Could some 200 line displays have been pushed to 240 lines?

Question

Reviewing Raffzahn's answer about CGA emulators for Hercules displays, and especially his initial (now corrected) note about 720x350 being PAL's natural resolution, I was wondering if 200/400 line displays of the era (IBM CGA, Atari ST) could theoretically have been made larger by an additional 40/80 lines and still be displayable on monitors of the era.

I'm well aware that 200 lines can fit nicely in 16 KB (in 640x200x1), and that it's of an overwhelming importance with some display controllers. But assuming a flat memory model, and a main memory shared by interleaving access between the CPU and the video controller, one could use a few additional KB to display more lines.

I'm also aware that before the Mac, nobody (or at least not many) felt the need for a 1:1 pixel aspect ratio, so I can understand that there was not much of a market for it.

But could it have been done physically with composite or TTL monochrome monitors with NTSC-like (60 or 30 Hz) frequencies? The standard allows for 480 visible lines on the 525 total, so it's at least plausible. I've also checked the timings of the 71.2 Hz Atari SM124, and adding 80 lines to get 640x480 doesn't seem too much. Looking on photographs of the monitor, it looks like there's enough space above and below the usual display area to expand it by 20% and get square pixels.

So: is it only related to memory size, or is there a problem with the monitors too?

Answer 1

If the monitor’s phosphor mask is fine enough to handle the increased vertical resolution, then the rest is mostly about timing, and typical components at the time were capable of handling a 20% increase, at least in vertical resolution — demonstrably so in fact since building CRTs capable of handling TV resolutions was common.

For example, the first IBM CGA monitor, the IBM 5153, can be modified to handle an EGA signal and display it; that involves handling 720×350 in a monitor officially capable of 640×200, so 640×240 is definitely possible.

All this suggests that the main constraint was memory, not the monitors. Bear in mind that increasing vertical resolution has multiple knock-on effects, beyond the framebuffer size: if characters increase in height, more room is needed for character ROMs, and if the number of text lines increases, more room is needed to represent the contents of a text-mode screen.

Answer 2

Just to summarize, some points which people may sometimes not fully realize about analog video:

1. The kind of analog video signals referred to in the question are based on the concept of raster scan, or scanline raster.

2. Except for the early experiments with mechanical scan television, scanline raster-based video signals are primarily designed to be displayed on cathode-ray tube displays, or CRTs for short.

3. CRT displays draw their picture on a phosphor coating using a “sweeping”, or “scanning” electron beam. The illumination provided by the phosphor coating will quickly decay — and we normally want it to quickly decay to reproduce good-quality motion animation — which means the image must be constantly refreshed (redrawn) on the screen.

4. The location of the electron beam on the phosphor screen is controlled by horizontal and vertical deflection yokes (electromagnetic coils) located on the neck of the tube. The circuitry in the TV drives these coils in a specific manner to create the raster scan pattern.

5. The electron beam is switched off (or greatly lowered in its intensity) at the end of each horizontal scanline, and also at the bottom of the screen, until the next scanline or the next frame/field will go live.

6. The beam needs time to get horizontally deflected back to the beginning of the next scanline, or vertically back to the top of the screen. It will not “reset” to these locations instantaneously, or by magic, but needs to be physically driven/guided there by the deflection yokes. This is why we have the recurring periods of “horizontal blanking” and “vertical blanking” in the generated video signal: it is to allow enough time for the horizontal and vertical retrace of the electron beam, while it is in the “blanked” state. There’s certain kind of electromagnetic inertia involved; even though the intensity of the beam is lowered during these periods (the beam is “blanked”) you can’t just instantaneously put it wherever on the screen. You need to smoothly guide it there.

7. Also, certain very interesting timing factors are at play which keep the horizontal and vertical scanning in sync with each other, in a manner that is sensible with regard to the kind of electronics that the TV manufacturers had at their disposal back in the day — and that also allow interlacing (a kind of CRT phosphors, persistence of vision, and scanning pattern / video raster-based “timings trick mode” which allows seemingly doubling the vertical resolution while keeping the bandwidth of the signal the same. Or doubling the vertical refresh rate and motion smoothness while nominally still keeping the full vertical resolution and the same bandwidth, whichever way you want to look at it.)

8. Given the above, all periods of the signal that are not spent in horizontal or vertical blanking are considered “active lines” (vertically) or “active period” on a scanline (horizontally) — free game for any content (“active picture”) you would like to display.

9. TVs adhere to broadcast standards set by national and international standardization bodies. Those standards (and also the aforementioned technical timing factors), such as CCIR/FCC System M, or CCIR System B define the timings used, and how much of signal is spent in blanking periods allotted for the horizontal and vertical retraces of the electron beam. CRT TVs and their controlling circuitry were built to implement those standards. They do have certain tolerance which allows them to lock in to signal sources not strictly adhering to the standard — such as the non-interlaced signals produced by many home computers or the sometimes very variable and wobbly time base of domestic VCRs — but in theory, they not need to sync to anything that deviates from the standard or support anything beyond what the broadcast standard defines about the timings and the blanking periods.

10. You can use the active periods of the signal for carrying image content to the fullest, but if you start modifying the time allotted for the horizontal and vertical blanking periods, or the actual scanline raster timings (the intervals at which the horizontal or vertical sync pulses occur in the signal) that’s a no-no. The CRT circuitry only has so much tolerance; it is designed with a certain agreed-upon target timing standard in mind. You would basically be “overclocking” your CRT.

11. That said, not all signal sources utilize the possible active picture area in the signal to the fullest. A case in point is the Commodore 64 which has these quite large, colorful borders around the readily controllable content area. This is just a design choice, though — the borders are definitely part of the active picture area defined for a (roughly) standards-compliant TV signal, and a video chip could have been designed that fills them up with content. The Commodore Amiga, on the other hand, has a graphics chip that can fill up the active picture area of a standard TV signal all the way to the edges of the blanking periods.

12. Then there’s the subject of (sometimes quite misunderstood) overscan and underscan. These are not properties of video signal but the effects of the chosen CRT TV/monitor image geometry adjustments. The video signal does not know or care whether the display displays the active picture area in an “underscanned” (full picture, or even extending to the blanking periods) or overscanned (zoomed-in, edges cut off) fashion.

    Overscanning means a CRT display is adjusted to “zoom in” to the active picture area present in the signal so that not all of the picture that is carried by the video signal is visible on the screen. The picture fills the screen but is somewhat cut off at the edges (the display only shows the middle/center part of the picture, due to local adjustments made on the CRT itself.)

    You can also adjust a CRT TV or video monitor to “underscan”: that is, display all of the active picture area in the signal, and even a bit beyond, so that you may actually be able to peek into the blanking periods.

    So why would we want to adjust the CRT to overscan at all, or waste some parts of the active picture that is present in the signal? This was originally because CRTs — the electronics and physics of driving an electron beam in a high-speed scanning pattern — are finicky devices, and were so especially back in the day when they were invented. It was difficult to keep the scanning electron beam pattern completely stable with old vacuum tube-based electronics and older transformer designs and analog circuitry only; the video raster often “breathed” (changed overall size) based on the brightness of the content, or the adjustments kept slowly crawling as the components aged. It was much easier to cut off some picture at the edges (which makes it look cleaner and the effects of “breathing” or “blooming” less noticeable) than to display the full active area of the signal on a CRT in a completely stable way, with clean, straight edges.

    Also, the old CRTs were quite roundish at their edges and especially in their corners, so there was no way to fully fill up the screen with content without “overscanning”; letting some of the active picture go over the edges of the phosphor screen.

    What is more, even the large, national TV broadcasters could not produce completely clean active picture area all the way to the very edges of the blanking periods, on analog gear. There was always some variance — for instance, some analog TV cameras scanned slightly wider than the others, or at least started blanking at a slightly different time. The active content in the signal could be visibly slightly wider in some shots and slightly narrower in others if you monitored it on a CRT that was set to underscan. By overscanning, domestic TVs hid these edge artifacts from the viewers.

    Broadcasters were always aware that the domestic TVs overscan and compensated for this by adhering to the so-called “safe area” practices, such as those recommended in ITU-R BT.1379-1 (see the illustrations on the later pages.) This means you need to frame any important action and all the titles within certain imaginary “safe area” borders, some distance away from the very edges of the the active picture area, to make sure all viewers can actually see those details on their overscanning domestic TV screen.

    The designers of TV-connectable computers and game consoles have also had to take this in account. In the early years, the go-to solution was usually a large, fixed border around the addressable graphics. In the later years — starting from the Amiga but continuing with the original Xbox etc. — the video chip or scaler allowed filling the entire active picture area with framebuffer content, but then it was the responsibility of the developer to keep importart details away from the edges and within the safe area. (I believe there was actually a chapter about the importance of designing for safe areas in the original Xbox SDK documentation.)

    Curiously, the practice of “overscanning” and the need for “safe area” adherence in TV/video production has followed us even to the era of digital TV broadcasts and non-CRT displays, which use discrete, addressable picture elements and have never suffered from the relative instability of the CRT raster scan pattern. This is mainly because digital TV standards (in so far as SD resolution content goes) still had to be made compatible with the analog video signal standards and CRT technology, to ease out the gradual transition from these to the digital world. If you now broadcast old SDTV archive material digitized from analog video tapes, they still have those edge artifacts caused by the production technology of the day; their active picture is not completely clean, at least not to the very edges. Hence, your LCD TV is usually still set up to emulate CRT overscanning by slightly zooming into the picture and cutting off the visibility to the very edges, which might be a bit unsightly in the older content. (The side edges may have all kinds of analog unevenness and other “stuff” going on, such as the picture from another video mixer input showing through as a narrow column, and the upper/lower edges may have VCR/VTR head switching noise, all still present in the digitized, archived content but usually not seen by the viewers since the TVs still “overscan”.)

    These days, professional-quality gear has gotten much cheaper, the barriers of entry have lowered, and also people who never lived through the analog era or who do not have professional background in the industry produce content for TV broadcasts. They often do not know about concepts such as safe areas, or overscanning, or the emulation of overscanning still found in modern TVs, and design their content for the “underscanned computer screen” (instead of designing it for “TV broadcast”) by placing their titles (e.g. those seen in ads, or on some low-budget cable channels) too close to the edges, which unfortunately leads them to become partially cut off on the TV screens of many viewers, even today.

13. There were at least three major types of dedicated computer monitors back in the CRT era:

    1. Those which were designed around TV standards-compatible timings (such as the RGB video monitors commonly used with the Amiga; Philips CM8833, Commodore 1084S and the like) and could even accept composite PAL or NTSC video signal on the side, and could also be utilized for video production or CCTV purposes.

    2. Those which were strictly “computer-only monitors” and adhered to the proprietary timing standards of a single manufacturer (usually designed to be used with a particular, specific system or display adapter product).

    3. Later on, “SVGA” etc. multi-scan computer monitors which were “generic” in their nature, fairly flexible, and generally allowed themselves to be driven using some practical set of the standard VESA Coordinated Video Timings modes, but also allowed custom, hand-crafted timings as long as they fell within their technical limits (usually expressed as the upper and lower bounds of acceptable horizontal and vertical refresh rates — but there might still be other, unstated practical limits on how much of the active picture in some “pathologically-crafted” video signal timing modes can actually be seen on the screen when adjusting the picture size from the control knobs etc.).

    So, depending on which category a particular monitor belongs to, its accepted timings (blanking periods and active picture area) are defined either by the national and international broadcast TV standardization bodies, a single manufacturer, or a computer or consumer electronics industry standards body (VESA).

---

However you generate the timings and a signal with picture content for a CRT-based display device, the video mode you created either follows the video timing standard the manufacturer of the CRT had in mind and tested when designing the circuitry of their product, or does not follow that standard. Once you deviate from that standard, all bets are off.

Custom video modes with slight deviations in timing and the number of active lines may be possible even on a nominally fixed-sync monitor — as long as they fall within tolerances of the device and the range of its picture geometry adjustments. But you may risk stressing the monitor in ways not tested (or safeguarded against) by the manufacturer.

Custom video modes are safer on multiscan monitors which the manufacturer has designed to support a wide range of horizontal and vertical scanning rates and where you can design a mode which falls within those boundaries.

In some implementations, the underlying video timings standard is somewhat underutilized by the video signal-generating device, by it not filling up all the possible (legal) active signal area with active content but generating large, static borders around the active content.

Some other times, the limiting factor is the display device, which is adjusted at the factory to cut off the edges of the active picture, in order to follow the accepted “norm” and best practice of the industry, and to alleviate various practical, technical shortcomings and concerns. This was (and still is) common for TVs but not so much for displays designed and marketed as computer monitors.

Sometimes the display device lets you choose which way you want it. For instance, the broadcast-grade CRT field monitors (usually made by Sony or Ikegami) have a separate “underscan” button which, when toggled on, displays all of the active picture area. Also, the “TV standards-compatible” RGB video monitors used with the Amiga and similar TV-compatible computers (but also for video production or industrial or security monitoring purposes) could be manually adjusted to underscan from the easily-accessible adjustment pots.

P.S. Even though the CRT needs the blanking periods for its physical beam retrace, standards bodies have defined ways to utilize the blanking periods in the signal for some other things, such as for carrying the colorburst synchronization waveform, Teletext data and Closed Captions. So, the blanking periods are not just “waiting for the retrace” periods but can still be utilized for carrying some extra content — just not the active picture content.

P.P.S. Since the modern LCD/OLED video displays do not actually need the blanking “pauses” in the signal (do not require or implement any kind of “horizontal or vertical retrace” due to the completely different technology that they’re based on), VESA has also defined some new, standard video signal timings with reduced blanking periods to optimize for the non-CRT use case.

Answer 3

Given sufficient memory to hold the larger frame buffer, the only limitation is the flexibility of the display controller.

As you alluded to in your question, the video timing standards (NTSC/PAL), and the monitors designed to work with them, easily supported 240 (non-interlaced) lines or 480 (interlaced) lines.

At least a few computers had flexible enough display controllers to simply display the extra lines. This was commonly referred to as "overscan", but only because the display controller was flexible and could be programmed to create scan lines outside the "default" range that it normally displayed.

The VDC chip in the Commodore 128 supports more than 25 lines of text via programming. It can also do interlace. I believe it's maximum vertical resolution is 491 (interlaced) lines. If you want to display bitmaps that large, you need 64KiB of video RAM, as was standard with the C128D. With the 16KiB video RAM on an un-expanded machine, you can still increase the lines of text (up to ~50), but will lack the necessary memory for full-size bitmap graphics.

These expanded modes can be displayed by the Commodore monitors of the time (1080/1084/1902), of course. But the picture below, taken today, shows a 47-line editor/text screen from a C128 with 16KiB video RAM running on a Sony PVM CRT.

The Amiga's display controller (Denise/Lisa) is noteworthy for its flexibility and ability to do "overscan" with standard NTSC/PAL output. It is perfectly normal to adjust the settings via software and display more than the "default" 200 or 400 lines. On an Amiga with 512KiB (or greater) graphics memory, you will not have any issues with supporting the larger bitmaps needed. There's even a "Preferences" setting for the Workbench to support this adjustment by the user, and it works perfectly well with the Commodore monitors of the same generation, while allowing over 280 (non-interlace) lines on most PAL monitors.

The Atari ST display controller wasn't designed to support "overscan" with NTSC/PAL output, and what it can do to draw in the default border regions is done using a complex software hack.

Answer 4

Yes - in fact in CGA basically a NTSC-compatible 240p RGB signal with 262 lines per frame is already sent to the monitor, but because the rest of the lines are known blank, they are zoomed out on the monitor to fill the screen with only the 200 active lines. The same RGB signal is directly converted to composite for standard NTSC TV connectivity. In theory, all it would take is to adjust the vertical size and hold knobs on the back of IBM 5153 CGA monitor to slow down the vertical scanning enough to fit 240 visible lines instead of the faster scanning speed of 200 visible lines. The monitor itself has already locked on to the signal, so it is just really a matter of if the size adjustment range is wide enough to allow adjusting to 240 visible lines.

Please also note that when you say NTSC TV has 480 visible lines and 525 total lines, it means in interlace, so the display is scanned with 240 visible and 262.5 total lines in one 60 Hz field, so it takes one odd and one even field to scan the display twice with 480 visible lines.

Longer answer:

It would most likely have been possible to extend the picture height of visible lines from 200 to 240 at least on an IBM 5153 CGA display. In fact, the IBM 5153 CGA display did come with vertical size adjustment behind the unit for picture adjustments.

CGA is rather compatible with standard NTSC signal. CGA uses 14.31818 MHz (4 times the NTSC color clock) master dot clock to generate the RGB video signal and composite output, so as CGA has 912 clocks per line, so their horizontal line rates are very close. The only difference is that NTSC is interlaced 525-line format and CGA is progressive 262-line format, so CGA has one half-line less per a 60 Hz field. So a CGA adapter actually outputs all the lines needed for a 240p signal on a TV, they are just blank due to memory not being enough.

The problem really is that while the NTSC video signal would have 480 to 483 active lines, or about 240 to 242 active lines per field, it does not mean they are all displayed as visible. Normal TV would overscan the image, so the visible area is horizontally and vertically less than the active image area.

That is why there is only 640 dots of active video per line on a CGA, plus border area.

So as the video width was fixed to 640 dots vertically, and that there was only 16 kilobytes of memory, it is only enough for 204.8 lines which would be guaranteed to be visible also on a NTSC TV set.

It would not have made sense to increase the amount of memory onboard the CGA adaper to get 240 lines, of which not all would be visible anyway. The frame buffer memory is dedicated video memory on the graphics card, so it can't be shared with system memory.

So to utilize the CRT screen area, they decided to not use square pixels on the CGA monitor and scan the 640x200 area to be fully visible (underscanned) in the 4:3 CRT monitor screen. The unused blank lines are still sent but zoomed out to fill the screen with only 200 lines.

On a NTSC TV, it would look slightly different due to more lines being visible in the same 4:3 aspect ratio screen area.

It might be possible to output less than 640 active dots, or rather, less than 80 active character clocks per line. It just leaves you with 68 character clocks, or 544 active pixels per line.

Answer 5

But could have it been done physically with composite or TTL monochrome monitors with NTSC-like (60 or 30 Hz) frequencies? The standard allows for 480 visible lines on the 525 total, so it's at least plausible. I've also checked the timings of the 71.2 Hz Atari SM124, and adding 80 lines to get 640x480 doesn't seem too much. Looking on photographs of the monitor, it looks like there's enough space above and below the usual display area to expand it by 20% and get square pixels.

So: is it only related to memory size, or there's a problem with the monitors too?

Others have provided detailed technical answers explaining how it was technically possible. But the simple answer is that it was not practical to have 240 lines - not because a suitably designed monitor couldn't do it, but because the signal had to be compatible with a standard NTSC TV set.

To see the extra lines on an 'NTSC-like' monitor as well as show a correctly sized picture in standard 640x200 resolution, the monitor would need a vertical size control that the user would operate depending on the screen mode. Any software that used the 240 line resolution would not work correctly on a TV or monitor that didn't have this control.

The monitor could have been designed to automatically switch screen height by using a different sync polarity or some other difference in the sync pulses, but if you are going to that (which is incompatible with NTSC) then why not just create a completely different resolution like EGA, with a dedicated monitor to go with it?

When the requirement for NTSC compatibility is combined with the extra memory and extra complexity of the video card circuit required to do 240 lines, it's obvious why they didn't do it. The original IBM Color Graphics Adapter was already jam-packed with ICs on a full-length card. Where would they fit the extra chips?

In those days 200 lines was the practical limit for an NTSC compatible computer display. 240p was not a thing, so they had no reason to consider it. even PAL compatible computers of the time didn't generally go above 200 lines because it would limit their market or require special versions of titles for different zones.

Another factor is that the Apple II (which the IBM PC was designed to compete with) and popular video controllers of the time such as the Motorola MC6847 (used in the Acorn Atom and Tandy Color Computer) and Texas Instruments TMS9918 (used in the TI99-4/A, Spectravideo, MSX etc.) which were designed to output NTSC video, only had a vertical resolution of 192 lines. So 200 lines was already better than the competition, and 640x200 produced 80x25 characters which was the standard for text terminals.

Of the few home computers in the 1980s that theoretically could do more than 200 lines in NTSC, again they didn't offer it as a standard resolution because the standard was 200 lines and they didn't want to risk hiding edges of the display in the overscan area. Also some machines could set the border to a color other than black, which looks ugly when using the maximum possible number of lines because the vertical blanking creates black bars at the top and bottom of the screen (which normally would be hidden in the overscan area).

To summarize,

1. Not compatible with NTSC TVs
2. Requires special monitor
3. Non-standard resolution
4. More complex and expensive video controller
5. Ugly display when colored border applied

Answer 6

(Answering my own question to supplement Stephen Kitt's 200/240 answer.)

For the Atari SM124, 420µs are needed for the vertical retrace, with vsync + vback providing 983.9µs. The vfront is 1855µs. So we have a theoretical total of 2418.9µs (without accounting for times that would be needed to settle voltage or whatever). That is, approximately 85 lines.

That would seems a little short. But with further search along the useful information provided by BrianH and Jukka Aho, I've been able to verify that the SM124 could indeed display 480 lines with a little help.

Answer 7

Yes. If you had a decent monitor, video memory, and flexible video controller/chip, many resolutions were possible. I wrote an interactive program for my Commodore 128 and its "80-column" (8563) chip. I had also hacked in double video RAM for this video (which I believe became standard with the 128-D). The program allowed me to get resolutions such as 120 char by 30 text lines. The chip could also interlace, so more than 50 text lines were possible. I tried to submit the program and article to RUN magazine, but they just sent me an offer of a job interview and plane tickets instead. 🤠

My monitor at the time was a "Sears Total Video System", which they sold in the mid 1980s. I really liked it. It was a television, but it also offered inputs for composite, monocrhome, and RGBI. So it was a perfect match for the various modes offered on a Commodore 128.

BTW, the Commodore Advance Basic had no problems with variable screen resolutions. I guess that was natural, since it already had to handle 40 and 80 column text.