CARD FEEDING 316 PROCEEDINGS--SPRING JOINT COMPUTER CONFERENCE, From the collection of the Computer History Museum (

Transcription

1 316 PROCEEDINGS--SPRING JOINT COMPUTER CONFERENCE, 1966 Figure I. IBM System-360 Model 20 showing the IBM 2203 Printer, the IBM 2020 Processor, and the IBM 2560 Mul1:i-Funcltion Card Machine. which detect the motion of magnets embedded in five separate timing discs. Card movement and location are monitored by nine solar cell sensors, rather than the more conventional mechanical card levers. The lower half of the MFCM contains two gates of electronics, power supplies, and the chip box. The upper half contains the card handling mechanism (see Fig., 2). An unusual: feature of the machine is the so-called "backbone" or vertically mounted plate on which are mounted most machine components. This backbone feature greatly improves accessibility of the card path and components over the more conventional side frame or "boxed-in" approach. The two card feed hoppers are mounted on the front or operator side of the backbone. The stacker assembly, which consists of a magnetic selector unit and five radial stackers, is also located on the operator side. All other components are mounted on the rear side of the backbone, including the entire serial card path, the read, punch, and print units, and the drive motor. The motor drives all units through toothed belts and pulleys. CARD FEEDING Card motion during reading or ejecting in the MFCM is controlled by the card feed clutch. From each parallel hopper the cards are fed into a cornering station by picker knives and continuously running feed rolls, as shown in Fig. 3. From the cornering station they move serially through the read, punch, and print stations before cornering again to move in a parallel path into one of the five stackers. The movement from the hopper through these operational stations is under complete processor control and is carried out by either cam operated or magnetically operated feed roll selection devices and the feed clutch. Cards are designated as primary or secondary according to the hopper from which they are fed. During a typical operation, one of each type of card is always in position to enter the read station, one of each type of card is always in position to enter th,e punch station and a single card is in position in the print station. As the card in the print station passes through the station to go on to the stacker, either

2 THE IBM 2560 MULTI-FUNCTION CARD MACHINE 317 PRIMAHY SECONDAHY Figure 2. Rear view of the MFCM showing location of major components. the primary or secondary card at the pre-punch station is fed through the punch unit to the print station and the corresponding primary or secondary card at the pre-read station is fed through the read unit to the pre-punch station. At the same time a card is being fed from the corresponding hopper to the pre-read station. All of this card movement occurs simultaneously during one clutch cycle, and thus all card stations are occupied on every cycle. Under processor control, cards from one hopper can be held indefinitely at the pre-read and prepunch stations while cards from the other hopper are processed through the machine. READING SYSTEM A magnetically operated pressure roll and the continuously running read inject roll grip the card and feed it into the read station from its pre-read position. The card is accelerated up to the reading velocity of 146 inches per second, which matches the peripheral velocity of the continuously running

3 318 PROCEEDINGS- -SPRING JOINT COMPUTER CONFERENCE, SECONDARY HOPPE PRIMARY HOPPER Figure 3. Schematic of the MFCM card path. read feed roll. As the leading edge of the card is gripped by the read feed roll, the trailing edge leaves the inject roll so that the card is under the control of the read feed roll for the entire read operation. The holes are sensed by 12 silicon solar cells located below the card path. Light is provided by 12 lens tip lamps located above the card path. The lamps are individually adjustable for position and intensity. Both the lamps and the cells are mounted inside sealed assemblies behind glass plates to prevent card dust accumulation from affecting light transmission. During a read operation, the CPU must be supplied with a series of pulses synchronized with card motion and timed so as to indicate the optimum time for sampling the output of the read circuits for each of the 80 card columns. The device which accomplishes this in the MFCM is the magnetic read emitter. A schematic of the emitter is shown in Fig. 4. The rotor or drum of the emitter is mounted on the read feed roll shaft. The periphery of the drum is plated with a thin, hard coating of cobalt. The stationary housing of the emitter contains 60 heads or probes equally spaced around the periphery of the drum. The housing also contains a write coil, a read coil, and an erase coil. As the leading edge of the card enters the read station, the trailing (or column 80) edge of the card uncovers a solar cell. The cell is located so that it becomes uncovered as column "zero" is directly over the read cells. The uncovering of the trailing edge cell causes the CPU to send a signal to the emitter write driver. This results in 60 magnetic bits being recorded on the drum due to the flux generated by the single write coil passing through the 60 heads. The read feed roll on the drum shaft feeds a card one column in 1/60 of a revolution so Figure 4. Schematic of the Magnetic Read Emitter. that as each column is passing under the read station, the 60 magnetic bits on the drum have rotated one position and are being sensed simultaneously by the 60 heads. The cumulative flux change of the 60 heads is sensed by the common read coil. Any air gap variation between the drum and the he:ads caused by eccentricity is of little consequence since the output signal of the emitter is the sum of the 60 signals. After the card has left the read station, the erase coil is energized and the 60 bits are erased so that the emitter may be resynchronized with the next card as the card is being read. PUNCHING To register a card for punching, the punch pusher clutch and the desired path selection solenoid is energized. The primary or secondary card s.elected is pushed into column 1 registration and the incre-

4 THE IBM 2560 MULTI-FUNCTION CARD MACHINE 319 mental pressure rolls are closed. Command. from the processor controls the number of columns punched and incremented. The chief difficulty in designing a high-speed serial punch is in maintaining accurate punching registration over 80 columns of movement. An error of only inch in column to column spacing would result in the punching being about one full column out of registration after 80 spaces. The basic problem is the high card acceleration necessary to achieve desired throughput. The MFCM solves this pi:oblem in two ways: 1) by keeping the acceleration as low as possible, and 2) by using feed rolls with a very high coefficient of friction to minimize card slippage. The punch unit is designed to allow approximately threefourths of the 6.25-millisecond punch cycle for card movement, thus minimizing the acceleration requirements. The unit utilizes three cams: one to drive the punch through the card, one for positively restoring the punch, and one for moving the interposer-armature assembly, as shown in Fig. 5. By optimizing each of these motions, the time the punch is in the card has been reduced to approximately one-fourth of the punch cycle. The punch mechanism is controlled by a magnet unit utilizing the so-called "no-work" principle. The magnets are only required to prevent the armature from moving rather than causing them to move; thus, the origin of the term "no-work" magnets. Several advantages are realized by using this style of magnet unit. 'Since reluctance of the magnetic circuit is low, the coil and core size may be reduced, resulting in a much more compact unit. Also, magnet driving requirements are lowered and heat dissipation is reduced. Figure 5. Cross section of the punch mechanism. A constantly energized single hold coil surrounding all 12 punch magnets prevents the armature from following the interposer cam. When a punch is to be operated, the appropriate punch magnet coil is energized. This cancels the hold flux, which releases the armature at the high point of the interposer cam, and the armature spring causes the armature to follow the cam. This moves the interposer between the punch bail and the punch, causing the punch to be driven through the carel. The restore lever pulls the punch out of the card. CARD DOCUMENT PRINTER Basic requirements for the MFCM included a card document printer that would provide an output of 100 fully printed cards per minute, with up to six lines per card. Existing printing mechanisms were capable of such output, but at an unreasonable cost, so an entirely new print unit was developed. The desired output is achieved by using a wire matrix print head for each print line and printing in the serial mode. These wire matrix print heads may be positioned manually by the operator in any of the 25 printing row locations from the top to the bottom of the card to accommodate any card format. Each of the printed lines may contain up to 64 characters spaced at 10 characters per inch. A wire matrix five wires wide by seven wires high forms the characters. The wires are driven against an inked ribbon, card, and platen to print the character on the carel. One matrix of 35 wires, hereafter called a print head, prints a complete line on a card. To print six lines on a card, six print heads are required. Print heads are installed in groups of two to the maximum of six. An incremental drive moves. the card from position to position, stopping the card in each position. Each print head prints a character on the card in a 7.23-millisecond cycle, which achieves the rated speed of 138 characters per second per line, or 828 characters per second for six print heads. A card ejected from the punch station is stopped in print position 1 registration by a magnet-operated print gate. During the first print cycle the card is gripped by incrementing rolls. An emitter on the incremental drive triggers each print cycle, which consists of setting up the characters for each print head, printing all of the characters in the position, and moving the card to the next column. Upon

5 320 PROCEEDINGS-SPRING JOINT COMPUTER CONFERENCE, 1966 completion of printing, the card is ejected from the print station during the following feed cycle. A ribbon is continuously driven past all the print heads at an average speed qf 7 inches per second. A spring-loaded reversing mechanism assists the ribbon-drive motor in achieving rapid ribbon reversal. Print Mechanism Figure 6 is a cross section of the print unit. Each print armature is fastened to a print wire. A print unit contains 70 wire and armature assemblies or two heads. A tube, which is securely fastened at Figure 6. ~--J)RINT TUBE.-IlZmaIIlWf--MAGNE:T MOUNTING BAR '---llrint CAM '----PRINT ARMATURE '----BUCK COIL '-----HOLD COIL '-----HOLD MAGNET Cross section of the print unit. the unit and at the print head, guides each wire. A compression spring drives each armature and wire against the ribbon, card, and platen to perform the printing. The print armatures are controlled by nowork magnet units similar to those used in the punch unit. Characters are' formed by energizing the required buck coils. The selected armatures drop onto the print cam which controls the time of impact of the wires, and restores the armatures to the magnet after printing. When not printing, the armatures are held away from the print cam by the hold magnets. Character Selection The printing circuitry in the MFCM consists of an array of 35 magnet drivers, each of which is connected, through isolation diodes, to the six buck coils associated with the same matrix position on all six heads. The common side of the 35 coils for each head are connected to the power supply through a separate silicon-controlled rectifier. This circuitry allows the 210 magnets required for six print heads to be controlled by only 35 drivers, since the six heads are set up sequentially rather than simultaneously. Two things make sequential operation possible: 1) a memory dwell built into the print cam, and 2) a motionless period for the card which is long enough to allow the three print cams to be timed for sequential operation. Figure 7 shows the sequential timing and the c:am profile for the three print unit cams. At high dwell on the print cam, the armatures have been restored from the previous cycle and are ready for setting up in the next print cycle. Setup begins when a pulse arrives from an emitter on the incremental driv~, timed with high dwell on the print cam for print unit one. The selected buck coils for print head one are energized for 350 microseconds, during which time their armatures are released. The armatures drop to the memory dwell which is an intermediate dwell on the print cam just following the high dwell. After the coils are de-energized and the holding flux increases, the armatures are not reattracted back to the magnets due to the air gap created by the memory dwell. After the 350-microsecond buck pulse for head one, there is a microsecond delay before head two is set up to allow the silicon-controlled rectifier for head one! to turn off. Head two is then set up in a similar manner. The second print unit, which controls head three and four, is timed so that the setup for its two heads comes just after the print circuitry has completed the setting up of heads one and two. The third print unit is similarly timed with respect: to the second. As a result of this method of timing, <!) z ~ UNIT I ~ ::E ~ UNIT 2 ~ z ~ UNIT 3 Cl. Figure 7. PRINT HEAD BUCK PULSE TIMINGS --" Electrical and mechanical timing diagram of the print unit.

6 THE IBM 2560 MULTI-FUNCTION CARD MACHINE 321 the impact from heads one and two occurs just after the card has come to rest in the dwell period, heads three and four print in the center of the dwell period, and heads five and six print just before the card starts to move. CONCLUSIONS The challenge of designing a machine that could, in a single pass, perform most of the functions required in a card-oriented processing system, has been met by the IBM 2560 Multi-Function Card Machine. This machine, as a part of System/360 Model 20, provides a dramatic reduction in the number of processing steps and the card processing time required by conventional punched card equipment. File maintenance or file updating is now possible with one machine. The unique concept of the MFCM has been likened to a "poor man's magnetic tape system." In this concept the two card hoppers simulate two magnetic tape input drives, the five card stackers simulate magnetic tape output drives, and the coordinated read-punch-print units simulate the read-write-copy function of magnetic tape. The wide acceptance and interest in the MFCM, since its announcement, is an indication of the basic soundness of the multi-function concept in card handling.

7

8 A NEW DEVELOPMENT IN THE TRANSMISSION, STORAGE AND CONVERSION OF DIGIT AL DATA R. P. Burr, John J. Rheinhold Photocircuits Corporation, Glen Cove, New York and Roy K. Andres RCA Communications, Inc., New York, New York INTRODUCTION There is a present trend in data processing systems toward the decentralization of computer systems as exemplified by time-sharing and related techniques. A consequence of this development is a requirement for digital communication systems capable of operating reliably at comparatively high speeds. One of the oldest known forms of such communication is the printing telegraph which operates at the lower end of the speed spectrum at rates ranging from 50 to 100 baud. There is some evidence to suggest that for routine digital communications purposes, particularly in the timesharing area, the economic maximum for a few years into the future will lie in the vicinity of 1000 to 2000 baud and that the volume of traffic at these rates will rapidly increase. Present practice is generally to transmit data at such rates in "real time" only over communication paths which are not ex. pected to fail during the transmission period. In the future, however, it is a virtual certainty that lowcost buffering devices will be required having the speed, versatility and storage capacity to handle system requirements at rates up to 4800 baud or more. The use of magnetic tape in preference to paper tape offers a preferred solution to the problem of high-speed buffering, provided that a technique can be devised which permits the use of magnetic tape so that it is in all operational respects equivalent to paper tape systems. Such a functional equivalence can be obtained if it is possible to: 1) record on the tape incrementally over a speed range extending to at least 300 characters per second on a character-bycharacter or bit-by-bit basis and 2) read the information recorded on the tape in an incremental manner through the same speed range. A device having these characteristics would be functionally equivalent to a paper tape system capable of operating at punching and reading rates up to 300 characters per second but would of course be physically smaller and mechanically simpler because of the improved packing densities available on magnetic tape. To additionally exploit the full advantages of magnetic tape in such an environment the device should also be capable or recording and reproducing digital information "on the fly" in a manner somewhat analogous to a conventional digital magnetic tape transport so that very high acceptance and transmission rates could be achieved if necessary. A capstan drive system for magnetic tape which is capable of satisfying these operating requirements can be achieved by the proper combination of a fast 323