11

I've just realized I have a gap in my understanding of the VGA CRT monitors of the 80386 era (1980s-1990s).

  1. What was the phosphor resolution? It turns out I did not have proper understanding of the shadow mask and having discrete R G B phosphor dots. But I also remember my monitor had analog controls to resize and move the image and possibly even do pillow correction. For that to work, the shadow mask resolution had to be significantly higher than the monitor image resolution. Is that correct? Did monitors typically use hole-type masks or slit-style masks (that could give infinite vertical resolution).
  2. How was control over the electron beams distributed between the monitor and the video card? Which part controlled the following aspects?
    • Start X,Y of the first scanline
    • Scanline width (was it always the same for all scanlines?)
    • Time to draw a scanline
    • Vertical distance between scanlines (was it always the same for all scanlines?)
    • Number of scanlines
  3. Could a (custom) video card technically output image with arbitrary resolution (up to some physical limit), say 1x1 to 1000x10 to 1000x1000?
1
  • 1
    It needs to be remembered that VGA monitors were adapted from standard TV monitors, and had most of the same controls and characteristics.
    – Hot Licks
    Dec 15, 2020 at 15:28

4 Answers 4

17

What was the phosphor resolution?

The usual specification quoted for CRT resolution is the 'dot pitch', or distance between groups of RGB phosphor dots. Here are some examples of IBM monitors produced from 1987 to 1993 (taken from here):-

Model Year Standard size viewable pixels   dot pitch (mm)    notes
8513  1987   VGA    12"   10.4"   640x480    0.28
8512  1987   VGA    14"   11.8"   640x480    0.41
8511  1992   VGA    14"   11.8"   640x480    0.39
9518  1992   VGA    14"   12.3"   640x480    0.28
8514  1987   XGA    16"   14.4"  1024x768    0.31
9515  1992   XGA    14"   12.3"  1024x768    0.28
9517  1992   XGA    17"   14.7"  1280x768    0.26 (stripe)  Trinitron
6314  1992  SVGA    14"   12.0"  1280x768    0.28
6318  1993  SVGA    14"   12.1"   800x600    0.39           "Low End"
6317  1993  SVGA    17"   14.7"  1280x1024   0.28

To fairly compare the dot pitches you also have to take into account the screen size, so for example a 14.4" diagonal (viewable) screen with dot pitch of 0.31mm should be equivalent to a 12.3" (viewable) screen with 0.27" dot pitch.

That's not the full story though, because actual resolution is also affected by beam size and video bandwidth.

The monitor doesn't need to know where each phosphor dot is because (unlike LCD screens) they are not lined up with pixels, there are simply (hopefully) enough of them that the beam will always land on at least a few dots.

However if the beam is narrowly focused relative to the dot pitch then single pixels may not be able to illuminate all phosphor colors equally, depending on where the beam lands relative to the dot triads or stripes. For smooth color rendering the dot pitch should be much smaller than the beam diameter. However a high resolution may still be achievable with large dot pitch by using a finer beam, at the expense of color rendering fidelity (pixels get fringes of different colors that vary depending on their exact location on the screen, and noticeable Moiré effects with fine pixel patterns).

CRTs with higher vertical resolution also need a narrower beam to separate the horizontal lines. However at lower resolutions this causes the scan lines to become visible due to the dark space between them, which makes the display a bit darker and harder to view. If the beam is fatter then the display looks better vertically, but the lines merge together when there are more of them, lowering the effective vertical resolution.

In the horizontal direction video amplifier bandwidth limits how fast the beam can change intensity, causing the image to smear sideways. The higher the pixel resolution and frame rate the higher the bandwidth needs to be. This is not a limitation of the CRT itself, but achieving higher bandwidth is difficult so it is generally matched to the dot pitch.

To achieve a small spot size the CRT guns must be designed to focus the beam narrowly, and the the deflection system has to accurately converge the 3 beams (one for each color) at the shadow mask. This makes the CRT and control electronics more expensive.

If the dot pitch is large relative to the pixel resolution then the display will show bad fringing and lose effective resolution. With triad dots the display will look blurry in all directions, but smoother horizontally than vertical stripes (Trinitron). However vertical stripes have higher brightness due to more of the beam getting through the slots in the shadow mask, and can have higher contrast by putting black stripes between the phosphors. This may make the screen appear sharper for photographic images, but not so much for text or fine lines (which is why few computer monitors used vertical stripe CRTs).

A well designed CRT monitor will have a tube with beam focusing and dot pitch matched, and the video bandwidth and viewable display modes matched to the tube. Some early 'Super VGA' monitors did not do a good job, due to trying to keep the cost down by using a lower resolution tube and/or poorer electronics. We can deduce that the 6318 was one of these, as the dot pitch is lower than other monitors with the same screen size and it could only display 1024x768 pixels in interlace mode. This explains its low price of only $318 (vs $480 for the 6314).

How was control over the electron beams distributed between the monitor and the video card? Which part controlled the following aspects?

Start X,Y of the first scanline
Scanline width (was it always the same for all scanlines?)
Time to draw a scanline
Vertical distance between scanlines (was it always the same for all scanlines?)
Number of scanlines

This is all determined by the timing of the sync pulses. The beam moves horizontally with sufficient magnetic deflection amplitude to go across the screen (from left to right) in the time between horizontal sync pulses, and vertically down the screen at a slower rate that scans the entire screen in one vertical field (time between vertical sync pulses).

Unlike an LCD screen the CRT controller does not look at the RGB video signals, so it has no idea what the pixel resolution is, only the sync timing. To accommodate different display modes the controller first detects the screen mode (by detecting sync pulse polarity and/or measuring the time between horizontal and vertical sync pulses) and adjusts the scan speeds to suit, then waits for the horizontal and vertical sync pulses and pulls the beam back between lines and at the end of the display.

To change the horizontal scan speed different capacitors are switched into the EHT transformer and scan coil circuit. Due to the high voltages involved, this is sometimes done with relays which you can hear 'clicking' when changing screen resolutions.

The number of scan lines and distance between scan lines is determined by the sync timing, as the beam moves across and down the screen at a constant rate until it sees the vertical sync. Because the beam is deflected by electromagnetic coils with analog control circuitry, the rate of movement and position on screen may vary with temperature and line voltage etc., so user controls are provided to adjust the height, width, and position of the display.

Could a (custom) video card technically output image with arbitrary resolution (up to some physical limit), say 1x1 to 1000x10 to 1000x1000?

Yes. Apart from possibly expecting a 'black' level at the display edges, The CRT monitor doesn't care what is in the video signals. So long as the sync timing matches a known resolution it will be quite happy scanning a completely blank screen or one that only has a single pixel spanning all displayed lines. This is how for example 320x200 can be displayed on a VGA CRT with the same timing as 640x400. The VGA card could just as easily put out 100x100 (with each line repeated 4 times) or 1000x400, and the monitor wouldn't know the difference so long as the sync timing was the same.

For 1000x1000 the monitor would have to scan at least 1000 lines, so it would need to support a screen with normal resolution of eg. 1280x1024 (with 24 'blank' lines not displaying pixels), and the VGA card would have to generate sync pulses equating to that screen mode. To get 1000 pixels to fill the screen horizontally the card would have to generate them at the appropriate rate, eg. at ~78% of the clock frequency required for 1280 pixels.

Early VGA monitors had fairly crude analog circuits that did not even count the number of lines or accurately measure sync timing. To synchronize with the signal they use simple PLL circuits that 'pull' the horizontal and vertical oscillators in phase with the sync pulses, so they often work with non-standard timing (though the screen may be off center or a different size than usual). If the timing is too far off the monitor might operate too fast or produce too much deflection, which could overheat components or blow up them up with high voltage.

6
  • 1
    The most credible reports about blown PC monitors are about the IBM Monochrome Display (the monitor that went with the MDA). It doesn't even have a PLL for the horizontal reflection, but drives the horizontal sync quite directly from the HSYNC pin. The PLL in sensibly designed monitors is only able to capture frequencies that are safe. Bad timings wont blow up the monitor, but just cause the PLL to run unsynced at a "random" safe frequncy. Dec 15, 2020 at 7:34
  • CRTs run at very high voltage (27kV on the CRT, ~1kV on the line output transistor) with not much margin below below abs. max. ratings. The components have enough stress on them at the best of times, and even more when frequencies are a bit off... Most people run standard video drivers so they aren't in any danger, but if you are running x-windows or an Amiga1200 it is easy to create bad timings, and a few people have blown their CRTs this way, eg. overclockers.com/forums/showthread.php/… Dec 15, 2020 at 10:35
  • 1
    Note that it is a matter of crude electronics. If I remember correctly: a flyback converter is used to get the high voltage for the tube anode. A flyback converter ramps up energy in a transformer and then suddenly dumps it into a capacitor on the other side of the transformer. Conveniently, this is also how the vertical deflection signal works - it gets stronger and stronger and then suddenly resets back to 0. So they decided to merge the two things... which means the flyback converter now has to run at the refresh rate... which means if the refresh rate is wrong it overloads and blows up.
    – user253751
    Dec 15, 2020 at 14:06
  • 1
    @user253751 useful info about how the HV is generated, and the issue when the refresh rate is wrong! However, the high voltage is almost universally generated from the horizontal deflection circuit, not the vertical. The horizonal scan frequency is 10s or low 100s of kilohertz with a blanking period (when the beam returns to the LH side) of microseconds or less, so the rate of change of current is very high, producing a useful high voltage across the deflection coil (V = L di/dt). On the other hand, the vertical frequency is 50 to maybe 120 Hertz, so pretty useless for producing high voltage Dec 15, 2020 at 15:33
  • 1
    (slightly more info: if the refresh signal is totally stopped, or too slow, then the current keeps increasing and never resets, which either overheats the transformer, or the transistor controlling the current)
    – user253751
    Dec 15, 2020 at 16:55
5

1a. Yes, the shadow mask should be at least double the monitor resolution in both X and Y directions to give a good picture.

1b. Most monitors had hole-style (shadow) masks, but Triniton used an aperture grille with slots and wires that extended the entire vertical height of the image, with one or two wires running horizontally across the screen to stabilize the vertical slots and wires.

2. VGA uses 5 discrete signals: Red, Green, Blue, Horizontal Sync, and Vertical Sync. H-Sync tells the monitor when to begin a new scanline, and V-Sync tells it when to begin a new frame. This is how the monitor and video card coordinate with each other to display the image. I believe they're both simple pulses. To determine (most of) the other aspects you listed, the monitor simply counts the pulses and measures the time between them. If you want more control than that, for example to drive a vector display, you would replace the digital H/V signals with analog X/Y signals.

I think the beam width is the same for all scanlines, but you can manually adjust the brightness.

3. The horizontal resolution is continuous so it can support whatever horizontal resolution you can throw at it, although the horizontal contrast may be too low to see every pixel. The vertical resolution is limited by the minimum and maximum timings that the monitor can support.

4

I've just realized I have a gap in my understanding of the VGA CRT monitors of the 80386 era (1980s-1990s).

It may be more a general need to consult basic information about basic analogue TV/composite video signal structure, independent of PCs or time.


What was the phosphor resolution?

There is no general (practical) phosphor resolution.

Resolution is an attribute of the digital signal source, not the analogue screen.

But I also remember my monitor had controls to resize and move the image and possibly even do pillow correction. For that to work, the shadow mask resolution had to be significantly higher than the monitor image resolution. Is that correct?

No, as those are not synchronized values.

Though, a the mask grid does imply a (fuzzy) lower end for usable details.

For any data about grid size, pleas check the screen used.

Did monitors typically use hole-type masks or slit-style masks

Depends on the manufacturer and series as both have been used at times and models.

(that could give infinite vertical resolution).

No, slits do as well have a defined height. otherwise the mask would not be stable.

How was control over the electron beams distributed between the monitor and the video card?

Beam control was due signal timing delivered by the controller (virsocard) which hadto be within the capabilities of the screen to decode / interpret it.

Which part controlled the following aspects?

All done by the CRT controller by generating the frame signal.

Or better, the controller creates a certain composite signal and the CRT electronics are adjusted to display this.

Start X,Y of the first scanline

The controller puts themat a certain position within the signal generated and the CRT electronics may offset them according to it's setup (remember the little potis to adjust the picture?

Scanline width

Like above.

(was it always the same for all scanlines?)

It better should. Having different line length may distort the decoding electronics.

Time to draw a scanline

By controller as it's the same as scanline width.

Vertical distance between scanlines

Defined by CRT electronics setup.

(was it always the same for all scanlines?)

Yes.

Number of scanlines

By controller.

Could a (custom) video card technically output image with arbitrary resolution (up to some physical limit), say 1x1 to 1000x10 to 1000x1000?

Sure. As long as it's within the limits the electronics of the CRT can adapt to/decode. Basic electronics of a CRT are build for a certain frequency range, were they could syncronize with the signal. Later/more expensive CRT had extended electronics able to sync within a wider frequency range. Ofen called 'Multisync' based on a brand name by NEC who introduced them first for a larger audience.

It's important to keep in mind that vertical and horizontal resolution limits are independent. While horizontal resolution is limited by the number of lines a screen electronic can display is horizontal resolution defined by bandwidth and signal type.


If you really want to get into video design, don't reach for the stars, but create a standard B&W composite signal first. Experiment with settings of your generator and reactions (and settings) of an according (simple) CRT screen. This will give you a firm grasp of the basics. Every further step like adding colour or higher/different resolutions will be quite clear and easy thereafter.

1
  • 1
    Might whoever voted this down add some comment why he considered this answer 'Not Useful' ? If possible with pointing out which of the information given is factual wrong?
    – Raffzahn
    Dec 14, 2020 at 23:03
4

Let me try a summary how analog CRT monitors work(ed).

Let's start with monochrome, later explain the color models, and then answer your questions.

How it works

The CRT produces a thin electron beam that produces light when it hits the phosphor. A deflection circuitry can bend that beam so that it can reach all of the sreen front area. This circuitry is driven by two ramp signals:

  • A low-frequency signal (e.g. 50Hz) for the vertical direction, so the beam starts at the top of the screen and continuously moves downward to the end, then quickly returns to the top, all this e.g. within 20ms.
  • A higher-frequency signal (e.g. 25kHz) for the horizontal direction, moving the beam from left to right, and then quickly returning (within e.g. 40µs).

Combining both circuits, the beam moves along a raster pattern of nearly horizontal lines covering the screen front.

The video card provides:

  • a clock signal (called VSYNC) telling the monitor when to vertically return ("retrace") to the top position,
  • a clock signal (called HSYNC) telling the monitor when to horizontally return to the left position,
  • an intensity signal commanding the current beam intensity to the monitor.

The deflection circuitry can only work within some given frequency ranges, so the video card isn't completely free to decide on these frequencies, but must create a timing appropriate for the monitor.

The intensity signal is zero during the times when the monitor is expected to do the horizontal or the vertical beam return.

So, the intensity signal is active (non-zero) during the major part of one horizontal cycle, e.g. during 32 of the 40µs. This time corresponds to the srceen width (the beam moves across the screen from left to right during that time), and the video card divides this time into slots corresponding to horizontal pixels, e.g. to fit 640 pixels into 32µs, each pixel occupies 0.05µs. In each of these pixel slots, the video card provides an intensity voltage corresponding to the desired screen brightness of that pixel.

Resolution

So, what's the resolution we get?

The vertical resolution is given by counting the horizontal cycles that fit into one vertical cycle, being the ratio between horizontal and vertical sync frequencies (minus the time for vertical retrace), e.g. 25kHz / 50Hz = 500, meaning that maybe 450 lines are usable.

So, the vertical resolution is given by the video card. But, as the acceptable sync frequencies are constrained by the monitor, the card isn't completely free in its decision.

The horizontal resolution also is defined by the video signal, by the aspect, how many different analog intensity values the card emits during one horizontal line. Here, the video card is completely free to produce whatever number of pixels it wants, but the monitor's beam controlling amplifier has its limits, so it won't be able to follow the signal changes exactly, if they come too fast, resulting in a horizontally blurry impression.

One additional aspect impacts the resolution: the diameter of the electron beam. If the beam size is too large, multiple scan lines overlap, so you can't distinguish the individual values from vertically-adjacent pixels. But the beam can also be "too thin", resulting in unpleasant gaps between the horizontal lines. And of course, large beam diameters also amount to horizontal blurriness. So, sending a high-resolution signal to a monitor with a blurry beam might be technically possible, but won't give the desired crisp image.

One remark on pixel aspect ratio:

As horizontal and vertical resolution are controlled by completely different timing aspects, it's easy to get pixels that aren't square, meaning different horizontal and vertical dpi values, resulting e.g. in circles appearing as ellipses.

Linearity

As the picture geometry is mainly controlled by the beam deflection circuitry, it's important that the beam moves with constant speed (at least during the active signal periods). Otherwise, you'd get different pixel scales over the screen size, resulting in distortions. CRTs never achieved perfect linearity, but many of them could fine-tune various aspects of that deflection, e.g.

  • Horizontal size: this modifies the amplification of the horizontal-deflection circuitry, making the beam cover a wider horizontal range within the given horizontal time.
  • Vertical size: this modifies the amplification of the vertical-deflection circuitry, making the beam cover a taller vertical range within the given vertical time.
  • Pillow correction: adding correction signals to the deflection system so that the corners get the same deflection ratio as the center (pillow error: there is too much deflection in the corners, barrel error: the center gets more deflection than the corners).

Color CRT

A color CRT works mostly the same, but:

  • There are three electron cannons, producing three beams (meant for R, G, and B). Some monitors arrange them in a row, others in a triangle configuration.
  • The beams have to go through a mask with holes before hitting the phosphor surface.
  • Coming from different electron cannons at different positions, after passing through one hole of the mask, the three beams hit the phosphor surface at different positions.
  • At these different positions, phosphors of different color are applied, so the electrons coming from the "R" cannon only hit red phosphor dots, the "G" electrons only green dots, and the "B" ones only blue dots.
  • The deflection circuitry deflects all three beams the same way, so they always hit the same spot of the screen front.
  • The video card provides three intensity signals, simultaneously controlling the three electron cannons.

With color CRT, we have one more aspect limiting the effective resolution: mask and the dotted phosphor surface need to be finely pitched to allow for the desired resolution. If a single RGB tripel is bigger than the desired pixel size, you don't get the impression of a "single mixed-color dot", but individual red, green, and blue dots.

Your questions

Phosphor resolution only applies to color CRTs, being the dot pitch. Dot pitch has to be clearly smaller than the desired size of a single pixel.

There were both dot-mask and slit-mask monitors. Dot-mask corresponds to a triangle configuration of the cannons, slit-mask to a linear one.

In theory, a dot mask imposes a limit on vertical resolution, but in reality other factors dominated: the beam diameter, and the difficulty of having three beams hit the same screen spot.

As explained above, the video card controlled the resolution by using an appropriate timing, but that had to fit within the limits of the monitor.

The position and size on screen could be controlled by both the video card and the monitor:

  • The card could start its intensity signal earlier or later (horizontally as well as vertically), resulting in the image moving left or right (or up and down).
  • The card could use more or less of the available beam-moving time, thus producing a bigger or a smaller image.
  • The monitor could adjust amplification and possibly offset of the deflection ramp signals, thus magnifying and/or moving the image, both horizontally as well as vertically.

From the video card's point of view, the scan line width was the same over the full screen. There was no reason to generate shorter and longer scan lines. But the analog beam deflection circuitry never was perfectly linear, so it was quite common that the image width varied slightly over the screen height (e.g. cushion or barrel distortion).

The time to draw a scan line was always constant, corresponding to the horizontal sync frequency. Monitors were designed for constant H and V frequencies, and always needed some time to adapt to different ones.

The vertical distance between scan lines was given by the amount of vertical deflection corresponding to one horizontal cycle. E.g. with a 50Hz / 25kHz setting, each scan line consumed 40µs, and within 40µs the vertical deflection advanced by roughly 1/450 of the screen height (not 1/500 to account for the vertical retrace).

The number of scan lines was given by the ratio between horizontal and vertical scan frequencies (minus the time needed for vertical retrace).

A video card could in theory produce any pixel resolution desired. Limiting factors were:

  • The maximum pixel clock rate, making e.g. 1000*1000 hard to achieve in this era (60MHz or more necessary).
  • The horizontal and vertical scan frequencies that the monitor was capable to handle.
  • The unpleasant appearance of low vertical resolution with thin electron beams, resulting in visible gaps between the lines.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .