Post by carlclaunch on Jul 29, 2014 3:31:51 GMT
The IBM Selectric typewriter mechanism was used as the console interface for the 360 era computers - including the IBM 1130 - plus many types of terminals used for banking, airline reservations, and timesharing access. The typewriter itself was not electronic, having only a power cord, power switch and motor as non-mechanical parts. IBM added solenoids and microswitches to control the mechanism and used it for the terminal and computer purposes. They also strengthened parts of the mechanism for more continuous use, rather than the sporadic operation experienced when used as a typewriter.
The engineering of electronic control was initially done for use with the previous mainframe generation, the 1401 and kin, as part of the 1050 system. It had paper tape readers and paper tape writers (1054/1055), plus typewriters with and without keyboards (1052 and 1053). Generically, these were the I/O Selectrics, but used in a variety of somewhat different enclosures and under a wide range of model numbers. The consoles for the 360 series were the 1052 (and sometimes 1053 when a print-only version was desired). The IBM 1130 mainframe used a 1053 for console output but leveraged the keyboard from the 029 keypunch as the keyboard.
Terminals built from I/O selectrics include the 2741 timesharing/access terminal, the 2970 airline/hotel reservations terminal, and quite a few other models for roles such as banking terminals that could print entries in a savings passbook. Later, IBM put the I/O selectric into the Mag Tape system, their first 'word processor' system.
These all used slight variations on the I/O Selectric mechanism.
IBM also licensed the use of the mechanism (and sold the mechanical core of the typewriter) to many companies for use in word processing systems and computers. Many of these had solenoids and microswitches added to a more standard typewriter mechanism, not the heavy duty version in the I/O selectric, and these adaptations were engineered by the various companies rather than IBM. These used the 88 character type elements as used with the Selectric I and II typewriters.
Somewhat later, IBM sold their Memory Typewriters which were selectric based, but with solenoids and microswitches, allowing users to edit and save documents in an onboard memory. They were not as heavy-duty as the I/O Selectrics since they would be used as typewriters instead of terminals, but they were a bit more rugged than standard typewriter models. These still used the round 'golfball' type elements and made use of the same 88 character Selectric I mechanism.
IBM evolved their Selectric typewriters to a new 96 character type element as the Selectric III range. No I/O Selectric models were built based on this type element. However, IBM did have an equivalent to the Memory Typewriter using this new type element, called the IBM Electronic typewriters.
Anyone looking to make a realistic 360 era computer system would want a Selectric based console like the 1052. These are extremely hard to find, the few that seem to come on the market are from word processing systems built and sold by other companies but that used an IBM mechanism inside. I found one that was sold as a Dura (later Itel) word processor, while Lawrence found one from a UK word processing maker. The Dura version had custom solenoids and microswitches, while Lawrence had a machine that was almost a pure I/O Selectric. Adapting these requires some logic to manage the interlocks and timing of the mechanism - holding off on typing the next character until the mechanical parts are done with their movement, for example - but is not a challenging task.
The IBM Memory typewriters, such as a Memory 100 model I acquired, have included electronic logic inside the typewriter to manage all the timing, simplifying the work of interfacing these to a computer. You simply toggle a small number of lines to encode the character you wish to type, including commands to backspace, tab, return the carrier and so forth. Very simple interfacing. Mechanically it still works the same way as an I/O Selectric and uses the same 88 character type elements.
The IBM Electronic typewriters, which I had first used to attempt a 1053 substitute for an 1130 system replica, uses the same kind of simple toggling of lines to request characters or other commands to be executed. However, these use the newer Selectric III 96 character type elements and a very different internal mechanism to tilt and rotate the ball. These are much more fragile and prone to wear or breaking, especially the nylon roller cams at the heart of the system. The earlier I/O and Memory mechanisms used metal levers, pulleys and cams.
The type elements come in two arrangements or coding schemes - bcd/pttc and correspondence. Correspondence is the name for the typewriter arrangement, where a particular key on the Selectric typewriter would tilt and rotate the ball to a specific place that had the letter desired. BCD/pttc placed the letters in a different arrangement. Pressing the same key on one of these machines would tilt and rotate to a different place on the ball compared to where a normal typewriter would pick for the same key. If you put a correspondence ball on a bcd terminal, or vice versa, you would not get the letters you expected as you typed.
The 360 and 1130 computers used the BCD arrangement. The 2741 terminals came in both types - thus users had to match their type elements to the terminal type. The timesharing software of the time would usually type out an initial welcome message; the manuals for users would contain both the intended string you would see and the string resulting frpm using the wrong ball type - if you got a string of seeming gibberish you read in the manual that this meant you had to switch the typeball and informed you of the type you needed.
Using a typewriter built for one arrangement with a computer in the other arrangement requires some trickery. It can also create unrealistic behavior - as for example when a given character you are typing is placed on the opposite side of the type ball between bcd and correspondence. This can cause extraneous shift actions - e.g. upper to lower - which are noticeable to users and may cause timing problems if a shift is not expected. A shift is an extra cycle of the typewriter mechanism which is interspersed between the cycles printing characters.
One could take a machine built for one arrangement and use it with the other, by translating the intended output letter to the letter on this machine that sits in the same tilt/rotate location as the intended letter inhabits on the opposite ball type. Similarly, you could take the tilt/rotate codes from pressing the keyboard keys and convert them to the tilt/rotate value of the other arrangement. Thus, I have taken a Memory typewriter that was wired for correspondence, drive it by the tilt/rotate codes from my 1130 which is in BCD mode, which twists the type ball to the BCD position. I stuck a BCD type element on it and get the output I want. Since I don't use the keyboard for my replica, I don't have to deal with the correspondence coding of keypresses, but if I were to wire it up for input I would just translate the codes so they became the proper BCD codes.
Another complication exists, however, when putting a BCD ball on a correspondence mechanism. A few characters are so small in cross sectional area that they make particularly deep and dark impressions - characters such as comma and period. Many of the newer models have the ability to moderate the velocity of the typeball striking the paper, using low velocity for those small characters. They do this, unfortunately, by the specific tilt and rotate value that is assigned to these characters. When you switch to a BCD ball on a correspondence typewriter, it will print the period and comma at full velocity but make certain regular characters faint because it will strike them at low velocity. These are the regular letters that inhabit a tilt and rotate position on a BCD ball where the same position has small characters on correspondence machines. The solution is to intercept and modify the velocity control signals, or to disable the low-velocity operation. The 1052 and 1053 were all mono-velocity mechanisms and would just print those small characters with a deep dark image, so disabling low-velocity produces more realistic printed output.
I will post examples of the connections to the Memory and Electronic typewriters, while Lawrence has documented how to connect to an I/O Selectric terminal. I am happy to help anyone interface, debug and adapt these typewriters to your 360 era computer reproductions.
The engineering of electronic control was initially done for use with the previous mainframe generation, the 1401 and kin, as part of the 1050 system. It had paper tape readers and paper tape writers (1054/1055), plus typewriters with and without keyboards (1052 and 1053). Generically, these were the I/O Selectrics, but used in a variety of somewhat different enclosures and under a wide range of model numbers. The consoles for the 360 series were the 1052 (and sometimes 1053 when a print-only version was desired). The IBM 1130 mainframe used a 1053 for console output but leveraged the keyboard from the 029 keypunch as the keyboard.
Terminals built from I/O selectrics include the 2741 timesharing/access terminal, the 2970 airline/hotel reservations terminal, and quite a few other models for roles such as banking terminals that could print entries in a savings passbook. Later, IBM put the I/O selectric into the Mag Tape system, their first 'word processor' system.
These all used slight variations on the I/O Selectric mechanism.
IBM also licensed the use of the mechanism (and sold the mechanical core of the typewriter) to many companies for use in word processing systems and computers. Many of these had solenoids and microswitches added to a more standard typewriter mechanism, not the heavy duty version in the I/O selectric, and these adaptations were engineered by the various companies rather than IBM. These used the 88 character type elements as used with the Selectric I and II typewriters.
Somewhat later, IBM sold their Memory Typewriters which were selectric based, but with solenoids and microswitches, allowing users to edit and save documents in an onboard memory. They were not as heavy-duty as the I/O Selectrics since they would be used as typewriters instead of terminals, but they were a bit more rugged than standard typewriter models. These still used the round 'golfball' type elements and made use of the same 88 character Selectric I mechanism.
IBM evolved their Selectric typewriters to a new 96 character type element as the Selectric III range. No I/O Selectric models were built based on this type element. However, IBM did have an equivalent to the Memory Typewriter using this new type element, called the IBM Electronic typewriters.
Anyone looking to make a realistic 360 era computer system would want a Selectric based console like the 1052. These are extremely hard to find, the few that seem to come on the market are from word processing systems built and sold by other companies but that used an IBM mechanism inside. I found one that was sold as a Dura (later Itel) word processor, while Lawrence found one from a UK word processing maker. The Dura version had custom solenoids and microswitches, while Lawrence had a machine that was almost a pure I/O Selectric. Adapting these requires some logic to manage the interlocks and timing of the mechanism - holding off on typing the next character until the mechanical parts are done with their movement, for example - but is not a challenging task.
The IBM Memory typewriters, such as a Memory 100 model I acquired, have included electronic logic inside the typewriter to manage all the timing, simplifying the work of interfacing these to a computer. You simply toggle a small number of lines to encode the character you wish to type, including commands to backspace, tab, return the carrier and so forth. Very simple interfacing. Mechanically it still works the same way as an I/O Selectric and uses the same 88 character type elements.
The IBM Electronic typewriters, which I had first used to attempt a 1053 substitute for an 1130 system replica, uses the same kind of simple toggling of lines to request characters or other commands to be executed. However, these use the newer Selectric III 96 character type elements and a very different internal mechanism to tilt and rotate the ball. These are much more fragile and prone to wear or breaking, especially the nylon roller cams at the heart of the system. The earlier I/O and Memory mechanisms used metal levers, pulleys and cams.
The type elements come in two arrangements or coding schemes - bcd/pttc and correspondence. Correspondence is the name for the typewriter arrangement, where a particular key on the Selectric typewriter would tilt and rotate the ball to a specific place that had the letter desired. BCD/pttc placed the letters in a different arrangement. Pressing the same key on one of these machines would tilt and rotate to a different place on the ball compared to where a normal typewriter would pick for the same key. If you put a correspondence ball on a bcd terminal, or vice versa, you would not get the letters you expected as you typed.
The 360 and 1130 computers used the BCD arrangement. The 2741 terminals came in both types - thus users had to match their type elements to the terminal type. The timesharing software of the time would usually type out an initial welcome message; the manuals for users would contain both the intended string you would see and the string resulting frpm using the wrong ball type - if you got a string of seeming gibberish you read in the manual that this meant you had to switch the typeball and informed you of the type you needed.
Using a typewriter built for one arrangement with a computer in the other arrangement requires some trickery. It can also create unrealistic behavior - as for example when a given character you are typing is placed on the opposite side of the type ball between bcd and correspondence. This can cause extraneous shift actions - e.g. upper to lower - which are noticeable to users and may cause timing problems if a shift is not expected. A shift is an extra cycle of the typewriter mechanism which is interspersed between the cycles printing characters.
One could take a machine built for one arrangement and use it with the other, by translating the intended output letter to the letter on this machine that sits in the same tilt/rotate location as the intended letter inhabits on the opposite ball type. Similarly, you could take the tilt/rotate codes from pressing the keyboard keys and convert them to the tilt/rotate value of the other arrangement. Thus, I have taken a Memory typewriter that was wired for correspondence, drive it by the tilt/rotate codes from my 1130 which is in BCD mode, which twists the type ball to the BCD position. I stuck a BCD type element on it and get the output I want. Since I don't use the keyboard for my replica, I don't have to deal with the correspondence coding of keypresses, but if I were to wire it up for input I would just translate the codes so they became the proper BCD codes.
Another complication exists, however, when putting a BCD ball on a correspondence mechanism. A few characters are so small in cross sectional area that they make particularly deep and dark impressions - characters such as comma and period. Many of the newer models have the ability to moderate the velocity of the typeball striking the paper, using low velocity for those small characters. They do this, unfortunately, by the specific tilt and rotate value that is assigned to these characters. When you switch to a BCD ball on a correspondence typewriter, it will print the period and comma at full velocity but make certain regular characters faint because it will strike them at low velocity. These are the regular letters that inhabit a tilt and rotate position on a BCD ball where the same position has small characters on correspondence machines. The solution is to intercept and modify the velocity control signals, or to disable the low-velocity operation. The 1052 and 1053 were all mono-velocity mechanisms and would just print those small characters with a deep dark image, so disabling low-velocity produces more realistic printed output.
I will post examples of the connections to the Memory and Electronic typewriters, while Lawrence has documented how to connect to an I/O Selectric terminal. I am happy to help anyone interface, debug and adapt these typewriters to your 360 era computer reproductions.