have used a single gun. They didn't have that level of beam positioning control, and had continuous-variable adjustment knobs that would have precluded lining up with a fixed grid without some fancy closed-loop control going on. – R.. Jul 24 at 13:03
  • 2
    And indeed the answer to the question @James_pic linked involves some fancy closed-loop control for the rare model that did operate with one beam. – R.. Jul 24 at 13:05
21

Before VGA was invented, CGA RGBI used 4 wires to get 16 colors and EGA used 6 wires to get 64 colors on the cable. Add 3 more for sync signals and signal ground, and this won't be an issue, simple cables and connectors exist for getting EGA's about 16.3 million pixels per second digital signals over to the monitor easily.

Amiga was released. It supports 4 bits per color component, or 12 bits of color for 4096 colors. It can be connected to a standard color TV directly (via SCART analog RGB), and the RGB is also used to generate color composite and RF outputs. It also has digital RGBI interface, but only 16 colors so why use it, analog RGB is much better, and even composite with single RCA connection is tolerable.

Enter VGA. You have up to 8 bits per pixel at about 28.3 million pixels per second. Digital video links such as LVDS or SDI have not been invented yet, because there has been no need for them yet. But RAMDACs exist. What any sane designer would do is to add one of these to offer a color look-up table as well, and to generate 6 bit output per component, or 18 bits per pixel, directly converted to analog RGB. This means no huge amount of digital wires are needed, just 3 coaxial cables for analog video, plus 2 wires for sync. Plus the monitor can just eat analog signals as digital signals have to be converted to analog anyway for driving CRT guns. This just made no sense to connect VGA digitally, or the number of colors or color palette lookup feature would have to be discarded.

Besides VGA, others use analog connections too (SUN/SGI, Apple..). Resolutions increase, refresh rates increase, color depths increase. Transmitting high bandwidth analog video needs more precision and engineering to have acceptable quality. Meanwhile, LVDS was invented, finds its use in LCD panels as a digital video link. Laptops need LCD panels so they have LVDS chips after color look-up table, so LVDS gets eventually integrated to chipsets. Finally CRT monitors start to be replaced with LCD displays. As the source and display device are digital, the high quality analog link between them could be eliminated cheaply with digital interface to avoid extra conversions. First using the laptop-based LVDS technology in various forms, but those were just external versions of internal laptop display links, so very early on TMDS-based DVI link replaced these. So more bandwith and quality is much more easily and cheaply gotten over digital link than doing analog conversions on both ends, since they are not even necessary. The rest is history.

  • 2
    "As the source and display device are digital, the high quality analog link between them could be eliminated cheaply" and yet in practice Digital interfaces were an extra feature, generally delivering marginally better quality. Analog VGA remained the baseline option for many years and to some extent it's still the baseline option even today. – Peter Green Jul 24 at 15:36
  • 2
    VGA uses an analog signal level for each color. Some video cards support 30 bit color, 10 bits per color per pixel. – rcgldr Jul 24 at 19:34
11

Older monitors had analog timing and, with one notable exception, were designed so that the signals presented on their inputs at any moment would fully describe the color to be displayed at that moment. The only monitors that used any sort of time multiplexing on their inputs were those that used the same analog composite color encoding methods used in broadcast television receivers (and typically used much of the same circuitry to decode the signals).

Older "digital" monitors would need to send video signals over enough wires to identify every possible color as a binary number. For a 64-color monitor like an EGA display, that required using six signals, and for a 262,144-color monitor like the VGA it would have required 18--too many to be practical. Using an analog RGB or YUV (today called "component video") signal reduced the number of signals required to three without imposing any constraints on pixel timing.

Newer standards like HDMI require sending high-speed serial bit streams with many bits of data per pixel. That in turn requires higher data rates and thus faster switching rates than could be accommodated with older technologies.

8

That question is built on somewhat weak ground. After all, most of the video formats you list as 'digital' aren't such - or at least not more as any of the analogue ones are. Just because an output like RGBI uses two levels per channel doesn't make it digital. They only feature a restricted number of signal levels.

Digital signalling only started with formats like DVI/HDMI/etc., where signal data was not transferred as levels but a data stream. Everything before isn't digital.


Now, the decision to use RGBI is one about saving components to encode the signals by some complex modulation (like FBAS or SVIDEO) just to add even more components to the display to decode it again.


And yes, as snips-n-snails suspects, it's all about the amount of data - just not so much about memory bandwidth but rather transfer bandwidth. Using multiple lines increases bandwidth without increasing data rate. Increasing values per step (and lane) also increases bandwidth without increasing signalling steps.

Here also hides the reason why it took so long to switch to real digital transfer - analogue already delivered an awesome amount of bandwidth. But with increasing bandwidth even further, hard-to-compensate signal distortion grew as well.

Going digital offers the solution by separating transfer from content. Distortion on transfer level are compensated or corrected) without influence to the data (and thus the image) transferred.


Addendum:

Bruce Abbot made an incredibly good point about using the terms "analogue" and "digital" by citing Wikipedia's definition of a digital signal:

"A digital signal is a signal that is being used to represent data as a sequence of discrete values; at any given time it can only take on one of a finite number of values."

Using this definition all of the mentioned formats are digital, as all of them only produce a finite number of values and only one of them at any given time. For example take VGA. By having 256 (=finite) values for each colour and three colours (still finite) and only one of these combinations is output for a pixel (at a time), it exactly fits this definition.

Taking this definition seriously, and it's not a bad one, then no computer generated output is ever analogue. It will always be a series of finite levels over time.

If that's valid, it may seam that this is just swapping the point above from before DVI everything is analogue to everything is digital, doesn't it? At least meaning that either way, there was no switching back and forth.

  • "A digital signal is a signal that is being used to represent data as a sequence of discrete values; at any given time it can only take on one of a finite number of values." - en.wikipedia.org/wiki/Digital_signal – Bruce Abbott Jul 23 at 0:28
  • @BruceAbbott You're making an interesting point. By using that definition all of them are digital, as all of them only produce a finite number of values and one one of them at any given time. – Raffzahn Jul 23 at 7:07
  • 3
    VGA is considered to be analog even though the RGB outputs have discrete levels, because the digital color values are converted to voltages by DACs, and the (analog) monitor makes no attempt to extract the original digital color values from those voltages. SVGA has no inherent limitation on the number of levels. At 256 levels or more per channel it is effectively pure analog. – Bruce Abbott Jul 23 at 7:52
  • @BruceAbbott Oh, so your comment wasn't about applying the wiki definition? Sorry for misinterpreting. Nonetheless it does make a great point. "considered to be" is equivalent to "by convention", usually avoiding logic. Further, the number of levels does not make it analogue in any way - as having levels is the very basic definition of digital. That's true for 256 or any higher number (like on SVGA) as long as it's a finite one. And it doesn't become analogue just by calling it that. Or does sloppy naming cancel logic? – Raffzahn Jul 23 at 8:23
  • In physics all voltages are 'digital' at the quantum level, but are treated as analog when large enough that the individual quantum levels cannot be resolved. In electronics the output of a digital to analog converter is considered to be analog even though there are a discrete number of steps. IBM VGA technical reference says "The circuitry that provides the VGA function includes a video buffer, a video digital-to-analog converter (DAC)... The video DAC drives the analog output to the display connector". – Bruce Abbott Jul 23 at 20:44
3

from digital to analog, then back to digital?

Simple answer:

It didn't go "back" to digital.

Why?

Although the EGA and the HDMI signal are both "digital" and the VGA signal is "analogue", the (digital) EGA signal is quite similar to the (analogue) VGA signal but it is completely different from the (digital) HDMI signal:

Both in a VGA and in an EGA signal the voltages of electrical signals directly reflect the color which is displayed at the position where the three electron beams of a CTR tube monitor currently point to.

  • EGA uses 6 digital signals for the color, two for each electron beam

    This means that 2^6 = 64 different colors are possible

  • CGA uses 4 digital signals for the color

    This means that 2^4 = 16 different colors are possible

  • VGA uses 3 analogue signals, one for each electron beam

HDMI uses 4 digital signals. If HDMI worked like EGA, 16 colors would be possible using 4 digital signals.

However, HDMI works completely differently:

The voltages of the four signals measured at a certain time do not directly reflect the color of a certain pixel.

Instead, the HDMI signal is more comparable to an Ethernet signal while you are streaming a video over the network: Data bits of a video file are transferred over the HDMI wire and some electronics in the monitor will perform the video file decoding.

The analogue "composite video" signal (for TV sets) works differently than all of the signal types I described above.

Why did computer video outputs go from digital to analog

In the 1980s, both analogue and "EGA-like" digital signals were used.

However, around 1990 higher color resolutions became standard. For 256 colors we would require 8 wires in a "EGA-like" digital signal; and for 16M colors we would require 24 wires. And of course we would need to change the monitor because a "256-color monitor" has only 8 input wires...

A VGA signal requires only 3 wires; and because a VGA monitor has an analogue input, it is theoretically able to display any color depth, so we don't have to change the monitor when changing from 256 to 16M colors.

... then ... to digital?

With higher resolutions "VGA-like" signals will have problems. Just one example: The frequencies of the signals rise and you will have signal crosstalk between the wires.

Using an "Ethernet-like" signal (such as HDMI) was the solution for these problems.

In the early 1990s such a technology would have been very expensive (unless it should only support a very low video resultion). For this reason nobody was thinking about such a technology.

  • Good write up, except the point about technology not being able in 1990. Of course it was. What was missing was the need to use it. Up to resolutions used back then a digital transport wouldn't bring any advantage, but additional cost for en/decoders. System switches only can happen if the new one offers a 'killer feature' like back when otherwise inferior petrol cars replaced electric due their ability to 'recharge' within minutes. – Raffzahn Jul 23 at 8:28
  • 1
    @Raffzahn I edited the last part of my answer. – Martin Rosenau Jul 23 at 9:37
  • Martin, great, except, it wouldn't be very expensive at all. The additional cost would be rather acceptable. A SVGA display would need only like 300 MBit transfer rate. Not a big deal in 1990s technology (remember, we talk short distance of 50 to 150 cm, not long distance network). But it would't bring any improvement at all to the user. It was the pressure on analogue transfer due ever increasing screen size that made the switch to digital viable. – Raffzahn Jul 23 at 9:46
  • 1
    @raffzahn 300 Mbps on an external interface would have been a big deal. That's higher bandwidth than Ultra SCSI with its 50-pin plug, higher than even most PCI busses, and those are internal. USB 2.0 didn't arrive until 2001, and "Full-Speed" 1.0 was 12 Mbps. DVI was the first external serial interface with enough throughput to deliver even VGA resolutions. – clacke Jul 24 at 0:01
  • @clacke You're mixing up different technologies and targets, so this comparison doesn't really work. But more important, you're missing the point by just mentioning implementations. Each of them has a target to deliver and only works in that context. USB wasn't about delivering maximum possible speed but an easy plug and play network. Similar, DVI wasn't because finally transmission technology was able to do the speed, but because at that point in time the existing technology (VGA) hit its limits. And switching to pure digital was less expensive than continuing to improve it. – Raffzahn Jul 24 at 0:10
2

I believe it was cost and terminal emulation. The RGB monitors were very expensive at the time and would not work (easily) as dumb terminals. IBM wanted a "low" cost display that was double the resolution of CGA and that could be used to emulate their mainframe 3270 terminals. EGA was a temporary stop-gap design while work continued on VGA - 640x480 graphics and sharp 80 col x 25 row display in "B&W" and color. Lotus 123 never looked so good as did the 3270 display on a VGA screen.

Business PCs sales really took off when VGA hit the streets. You could now use one display to do PC work and emulate an IBM mainframe 3270 terminal. Dec VT100/200 emulation and Burroughs (Unisys) B6000/A10 TDI emulation soon followed. VGA was the key to having only one terminal at your desk. I was in banking in Chicago at the time and we couldn't get enough IBM XTs with VGA and 3270 cards. We made a lot of money using those 640x480 VGA monitors for PC work and dumb terminals. Networks based on Ethernet were not mainstream yet - Mainframes paid the bills.

1

This all was driven by cost. The vertical and horizontal synchronization signals as well as the separate RGB signals had to be combined on the graphics card if using composite video. This greatly complicated the card and ran up the cost. It also complicated the graphics display as the display had to use circuitry to separate out the component parts of the signal inside the monitor. As wideband low cost video amplifiers were an oxymoron at the time the reduced complexity of the already expensive graphics card and reduced component count of the monitor which did not need as many parts to parse out a composite video the engineers was a win-win. It was simply simpler to make a cable with extra wires and less expensive.

  • Thanks for the answer. When you say "at the time", are you referring to the switch from digital to analog, or the later switch back to digital? – Kaz Jul 25 at 19:00
  • I can appreciate how a composite sync requires a more complex display, but the same needn't be true of the computer. The BBC Micro used a 6845 CRT controller chip that generated separate horizontal and vertical sync; it simply combined them using a NOR gate to generate an (inverse) composite sync signal. – Kaz Jul 25 at 19:05

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.