15

This may seem like an absurd question at first, but I've been giving it some thought and I'm genuinely curious about the design details of these devices.

I was reading an answer on an unrelated Stack Exchange site about the retrace/flyback details on old CRTs, and the image attached to that answer piqued my interest:

NTSC scan lines

I understand that both the horizontal and vertical deflection of the electron beam is controlled by two sawtooth waves, the vertical running at the refresh rate and the horizontal running a few hundred times faster than that. I also understand that both sawtooths are constant sweeps, and not "stair stepped" to hold at any particular voltage to accommodate the viewable lines or retrace periods.

Here's the premise of my question: In the context of one single horizontal trace across the screen, the vertical position is also constantly increasing (towards the bottom) in preparation for the next scan line. It then follows that the scan line's vertical position at the left edge of the screen is slightly higher than the position at the right edge, and the whole screen is a parallelogram with left and right edges perfectly vertical, and top and bottom edges both slanted down towards the bottom right.

Assuming the premise is correct, was it common (or even feasible) for the designers of CRT computer displays to counteract this effect and make the screen and its contents perfectly square? Would such a compensation have even been worth the effort?

  • 3
    It's always been a (very) slightly distorted rectangle. ...and nobody cared. – tofro Mar 26 at 9:11
  • 4
    If the deflection grids/magnets are not mounted exactly perpendicular to each other -- and there doesn't seem to be any reason why they'd have to be -- the "horizontal" scan can be given a slight upwards slant that compensates for the steady drop of the vertical deflection during the line. (That is, if one cares, which I'm not sure one does). – Henning Makholm Mar 26 at 12:52
  • The other imperfections in CRT geometry are likely to drown out the slight parallelogram effect anyway. – rackandboneman Mar 27 at 9:17
  • And, of-course, PAL and SECAM might just do this all a little different. – Mast Mar 27 at 16:40
20

Canonically in NTSC standards the drawn lines are tilted slightly so that the start of the odd field starting at the left is perfectly level with the top of the even field starting in the middle horizontally. The actual angle though is so tiny, ~0.09 degrees, that it's negligible compared to the rotation error you'd get just from the earth's magnetic field and how well the factory calibration was done. TVs were generally set up with ~5-10% of the image being drawn outside of the edges of the screen anyway which would effectively crop the edges to be a rectangle no matter how far off the scanning pattern was.

  • 2
    "[...] edges of the screen [...] crop the edges to be a rectangle [...]" - not exactly, the CRT screens I remember weren't rectangular, but had quite rounded edges and corners, even adding to the validity of your argument. – Ralf Kleberhoff Mar 26 at 20:15
16

IIRC, the electron gun was actually installed in a position where it was rotated slightly relative to the tube, to compensate for this effect, so the scan lines did end up being horizontal.

  • 5
    ... which means that it is still a parallelogram, it's just the sides that aren't vertical rather than the top and bottom that aren't horizontal! – Tommy Mar 26 at 2:58
  • 14
    @Tommy: I think that's arguable. If you consider the screen (i.e. what the user sees) as a rectangular mask that clips a larger parallellogram shape (the CRT projection), then I would still say that "the picture area is rectangular". Similarly, I would say that a modern day projector produces a rectangular shape even though its lens could project a larger circle/disc. You're right that the CRT could/would produce a parallellogram when not being clipped, but since it's being projected on a rectangular shape, it can only fill that rectangular shape and anything else is lost/not registered. – Flater Mar 26 at 9:48
  • @Flater you've bested me there. I agree — one rectangular image, with horizontal scans, is produced rather than another. – Tommy Mar 26 at 11:33
  • 1
    @smitelli: It's possible that the electron gun was set up to not light up a pixel for any part of the paralellogram that is outside of the requested rectangle, I'm not sure how they handles the screen adjustments tbh. But I think that doesn't quite change the point that the screen output (i.e. the actually visible pixels) is rectangular. – Flater Mar 26 at 14:44
  • 1
    @smitelli You could change the "background color", yes. It wasn't used much in DOS times, though (you're probably confusing DOS with systems like Commodore 64), since you usually calibrated the display to show as close to the full extent of the display as possible. One thing people tend to forget is that CRTs were analog and highly tweakable - you could easily change the shapes and extents of the picture being drawn. Indeed, CRTs aren't even rasterized at all - pixels are in the graphics card, not the screen. You could easily draw vector graphics directly on the CRT :) – Luaan Mar 27 at 8:57
12

Yes there are a lot of compensations in a CRT like:

  • magnets counteracting background magnetic fields
  • circuits counteracting curvature of CRT screen surface
  • circuits counteracting different length of the beam (edges/center)
  • "linearizations" of brightness (gamma correction)

and probably much more I can not think of right now...

But back to your question the stuff you are describing is applicable only on monochromatic CRTs with analog sweeping. As the color ones got a luminofor masks presenting form of a "pixelated" or "lineated" grids preventing of any skew. But still the compensations must be done (by rotating the sweeping slightly) otherwise the beam could jump a line ...

However in digital era more recent CRTs are usually driven by a CPU and the sweeping is controlled digitally by DAC so no more skew as the vertical coil is no more changing during the vertical sweep ...

5

To expand further, it actually wasn't especially feasible — there is no easy solution that doesn't eliminate the interlacing.

Interlacing works because the timing of the vertical retrace varies. On odd fields it is triggered so that scanning resumes at the beginning of a line. On even fields it is triggered so that scanning resumes in the middle. Because of the diagonal scan, that sets one field 0.5 lines higher than the other. If the scan weren't diagonal then the two fields would not enmesh in that manner — they would instead sit exactly on top of each other, just starting in different places.

On a classic TV it's undesirable to make the diagonal scan anything other than diagonal because the flying spot during the capture process was diagonal. So you wouldn't be unskewing the image, you'd be skewing it.

On a monitor life is slightly different, and true horizontals are likely accurately to reflect the image. But it's also generally the case that monitors have smaller scan lines in order to output a higher resolution, so the effect is less visible anyway — on a 14" 800x600 monitor you're already talking about the right hand side being less than 0.3mm lower than the left, but being almost 28.5cm to the right. With a multi-sync monitor, how far down the right is compared to the left is a variable function of the resolution.

4

The effect you are describing did not matter, mainly because in a television the effect is so small, and the TV camera also had a CRT tube that converts light to video signal in a matching scanning pattern so the picture is in fact not tilted due to the scanning. Computer systems usually used progressive scanning on CRT monitors so for example VGA has twice faster line rate than NTSC TV so the lines are more horizontal. Even with earlier computer systems such as CGA that was NTSC compatible, most likely other distortions of the CRT were more noticeable than scan lines not being exactly horizontal.

1

If the horizontal deflection circuit had no effect on the vertical positioning of the beam, and if the vertical deflection circuit had no effect on the horizontal positioning, then the shape would indeed be a parallelogram.

In practice, however, it would be very difficult to design a CRT-based display in which the effects were cleanly isolated in that fashion. Horizontal and vertical deflection circuits interact with each other for a variety of reasons, and instead of trying to prevent such interactions, most displays instead try to compensate for them with a mixture of adjustable and non-adjustable compensation circuits that can independently adjust the width of the screen at the top, middle, and bottom. Further, the angle of the yoke assembly is often adjustable. While vertical motion during each scan line would cause a slight skew if horizontal and vertical deflection were independent, consumer-grade equipment is seldom close enough to perfect calibration to make that significant.

0

Consider the case of rear projection CRT based HDTVs. These use 3 monochrome CRT displays, one red, one green, one blue, none of which have any mask. There is no parallelogram effect on these HDTVs. There are all sorts of computer controlled dynamic calibration parameters, and generally two "native" modes, 480p and 1080i. The most common calibration is convergence setup that displays a grid, used to adjust one of the colors as a baseline, then the baseline color and one other color to converge the other color. Any parallelogram effect would be noticed on the baseline grid setup. You can use a measuring tape to confirm that there is no parallelogram effect using the baseline grid setup.

In the case of a Mitsubishi 65 inch (diagonal) HDTV, the 3 monochrome CRTs are 9 inch tubes, and the screen size is ~56.6 inches by ~32 inches. That's enough of a zoom factor that any parallelogram effect would be noticable. For the Mits, there are 64 points on the grid used for convergence. Each point on the grid can be adjusted left/right and up/down. The base line grid is green, and that is setup first to get a proper grid. Then red or blue are converged on top of the green grid, with green + red showing as yellow, and green + blue showing as aqua. I think 2006 was the last year Mitsubishi made consumer CRT based rear projection HDTVs. Calibration and convergence procedure has to be done twice, since there are two native modes, 480p and 1080i with independent settings.


I'm an old guy, and I don't recall any parallelogram effect on computer oriented monochrome CRT monitors, at least not by the time of Dec VT52, Dec VT100 (Ascii monitors), or IBM 3270 (block oriented monitor) (early 1970's).

I still have a Viewsonic G225FB computer color monitor, which includes a computer controlled rotate adjustment (despite using a mask, the phosphors can be partially "painted") which can be used to get rid of any parallelogram effect. I also have a Sony KV-1311CR color TV which can be used as a 640x480 computer color monitor (for Atari ST), and it doesn't have any parallelogram effect.

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.