The Rand Tablet: A Man‐Machine Graphical Communication Device

Present-day user-computer interface mechanisms provide far from optimum communication, considerably reducing the probability that full advantage is being taken of the capabilities of either the machine or of the user. A number of separate research projects are underway, aimed at investigating ways of improving the languages by which man communicates with the computer, and at developing more useful and more versatile communication channels. Several of these projects are concerned with the design of “two-dimensional or ”graphical“ man-computer links. Early in the development of man-machine studies at RAND, it was felt that exploration of man's existent dexterity with a free, penlike instrument on a horizontal surface, like a pad of paper, would be fruitful. The concept of generating hand-directed, two-dimensional information on a surface not coincident with the display device (versus a “light pen”) is not new and has been examined by others in the field. It is felt, however, that the stylus-tablet device developed at RAND (see Fig. 1) is a highly practical instrument, allowing further investigation of new freedoms of expression in direct communications with computers. The RAND tablet device generates 10-bit × and 10-bit y stylus position information. It is connected to an input channel of a general-purpose computer and also to an oscilloscope display. The display control multiplexes the stylus position information with computer-generated information in such a way that the oscilloscope display contains a composite of the current pen position (represented as a dot) and the computer output. In addition, the computer may regenerate meaningful track history on the CRT, so that while the user is writing, it appears that the pen has “ink.” The displayed “ink” is visualized from the oscilloscope display while hand-directing the stylus position on the tablet, as in Fig. 1. Users normally adjust within a few minutes to the conceptual super-position of the displayed ink and the actual off-screen pen movement. There is no apparent loss of ease or speed in writing, printing, constructing arbitrary figures, or even in penning one's signature. To maintain the “naturalness” of the pen device, a pressure-sensitive switch in the tip of the stylus indicates “stroke” or intended input information to the computer. This switch is actuated by approximately the same pressure normally used in writing with a pencil, so that strokes within described symbols are defined in a natural manner. In addition to the many advantages of a “live pad of paper” for control and interpretive purposes, the user soon finds it very convenient to have no part of the “working” surface (the CRT) covered by the physical pen or the hand. The gross functioning of the RAND tablet system is best illustrated through a general description of the events that occur during a major cycle (220 μsec; see timing diagram, Fig. 2). Figure 3 is the system block diagram with the information flow paths indicated by the heavier lines. The clock sequencer furnishes a time sequence of 20 pulses to the blocking oscillators. During each of the 20 timing periods, a blocking oscillator gives a coincident positive and negative pulse on two lines attached to the tablet. The pulses are encoded by the tablet as serial (x,y) Graycode position information which is sensed by the high-inputimpedance, pen-like stylus from the epoxy-coated tablet surface. The pen is of roughly the same size, weight, and appearance as a normal fountain pen. The pen information is strobed, converted from Gray to binary code, assembled in a shift register, and gated in parallel to an interface register. The printed-circuit, all digital tablet, complete with printed-circuit encoding, is a relatively new concept made possible economically by advances in the art of fine-line photoetching. The tablet is the hub of the graphic input system, and its physical construction and the equivalent circuit of the tablet itself will be considered before proceeding to the system detail. The basic building material for the tablet is 0.5-mil-thick Mylar sheet clad on both sides with ½-ounce copper (approximately 0.6 mils thick). Both sides of the copper-clad Mylar sheets are coated with photo resist, exposed to artwork patterns, and etched using standard fine-line etching techniques. The result is a printed circuit on each side of the Mylar, each side in proper registration with the other. (Accurate registration is important only in the encoder sections, as will be seen later.) Figure 4 is a photo of the printed circuit before it has been packaged. The double-sided, printed screen is cemented to a smooth, rigid substrate and sprayed with a thin coat of epoxy to provide a good wear surface and to prevent electrical contact between the stylus and the printed circuit. The writing area on the tablet is 10.24 × 10.24 in. with resolution of 100 lines per inch. The entire tablet is mounted in a metal case with only the writing area exposed, as can be seen in Fig. 1. This research was supported by the Advanced Research Projects Agency under contract No. SD-79. Any views or conclusions should not be interpreted as representing the official opinion or policy of ARPA or of the RAND Corporation. Although it would be very difficult to fully illustrate a 1024 × 1024-line system, it does seem necessary, for clarity, to present all the details of the system. Thus an 8 × 8-line system will be used for the system description and expansion of the concept to larger systems will be left to the reader. Figure 5 shows the detailed, printed circuit on each side of the 0.5-mil Mylar for an 8 × 8-line system. The top circuit contains the x position lines and the two y encoder sections, while the bottom circuit has the y position lines and the two x encoder sections. It should be noted that the position lines are connected at the ends to wide, code-coupling buses. These buses are made as wide as possible in order to obtain the maximum area, since the encoding scheme depends on capacitive coupling from the encoder sections through the Mylar to these wide buses. It should be further noted that the position lines are alternately connected to wide buses on opposite ends. This gives symmetry to the tablet and minimizes the effect of registration errors. With reference to Fig. 5, at time t1 encoder pads p1 + are pulsed with a positive pulse and pads p1 – are pulsed with a negative pulse. Pads p1 + are capacitively coupled through the Mylar to y position lines y5, Y6, Y7, and y8, thus coupling a positive pulse to these lines. Pads pi — are capacitively coupled to y position lines y1, Y2, Y3, and Y4, putting a negative pulse on these lines. At time t2, encoder pads p2 + and P2 – are pulsed plus and minus, respectively, putting positive pulses on y position lines y3, Y4, Y5, and y6, and negative pulses on y position lines y1, Y2, Y7, and y8. At the end of time t3, each y position line has been energized with a unique serial sequence of pulses. If positive pulses are considered as ones and negative pulses are zeroes, the Graypulse code appearing on the y position wires is as follows: The x encoder pads are now sequentially pulsed at times t1, t5, and t6, giving unique definitions to each x position line. If a pen-like stylus with high input impedance is placed anywhere on the tablet, it will pick up a time sequence of six pulses, indicating the (x,y) position of the stylus. It should be pointed out again that the stylus is electrostatically coupled to the (x,y) position lines through the thin, epoxy wearcoat. If the stylus is placed on the tablet surface at a point (X1,Y5), the pulse stream appearing at the pen tip would be as indicated in Fig. 6. This detected pulse pattern will repeat itself every major cycle as long as the stylus is held in this position. If the stylus is moved, a different pulse pattern is sensed, indicating a new (x,y) position. Since there are 1024 x position lines and 1024 y position. lines, 20 bits are required to define an (x,y) position. The actual timing used in the RAND system was shown in Fig. 2. Timing pulses t21, t22, and 223 are additional pulses used for bookkeeping and data manipulation at the end of each major cycle. The position lines on the full-size tablet are 3 mils wide with a 7-mil separation. The code-coupling pads are 16 to 17 mils wide with a 3- to 4-mil separation. Figure 4 shows that the encoding pads which couple to the lower set of position lines (y position lines) are enlarged. This greater coupling area increases the signal on the lower lines to compensate for the loss caused by the shielding effect on the upper lines (since they lie between the lower lines and the stylus pick-up). The encoding pad for the two least-significant bits in both x and y was also enlarged to offset the effect of neighboring-line cancellations. With these compensations, all pulses received at the stylus tip are of approximately the same amplitude. Figure 7 is an illustration of the approximate equivalent circuit of the encoder-tablet-stylus system, along with typical system parameter values. It is clear that the values of C1 vary with encoder-pad size, and the value C4 varies according to whether top or bottom lines are being considered. The value of C4, is also dependent on the stylus-tip geometry and wear-coat thickness of the tablet. The signals arriving at the input to the stylus amplifier are approximately 1/300 of the drive-line signals. The character of the signals at the stylus input is greatly dependent on the drive-pulse rise time. Figure 8 is an oscilloscope pattern of the amplified signals at the stylus output.† These signals are amplified again and strobed into a Gray-code toggle. An x bit at t8 and a y bit at t17 are smaller than the rest. This indicates that the stylus tip is somewhere between lines and these are the bits that are changing. Since the final stages of the amplification and the strobing circuit are dc-coupled, the system is vulnerable to shift in the dc signal level. For this reason, an automatic level control (ALC) circuit has been provided to insure maximum recognizability of signals. During the first 180 μsec of a major cycle, the stylus is picking up bits from the tablet. During the last 40 μsec, the tablet is quiet. – i.e., the stylus is at its quiescent level. During this 40μsec interval, the quiescent level of the pen is strobed into the ALC toggle. If the quiescent level is recognized as a zero, the ALC condenser changes slowly into the proper direction to change the recognition (via a bias circuit) to a one, and vice versa. For a perfectly balanced system, the ALC toggle would alternate between 1 and 0 with each major cycle. Time-wise, the bits are received from the stylus in the order N,…, j + 1, j,…, 1. When all 20 bits have been assembled in the shift register, they are gated to the output register. As a new (x,y) value is being converted to binary and shifted into one end of the shift register, the old binary value is being shifted out the other end. This old binary information is serially reconverted to Gray and compared to the new, incoming Gray value, one bit at a time. If the old Gray number and incoming Gray number differ in more than one bit in either x or y, a “validity” toggle is set to indicate an error. If the two Gray-code series differ in more than one bit, this indicates that the pen has moved more than one line during the 220-μsec interval. As this is not probable during normal usage, it is assumed that an error has occurred. If a set of data are determined as not valid, the output register is left with its previous value, and an “old-value” toggle is flagged. In practice, the validity check rarely detects errors while the pen is in contact with the tablet. The pen validity check is used to suppress the display of the pen position as the pen is lifted off the tablet. The logic and clock systems are made up principally with state-of-the-art NOR circuits and univibrators. The blockingoscillator circuit shown in Fig. 9 was designed to drive the encoder pads. This use of transformer coupling was found to be important since well-matched positive and negative pulses were required to obtain proper cancellation at the tablet surface. The stylus amplifier has a gain of approximately 30 db with an additional 30-db gain in the principal electronic package. The total electronic system is assembled in a 5“ × 5” × 19“ printed-circuit card cage and contains some 400 transistors and about 220 diodes; however, little attempt has been made to minimize the number of components. Also, the electronics could be shared with a number of tablets in a multiple-tablet system. Figure 10 is a block diagram showing the graphic inputoutput system as used at RAND for the evaluation of hardware, human engineering studies, and investigation of programming implications. The computer used was the JOHNNIAC, a tube machine of the Princeton class. Preliminary studies indicate that with a great amount of care in construction, a 200-line-per-inch tablet could be achieved. The resolution of this line density would not present a major problem; on the other hand, 100 lines per inch is adequate for all current intended applications. It is certainly within the state of the art to decrease the major cycle time; however, in usage at RAND, the 4.5-kc rate has been adequate. When the stylus is swept rapidly across the surface of the tablet, it has been found that an average of two or three complete sets of position data are obtained for each line. Setting the multiplexing switch (Fig. 10) to display the stylus position on the scope every 10 msec has proved adequate, and since only 50μsec are required to display the point, 99.5 per cent of the display scope time is left for the computer. The tablet currently is in regular use at RAND in studies toward the development of on-line graphical programming languages and on-line interaction with problem parameters. In addition to its use at RAND, several copies of the tablet have been supplied to other researchers in the field. The tablet has been found to be particularly valuable in applications where excellent linearity and accuracy are important. Normal-thickness C.G.S. maps have been placed over the tablet to digitize contours by manual tracing with the pen. Development of the stylus-tablet device has been carried to the point where, we feel, it represents a practical and economical tool for use in many applications. Additional application areas might be served by more development effort in directions such as providing for rear-projection of images onto the (translucent) tablet panel, provision for use of more than one sensing element, extension of the surface dimensions, etc. MALCOLM R. DAVIS received his B.S. degree in Engineering from the University of California at Los Angeles in 1954. He attended graduate school at U.C.L.A. for one year prior to joining The RAND Corporation where he supervised the Analog Computer laboratory. Since 1958, he has been active in the digital field, doing design of digital systems for research in man-machine interaction. He is currently head of the Hardware Group of the Computer Sciences Department at RAND. Mr. Davis is a member of Sigma Xi, IEEE, and is a past chairman of the Los Angeles chapter of the IEEE Computer Group. THOMAS O. ELLIS received the M.S. degree from the University of California at Los Angeles in 1953. He joined the Computer Sciences Department at The RAND Corporation in 1953, where has has since been active in the research and design of computers and on-line systems. These have included such projects as JOHNNIAC, JOSS, the RAND Tablet, several display systems, and GRAIL — an experiment in graphical man-machine communication techniques. Mr. Ellis is a member of Tau Beta Pi, Sigma Xi, and IEEE.