INVESTIGATION INTO TYPEBALL FABRICATION
Quite a few problems would disappear if I could get my hands on the two typeballs used with the 1130 system - one specially made for running APL on the machine and the other for operation with the usual operating system DMS2.
These are rarer than hen's teeth, as the saying goes, which makes getting an authentic one quite unlikely. On the other hand, if I could make or modify a typeball to match the character layout of the 1130, my interface to the Selectric Memory 50 typewriter would be trivial to accomplish. Same for the APL typeball, even harder to find than the regular typeball.
As well, several friends who are building 360 system replicas and want to make use of IO Selectric or Memory Typewriters as the console will face the same issue - the so called 'correspondence' typeball is for letters on typewriters, but they really need the computer oriented one. I am sure that each would be very happy to get their hands on a proper typeball for their projects.
Finally, anyone running an IBM made 1130 system (just a handful are still operational) would love to have a source for a backup typeball and the APL ball would be appealing as few have one.
The typeball is roughly spherical in shape, made of nylon with a metallic coating added, reportedly to make the new Selectric typewriters appear less 'cheap'. It has 88 locations around the waist of the ball, each with one character that will be slammed into a ribbon, which in turn strikes the paper rolled partway around a cylindrical 'platen'.
The characters are arranged in four circumferential bands of 22 each. The typeball is fastened to a vertical axle and that axle can be tilted to one of four levels. In addition to the tilt, the mechanism can rotate the ball to 11 positions, five to the left of its rest position and five to the right. A final operation an spin the ball 180 degrees to give the other half sphere its 11 positions.
The spin of 180 degrees corresponds to the 'shift' key on the typewriter - one side is generally for upper case (capital) letters and the other side for lower case. The spin from one case to the other takes its own cycle that is done before typing the desired characters from that face.
When a character is printed, the ball will rotate to one of the 11 positions while tilting to one of the four bands, thus selecting which of the 44 characters on this side of the ball will be typed. The face of the font is concave, higher at the top and bottom of the character than in the mid-line, because the paper and platen are curved away above and below the center of where the ball will strike.
In about a hundredth of a second the ball rotates, tilts, stabilizes in position and then is thrown forward to strike the ribbon, paper and platen. As it bounces back, the carrier assembly that holds the typeball starts its move one space to the right, finishing all movement before 65 milliseconds have elapsed.
For the type to look good on the paper, each character on the ball has to be lined up with the same baseline as the others, else the line of print will look wavy. The character can't be too far left or right of the ideal target spot on the line, or letters will crowd together or gaps will be noticeable. Thus, the precision of placement for the type on the ball is fairly exacting. The ball has enormous acceleration and deceleration to accomplish its movements about 16 times per second. It must not deform in order to keep the character shaped properly.
This is why it is made of fairly thin nylon, with the approximate spherical surface actually composed of 88 rectangles, slightly concave top to bottom and flat side to side, which gives a slightly dimpled look that justifies the nickname 'golf ball'. Had it been left white or off white nylon, the similarity would be even more stark.
About the bottom 80-90 degrees of the sphere is open, so it can be placed on its vertical axle. The bottom edges are jagged teeth, used by the typewriter mechanism to lock the typeball securely in position before it strikes the ribbon and then releasing the tooth before the ball has to tilt and rotate for the next letter.
The top of the ball is flattened, with the hole drilled down through it for the axle, then a rounded plastic cap is attached with a lever mechanism that secures the typeball on the axle. It seems probably that the typeball body itself is made by injection of molten nylon into a metal mold, then machining of the axle hole and rotation point up inside the hollow of the ball.
Injection molding is done with two mold halves that are held together with tons of pressure while molten plastic is forced in at very high pressure and held pressurized until the plastic solidifies. The top and bottom of the mold are separated and the new part comes out. The line where the top and bottom molds meet is called the parting line - it is visible on any injection molded object if you look carefully.
The incomplete sphere of the typeball would be easy to make in a mold if it didn't have type characters where it does. If the widest point of the ball - the midline around the sphere - is where the mold halves meet then the molds will pull apart easily and the part will pop out readily. However, there will be a narrow line around that circumference. If you look at any typeball, you will see that the midline is right in the middle of a band of 22 characters. That would put a tiny line artifact across all 22 of those characters. That clearly doesn't happen.
Instead, you can see subtle parting lines vertically separating each of the 22 rotation points around the ball. This means IBM probably used sliding inserts in the mold. An insert would form the four characters of the four tilt bands for one of the 22 rotational positions. A typeball mold would have 22 inserts that form a vertially concave arc on the inside, these inserts slide on tracks radially inward to meet together forming the shape of the ball. They slide away from the center when the mold is opened, releasing the finished object. A smaller number of inserts is possible, as few as two sliding together as half-spherical concave openings.
It is challenging but not impractical to mill a mold with several sliding inserts shaped to form the typeball. The difficulty is forming the shape of the characters as depressions in the sliding insert. Look closely at a typical letter from a typewriter and you will see that many of the inside edges of the letter have sharp ninety degree turns. Milling uses circular cutters rotating rapidly to make patterns in metal - but it is very hard to form a 90 degree turn with a circle have a non-zero radius.
You only need the top line of the font to be shaped properly, the rest doesn't cause an impression through the ribbon to the paper. A 5D mill could concievable bring the mill in 'sideways' so the flat bottom of the mill formed the almost vertical wall of the font edge, but it is still extraordinarily challenging if the shape of the characters didn't make it impossible for geometric reasons.
I think it possible to make a mold for the inserts based on impressions from an IBM (or GP who made compatible typeballs) ball, so that an insert could be cast out of molten aluminum. The sequence would be:
- First take a real typeball that will be copied and embed it in silicone mold making solution. That gives a 'negative' mold of the typeball but captures the shape of the font perfectly.
- Second, plastic resin is poured into the mold to make a 'positive' replica of the typeball with its fonts - all perfectly detailed but made of brittle polyurethane resin.
- Step three is to cut the replica to get a half sphere that will form one of the two inserts. It probably will take two polyurethane positive copies to be sure to have 100% of the half sphere for each insert.
- Fourth put the resin half-sphere in a container shaped like the eventual slider insert, but oversize to allow milling to precisely the final dimensions.
- Fifth, pour hot wax into the container and let it solidify, making a wax 'negative' of the original typeball but in the shape of the desired metal
- Sixth, use a solvent that will melt the resin sphere but not disturb the wax around it. This leaves the wax negative without ruining the depressions forming the type.
- Seventh, built a mold of refractory material around this wax, using something like sand and plaster of paris, that can set into a shape that has the wax insert shape totally embedded inside it. There will be some wax channels formed to allow access of molten metal down into the mold, once the wax disappears.
- Eighth, heat the mold to cause all the wax to be melted and evaporated, leaving a hollow inner space in the mold that is shaped like our desired slider insert.
- Ninth, pour molten aluminum into the mold to cast the insert. It will have the exact shape of the typeface and ball surface, but the remaining edges are oversized to allow us to mill it to exact dimensions.
- Tenth, mill down the insert to fit into the two part mold, slide in and out radially as the mold is closed and opened, and to meet exactly to form the two hemispheres of the typeball. The inserts meet each other, but also meet the top mold halve which forms the top part of the typeball and the inserts meet the bottom half which forms the teeth and hollow underside.
Now, if all of that is done well and the quality and fit seem good enough, it is time to test it with some nylon in the injection molding machine.
As you can see, this is a complex process that should result in well formed type in a nylon part of the proper shape. Finishing if off with the drilled hole for the axis and some other finesses gives us a body ready to have a cap attached.
Not sure it is worth it, but it seems like it is achievable within the tools and precision available to me - although just barely and probably with several wasted attempts. As a result, I will continue ahead with the typewriter and its existing 'correspondence' typeballs but this will be in my back pocket.