types of information to types of devices, giving the system designer a framework
... the graphics system designer can be compared and contrasted according to ...
A simple classification of graphics input requirements relates types of information to types of devices, giving the system designer a framework for overall design decisions.
System Design Considerations for Graphics Input Devices Mark Ohlson Texas A&M University
The development and utility of computer graphics applications have been greatly increased by interactive program communication. The quality of the interaction between user and program depends on available input and feedback devices, which should be well suited to the user's communication needs. The universe of discourse in computer graphics is, broadly, the set of visual objects on one, two, or three dimensions that can be represented on two-dimensional media, independently of whether they actually exist or have counterparts in the real world. Some common modes of communication set the needs of graphic I/O equipment apart from the needs of other interactive applications; at the same time, some needs are common for all interactive work. The variety of physical input devices available to the graphics system designer can be compared and contrasted according to their functionality, complexity, precision, and potential modes of use. The structure of input devices tends, of course, to reflect human modes of expression. For example, humans commonly communicate position in a particular space or identify objects in that space by pointing, and graphics input devices are designed to implement pointing in ways that are feasible from an implementation standpoint and that please the user by reflecting his natural communication tendencies. The reason for pleasing the user is economic as well as aesthetic. To the extent that an input device satisfies the user's natural modes of expression in a particular activity, he will view his interaction with the computer as being more natural, and will adapt more easily to a computer graphic environment while thinking and reacting in terms of the application environment. Consequently, he will be able to do more productive and hence more cost-effective work. November 1978
Attempts to provide users with natural input devices have led to the development of prolific quantities of gadgets, each with particular (and sometimes peculiar) installation and use requirements. The graphics system designer has a complex task. In order to make judgments for a system design, he must understand functional requirements, precision requireiments, cost considerations, and the elements of user satisfaction. The first three can be quantified sufficiently for decision making, but the last is difficult to address in any absolute fashion.
The functional design framework Functional requirements arise from the types of information required in a particular environment and the functional capabilities of the available input devices. Information types. Deecker and Penny 6 identify six common input information types: (1) positioncoordinate sets, (2) vector of position-coordinate sets, (3) text, (4) numeric value, (5) object identification, and (6) control-function identification. Position information may be expressed in relation to a changeable or fixed reference or coordinate system. The number of coordinates needed to specify a position is determined by the universe of application. A simple x,y value pair is sufficient for a twodimensional Cartesian world, and a radius-angle pair is sufficient for a polar world.. A threecoordinate scheme is necessary for a three-dimensional world, Cartesian or otherwise. More coordinates may be needed to specify a position for some applications, as in spatial navigation, where three coordinates specify a 3D location and three more
0018-9162/78/1100-0009$00.75 '(.
1978 IEEE
specify orientation (yaw, pitch, and roll). A vector of position-coordinate sets simply represents a series of points, which may be indicative of a set of connected line segments or simply a set of sample points. Text input usually means alphanumeric character codes originating from a keyboard, but ultimately means any mode of entry that results in character information. Similarly, value input may be simply text input of type numeric, but might also indicate entry of numeric values by other means. Objectidentification input conveys the name of some visual object of interest to the user. It usually relies on displaying to the user the set of possible objects that may be identified. Typically, a user is presented with a screen on which a complex picture is shown, and from which he desires to choose a particular object for further operations. Control-function identification (or procedure identification) conveys the name of some function (or procedure) that the user wishes called, initiated, or otherwise executed. Through this type of input, the user can exert control over program transfers directly, rather than indirectly through data (text) input. The purpose of these information-type definitions is to identify and separate information requirements for input; no particular type of device is specified to accomplish the input, although in some cases it seems obvious (as text input through a keyboard). In fact, all six types could be implemented through a keyboard-though with a significant loss of naturalness.
Device types. Just as input information can be grouped into types, so input devices can be grouped according to their functional similarities. Several such classifications have been proposed.38'18'19,23'27'28 That of Foley and Wallace8 presents four device types, to which a fifth is added by others owing to its universality: (1) pick, (2) button, (3) locator, (4) valuator, and (5) keyboard. Devices of the pick class are used to process object-identification information. If, as is generally
the case, this process is based on visual feedback information on an output device such as a screen, then its behavior is highly dependent on the software and hardware of that output device. The pick device must be able to examine the contents of the screen, either directly or through the screen's display file,20 in order to determine what object is being referenced. The result of a pick is a reference to that object, whether the reference is a name, an index, or other identifying tag information used by the system. Button devices are used to process functionidentification information. In the simplest case, a button may be a hardwired circuit automatically activating. a fixed function. If the identification of functions in a particular system is done in software, it may be done by name, by an index into a branch table, or by a (possibly integer) valued casestatement implementation. This device type is used to give the user strong control over program behavior, allowing him to identify functions, causing direct transfer to the named function. Such a button may or may not be software programmable. Locator devices are used to gain position information in a coordinate space. Some devices in this class can return values in only two dimensions, while others can be extended to three dimensions in a simple manner. If higher dimensionality is needed, it can be gained by coupling these devices at the hardware or software level; the problem in these cases is to make the devices reasonably facile for human interaction. Valuators are used to process value or numerical information. In this respect they behave as (and in fact are) one-dimensional locators. Used alone, they are a convenient mode of data entry compared to keyboard entry. In pairs, they can be used as a twodimensional locator. It is also common to use them in conjunction with 2D locators as 3D locators. Keyboard devices process character or text information. A keyboard is, in a sense, a collection of buttons, but because of its accepted standard design and function it is regarded as an entity in itself.
Figure 1. Information-device mapping.
10
COMPUTER
A system employing at least one of each of the above device types may gather information types in a simple and direct mapping, as shown in Figure. 1.
Integrating information and devices. This initial consideration of functionality defines a classification system that has implications not only for design methodology but also for device standardization. Interest in standardization for portability of software and interchangeability of hardware in graphics is strong,"'23 and standardization is an important goal in the design of new graphics systems. It should not limit designers' intentions, but rather should concentrate them on feasible goals. So far, we have defined two sets: the set of information requirements as primitives, and the set of available input devices as a pool of virtual devices. The realization of a graphics system results from formulating a mapping between the two sets that reveals which virtual devices are necessary in a particular application environment and that should indicate the relative importance of each device by virtue of how and how much it is to be used. This information in turn becomes the basis for considering the actual physical devices that might be chosen to further define a system design. The factors yet to be considered-namely, precision, cost, and user satisfaction-cannot be applied to virtual device definitions, but can be examined once particular physical devices have been chosen. User satisfaction is in large degree a measure of the similarity of the required input device behavior
to the user's customary communication behavior. Figure 2 illustrates four input styles for commonly available devices, tagged for their mode of input and mode of feedback. Direct feedback occurs at the location of input, whereas indirect feedback involves separate locations for input and feedback. Input mode here is either graphical (drawing) or tactile (touching). None of the illustrated input modes is inherently superior to any other. The world of drafting would prefer the indirect graphic mode (Figure 2c), while flight simulation would opt for the indirect tactile (Figure 2d).
In addition to functional requirements dictated by information types, the designer must consider precision, cost, and user satisfaction. Precision requirements for an application system are usually simpler to define than user-satisfaction requirements. Precision requirements tend to render some of the devices under consideration unacceptable regardless of their other properties. If an input device is to be used as a feedback mechanism for a display, it should equal or exceed the display precision if point addressability is required. Some input devices are used for data input, where feedback-display requirements are secondary to those for storing and using data. In these cases, precision should be matched to the intended use of the data and the
(d) INDIRECT TACTILE
I
Figure 2. Behavioral characteristics. November 1978
11
Table 1.
Graphics input device characteristics. PRECISION FUNCTIONALITY DEVICE INFORMA- RESOLUTION & CLASS TION ACCURACY Tablet or digitizer
Locator
Touchpanel
Locator
Joystick
Locator
Mouse and trackball
Locator
CLASSES Position, Vector, Function ID, Object ID Position, Function ID, Object ID
SPECIAL CONSIDERATIONS
Medium to high
Medium
Low
1/4" to 3/4",
Low to medium
Low
Low
Direct tactile
Independent display
Gross resolution makes use for detail work impossible.
Position, Vector, Object ID
Low Resolution to dependent on A/D converter, medium ±10% to ± .1%
Low
Low
Indirect tactile
Tactile and/or independent
Large variety ot types to fit most applications.
Position, Vector,
As for joystick, 10-1000 accurately differentiable points As for joystick
Low to medium
Low
Low
Indirect tactile
Tactile and/or independent display
Relative positioning of mouse. Slewing capabilities of trackball.
Low
Low
Low
Indirect tactile
Choice of dials, slides, levers, and other styles.
Low to medium Medium
Medium
Low
indirect tactile
Tactile and/or independent display Visual (switch
High
Medium
Direct graphical
Independent display
Object ID
Pick
USER SATISFACTION INPUT FEEDBACK TYPE
.05" to .001", up to ± .005"
Potentiometers Valuator Value 2D pos. if used in pairs) Button Function Function ID Switches Light Pen
COST HARDWARE SOFTWARE SOFTWARE DEVELOP. OVERHEAD COST
accuracy much higher than resolution
NA
Object ID, Dependent Position, on display device Vector, Function
Indirect (some- Digital position Some units are not readout or suitable for online independent graphical interactive use. display
times direct)
display
ID
to
high
position)
Possible use of software-programmable buttons. Only usable as simple pick on refresh displays, or locators on raster-scan displays.
machine-dependent storage accuracy. If the system considered in absence of other graphics system comwill store input as 32-bit floating-point values, then ponents. The types of output devices, especially input precision of only three significant digits can displays being considered as interactive terminal give false indication of accuracy during computa- sites, will affect both potential input device pos-
tions. Cost is the requirement, or restriction, against which all else is measured. Getting the best possible system within the given cost constraints is an obvious goal, but accurate measurement of cost must take into account several factors beyond the hardware costs of adding an input device to a system. There are at least two software costs: the cost of developing and maintaining software to support the device, and the cost of software overhead-and hence system degradation-incurred while the system is using the device. How these costs should be blended is difficult to say, although current trends indicate that software maintenance is becoming the overriding cost in most systems.
Graphics within the total system context However, in considering cost and other factors, it emphasized that input devices cannot be
must be 12
sibilities and cost. The three common display types are storage-tube, vector-refresh, and raster-scan screens. Since these devices usually represent the most highly interactive modes of information feedback to the user, they heavily influence a system's behavior as viewed both by the user and by the implementer. Storage-tube displays require that output information be traced only once on the screen for viewing purposes. Lines remain visible for a relatively long time-on some models, indefinitely. Vector drawing on such a display is relatively slow, and the ability to selectively erase line segments may or may not be available. Erasing the entire screen and redrawing a new picture can be a slow process, making these displays suitable for static drawing applications but unsuitable for dynamic applications. The choice of such a display will affect the choice of input devices. Light pens will obviously be excluded because they do not function on storage-tube displays. If direct COMPUTER
graphical input is important, then some other device will have to be chosen. Dynamic displays, of both vector-refresh and raster-scan types, must completely regenerate the screen contents, usually at a rate of 30 times per second or faster, to maintain an acceptable picture image, The sizable hardware and software cost to accomplish this task must immediately become a consideration. Vector-refresh displays regenerate the image as a series of vectors, interpreted from a display file.20 The current low level of manufacture of these displays makes them expensive. Raster-scan displays (television tubes) regenerate images by scanning horizontally across the screen with an electron beam, intensifying points lying on vectors passing through the horizontal scan lines. Most raster-scan displays scan first the set of even lines, then the odd, for reasons of picture quality as perceived by the user. This method produces an apparent doubling of the speed of the screen refreshing capabilities, resulting in less flicker. At the, hardware level, this mode of display is not natural from the user's point of view. What he views as a set of connected and related points, a line, must be stripped apart into scan-line format for the display, which then reassembles it into visual feedback on the screen. Extensive hardware and software are needed to accomplish this transformation effectively, hiding it from the user. The redeeming cost factor for raster-scan devices is that television tubes are manufactured in great quantities, and the associated technology is highly developed. The requirement for a dynamic display in any system probably arises from the nature of the application. Consequently, input devices for these systems should be picked for their ability to perform in highly interactive modes-modes that take advantage of the display characteristics.
Characteristics of input devices
backlighted tablets (for inputting data from translucent film). The various types of tablets are differentiated by the principle used to determine location on the tablet. Matrix-encoded tablets, as the Rand tablet,5,20 associate a unique code with each point on the surface. Usually, two sets of parallel wires, vertical and horizontal, are inlaid beneath the surface of the tablet, each carrying a unique code. The stylus tip acts as a receiver, picking up a pair of unique line codes that can be decoded to a unique grid crossing, and thence to an x,y position pair. The resolution of such a tablet is limited by the density of the parallel wires and the receiver's differentiation of signals, while absolute accuracy is limited by the linearity of the inlaid wires. A minimal configuration of such a tablet is shown in Figure 3.20 This type of tablet generaUy provides the greatest precision possible. Other tablets rely on measuring the physical distance of an addressed point from the x and y axes along two adjacent sides of the tablet, using some physical phenomena. The voltage gradient tablet20 uses a conductive plate as the tablet surface; by applying a potential to the stylus tip, it is possible to measure a drop in potential or apparent resistance across the plate from the stylus tip to any two adjacent edges, and from this deduce an x,y position. This has two drawbacks. One is the need for high conductive uniformity of the plate to get accurate positions. The other is that the stylus must contact the plate directly, disallowing many tracing operations. An improvement over this is the use of sound waves as a medium of measurement. Acoustic tablets20 use a stylus that generates a spark or sound wave at its tip to be picked up by strip microphones located on two adjacent edges of the tablet. The time delay between sound generation and reception registered at the two edge microphones can be translated into distances, and thence to an x,y position pair. This method leads, by a simple extension, to three-dimensional input, us-
We can now take a more detailed look at input devices that are currently available. Table I collects the salient properties of the devices for reference and comparison.
Tablets. The tablet (or digitizer) is one of the most common of locator devices. Although the two names are often used synonymously, "tablet" usually refers to the basic, unattached input device of concern here, and digitizer to a system of connected devices, of which a tablet is a central part, used to input information, sometimes in a standalone configuration. The advantage of using a tablet, as pictured in Figure 2b or 2c, is its behavioral similarity to drawing. Tablet devices are available in almost any size, from five inches square to five feet square, either as movable boards or as built-in tables. Depending on the particular model, drawings to be input may be made on top of opaque tablets, underneath transparent sheets or through November 1978
Figure 3. Matrix-encoded tablet.
13
ing a third edge microphone, as shown in Figure 4. Because this design does not require a surface, it suggests a great variety of installation possibilities. The major drawback of both two-dimensional and three-dimensional acoustic input is the lack of fine resolution and accuracy. Finally, there is the touch-sensitive tablet. Its advantage is that one can use an unsensitized stylus or even a finger. The edges of this tablet generate highfrequency vibration waves on the glass tablet surface. Any object contacting the plate will cause reflection of waves back to the edges, where wavedetection hardware is mounted. The time lapse between the generation of a surface wave and the reception of its reflection is used to locate the point of reflection. Drawings to be traced can be put under the glass surface, or the tablet can be mounted over an existing display device for direct-feedback operation. A device very similar to a touch-sensitive tablet is. the touchpanel. It consists of a series of light-beam emitters mounted in two adjacent edges of a frame, and a similar series of light-beam detectors mounted in the opposite edges. A pointing operation is
XI
+
y2
=
D2
X
J DI + DI
=
2D'A
2 y2
+
Z2 =
D'
z2
+
X2
DI
=
Y=
I
Figure 4. Three-dimensional tablet. 14
, +
D2 D_2\_
77-77n2,
V722, + D2,,
2
D
recognized when two perpendicular beams are interrupted simultaneously. This information forms an immediate x,y position pair. This device is well suited to display-screen pointing operations, but is greatly limited by its gross resolution capabilities. Analog devices. Tablets and touchpanels are basically pointing devices, whether feedback is direct or indirect. Analog devices are fundamentally different. Most make use of resistive devices, either potentiometers or pressure-variable resistors (transducers), as measuring devices. The variable resistance is used as a voltage divider, forming a measuring ratio that usually varies from 0 to 1 over the range of the device. The accuracy of these devices is dependent on the quality of the variable resistors used, and is measured as a percent of full range. Since the devices measure continuously, their resolution is infinite. However, in order to be used by a digital system the measurement must be passed through an analog-to-digital converter, so the effective resolution is determined by the bit width of the converter (usually on the order of:8 to 14 bits). In any case, the lack of accuracy may make the resolution question of small interest. Typical device accuracy ranges from 10 percent of full throw to 0.1 percent. The characteristics of these indirect tactile devices will affect user satisfaction and the ease with which users become accustomed to them. The most common device is the joystick, pictured in Figure 2d. Usually, the base of a movable joystick contains two potentiometers that vary directly with the motion of the armature. In these models, the arm moves physically with applied force, where the force is either a constant friction force or increases with distance from the center. This friction force, or spring force in center-return types, is adjustable in many models to suit the user. Other joysticks have immobile arms, values being obtained from the force applied by the user; these devices do not usually exhibit a linear relation between force and digital readings, which must be taken into account. What type is chosen-force or motion actuated-depends on the environment to be simulated. An added advantage of joysticks is their threedimensional capability. Many models-are available with a third resistive device, mounted in the handle, that may be activated by torque about the handle axis, downward pressure or movement of the handle, or thumb-actuated slide resistors. The variety of styles, grips, and sizes is due in great part to military vehicle control needs, but translates to a large selection of reliable controls for graphics applications. A similar device is the trackball, used for cursor control in many radar-type installations. It consists of a spherical ball mounted in a fixed-base housing. The sphere spins freely in all directions, with frictional forces being adjustable in many versions. Either the spin of the ball causes the movement of internally-mounted potentiometers, or optical encoders generate signals used for up-down digital COMPUTER
counters in x and y directions. Most models of trackballs give an added direct tactile feedback to the user by way of the momentum acquired by the sphere; this slewing effect can be advantageous in speeding user response in some environments. Generally speaking, trackball devices are more accurate than joysticks because of their greater freedom of movement, but do not have the natural return to center that most joysticks do. The mouse device, developed by SRI,20 is a selfcontained handheld device that can be used on any flat surface by virtue of two supporting wheels that drive internally-mounted potentiometers. The wheels are mounted at right angles to record motion in x and y directions relative to the orientation of the mouse. As the mouse is moved across the surface, the wheels cause a directly related change in the potentiometer settings showing cumulative movement. It provides an inexpensive tracing mechanism with no fixed origin (since it can be lifted above the surface), though its accuracy is gross. Errors can be especially noticeable as the wheels slip during diagonal movements, more or less depending on the pressure applied and the type of surface. Also, changes in orientation of the mouse change the coordinate system of recording data, but not the coordinate system of the data itself. It should be noted that valuator devices are usually potentiometers or other variable-resistance devices working on the same principles as singledimension locators. Typical valuators are slide potentiometers, dial potentiometers, or lever potentiometers. As in other locators, models are available in immobile, force-actuated models and in sizes from miniature to very hefty.
Buttons. Button devices consist of simple switches or multiple-position switches, along with the necessary interface hardware. They may be eventdriven devices operated like keyboards, or they may be sampling devices on fixed-position switches. A set of buttons, or even a single button, may be interfaced as a single input device or may be multiplexed over a common device line, as is common with keyboard and function-key devices. In each case, each key must generate a unique code or be otherwise uniquely identifiable. In addition, the hardware implementation of the switches may fix the consequences (or function) of the switches or, more generally, allow system or user software to assign consequences. The only drawback to software control over button devices is that the functional definition of a button may be changed without any physical change in the button that would alert the user to altered function. Although buttons are clearly separated here from other devices, they are commonly attached to other types of devices, making their logical functions closely interwoven with the devices they are attached to. It is common, for example, to attach a finger-actuated switch to a tablet stylus to allow the user to signal his intention to use the tablet.
November 1978 Reader Service Number 6
Light pens. The light pen is the only natural pickclass device currently accepted, though even its degree of naturalness is questionable. It is used as a direct graphical device (see Figure 2b), but at the hardware level, it is an interrupt-driven sensor, actuated only if a light beam is detected by the tip of the pen. Its basic piece of information is timing information, useful only if it can be determined what was occurring in the display device at a particular instant. This makes the use of a light pen extremely dependent on other system hardware and software. First the display device being used must generate its picture in a predictable timing sequence. This disallows use of a light pen with storage-tube or plasma-screen displays, where images need not be refreshed; picking functions on these devices must be simulated in software or hardware, using locator devices. Therefore, light pens can be used only- with refresh and raster-scan displays.
proximately how far along that line the beam was when the interrupt occurred (taking into account any fixed time lag). This, of course, must be handled with cooperation of built-in display-scanning hardware. Once a position is known, however, the problem reverts to one of using a locator device to simulate a pick operation.
Device simulation
The behavior of the various input devices at the lowest level defines their basic capabilities, but does not limit their use to a particular class. For instance, locators are commonly used to simulate buttons, in conjunction with a display device. A menu, or list of functions available, is listed on the display device, and the locator is used to indicate which of the alternatives a user wishes. This not only allows softwareprogrammable buttons but also relieves the user of Though used as a graphical device, the responsibility of remembering which functions are associated with which buttons at any point in the light pen is actually an time. In fact, many common digitizing systems ininterrupt-driven sensor. clude a small, low-resolution tablet for use as a function menu and a large, high-resolution tablet for In refresh display systems, the picture is redrawn highly accurate data input. As alluded to above, location devices can also be at least 30 times a second from a display file of drawing instructions.20 A light-pen interrupt can be used to simulate pick-class devices, though softused to stop the display-file interpreter and locate ware/hardware support is more demanding than it is what line segment (display-file instruction) was just for other device simulations. It involves simulating being drawn that caused the event. This process the function of drawing the vectors of the current may be handled in hardware, firmware, or software, picture from a pseudo display file or other picturedepending on how the display-file interpreter is im- data structure, through the use of a comparator plemented. The process may be complicated if the function. The locator-position input pair are loaded drawing speed of the display is fast relative to the into two comparator registers, and simulated picresponse of the light pen, making it unclear which of ture drawing is started, vector by vector. For each several recent vectors being drawn may have caused line, the position pairs of the two endpoints are loadthe interrupt. (It is possible to retry a limited ed into the comparator, which checks to see whether number of potential vectors for a confirming inter- the line joining those endpoints lies on or sufficientrupt without causing appreciable picture fading or ly near the original position pair. If so, the comflicker.) Further problems may be encountered if parator sets a condition code indicating a hit. branching instructions are allowed in the display- Though this process may be slower than most disfile code, as the retracing of the most recently exe- play-file processors used in vector-refresh systems, cuted instructions becomes more difficult. To en- it need be done only upon user request, rather than sure proper operation, display-code branches can be by the continuous iteration of refresh systems. delayed sufficiently to guarantee that a light-pen interrupt is not currently being initiated, or a limited back-tracing facility can be provided, such as an instruction-location queue of fixed size maintaining Microprocessor applications the most recent instruction locations. When the vector causing the interrupt is identified, its name, or With all of the devices we have been considering, the name of the object of which it is a part, may be special considerations are required to present input returned as the object-identification information. information to the central processing site of the This will involve many system-dependent imple- system in a standard format. Many current graphmentation factors, including naming conventions. ics-system designs use an intelligent terminal to In the case of raster-scan displays, the beam take some of the burden from the central site. As the always covers a fixed pattern on the screen, and the price and versatility of microprocessors improve, beam's intensity is varied to generate a picture. new opportunities are being created to further Here, the system can decode a light-pen interrupt lighten this load. The current price of microevent to an approximate x position along a known processors and associated circuitry indicate that it horizontal line (y position) by virtue of knowing is worthwhile to waste hardware if it can lead to which horizontal line the beam was scanning and ap- simplified and structured systems. At the simplest
16
COMPUTER
level, microprocessors with permanent firmware can be used as device controllers, to guarantee that devices will present their information to the receiving processor in a standard format. This does not take full advantage of the microprocessor's potential, but does allow plug-in substitution capability for widely different physical devices, and a degree of insulation of internal processing from devicedependent requirements. For extended capabilities, these controllers can allow software loading by the controlling processor, giving functional substitution capabilities under system or user control. This can allow not only device standardization, but device-type simulation as well. Using a microprocessor in this way suggests a hierarchical networking structure. At the top is the central processing site, with large-volume peripherals assignable to users either directly or as spooled devices. These would include large backup stores, high-quality plotters, special film-recording facilities, fast printers, and other standard devices. At the middle level would be a local processor to accomplish limited processing such as text editing, picture editing, picture previewing, information gathering, and program generation. At the lowest level would be the individual controllers for I/O devices, interfacing to the local processor at the information-class and device-class level.
Potentially, each user station can be implemented as a local network of device handlers and information processors. To the degree that responsibility and power are delegated to the lowest level, new microprocessor potential is exploited and more computational power is available to the local processor. This leads to less responsibility and demand on the central site, and greater possibilities for a distributed graphics network. A group of local processors can be fully or partially connected in various network configurations, or loops, rather than in the hierarchical star. Each node can be either a specialized node controlling use of shared devices, or a generalpurpose node supporting a single user or possibly a group of users. In such a situation, the functions of the operating system are distributed among the nodes. This networking facility can be expanded to offer each user the full potential of power available at each node. In this situation, even the individual device handlers interface at the operating-system control level, routing input graphic information through the network to the user process that requires or requested the information. Potentially, as microprocessor prices decrease and capabilities increase further, each user station can itself be implemented as a local network of device handlers and information processors, making the entire system a two-level network.
November 1978 Reader Service Number 7
Conclusion In considering input-device choices and functions
for a computer graphics system, user satisfaction and satisfied information requirements are the designer's goal. Several factors enter into these decisions, especially the effects of other graphicssystem components. In order to weigh the choices reasonably, a structure is needed within which
devices can be compared. Information and device classes are the basic elements of system design. The
simplest-level classification is viable whether devices interface to a central site, to local processors, or to immediate device handlers. It is most important that the classification be amenable to being carried over into more complex system designs before the proliferation of more complex devices becomes unmanageable. a
References and Bibliography 1. Brenner, A. E. and P. deBruyne, "A Sonic Pen; A Digital Stylus System," IEEE Trans. Computers, Vol. EC-19, No. 6, June 1970, pp. 546-548. 2. Capowski, J. J., "Remote Manipulators as a Computer Input Device," MS Thesis, Dept. of Computer Science, Univ. of North Carolina, Chapel Hill, 1971. 3. Cotton, I. W., "Network Graphic Attention Handling," Proc. Online 72 Conf., pp. 465-490. 4. Curry, J. E., "A Tablet Input Facility for an Interactive Graphics System," Proc. Joint Int'l Conf on Artificial Intell., May 1969, p. 33. 5. Davis, M. R. and T. 0. Ellis, "The Rand Tablet: A
Man-Machine Graphical Communication Device," AFIPS Conf. Proc., Vol. 26, Part I, 1964 FJCC, pp.
325-332. 6. Deecker, G. F. P. and J. P. Penny, "Standard Input Forms for Interactive Computer Graphics," Computer Graphics, Vol. 11, No. 1, Spring 1977, pp. 32-40. 7. Foley, J. D., "An Approach to the Optimum Design of Computer Graphics Systems," CACM, Vol. 14, No. 6, 1971, pp. 380-390. 8. Foley, J. D. and V. L. Wallace, "The Art of Natural Man-Machine Conversation," Proc. IEEE, Vol. 62, No. 4, Apr. 1974, pp. 462-471. 9. Foley, J. D., Hogan, R. and C. Dunham, DIGS-Display Independent Graphics System, Dept. of Computer Science, Univ. of North Carolina, Chapel Hill, Sept. 1975. 10. Foley, J. D., "Picture Naming and Modification: An Overview," (Proc. ACM Symposium on Graphic Languages, Apr. 1976), Computer Graphics, Vol. 10, No. 1, Spring 1976, pp. 61-63. 11. "Graphics Standards Planning Committee," (SIGGRAPH Activities), Computer Graphics, Vol. 8, No. 4, Winter 1974, pp. 7-9. 12. Hansen, W. J., "User Engineering Principles for Interactive Systems," AFIPS Conf Proc., Vol. 39, 1971 FJCC, pp. 523-532. 13. Hornbuckle, G. D., "The Computer Graphics User/Machine Interface," IEEE Trans. Hum. Factors Electron., Vol. HFE-5, No. 1, Mar. 1967, pp. 17-20. 18
14. Konkle, K. H., "An Analog Comparator as a PseudoLight Pen for Computer Displays," IEEE Trans. Computers, Vol. C-17, No. 1, Jan. 1968, pp. 54-55. 15. Meadow, C. T., Man-Machine Communication, Wiley, New York, 1970. 16. Michael, G. A. and R. K. Cralle, "A Survey of Graphic Data Processing Equipment for Computers," Computer Oriented Circuit Design, Prentice-Hall, 1969, pp. 381-442. 17. Newman, W. M., "A System for Interactive Graphical Programming," AFIPS Conf. Proc., Vol. 32, 1968 SJCC, pp. 47-52. 18. Newman, W. M., "A Graphical Technique for Numerical Input," Computer Journal, Vol. 11, No. 1, May 1968, pp. 63-64. 19. Newman, W. M., "An Approach to Graphics System Design," Proc. AFCET/IRIA Journees Graphiques, 1973, pp. 23-26. 20. Newman, W. M. and R. F. Sproull, Principles of Interactive Computer Graphics, McGraw-Hill, 1973. 21. Newman, W. M. and R. F. Sproull, "An Approach to Graphics System Design," Proc. IEEE, Vol. 62, No. 4, Apr. 1974, pp. 471-483.
22. Roberts, L. G., "The Lincoln Wand," AFIPS Conf
Proc., Vol. 29, 1966 FJCC, pp. 223-228. 23. "Status Report of the Graphics Standards Planning Committee of ACM/SIGGRAPH," Computer Graphics, Vol. 11, No. 3, Fall 1977, pp. 1-1 - 1-19, II-1 -
11-117.
24. Stotz, R., "Man-Machine Console Facilities for Computer Aided Design," AFIPS Conf. Proc., Vol. 23, 1963 SJCC, pp. 323-328. 25. Sutherland, 1. E., "Computer Inputs and Outputs," Scientific American, Vol. 215, No. 3, Sept. 1966, pp. 86-96. 26. Teixera, J. F. and R. P. Sallen, "The Sylvania Tablet: A New Approach to Graphic Data Input," AFIPS Conf. Proc., Vol. 32, 1968 SJCC, pp. 315-322. 27. Trambacz, U., "Towards Device-Independent Graphics Systems," Computer Graphics, Vol. 19, No. 1, Spring 1975, pp. 49-52. 28. Villalobos, L., "Virtual Graphic Device," Computer Graphics, Vol. 8, No. 4, Winter 1974, pp. 14-21. 29. Wallace, V. L., "The Semantics of Graphic Input Devices," (Proc. ACM Symp. on Graphic Languages, Apr. 1976), Computer Graphics, Vol. 10, No. 1, Spring 1976, pp. 61-65.
Mark Ohlson is an instructor at Texas A&M University. His research interests include language design and
translation, graphics system design,
distributed systems, and computer architecture. He is currently involved in design and implementation of a distributed graphics operating system. A member of the IEEE Computer Society and of the ACM and president of the local chapters of each, Ohlson received the BS degree from the State University of New York at Brockport in 1975, and the MS from The Ohio State University in 1977.
COMPUTER
