The Typewriter

In the United States, the majority of mechanical typewriters produced type in two sizes, "pica" and "elite"(1) We can, for the purposes of this short article, define the two sizes thus: pica typewriters fit ten characters per horizontal inch (or, 10CPI), while elite machines fit twelve per inch (12CPI). Both types of machines fit six lines per vertical inch (6LPI), or three when using double line-spacing (3LPI).

The most common paper size in North America, what we call "letter size", measures eleven inches on its long side, and eight-point-five inches on its short. The stylistic convention in typewritten documents, the "standard format"(2) dictates that one inch of margin be given on all edges of the page, leaving an "active area" of no larger than 6.5 x 9".

On an elite typewriter, a 6.5" line of text contains 6.5 x 12CPI, or, 78 characters. A 9" column of single-spaced lines of text holds 9 x 6LPI, or, 54 lines. While single-spacing is fine for business and personal correspondence, in science, academia, and publishing, the aforementioned standard format dictates that documents be prepared with double line spacing to facilitate editorial marking, notations, comments and so-on. In addition, pages may have headers or footers with page numbers, dates, or what-have-you, which further reduces the amount of "body" text on the page. Such a double-spaced document would have 9" x 3LPI, or, 27 lines of text; subtracting two or three lines for the header/footer, we're left with 25 or 24 lines of text.

Thus, long before the interactive computer screen, a convention was already in place for a properly formatted page of text to contain roughly 25 lines, with up to about 80 letters and spaces per line. Coincidentally, this convention was strangely compatible with the new technology of the television screen.

The TV

While there were some timing differences between European television and North American television, TV/video equipment was of equal capability and fairly-well standardized worldwide by the mid 20th century. The first "low-cost" (that is to say, low-cost to the eyes of corporate management) CRT computer displays were direct descendants(3) of TV/video equipment(4), and the standards that early video equipment were designed to meet defined the capabilities of early computer terminals.

TV-type CRT displays trace an electron beam across the face of a screen roughly 15,000 times per second (a bit less than this, actually – only 92% of each second is spent drawing the picture on the screen; the remainder is spent lazily angling the deflection coils from the bottom-right corner of the screen to the top-left), so that every 1/60th of a second, something in the neighbourhood of 200 to 240 horizontal lines(5) of image information would be beamed into the phosphor-coating on the inside face of the vacuum tube.

Letters and numbers on a raster screen are made-up of arrays of dots, or pixels. It takes an array of no less than 6x8 pixels to legibly render all the upper- and lower-case letters of the English alphabet(6). You can get away with a slightly smaller array if you don't care about your descenders (the hangy-down bit on lower-case Ps and Ys and Qs) being above the baseline instead of below it, but ultimately people do care. So, if we need eight vertical pixels for a legible character matrix, and we've only got 200-240 vertical lines in which to put pixels on our screen, then we're limited to drawing no more than 25-30 lines of text, right? Right.

The limitations on how many horizontal pixels a TV screen can show is really a matter of quality and expense of equipment, rather than standardized timing signals. You can think of it this way: the dots that a CRT monitor draws on its surface are caused by an increase in the current flowing to the electron gun, just like the sounds that come out of a loudspeaker are caused by increases in current flowing to the speaker's voice coil. Even if you have a really great pair of speakers, a crummy amplifier will attenuate high frequencies, and you'll never get your music sounding as crisp and lovely as it otherwise should. CRT displays are analog devices, and you can feed them as high-resolution/high-bandwidth a signal as you want, but the quality of the internal components of the display set are what determine the maximum horizontal resolution.

With that flexibility in horizontal resolution, why choose 80 characters per line rather than 72, or 95, or any other number? The blame for that is usually laid upon IBM. IBM's 2260 Model 3 was an important early video display terminal that could display 12 lines of 80 characters on its screen. Why 80x12? Because before IBM was the computer company, they were The Tabulating Machine Company, and their business was punchcards. Various punchcard formats were in use through the late 19th and early 20th centuries, but in 1928, The Tabulating Machine Company introduced an 80-column/12-row card, and it became the dominant punchcard format used by tabulating machines the world over. The company's name soon changed to International Business Machines, and although two decades later they got into the computer business in a big way, in the mid-1960s when IBM introduced the 2260 terminal, punchcards were still a huge part of their business – tabulating machines may have been out, but the punchcard stayed relevant as a low-cost data input/output and storage medium for computers. It was natural that IBM's video display terminal would, despite being able to do a lot more than just show the holes punched on a card, mirror the card's dimensions.

Closing The Loop

That's usually where the story stops: we have 80-column displays because IBM had 80-column punchcards. But I want to go just a bit further back and tie this article up by returning to typewriters.

The question I'm curious about is, why did IBM move from 45-column cards to 80-column cards in 1928? Why not double card density and move to 90 columns (Burroughs Corporation did this — sort-of — in a crufty way, by storing numerical information on the existing 12-row/45-column cards as two parallel sets of 6 rows). Why not go to a nice, round number like 100, or upgrade more modestly to 60? What follows is entirely my own conjecture; it may not have any legs at all.

IBM manufactured a great variety of keypunches over the years, some remarkably sophisticated, and some of Spartan utility. The most common, by far, in the 1930s was the straightforward mechanical keypunch, the Type 1, which consisted of a heavy base with a moving platform/carriage onto which the card was placed, and an overarching set of 12 keys, which actuated 12 punches aligned with the 12 rows of the card. The punch operator would press the keys corresponding to the holes that they wanted punched in a column, and the carriage would advance the card by one column beneath the key/punch assembly.

The carriage and skip-bar mechanisms on these mechanical keypunches are actually very nearly identical to that of a typewriter, and I don't know how much of a stretch it is to guess that IBM chose the 80-column format because the column spacing is very, very close to the "elite" character spacing on typewriters. I'm not suggesting that IBM chose 80 columns because it meant they could build keypunches with off-the-shelf parts — the spacing on 80-column punchcards is actually slightly narrower than the letter spacing on my "elite" Royal typewriter — but it's certainly not inconceivable that IBM could have taken advantage of the more-or-less off-the-shelf design of existing typewriters for the card transport mechanisms in their punches. Also, alphanumeric keypunches (like the mysterious Type 34, of which there is a tiny photo on page 4 in this riveting document from 1936, Machine Methods of Accounting – I cannot find a lick of information about it elsewhere), which were operated like typewriters and not only punched the appropriate holes on the card, but printed the corresponding alphanumeric character atop the column, further suggest convergence of typewriter and punchcard.

Ultimately, these things all reinforced one another in the great web of feedback loops that is technological development. Page formatting conventions that predate the computer influenced the screen formatting conventions of the video display terminal, which was conveniently just within the capability of mid-century video technology, while mechanical conventions of the typewriter influenced the design of punched-card machines which, in turn, set the design goals of early video display systems whose success influenced the design of subsequent systems, and so-on, and so-on.

If this were a proper essay, I'd go back and fix my thesis and do one more draught, but this isn't a proper essay, this is an entry on my blog. So I'm calling it quits right …here!

—L.

Notes

1. The usage of the term pica is mysterious to me. Pica, for those not familiar with typographic measurements, is equal to 12 printers' points, or 1/6th of an inch. Both pica and elite typewriters produce glyphs that are 1 pica in height, but in neither case are the glyphs one pica wide.

2. The "standard" appears to have been relayed from God himself to the high priesthood of editors and academics in the primordium of the 19th century. I have had no success in determining the origins of the standard.

3. The IBM 2260 series of displays/terminals were …maybe something less than "direct descendants" of TV technology. For the technically inclined, there's an interesting description of the vertical-scanning electro-mechanical weirdness of their design here: "The IBM 2260 Display Station".

4. There were CRT displays based on technology not so closely related to TV/video in the early days, too; what we now call vector displays, just like video-type raster displays, use electro-magnets to steer (deflect, in the proper jargon) an electron beam inside of a vacuum tube, but that's where the similarity ends. Instead of just drawing a whole bunch of rows of dots and lines in parallel across the screen, vector displays can draw lines and curves of any length, and at any angle, and are more closely related to oscilloscopes than TVs. Their deflection apparatus and control circuitry are different, and more expensive, but for certain applications they produce a much higher quality of images than raster displays.

5. Some of the knowledgeable among you are thinking, wait a minute! Standard Definition North American TV has 480 horizontal lines, not 200-240! Well, sort-of. Remember, I said 60 times per second, not 30. Analog TV signals are interlaced, so that the odd-numbered rows of picture data are drawn 1/60th of a second, then the even-numbered rows are drawn the next 1/60th of a second – so in each 1/60th, only 240 lines are being drawn (if we really want to pick nits, it's 241.5 – there are 483 picture lines in NTSC video). Interlaced video like that isn't too hard on the eyes when you're looking at moving pictures, but for static, high-contrast, high-detail images like text, it's flickery and headache-inducing. So, rather than interlacing to take advantage of higher vertical resolution, the old computer terminals just drew fewer lines twice as often for a nice, steady, easy-on-the-eyes picture.

6. Some early terminals used smaller arrays and could only legibly represent upper-case letters. Unfortunately, COMMUNICATING IN UPPER-CASE ALL THE TIME IS SOMEWHAT LACKING IN NUANCE. As mixed-case terminals became available and affordable, the shouting ceased.