Below is a diagram showing the keyboard layouts and the arrangements of characters on the elements of the original IBM Selectric typewriter, the IBM Selectric Composer, and the later 96-character Selectric III as well as various IBM electronic typewriters:
When I first encountered the arrangement on a 96-character element, it seemed to be very similar to that of an 88-character element, only rearranged; for example, the letter Z is still in the top row along with the numerals. But when I looked at the arrangement of characters on Composer elements as well, I saw that indeed IBM did take into account of what it had learned in designing the Composer element; except for the letter U, the same letters are on the third row as on the Composer element, and the narrow letter I has been moved off that row.
The typestyles provided on 96-character elements were the same ones previously provided with the 88-character elements, including the proportionally-spaced ones. Although no commonplace IBM typewriter product, if one excludes everything related to the Selectric Composer, using 88-character elements offered proportional spacing, those typestyles were first introduced on 88-character elements for the IBM Mag Card Executive word processing system.
This system was used by large offices; I believe that there was one used to prepare correspondence on behalf of the President of the United States of America.
Since the proportionally-spaced typestyles of the Mag Card Executive worked with the existing 88-character element arrangement, why was there any need to take the lessons learned from the Selectric Composer into account when designing the 96-character element?
An IBM Selectric element is shaped like a portion of a sphere. The third row of characters, as shown in the tables in the diagram above, is on the "equator" of that sphere.
The second row of characters lies above those characters, and the fourth row of characters lies below them, with their centers at 16 degrees north and south latitude respectively. So the characters in those rows are somewhat more closely spaced on the element.
The first row of characters is placed on the element centered at 32 degrees north latitude, so those characters are spaced together the most closely.
At least, I currently think that the tilt angle on the IBM Selectric is 16 degrees. At the moment, I don't have an actual element in front of me to study; my examination of photographs of such elements on the web has produced... conflicting results, ranging from 13 degrees to 17 degrees. But recently I found an image that was photographed from the right angle to enable me to be fairly confident that 16 degrees is the right angle.
As to the size of the element, that has proven even more difficult to estimate, but the best image I've found suggests a diameter of 1.33 inches; another suggests one somewhere under 1.38 inches.
As to the radius of curvature of the characters on the element, though, definitive information was available: the circumference of the platen is 4.5", and so a likely value for the radius would be in the neighborhood of 0.7162".
The important distinction to take into account, though, is that on an 88-character element, each of those four circles of characters has 22 characters in it; on a 96-character element, each of them contains 24 characters. So the characters are more cramped on a 96-character element; this made taking character width into account more critical for those elements.
On the Mag Card Executive, the widest characters were allocated 7/60" of width along the page, or about 0.1167 inches. On the Selectric Composer, the widest characters were allocated 9/72" of width along the page - or 1/8", or 0.125 inches. Thus, the fact that the Selectric Composer aimed to approach fine typesetting in quality did lead to it using even wider characters than the Mag Card Executive did, thus making the width of characters more critical for the Selectric Composer.
And it's not necessary to guess; a paper published about the design of the Selectric Composer noted that the arrangement of characters on the element was modified in order to avoid impressions from adjacent characters on the element marking the page.
Also, while the Selectric III typewriter using 96-character elements came out in 1980, 96-character elements were first used with two models of the IBM Electronic Typewriter that came out in 1978, the Model 50 and the Model 60; the Model 50 was one of the three models that supported proportional spacing, and so proportional spacing was not an afterthought in the case of the 96-character element. Therefore, considering it in the design of the 96-character element would not have required any anticpation of possible future requirements.
The IBM Electronic Composer (from 1975) was similar to the original IBM Memory Typewriter (from 1974) in appearance, but it was a version of the Selectric Composer. The mechanism of the typewriter could be turned off to allow changing elements without turning off the electronics. Motion of the element across the page was regulated according to the three possible unit sizes by a disk with rectangular holes in it which were sensed photoelectrically.
As the first step in the direction of a "perfect" typewriter, I would propose adding a fourth circle of holes to that disk, so that in addition to the three Composer escapements of 1/72" (red), 1/84" (yellow), and 1/96" (blue), the unit value of 1/60" could be handled, to allow the use of Pica, Elite, and Mag Card Executive elements.
This would mean, though, changing much of the design of the machine; the Electronic Composer still was partly mechanical, and a fully electronic design would be required to cope with the fact that the keys on the keyboard correspond to different positions on the element on the Selectric Composer versus Selectric typewriters. So a design similar to that used with 96-character elements in the later IBM Electronic Typewriters, or to that used in typewriters from Silver-Reed which used Selectric elements, would be needed.
The IBM Electronic Composer allowed tab stops and margins to be set at locations separated by pica increments across the page. This worked with all the available type sizes. However, 60, 72, 84, and 96 are not just all multiples of six; they're also all multiples of twelve.
Thus, one could imagine offering additional escapements between the existing values; 66, 78, and 90 units to the inch as well.
This could potentially serve two purposes.
To meet diverse printing requirements, many different point sizes of each typeface, usually somewhere in the range from 12 point to 8 point, were offered for the Selectric Composer. For any given typeface, often more than three sizes were offered. And so it would happen that two different point sizes of the same typeface would use the same escapement, and thus their characters would be identical in width - only the characters in the larger point size would be taller.
A likely result of that would be that some point sizes would be a bit too wide or too narrow.
More possible alternative escapement values would clearly ameliorate that issue. And, indeed, as noted in a table on the previous page, it appears that adding these intermediate sizes would fully resolve it, and nicely provide good optical scaling as well, for at least those typefaces with the same natural proportions of height to width as Press Roman. A narrower typeface, for which 12 points, rather than 11 points, corresponded well to a 1/72" escapement, or a wider one, where the best appearance with a 1/72" escapement is found at 10 points, might not fit the available escapement values quite as neatly, of course, but having more alternatives would still be likely to help somewhat.
But there is also a second purpose that could be served.
Long before the Mag Card Executive (1972) and the Selectric Composer (1966), back in 1944, IBM brought out a typebar electric typewriter with proportional spacing, the IBM Executive typewriter.
The proportional spacing offered by these typewriters was relatively coarse, two units for the narrowest characters, and five units for the widest, but it was still a great improvement over ordinary monospaced typewriter type.
Several different typestyles were offered with the Executive typewriter, and in different sizes. Possible unit sizes were 1/32" (Documentary, Bold Face), 1/36" (Mid-Century, Heritage), and 1/45" (Text, Charter). 1/32" is three times 1/96". 1/36" is two times 1/72". 1/45", however, is two times 1/90", so it is only attainable by means of one of the escapement sizes I propose to add.
The Selectric Composer printed characters that were up to 9 units in width.
IBM Executive typewriters printed characters that were up to 5 units in width; thus, a 5 unit character with a 1/36" escapement would take up 10 units of 1/72", the largest Selectric Composer escapement. And a 5 unit character with a 1/32" escapement would be even wider.
So if I were thinking of transferring the typestyles of the IBM Executive typewriter to the Selectric typeball, it would seem clear that they would not fit. And, in addition, as I noted on the previous page, the unit system of the Selectric Composer involved a compromise; the widest characters, when considered in proportion to the other characters, should really have been 11 units wide instead of 9 units wide in order to most closely approximate the relative proportions found on typefaces used in the standard methods of typesetting.
But that's not an insuperable problem. There's no reason why the perfect typewriter couldn't be a daisywheel typewriter instead. After all, those printed considerably faster, and thus, while many companies made their own typewriters with interchangeable typeballs once the IBM patents ran out, the daisywheel drove them all, along with the Selectric itself, from the market when it became an option.
While IBM's own daisywheel typewriters, the Wheelwriter series of typewriters, offered the same typefaces as the previous Selectric typewriters - including the proportionally-spaced ones of the Mag Card Executive and the Electronic Typewriter 50, 65, and 85 - no attempt was made to offer a daisywheel version of the Selectric Composer.
There could have been many reasons for that, and one obvious one, even at that time, would have been the laser printer, even if, in 1984, when the first Wheelwriter typewriters came out, laser printers weren't quite as inexpensive as they are nowadays.
But it is also true that the quality of printing, including specifically the registration of characters, of the Selectric typewriter was considered to be superlative, and that daisywheel printers, while quite acceptable for typed documents, might have been, even if ever so slightly, not up to the high standard set by the Selectric.
Switching to using daisywheels means that one couldn't use the pre-existing golfball typewriter elements. If that is acceptable, then if one still wanted to use a typeball, clearly the same basic design could continue to be used, but with larger characters, by using a typeball larger in size.
But in addition to the obvious solution, is there some way that a modified type element could share the same diameter as the original IBM Selectric element, so that one could use either an element of the original type, or one of the new ones, and, with the new ones, enjoy the ability to use wider characters?
I considered a number of elaborate possibilities for doing this.
Given that the type element already had a band of characters at an angle of 32 degrees up from the "equator" of the type sphere, why not add a band of characters at an angle 32 degrees down?
Then, instead of 11 rotate positions per side for 44 characters, one could have 9 rotate positions per side for 45 characters, or one could retain the 11 rotate positions, and omit some of the 55 characters to create double-width positions for extra-wide characters.
However, allowing the type element to go through a larger tilt angle creates difficulties.
The teeth on the bottom of the element are used as a rotate detent, to ensure the accurate horizontal alignment of characters during typing. If they are tilted further, it would be more difficult for the notch between teeth to align with the railing used as the tilt detent.
Also, the post on which the element is placed tilts at its center; a collar, held up by two supports, pivots to tilt it. If the element tilts further, the opening at the bottom of the element might not provide room for those two supports.
These obstacles aren't insuperable.
The rotate detent could tilt with the element, and interlock with a guide rail below.
The post on which the element is placed could be put into a cradle, which is pivoted outside the element to the right and left.
In addition to allowing an extra row of characters at the bottom, a further step might be to allow an extra row at the top. Since that row would have its positions packed more closely together than any existing row on the element, one might use only every second character position on that row.
Five and a half rows of characters... is half of eleven. So if one uses 24 rotate positions around the element, as with a 96-character element, one would have three groups of four pairs of columns, each pair of column having 88 characters. So one could either have three shifts of characters for typestyles with no width issues, or one could make all the characters half again as wide and provide 88 characters on the element.
In addition to the top row of characters being narrower, so that only every second position could be used, those characters would be at a shallower angle, and a smaller circle would be above them. And that circle would be higher on the element.
So the element could no longer attach to the post with the normal lever and spring mechanism. I felt that the best way to connect such elements to the typewriter would be to have a hole in the post, and a rod extending down along the center of the element to insert into that hole. A little pin in the center of the rod, which a push-button at the top of the element, suitably recessed, could extend could be used to release the element from the typewriter.
But upon reflection, it seemed that there was a much simpler alternative available.
Simply take the 96-character element, and omit eight characters from it, in two columns on opposite sides.
This would allow four characters, and their shifted counterparts, to be extra-wide.
This works for allowing M and W to be 11 units wide instead of 9 units wide on elements bearing modified versions of the Selectric Composer typefaces.
But another goal I had in mind was to permit the typestyles for the typebar Executive typewriters to be transferred to elements, so that they could be used. As noted above, Executive typestyles used unit sizes of 1/32", 1/36", and 1/45", and these unit sizes are, respectively, 3 units of 1/96" (blue), 2 units of 1/72" (red), and 2 units of the new 1/90" unit size I proposed to add (green).
M, m, and W are 5 units wide in several IBM Executive typestyles; in some, m is only 4 units wide.
But some of the special characters are also 5 units wide in some of those typestyles. As one might expect, @ is 5 units wide in three typestyles; Mid-Century, Charter, and Text. In Mid-Century, % is 5 units wide as well.
As it happens, Mid-Century uses a 1/36" unit, and Charter and Text both use the smallest 1/45" unit.
Unfortunately, though, Arcadia, which uses a 1/32" unit, and which therefore must be accomodated with a provision for its widest characters, allocates 5 units of width to the & character.
Thus, five keys on the typewriter keyboard would seem to have to be allocated to the wide column on the element: M, W, @2, %5, and &7.
Of course, one could just use a different arrangement of characters on the element for Arcadia than is used for other typestyles requiring provision for wide characters, thus in effect just swapping % and & on the keyboard when an element for that typestyle is in use! Or, if the typewriter can detect when such an element is in use (and, indeed, it has to be informed of the same for the assignment of unit widths to each character to be suitably modified for the given typestyle), of course the &7 key could be used for the position usually requested by the %5 key for this category of type element.
However, with the Selectric Composer, the ultimate maximum width for a character, on the 88-character typeball, was 9 units with a unit size of 1/72", which amounts to 1/8". With a unit size of 1/32", four units would give a character width of 1/8", and as that is the width of most of the capital letters, it would not be possible to place all such characters on the "equatorial" band of the element which has the most space. This seems to indicate that the simple scheme of taking a 96-character element, and omitting the characters in one column, as a way to avoid wide characters making unwanted marks on the paper when the characters beside them on the element are typed, which is called "shadow printing".
This problem is even worse than it might seem, as the space between characters on an element, called the "side clearance", for the Selectric Composer was allowed to be smaller than for ordinary Selectric typewriter elements, because while a typewriter might be used to type multiple carbon copies, as the Selectric Composer was to be used to create camera-ready copy for printing, it would not be unreasonable to advise users that it would not work properly if used to type on surfaces which receive too deep an impression. But an element which reproduces typestyles from the IBM Executive typewriter would be expected to work under typewriter conditions, not just under cold-type composition conditions.
A close estimate of the actual dimensions of the problem can be obtained from looking at the typestyles for use on the Mag Card Executive, which made the transition to the 96-character typeball. The widest characters in these typestyles had a width of 7/60", but this width was used for many capital letters, not just the very widest characters M and W, and the proportional elements here were made for typewriter use, not cold-type composition use. The desired width of 1/8" is 7.14% wider than that, or one part in fourteen. Which indicates that if those typestyles work on a 96-character element, 1/8" would not cause problems on an 88-character element after all, but a 96-character element with an extra column of characters is needed because there are also a few five-unit characters in addition to the many four-unit characters that also cause problems.
There is, however, an apparent fallacy in this reasoning. Remember that on the 88-character elements for the Selectric Composer, 9-unit characters in the 1/72" unit size, which were 1/8" wide, only became possible through relaxing the requirements for avoiding shadow printing from those of the Selectric. So 1/8" characters are not something even an 88-character element can normally support. So if 7/60" characters on the 96-character element are equivalent to characters even wider than 1/8" on an 88-character element, that cannot contradict what is already established as fact; instead, the conclusion must be that because the IBM Electronic Typewriter model 50, 65, and 85 were electronic typewriters, it was considered acceptable for the proportionally-spaced typestyles not to work properly if multiple carbon copies were being made - because on an electronic typewriter, at least a later model that has enough memory to store a whole page, one can just tell it to retype a page instead of making carbon copies. Or, it may just be that IBM had been overly cautious in specifying what is required to avoid shadow printing, but that is not a conclusion I would consider it safe to reach without definite evidence.
Given the capability to handle both 88-character and 96-character elements, however, an exotic possibility occurs to me that may yet make it possible to avoid the need for an enlarged element or one with a wider range of tilt positions. Since nearly all the four-unit characters are upper-case, with w as an exception being no problem, as W is one of the five-unit characters provided for, if an empty half-column is sufficient... have the shifted side of the element follow the 88-character spacing of columns, and the unshifted side follow the 96-character spacing of columns, with half a column being left on either side!
However, the Directory typestyle for the IBM Executive has a 1/32" unit size, and it has the additional lower-case letters h and n as having four units. But it has no five-unit characters other than M, m, and W, and so, as in the case of the & in the Arcadia typestyle, the problem could be solved by a different ordering of characters on the typeball - but in this case, the typewriter would have to actually re-assign positions on the type element to keys on the keyboard.
Note that since the valley between teeth used for the tilt detent for a given character is two character positions over from that character, the transition in tooth sizes would be skewed relative to the transition in the widths of character columns, giving the type element an unusual appearance in that respect.
Would an extra half column be enough? In each character column, one can imagine as a starting point a normal pica character, 1/10" wide, or 6/60" wide. On a typeball with proportional spacing, the left, starting, edge of every character as printed would be in the same position; the width of the character changes how the typehead advances after printing. Incidentally, note that this means that on an element which follows the 88-character positioning of columns on one side, and the 96-character positioning of columns on the other, the width available on the left half of the column (left from the perspective of the paper) would be governed by the 96-column standard, reducing the benefits from 88-character spacing.
That extra half column, however, adds only a quarter-column of width to what may be used by the previous character, since otherwise there would be shadow printing due to an intrusion into the plane (or, rather, the platen-shaped cylinder) defined by the surface of the following character.
In the top row of the element, that would seem to be only enough to permit a character with a width of 1/8", which is both 7 1/2 units of 1/60" and four units of 1/32". Given that the middle three rows can already handle characters 7/60" wide, however, we should have 17/120" (0.14167) available at least. But we need 5/32" (0.15625), and we need to have four positions, not just three, with that width. Thus, without a more detailed calculation, it is not certain that this approach, despite its apparent effectiveness in squeezing out every scrap of available character width on the element, would be successful.
Further thought suggests another possible approach.
The amount of space on the left half of a character is determined by the characters with the least available space on the element, as noted above. If one were to omit every second character in the top row of the element, then that minimum would be determined by the second and fourth rows of the element, not the top row. Doing so on a 96-character element, though, would leave only 84 usable characters. While many typewriters did omit the !1 and += keys, to have only 42 keys and 84 characters, that would not really be desirable. Still, one could omit += and 1/4 1/2 instead.
The gain from doing this, though, given a 16-degree tilt angle, can be expressed by the ratio cos(24°)/cos(40°), which is 1.1925..., a fairly large amount.
Increasing the number of columns of characters on each side of the element from 12 to 14 would still leave a net gain in character width compared to the 96-character element, but the characters would still be narrower than on the 88-character element.
But another possibility still remains. If one used smaller teeth at the bottom of the element, one could have 54 teeth around the base; 27 teeth on each side. That would provide thirteen and one-half columns, and omitting half a column might be enough to permit extra-wide characters.
The ratio 27/22 is 1.227..., however, and so this increase in the number of columns still more than offsets the gain in width obtained through omitting every second character in the top row, thus removing the top row as the constraint to width.
Going to 12 1/2 columns instead of 13 1/2 columns, having 50 teeth around the base of the element, would leave one limited to 42 characters, but at least now that there's an omitted half column, in addition to all 42 characters being wider, up to four of them could be extra-wide. Since 25/22 is 1.136..., now ordinary characters can be more than 5% wider than on an 88-character element.
If one put the characters for the following five keys on the typewriter: . , :; "' _- next to each other on the top row, thus filling in two of the alternate spaces otherwise omitted, one could perhaps still retain the extra space on the left side of characters and yet have 88 characters on such an element. However, that would lose the possibility of having compatible elements for foreign languages with slightly different keyboard layouts.
Since the printing face of a character is cylindrical, to match the platen, it extends further out from the printing ball on its sides. Thus, it isn't possible, for example, to have 48 teeth around the base of the element, with two teeth defining the space for a column of characters in the wider three rows of the element, and three teeth giving 1 1/2 column's worth of space to characters in the top row; then, a character in the top row would pose a severe shadow printing hazard to a character in the row below it where the center of that lower character is directly below the left or right edge of the character in the top row. If this weren't so, 48 characters minus 4, with only eight characters in the top row instead of twelve, would work out perfectly.
However, the characters normally placed in the top row of an element are numerals and the letter Z; neither those characters, nor the special characters which are shifts of the digits, have descenders. Thus, as long as the typestyle being used isn't something like Orator, perhaps one could get away with it after all?
Ah, but the characters in the row below will frequently have ascenders in lower case, and will nearly always reach to the full cap height in upper case, so the same shadow printing hazard described as happening from above when characters in the second row are printed would also exist from below when those characters in the top row in half-column positions are printed, and in this case, there would be no remedy available.
Still another option might be to add a mechanism that shifts the print assembly slightly to the right before printing, so that the limitation that the left edge of all characters has to be at the same position on the element is removed. That would add significantly to the space available for characters on an element, but there is the concern as to whether it could be done without slowing down printing speed.
On further consideration, on the one hand, while the earliest IBM Executive typebar typewriters did have 42 keys, later ones had 44 keys. But on the other, the Vari-Typer let one put two type fonts on the machine, and easily switch between them. Given that a Selectric type element is about 1 1/3" in diameter, putting two of them next to each other wouldn't make the moving printhead and the ribbon cartridge too huge; between them, one could place a hammer, and thus have room for one daisywheel as well. Let's say the print points of the two elements are 2" apart, with that for the daisywheel being 1" away from either, in the middle, and the ribbon cartridge moves normally across the paper, with the assembly with three print options moving within so as to avoid any need to back up the ribbon. And place things like tall integral signs on the daisywheel - the "perfect" typewriter might indeed be attainable.
As envisaged so far, the typewriter would have an interesting array of control levers.
One lever would be used to select the unit size, and it would have a series of seven colored triangles next to it:
White: 1/60" Purple: 1/66" Red: 1/72" Orange: 1/78" Yellow: 1/84" Green: 1/90" Blue: 1/96"
with Red, Yellow, and Blue being the three unit sizes the actual Selectric Composer offered.
No attempt would necessarily be made, though, to allow 11-unit, or even 9-unit, Selectric Composer typefaces to be used with the 1/60" or 1/66" units, as it's unlikely they could fit on an element without radical modifications.
Another lever would select which spacing system would be used:
Filled triangle: 6 units each character (Pica with 1/60") Open triangle: 5 units each character (Elite with 1/60") Filled square: 4 units each character (15 characters per inch with 1/60") Filled circle: Mag Card Executive-style proportional Striped triangle: Selectric Composer-style proportional
And then a knob to the lower left of the keyboard, operated by reaching under the typewriter, would select the typestyle to match the different spacings of the IBM Executive typestyles.
This would also allow Selectric Composer-style typefaces to have different widths assigned to the characters depending on the typestyle, instead of all using the same unit scheme.
To allow access to an even larger library of type designs, though, I considered adding another lever to the controls of this ultimate typewriter.
It would have five positions, and it would be the Unit Size Multiplier lever:
A 23 23/900" B 21 21/900" or 7/300" C 19 19/900" D 17 17/900" - 15 Off, unit size 1/60"
This would use gears to expand the unit size by various ratios. Starting from the typewriter unit size of 1/60", this would provide the unit sizes used by the Vari-Typer, allowing its library of proportionally-spaced typefaces to be put on Selectric-style golfball elements!
Additional spacing system options would be needed, making the assortment expand to:
Filled triangle: 6 units each character (Pica with 1/60") Open triangle: 5 units each character (Elite with 1/60") Filled square: 4 units each character (15 characters per inch with 1/60") Open square: 3 units each character 20 characters per inch with 1/60", but also used for Vari-Typer fixed pitch Filled circle: Mag Card Executive-style proportional Two concentric circles: Proportional spacing where M and W, given their full proportional width, are still only 7 units wide. Similar to the unit system used for Linn Boyd Benton's Self-Spacing Type. Used for superscript and subscript fonts. Open circle: Typebar Executive-style proportional. But with a unit size of 1/60", the characters would be tiny, and would serve as second-level superscripts and subscripts. Circle with triangular opening: 5-unit proportional, as with the open circle, but with the unit size of each character multiplied by three. Used with the Blue (1/96") unit size for typebar Executive typestyles with the 1/32" escapement. Note that a digit, indicating the setting for the typestyle knob, would typically appear in the center of this symbol. Circle with square opening: 5-unit proportional, with the unit size of each character multiplied by two. Used with the Red (1/72") unit size for the 1/36" escapement, and with the Green (1/90") unit size for the 1/45" escapement. Note that a digit, indicating the setting for the typestyle knob, would typically appear in the center of this symbol. Hexagon: 4-unit proportional, typically used with the 1/60" unit size, but with the Unit Size Multiplier lever in one of the four lettered positions. Typically, a letter, indicating the Unit Size Multiplier lever setting, would appear in the center of this symbol. Used for Vari-Typer typestyles. Striped triangle: Selectric Composer-style proportional
And, of course, since the Blickensderfer didn't offer proportional spacing, there would be no real difficulty in transferring its library of typestyles, as well as the fixed-spacing Hammond ones, to elements as well, not counting, of course, copyright and trademark issues; there is still a company around which bought the company that bought Vari-Typer.
Presumably, it would not be difficult to include, as a standard feature in the programming of an electronic typewriter, the ability to compose characters after the manner of a Korean typewriter, although, with at most 96 characters on an element, it might be necessary to accept the sort of limitations experienced on ordinary Korean manual typewriters of the 3-set type, as opposed to the higher-quality printing done by typewriters of the 5-set type.
With the Unit Size Multiplier giving the unit sizes for the Vari-Typer, however, it has occurred to me that a feature of this nature can extend the typewriter's capabilities even further.
Instead of expanding the 1/60" unit in ratios of 15:17, 15:19, 15:21, and 15:23, shrinking the 1/72" unit in ratios of 18:n, where n is anything from 17 3/4 down to 4 in steps of 1/4 would allow the unit size for every set width from 17 3/4 points to 4 points, and the plain 1/72" unit with the feature disengaged would be the unit size for a set width of 18 points.
With that feature added, one could add the diamond, usually colored red for the 1/72" unit used with it, to indicate 18 unit typestyles... and thus a mere typewriter would now be able to do the work of a Monotype caster!
Finally, to round out the insanity of trying to make a single element typewriter do what laser printers can do much better, I envisaged having the mechanism handle cylindrical elements, such as those of Adler (Royal, Triumph) typewriters which involved (I had mistakenly thought) an up-and-down motion instead of a tilt motion in addition to a rotate motion.
This was not because I was aware of an important family of typefaces offered by these typewriters to which access was important. Instead, it was so that elements providing extra-tall characters could be provided.
So one could have initial letters to begin chapters!
But that, of course, was not the real reason this was tempting. Rather, a special element with different sizes of integral signs, product signs, summation signs, contour integral signs, and so on was what excited my desire.
And, incidentally, that would also solve the issue of making enough space available on the element for wider characters without a hazard of shadow printing in a simple and complete fashion, so the typestyles for typebar Executive typewriters, for example, would pose no issue whatever.
However, on closer examination of images of printing elements for those typewriters, I see they did not have a true up-and-down motion like the type cylinders in a Model 33 Teletype, or like the elements in many old typewriters. Instead, they had a shape like that of the elements for the Facit single-element typewriters, like the middle part of a football - there was still a tilt motion, but it was about a center that lay behind the element, away from the paper, instead of in the center of the element, as yielded a spherical shape for the IBM golfball.
More precisely, the shape of the base surface on an Adler element was cylindrical, which is what led to my confusion, but the characters that projected outward from that surface still projected less far at the top and bottom - and at an angle as well, so that the base being cylindrical instead of a football-like one from which all the characters would extend by an equal amount was, basically, a cosmetic choice; somewhat in the same category of IBM elements having the characters on a spherical base, while GP elements had the characters on each row on a base that was in the shape of part of a cone.
That allows wider characters, but not tall characters that occupy multiple horizontal rows on the element.
But that just means that the complexity of providing compatibility to existing Adler elements is avoided; one could still design cylindrical elements that fit in the typewriter as an alternate type of element. And, as that would not require changing over to using a cradle for the tilt motion and other such changes, it would in many ways be simpler than trying to use a modified spherical element with a greater range of tilt motion, even though it involves the radical step of adding provision for a completely different motion as an alternative to the tilt motion.