IBM 1620


The IBM 1620 was announced by IBM on October 21, 1959, and marketed as an inexpensive "scientific computer". After a total production of about two thousand machines, it was withdrawn on November 19, 1970. Modified versions of the 1620 were used as the CPU of the IBM 1710 and IBM 1720 Industrial Process Control Systems.
Being variable-word-length decimal, as opposed to fixed-word-length pure binary, made it an especially attractive first computer to learn on and hundreds of thousands of students had their [|first experiences with a computer] on the IBM 1620.
Core memory cycle times were 20 microseconds for the Model I, 10 microseconds for the Model II. The Model II was introduced in 1962.

Architecture

Memory

The IBM 1620 was a variable "word" length decimal computer with a magnetic-core memory that could hold:
The above was in the Model I. The Model II deployed the IBM 1625 core-storage memory unit, whose memory cycle time was halved by using faster cores, compared to the Model I's : to 10 µs.
While the 5-digit addresses of either model could have addressed 100,000 decimal digits, no machine larger than 60,000 decimal digits was ever marketed.

Memory access

Memory was accessed two decimal digits at the same time. Each decimal digit was 6 bits, composed of an odd parity Check bit, a Flag bit, and four BCD bits for the value of the digit in the following format:
C F 8 4 2 1
The Flag bit had several uses:
In addition to the valid BCD digit values there were three special digit values :
C F 8 4 2 1
1 0 1 0 Record Mark
1 1 0 0 Numeric Blank
1 1 1 1 Group Mark
Instructions were fixed length, consisting of a 2-digit "op code", a 5-digit "P Address", and a 5-digit "Q Address". Some instructions, such as the B instruction, only used the P Address, and later smart assemblers included a "B7" instruction that generated a 7-digit branch instruction.
Fixed-point data "words" could be any size from two decimal digits up to all of memory not used for other purposes.
Floating-point data "words" could be any size from 4 decimal digits up to 102 decimal digits.
The Fortran II compiler offered limited access to this flexibility via a "Source Program Control Card" preceding the fortran source in a fixed format:
*ffkks
The * in column one, ff the number of digits for the mantissa of floating-point numbers, kk the number of digits for fixed-point numbers and s is to specify the memory size of the computer to run the code if not the current computer: 2, 4, or 6 for memories of 20,000 or 40,000 or 60,000 digits.
The machine had no programmer-accessible registers: all operations were memory to memory.

Character and op codes

The table below lists Alphameric mode characters.
The table below lists numeric mode characters.

Invalid character

The Model I used the Cyrillic character Ж on the typewriter as a general purpose invalid character with correct parity. In some 1620 installations it was called a SMERSH, as used in the James Bond novels that had become popular in the late 1960s. The Model II used a new character ❚ as a general purpose invalid character with correct parity.

Architectural difficulties

Although the IBM 1620's architecture was very popular in the scientific and engineering community, computer scientist Edsger Dijkstra pointed out several flaws in its design in EWD37, "A review of the IBM 1620 data processing system". Among these are that the machine's Branch and Transmit instruction together with Branch Back allow a grand total of one level of nested subroutine call, forcing the programmer of any code with more than one level to decide where the use of this "feature" would be most effective. He also showed how the machine's paper tape reading support could not properly read tapes containing record marks, since record marks are used to terminate the characters read in storage.

Software

IBM supplied the following software for the 1620:
The Monitors provided disk based versions of 1620 SPS IId, FORTRAN IId as well as a DUP. Both Monitor systems required 20,000 digits or more of memory and one or more [|1311 disk drives].
A collection of IBM 1620 related manuals in PDF format exists at bitsavers.

1620 Non-decimal arithmetic

Since the Model I used in-memory lookup tables for addition/subtraction, limited bases unsigned number arithmetic could be performed by changing the contents of these tables, but noting that the hardware included a ten's complementer for subtraction.
To do fully signed addition and subtraction in bases 2 to 4 required detailed understanding of the hardware to create a "folded" addition table that would fake out the complementer and carry logic.
Also the addition table would have to be reloaded for normal base 10 operation every time address calculations were required in the program, then reloaded again for the alternate base. This made the "trick" somewhat less than useful for any practical application.
Since the Model II had addition and subtraction fully implemented in hardware, changing the table in memory could not be used as a "trick" to change arithmetic bases. However an optional special feature in hardware for octal input/output, logical operations,
and base conversion to/from decimal was available.
Although bases other than 8 and 10 were not supported, this made the Model II very practical for applications that needed to manipulate data formatted in octal by other computers.

Model I

The IBM 1620 Model I
was the original. It was produced as inexpensively as possible, to [|keep the price low].
The IBM 1620 Model II was a vastly improved implementation, compared to the original Model I. The Model II was introduced in 1962.
While the Lower console for both the Model 1 and the Model 2 IBM 1620 systems had the same lamps and switches, the Upper console of the pair were partly different.

Upper console

The balance of the Upper console was the same on both models:
The Model I console typewriter was a modified Model B1, interfaced by a set of relays, and it typed at only 10 characters per second.
There were a set of instructions that wrote to the typewriter, or read from it. The general RN and WN instructions had assembly language mnemonics that supplied the "device" code in the second address field, and the control code in the low-order digit of the second address field.
To simplify input and output, there were two instructions:
The Model II used a modified Selectric typewriter, which could type at 15.5 cps a 55% improvement.

Peripherals

Available peripherals were:
The standard "output" mechanism for a program was to punch cards, which was faster than using the typewriter. These punched cards were then fed through an IBM 407 mechanical calculator which could be programmed to print two cards, thus being able to use the additional print columns available on the 407. All output was synchronous, and the processor paused while the I/O device produced the output, so the typewriter output could completely dominate program running time.
A faster output option, the IBM 1443 printer was introduced May 6, 1963, and its 150–600 lines/minute capability was available for use with either model of the 1620.
It could print 120 or 144 columns. The character width was fixed, so it was the paper size that changed; the printer printed 10 characters to the inch, so a printer could print a maximum of 12 inches or 14.4 inches of text. In addition, the printer had a buffer, so the I/O delay for the processor was reduced. However, the print instruction would block if the line had not completed.

Operating procedures

The "operating system" for the computer constituted the human operator, who would use controls on the computer console, which consisted of a front panel and typewriter, to load programs from the available bulk storage media such as decks of punched cards or rolls of paper tape that were kept in cabinets nearby. Later, the model 1311 disc storage device attached to the computer enabled a reduction in the fetch and carry of card decks or paper tape rolls, and a simple "Monitor" operating system could be loaded to help in selecting what to load from disc.
A standard preliminary was to clear the computer memory of any previous user's detritus being magnetic cores, the memory retained its last state even if the power had been switched off. This was effected by using the console facilities to load a simple computer program via typing its machine code at the console typewriter, running it, and stopping it. This was not challenging as only one instruction was needed such as 160001000000, loaded at address zero and following. This meant transmit field immediate to address 00010 the immediate constant field having the value 00000, decrementing source and destination addresses until such time as a digit with a "flag" was copied. This was the normal machine code means of copying a constant of up to five digits. The digit string was addressed at its low-order end and extended through lower addresses until a digit with a flag marked its end. But for this instruction, no flag would ever be found because the source digits had shortly before been overwritten by digits lacking a flag. Thus the operation would roll around memory filling it with all zeroes until the operator grew tired of watching the roiling of the indicator lights and pressed the Instant Stop - Single Cycle Execute button. Each 20,000 digit module of memory took just under one second to clear. On the 1620 II this instruction would NOT work. Instead there was a button on the console called Modify which when pressed together with the Check Reset button, when the computer was in Manual mode, would set the computer in a mode that would clear all of memory in a tenth of one second regardless of how much memory you had; when you pressed Start. It also stopped automatically when memory was cleared, instead of requiring the operator to stop it.
Other than typing machine code at the console, a program could be loaded via either the paper tape reader, the card reader, or any disk drive. Loading from either tape or disk required first typing a "bootstrap" routine on the console typewriter.
The card reader made things easier because it had a special Load button to signify that the first card was to be read into the computer's memory and executed this is the "bootstrap" process that gets into the computer just enough code to read in the rest of the code that constitutes the loader that will read in and execute the desired program.
Programs were prepared ahead of time, offline, on paper tape or punched cards. But usually the programmers were allowed to run the programs personally, hands-on, instead of submitting them to operators as was the case with mainframe computers at that time. And the console typewriter allowed entering data and getting output in an interactive fashion, instead of just getting the normal printed output from a blind batch run on a pre-packaged data set. As well, there were four program switches on the console whose state a running program could test and so have its behavior directed by its user. The computer operator could also stop a running program then investigate or modify the contents of memory: being decimal-based, this was quite easy; even floating-point numbers could be read at a glance. Execution could then be resumed, from any desired point. Aside from debugging, scientific programming is typically exploratory, by contrast to commercial data processing where the same work is repeated on a regular schedule.

Console

The most important items on the 1620's console were a pair of buttons labeled Insert & Release, and the console typewriter.
The typewriter is used for operator input/output, both as the main console control of the computer and for program controlled input/output. Later models of the typewriter had a special key marked R-S that combined the functions of the console Release & Start buttons. Note: several keys on the typewriter did not generate input characters, these included Tab and Return.
The next most important items on the console were the buttons labeled Start, Stop-SIE, and Instant Stop-SCE.
For program debugging there were the buttons labeled Save & Display MAR.
When a Branch Back instruction was executed in Save mode, it copied the saved value back to the program counter and deactivated Save mode.
This was used during debugging to remember where the program had been stopped to allow it to be resumed after the debugging instructions that the operator had typed on the typewriter had finished. Note: the MARS register used to save the program counter in was also used by the Multiply instruction, so this instruction and the Save mode were incompatible! However, there was no need to use multiply in debugging code, so this was not considered to be a problem.
"Breakpoint" procedureNotes
Press Stop-SIEStop the computer at the end of the current instruction.
Press SaveSave the address to resume execution at.
Press InsertUnlocks typewriter keyboard and shifts into numeric mode.
Type 35xxxxx0010036xxxxx0010042
xxxxx is the address that you plan to set the breakpoint at.
Press ReleaseLocks typewriter keyboard.
Press StartBegin execution. Allow the 12 digit instruction to print out.
Press ReleaseStops the Dump Numeric.
Press StartBegin execution.
Type 48Replace the opcode of the instruction to "break" at with a Halt opcode.
Press ReleaseLocks typewriter keyboard.
Press StartResume execution. Wait until the computer halts at the "breakpoint".
Press InsertUnlocks typewriter keyboard and shifts into numeric mode.
Type 36xxxxx0010049xxxxx
xxxxx is the address that you previously set the breakpoint at, you are now going to clear it.
Press ReleaseLocks typewriter keyboard.
Press StartBegin execution.
Type oooo is the 2 digit opcode the original 12 digit instruction previously printed out.
Press ReleaseLocks typewriter keyboard.
Press Stop-SIEThe machine is now ready to resume execution from the location of the "breakpoint". You can perform any required debugging actions now, before continuing.

All of main memory could be cleared from the console by entering and executing a transfer instruction from address to address +1, this would overwrite any word mark, that would normally stop a transfer instruction, and wrap around at the end of memory. After a moment, pressing Stop would stop the transfer instruction and memory would be cleared.

IBM 1621/1624 Paper Tape reader/punch

The IBM 1621 Paper Tape Reader could read a maximum of 150 characters per second;
The IBM 1624 Paper Tape Punch could output a maximum of 15 characters/second.
Both units:
The 1621 Tape Reader and 1624 Tape Punch included controls for:
Bootstrap procedureNotes
Press InsertUnlocks typewriter keyboard and shifts into numeric mode.
Type 36xxxxx0030049yyyyy
xxxxx is the address to load the tape into.
yyyyy is the address to begin execution.
Press ReleaseLocks typewriter keyboard.
Press StartBegin execution.

IBM 1622 Card reader/punch

The IBM 1622 Card reader/punch could:
The 1622's controls were divided into three groups: 3 punch control rocker switches, 6 buttons, and 2 reader control rocker switches.
Punch Rocker switches:
Buttons:
Reader Rocker switches:
The 1311 Disk drive controls.
Bootstrap procedureNotes
Press InsertUnlocks typewriter keyboard and shifts into numeric mode.
Type 3400032007013600032007024902402
x
y1963611300102		x Specifies source of Monitor control cards: 1=typewriter, 3=paper tape, 5=cards
y Specifies disk drive on which Monitor resides: 1, 3, 5, 7
02402 is the address of the entry point of the Monitor program.
Press ReleaseLocks typewriter keyboard.
Press StartBegin execution.

Restart procedureNotes
Press InsertUnlocks typewriter keyboard and shifts into numeric mode.
Type 490225FLG6
02256̅ is the address of the location containing the address of the restart point of the Monitor program.
Note: this procedure assumes the Monitor is already loaded in memory
Press ReleaseLocks typewriter keyboard.
Press StartBegin execution.

General

The FORTRAN II compiler and SPS assembler were somewhat cumbersome to use by modern standards, however, with repetition, the procedure soon became automatic and you no longer thought about the details involved.
FORTRAN II compilation procedureNotes
Set the Program Switches as follows:
  1. OFF
  2. OFF
  3. OFF
  4. OFF
Pass I options
Set Overflow Check switch to Program and all others to Stop
Press Reset
Load blank cards into the Punch hopper then press Punch Start
Load Pass I of the compiler into the Read hopper then press LoadWait for Pass I to load and print on the typewriter "ENTER SOURCE PROGRAM, PRESS START"
Remove Pass I of the compiler from the Read stacker
Load the program source deck into the Read hopper then press StartWait for Pass I to complete and print on the typewriter "TURN SW 1 ON FOR SYMBOL TABLE, PRESS START"
Turn Program Switch 1 OFF then press StartIf a symbol table listing is desired for debugging, turn Program Switch 1 ON instead. The symbol table listing will be printed on the typewriter.
Wait for Pass I to print on the typewriter "END OF PASS 1"
Set the Program Switches as follows:
  • OFF
  • OFF
  • OFF
  • OFF
    		• Pass II options
    Set Overflow Check switch to Program and all others to Stop
    Press Reset
    Load blank cards into the Punch hopper then press Punch Start
    Load Pass II of the compiler into the Read hopper then press LoadWait for Pass II to load
    Remove Pass II of the compiler from the Read stacker
    Remove the intermediate output of Pass I from the Punch stacker, then load it into the Read hopper and press Reader Start then StartWait for Pass II to complete and print on the typewriter "SW 1 ON TO PUNCH SUBROUTINES, PRESS START"
    Remove the intermediate output from the Reader stacker
    Turn Program Switch 1 ON, load the Subroutine deck into the Read hopper, then press Reader Start then StartWait for Pass II to print on the typewriter "END OF PASS II"
    Remove the Subroutine deck from the Reader stacker and the completed Object deck from the Punch stacker
    GOTRAN was much simpler to use, as it directly produced an executable in memory. However it was not a complete FORTRAN implementation.
    To improve this various third-party FORTRAN compilers were developed. One of these was developed by Bob Richardson, a programmer at Rice University, the FLAG compiler. Once the FLAG deck had been loaded, all that was needed was to load the source deck to get directly to the output deck; FLAG stayed in memory, so it was immediately ready to accept the next source deck. This was particularly convenient for dealing with many small jobs. For instance, at Auckland University a batch job processor for student assignments chugged through a class lot rather faster than the later IBM 1130 did with its disk-based system. The compiler remained in memory, and the student's program had its chance in the remaining memory to succeed or fail, though a bad failure might disrupt the resident compiler.
    Later, disk storage devices were introduced, removing the need for working storage on card decks. The various decks of cards constituting the compiler and loader no longer need be fetched from their cabinets but could be stored on disk and loaded under the control of a simple disk-based operating system: a lot of activity becomes less visible, but still goes on.
    Since the punch side of the card reader-punch didn't edge-print the characters across the top of the cards, one had to take any output decks over to a separate machine, typically an IBM 557 Alphabetic Interpreter, that read each card and printed its contents along the top. Listings were usually generated by punching a listing deck and using an IBM 407 accounting machine to print the deck.

    Hardware implementation

    Most of the logic circuitry of the 1620 was a type of resistor–transistor logic using "drift" transistors for their speed, that IBM referred to as Saturated Drift Transistor Resistor Logic. Other IBM circuit types used were referred to as: Alloy, CTRL, CTDL, and DL. Typical logic levels of all these circuits were high: 0 V to -0.5 V, low: -6 V to -12 V. Transmission line logic levels of SDTRL circuits were high: 1 V, low: -1 V. Relay circuits used either of two logic levels high: 51 V to 46 V, low: 16 V to 0 V or high: 24 V, low: 0 V.
    These circuits were constructed of individual discrete components mounted on single sided paper-epoxy printed circuit boards with a 16-pin gold-plated edge connector, that IBM referred to as SMS cards. The amount of logic on one card was similar to that in one 7400 series SSI or simpler MSI package.
    These boards were inserted into sockets mounted in door-like racks which IBM referred to as gates. The machine had the following "gates" in its basic configuration:
    There were two different types of core memory used in the 1620:
    The address decoding logic of the Main memory also used two planes of 100 pulse transformer cores per module to generate the X-Y Line half-current pulses.
    There were two models of the 1620, each having totally different hardware implementations:

    A computer for the "small scientific market"

    In 1958 IBM assembled a team at the Poughkeepsie, New York development laboratory to study the "small scientific market". Initially the team consisted of Wayne Winger, Robert C. Jackson, and William H. Rhodes.

    Requirements and design

    The competing computers in this market were the Librascope LGP-30 and the Bendix G-15; both were drum memory machines. IBM's smallest computer at the time was the popular IBM 650, a fixed word length decimal machine that also used drum memory. All three used vacuum tubes. It was concluded that IBM could offer nothing really new in that area. To compete effectively would require use of technologies that IBM had developed for larger computers, yet the machine would have to be produced at the least possible cost.
    To meet this objective, the team set the following requirements:
    The team expanded with the addition of Anne Deckman, Kelly B. Day, William Florac, and James Brenza. They completed the [|CADET] prototype in the spring of 1959.
    Meanwhile, the San Jose, California facility was working on a proposal of its own. IBM could only build one of the two and the Poughkeepsie proposal won because "the San Jose version is top of the line and not expandable, while your proposal has all kinds of expansion capability - never offer a machine that cannot be expanded".
    Management was not entirely convinced that core memory could be made to work in small machines, so Gerry Ottaway was loaned to the team to design a drum memory as a backup. During acceptance testing by the Product Test Lab, repeated core memory failures were encountered and it looked likely that management's predictions would come true. However, at the last minute it was found that the muffin fan used to blow hot air through the core stack was malfunctioning, causing the core to pick up noise pulses and fail to read correctly. After the fan problem was fixed, there were no further problems with the core memory and the drum memory design effort was discontinued as unnecessary.

    Transferred to San Jose for production

    Following announcement of the IBM 1620 on October 21, 1959, due to an internal reorganization of IBM, it was decided to transfer the computer from the Data Processing Division at Poughkeepsie to the General Products Division at San Jose for manufacturing.
    Following transfer to San Jose, someone there jokingly suggested that the code name CADET actually stood for "Can't Add, Doesn't Even Try", referring to the use of addition tables in memory rather than dedicated addition circuitry. This stuck and became very well known among the user community.

    Implementation "levels"

    Notable uses

    An IBM 1620 model II was used by Vearl N. Huff, NASA Headquarters to program a three-dimensional simulation in Fortran of the tethered Gemini capsule Agena rocket module two-body problem at a time when it was not completely understood if it was safe to tether two objects together in space due to possible elastic tether induced collisions. The same computer was also used to simulate the orbits of the Gemini flights, producing printer-art charts of each orbit. These simulation were run over-night and the data examined the next day.
    In 1964 at the Australian National University, Martin Ward used an IBM 1620 model I to calculate the order of the Janko group J1.
    In 1966 the ITU produced an explanatory film on a 1963 system for typesetting by computer at the Washington Evening Star, using an IBM 1620 and a Linofilm phototypesetter.

    Use in film and television

    CADET

    Many in the user community recall the 1620 being referred to as CADET, jokingly meaning "Can't Add, Doesn't Even Try", referring to the use of addition tables in memory rather than dedicated addition circuitry.
    See [|development history] for an explanation of all three known interpretations of the machine's code name.
    The internal code name CADET was selected for the machine. One of the developers says that this stood for "Computer with ADvanced Economic Technology", however others recall it as simply being one half of "SPACE - CADET", where SPACE was the internal code name of the IBM 1401 machine, also then under development.