LI B R.AFLY OF THE UNIVERSITY Of ILLINOIS 510.84 U6r no. 349-354 cop. 2. Digitized by the Internet Archive in 2013 http://archive.org/details/vistasystemforco354roll lB7 3f mm * * n ««$g WO REPORT NO. 35^ ??U^ coo-ii*6y-oiU7 THE VISTA SYSTEM FOR COMPRESSION OF TELEVISION SIGNAL BANDWIDTHS by DAVID CHARLES ROLLENHAGEN October, I969 THE LIBRARY OF THE OCT 28 1969 mUVEKSITYOFIUiW Report No. 35U THE VISTA SYSTEM FOR COMPRESSION OF TELEVISION SIGNAL BANDWIDTHS* by DAVID CHARLES ROLLENHAGEN October, 1969 Department of Computer Science University of Illinois Urbana, Illinois 618OI Supported in part by Contract Number U.S. AEC AT(ll-l) lk6y and submitted in partial fulfillment for the degree of Doctor of Philosophy in Electrical Engineering, at the University of Illinois, October, 1969- THE VISTA SYSTEM FOR COMPRESSION OF TELEVISION SIGNAL BANDWIDTH'S David Charles Rollenhagen, Ph.D. Department of Electrical Engineering University of Illinois, 1969 The VISTA (Variable Interlace System for Television Applications) System compresses the bandwidth of television video signals in accordance with the amount of information change between successive television frames . The system automatically determines the information change and selects one out of sixty- four possible scanning speeds commensurate with this change. If the television information is constant, a large compression is possible, whereas a sudden change requires the full bandwidth of four MHz for a faithful reproduction of the television image- The system thus realizes the most efficient use of the television bandwidth. Flicker and line drift effects at reduced scan rates are overcome by means of a variable persistence display device and by interlacing the fields in a programmable sequence. These techniques enable high quality reproductions of printed matter at bandwidths of up to eight-to-one. A variable scan rate television camera is described and the signal current output of the vidicon as a function of scan rate is analyzed in detail. Also described in detail are the variable rate horizontal ramp generator and the automatic information change detector. ACKNOWLEDGEMENT The author is greatly indebted to Professor W. J. Poppelbaum, Principal Investigator of the Hardware Research Group and co- inventor of the VISTA system. His good counsel and friendship have led to a most enjoyable and rewarding four years at the University of Illinois. Few men undertake the combined roles of researcher and educator with more dedication. IV TABLE OF CONTENTS Page 1. INTRODUCTION 1 2. VISTA SYSTEM OPERATION 6 3. SUBSYSTEMS 10 3.1 Delta Density Detector 10 3.2 Variable Rate Horizontal Ramp Generator 14 3-3 Frame Control 20 3.h Vertical Ramp Generator 23 3.5 Scan Program Control 26 3-6 Variable Scan Rate Television Camera 28 3.6.1 Signal Level as a Function of Scan Rate 30 3.6.2 Video Preamplifier 1+2 3.6.3 Range Discriminator Circuit k'J 3»6.k Vidicon Protection in the Event of Scan Failure ... ky 3.6.5 Deflection Amplifiers 51 h. CONCLUSION 5k APPENDIX 59 LIST OF REFERENCES 73 VITA 75 LIST OF FIGURES Figure Page 2.1 Simplified Block Diagram of the VISTA System 7 3.1.1 Block Diagram of the Delta Density Detector 11 3.1.2 Compare and Hold Circuit 12 3.2.1 Variable Rate Horizontal Ramp Generator 15 3.2.2 Retrace Circuit 16 3.2.3 Waveform Illustrating Self -Correcting Nature of Horizontal Ramp 18 3.2.4 Derivation of Binary Numbers for Horizontal Ramp Generator 21 3.3 •! Block Diagram of Frame Control 24 3.4.1 Vertical Ramp Generator 25 3.5-1 Scan Program Control 27 3.6.1 Block Diagram of Variable Scan Rate Camera ........ 29 3.6.1.1 The Vidicon Target and Associated Circuitry 31 3.6.1.2 Complete Circuit for Each Element of the Target Surface . 34 3.6.1.3 Beam Current at a Point as a Function of Time shown for a frame time of T seconds and for a frame time of xT seconds 35 3.6.1.4 Waveforms Illustrating Derivation of Signal Amplitude . . 37 3.6.1.5 Beam Current at a Point Assuming the Distribution [1 + cos (I)] 38 3.6.1.6 Normalized Signal Amplitude as Function of Scan Rate with Constant Illumination 4l 3.6.2.1 Simplified Diagram of Vidicon-Preamplifier Circuit .... 43 3.6.2.2 Gain Control Circuit 45 3.6.2.3 Video Preamplifier 46 3.6.3.1 Gain Range Discriminator 48 3.6.4.1 Horizontal Deflection Failure Detector 50 vi Figure Page 3.6.4.2 Vertical Deflection Failure Detector 52 3.6.5.1 Deflection Amplifiers 53 A. 1.1 Disc Video Control 6l A. 2.1 Analog- to-Digital Converter 62 A. 3.1 Partitioning Procedure Applied to a Hypothetical Gain Curve 6k A. k.l Error Correcting Circuit 69 A.U.2 Per cent Error Versus Number of Cycles Illustrating Convergence of Ramp Voltage Error 72 1. INTRODUCTION The desirability of compressing the bandwidths of television video signals certainly requires no justification in this day and age where the over-crowded frequency spectrum is forever expanding to accommodate new bands . In fact, bandwidth compression has been the object of rather intensive study for some forty years. Evidently, it was clear at the outset that sufficient redundancy existed in television video signals to warrant such study. To date, however, there exists a wide gap between the recognition and actual quanti- tative measure of redundancy, and the practical means for utilizing this redundancy to achieve some economy of bandwidth. With the advent of information theory and in particular Shannon's work, means are now available for the measurement of redundancy and the theoretical determination of an upper bound to the possible bandwidth compression of television signals. Work in this area has now progressed to the extent that further advantage may be taken of the psycho-physical characteristics of vision beyond simply minimizing flicker. That is, redundancy exists due to man's inability to concentrate fully on the spatial, contrast, and motion characteristics of a television picture simultaneously. Hence the recent interest in the statistical nature of typical television pictures. Again, the fact remains that in spite of the theoretical advances with regard to redundancy and the possible bandwidth compression associated therewith, the practical realization of such bandwidth compression has lagged behind; not because of insufficient interest or talent in this area, but because of certain apparently insurmountable technical obstacles. Nearly every system conceived thus far for the purpose of reducing television bandwidths realizes one or more of three basic types of redundancies: redundancy from element to element along a horizontal line, redundancy between 2 horizontal lines, and redundancy from frame to frame. The first two deal with picture areas, and the latter deals with picture motion between frames. It has been shown that a high degree of correlation exists in typical pictures 3 for a run length of up to 50 elements per line, for kO lines, and for 30 frames. At first glance these figures could, theoretically, result in a tremendous economy of bandwidth. Only the most complex, and hence prohibitively costly, scheme could realize all three types of redundancy in an on-line fashion. To the extent of the author's knowledge, any workable scheme conceived thus far has dealt with, at most, only one of the three types of redundancy. The type most amenable to practical application is element-to-element redundancy along horizontal lines, though it has been suggested that the scheme due to Cherry ' k and Gouriet could easily be extended to cover element-to-element redundancy 5 6 in both the horizontal and vertical directions. ' Frame to frame redundancy is, perhaps, the most difficult to treat because some intermediate storage facility is required at the receiving end of the television system. The most successful type of system involves detection of brightness changes along horizontal lines. The only video information required is the position and extent of areas of constant brightness along horizontal lines. Of course this in itself does not realize any economy of bandwidth, since many brightness changes may be concentrated in a small area of the picture, and the required bandwidth in such cases may be the full bandwidth. The brightness information must be redistributed along the time scale to effect a constant average flow of information. The value of this information rate depends on the statistical nature of the picture. The particular scheme referred to above has led to a three-to-one bandwidth compression consistent with reasonably- good quality pictures. One inherent problem of nearly any element-to-element redundancy scheme is the enhancement of the noise level during detection of brightness changes. As Professor Cherry so correctly points out, statistical methods must be employed in detecting brightness changes to reduce the effect of additive noise. Simple element-to- element differentiation schemes, for 7,8 example, are not practical in any real communication link. Thus far, it appears that a three-to-one compression is the upper limit commensurate with a faithful reproduction of typical television broadcasts. Considering the inherent complexity of any practical system, it is questionable as to whether a three-to-one compression is economically worthwhile. Even more discouraging is the fact that such systems have realized 9 little, if any, improvement over a simple high cut-off filter. The picture may not be so bleak, however, and surely the discouraging practical aspects do not call for a termination of bandwidth compression studies. Television is finding greater and greater usage in areas which lend themselves to bandwidth compression schemes far better than does commercial television. Reproduction of printed matter, drawings, picture phone applications are good examples, to name a few. Many ham radio operators will testify to the usefulness of band- with compression schemes. A completely different class of bandwidth compression schemes aims not at a digital dissection of television pictures, but rather at a modification of the basic scanning process with no alteration of the resulting video signal. The pseudo-random scan, the dot interlace scheme, ' the four-to-one interlace 12 scan are systems which fall into this class. For systems of this class it is h no longer meaningful to speak in terms of redundancies , since the processes involved here are more adequately expressed in terms of the frequency domain rather than in terms of the time domain. Generally speaking, systems in this class realize a bandwidth reduction directly, usually through reduction and modification of the scanning velocity. Of course, these schemes, too, are not without their drawbacks. Two inherent problems are the inability to transmit picture motion without blur or jumpiness and the flicker problem associated with reduced scanning speeds. The system described herein falls into the "frequency class" rather than the "time domain class" of systems. The objective of this project is to transmit the full signal bandwidth only when severe picture motion warrants its use. At other times the bandwidth is compressed, the exact amount of compression (and hence the choice of scan rates), depending on the severity of the picture motion. Should the entire picture change between successive television frames, the full (h MHz) bandwidth is called for to reproduce this change without blur. Should only a small segment of the total picture area change between successive frames, the bandwidth may be reduced significantly without picture degradation. The bandwidth is compressed by reducing the scanning velocity in the television camera and receiver simultaneously. At reduced rates, the flicker problem is eliminated through the unique application of a variable persistence display device. In addition, the television fields are interlaced to help eliminate the downward drift of horizontal lines. The number of fields per frame depends on the scan rate, and hence the bandwidth. The exact interlace sequence is programmable to ascertain that sequence which optimizes the picture quality. A question of economics presents itself in that the system must be capable of the full television bandwidth in the event of a complete picture change between successive television frames. This system in itself represents no economic gain, although its application in a larger system could realize a substantial gain. An observation of nearly any commercial television broadcast leads one immediately to the conclusion that very seldom is their sufficient motion, or enough complete picture changes to warrant the use of the full bandwidth. Possibly, many television signals could be monitored and statistically sampled for their respective bandwidth requirements. Each could then be shunted to an appropriate channel having the required bandwidth, and no more. It is not difficult to imagine many such schemes, but an elaboration beyond the immediate design and construction of the VISTA system is outside the scope of this project. Perhaps the most important motivating goal is to determine to what extent the bandwidth of the commercial television signal can reasonably be compressed using the reduced scan rate method and employing the variable persistence and programmable interlace features. A significant out- some of the project will be that particular interlace sequence most acceptable to the viewer. In addition, the project provides a valuable insight into technical problems associated with generating variable-rate waveforms and designing a camera operable at any one of many different scan rates. 2. VISTA SYSTEM OPERATION A central problem in the design of the VISTA system was to derive a signal representing a measure of the amount of information change between successive television frames, hereafter referred to as the delta density signal, and to generate deflection voltages based on this measure. The deflection signals are then applied simultaneously to the deflection inputs of the television camera and monitor constituting the closed circuit system. Additional complexities centered around the primary objective of this study, namely the optimization of the television picture quality at reduced bandwidths by virtue of the persistence control and the programmable raster shift sequence. Sixty-four different scanning speeds are provided, the fastest corresponding to the conventional rate, and the slower speeds derived by dividing the fastest rate by the integers, 1, 2, . . ., 6U. If the delta density signal indicates a complete change of information, the fastest rate is selected. Otherwise one of the slower rates is selected depending on the relative magnitude of the delta density signal. Figure 2.1 presents a simplified block diagram of the VISTA system. The television camera is used for two different purposes. It provides a signal to enable the delta density detector to measure the information change between frames. In addition, it is used as a slow scan camera during the bandwidth compression process. Thus the system alternates between a fast scan mode during which the information change is sampled, and a slow scan mode during which the bandwidth is compressed on the basis of the sample. This alternation is not observed by the viewer since the same picture is displayed regardless of the rate at which it is scanned. One of the many unknowns of the system was the number of frames required for the vidicon tube in the camera to settle a c H c s t a. c .0 CD -P o CO *H bO co" •H Q M o o rH PQ (D o o QJ O w •H O a h + ■ P a) •H P CO p r-. c a CJ a.' '^ Tj ti H all a; o P a H bD(= (i) •H n CO < •H 3= O o H o w o -p o a> -P cu O >> -p •H ra id P /N o H o H to ■t CM UJ_I or \- I- z ujo a. u u CO LJ I a H if) ©" 12 a: >- OD a: SjJ t- z 3 o o tr LU o o < o o z o u p A *& AAA AAA O H NDI t ID CM CM (NJ CO CM CM i: O 13 0) O & H cS -P o ■H O w •H H CM en a> I •H 16 o u •rl O CD O OJ *H -P CD « OJ on CD •H 17 This unique scheme for generating the horizontal ramp enjoys two distinct advantages over any other arrangement. First, the retrace period is always a fixed proportion of the entire ramp period (namely 15. 8 percent) regardless of the scan rate. This follows from the fact that regardless of the magnitude of the capacitance, all capacitance values are charged and discharged by the same two constant current sources. Second, for sudden disruptions in the system which may cause a faulty count or any other spurious non-periodic error, the circuit is self -correcting. This is a direct result of the geometry of the waveform which is illustrated in Figure 3-2.3. Here the period is designated T, the ramp slope a, the retrace slope fa, and the voltage amplitude V. Assume a sudden displacement at the beginning of the ramp (so chosen for the sake of convenience) of magnitude 8T. When retrace is initiated by the coincidence circuit, a voltage error &V-, = a5T results. During the second period the error in the ramp period is 5T Q = - ■$ — which leads to a voltage <— i a aST error 8V = - — — . Continuing in a similar manner, aB T SV 3 = + — 8V = (-l) 2n+1 a6T n n-1 Thus after n periods, the voltage error has been reduced by a factor of f Since f = 5-35? the initial voltage error is reduced by a factor of 28.6 after three complete periods . 18 & « H CO •p § N •H is O o 13 a - £ •H •P o CD u u o o «h H CD CO bD a •H -P W o 2 + nl2 5 + n,»2 4 + n'2 3 + n'P c switches transformed to the number N' = n>2 + n'2 + n,' 2 + n'2 + n'2 + n'2 6 5 4 3 2 1 + n'2 which equals n^2 + n, 2 + n 2 + n^2 + n n 2 + (n^ + 1)2 . Both numbers 5 1 +3 21 v y N and N ' posses 6k possible states, but their respective decimal equivalents are zero to 63 and one to 6k. An addition of the binary digit 1x2 to the number N, following an inversion, results in a signal N' which may be used to control the binary weighted capacitors directly. Otherwise the zero state of N, while being a perfectly admissible binary number, would not be a useful state in the switching of a capacitor. The result is, of course, a seven bit number. If t is the ramp period, t may be simply expressed in terms of the circuit parameters and the number N' as 20 vc t ■ ^(i + 5 i ) ^ n 6 + 32n ; + 16n i + 8n 3 + S + 2n i + n o ] = t . [6Un*. + 32n; + l6n,' + 8n! + kn' + 2n* + nl] min L 65 4 3 2 10 J Table 3-2.1 lists several exemplary "binary numbers N, the respective conversions N', and the corresponding ramp period t. The NAND gates immediately following the buffer register in Figure 3-2.4 were originally intended simply as inverters. However, the gates were found ideally suited as an input for switching the system between the slow and fast scan modes. With the fast scan gate at a logical "0", the outputs of the WAND elements must be logical "l"'s regardless of the number N. The 6 5 h 3 2 1 1 resulting binary number is N' = 0-2 + 0-2 y + 0*2 + 0-2 J + 0*2 +0*2 +1^2 corresponding to t = t . , the desired horizontal ramp period in the fast scan nan mode. On the other hand, with the fast scan gate at a logical "1", the NAM) elements behave as simple inverters in the conversion from N to N' during the slow scan mode. 3.3 -Frame Control The number of frames scanned in each of the slow and fast scan modes is initiated by the operator at the control panel and implemented by the frame control subsystem. The scan mode control flip-flop, one of the most important single elements in the system, is contained within this subsystem. Figure 3«3«1 illustrates the mechanism by which the scan mode control flip-flop switches the entire system between the slow and fast scan modes at the conclusion of the preselected number of frames. The line counter and the 111111111 count detector determine a complete frame of 512 scan lines. Each pulse corresponding 21 ft s o3 w o 0- T3 <3 CD M EH U -P^~ •H TC* o o AAAA CD -a o. •P CO < ^ 0) 0) P> <+H W Ch -H 3 bo CD -P a3 bO a5 o w -P 03 03 u CD ^H CD > u CD P 1 U P> CD \. > < a o o o -p oj CD c CD ft « 03 -P C O tq •H *h o o w CD £2 >s •H PQ <+H o o •H -P a3 > CD Q OJ CD •H 22 I 000000 000001 000010 EL 1000000 linn 111110 6k T . mm 63 t 62 nan min 111110 min ooooio oooooi 2 T min T . min Table 3- 2.1 Exemplary Binary Numbers Used in the Derivation of the Horizontal Ramp 23 to the completion of one frame then increments the 6-bit frame counter. When the output of the frame counter coincides with the scan duration signal set by the operator for each mode, a coincidence pulse occurs indicating that the prescribed number of frames has been scanned. At this time the scan mode control flip-flop changes states for the beginning of a new frame count in the alternate mode. A Q output of "1" corresponds to the fast scan mode during which time the fast scan gate is actuated and the fast scan duration signal enters the coincidence detector circuit. When the Q, output becomes a logical "0" the slow scan mode begins at which time the A/T) register containing the digital delta density signal is strobed, and the slow scan duration signal enters the coincidence detector circuit. 3'^ Vertical Ramp Generator Among other responsibilities demanded of the vertical ramp generator is that it shift each successive raster up or down in accordance with instructions received from the scan program control. The means for achieving this result is the 9-bit adder illustrated in Figure 3.^.1. A digital count corresponding to the specified starting line of each field is added to the output of the 9-bit counter. The addition of the starting line count simply shifts each raster down by a specified number of lines at the beginning of each field. As this number changes, the level shift of the raster changes accordingly. The input register is strobed when the counter overflows and when the scan mode control flip-flop is in that state corresponding to the slow scan mode. The 9-bit counter produces a stair-step voltage ramp at the output of the D/A converter. The length and height of the steps is determined by the periodicity of the horizontal synchronization pulse which strobes the output •rl P _ H 2k a a cu g o d U a5 Tl <+H Ph fl H H Ctj 03 O O d Ih CO bO-P •H C! P co O CO o cr! ^H p CD co P u a! CCJ CD I*< bO p cu CO ,£> O £ •H O P 03 M ^ o CD P CO ^ w H O U P o o u Pn o •H « O o H co on CD •H fa ■p O & & 25 H o3 O •H -P u cd > O H -h n3 -P -P o3 H C! to c3 O -H £ n £ b£ •H O -H ?h ?h W ,2 CD o -P CO W o ra Jh CD -P Jh CD < > Q' O o Fh -p CD •P 3 -P •H ft W pq -p -h I 3 bO ON O m CD bO •H En 26 register at the end of each horizontal line. If the register is strobed every nth clock period, for example, corresponding to an n-to-one compression, the steps will be n times as long and n times as high as those for a one-to-one compression. The result is that every nth line is scanned. Except in the line sequential type scan, the ramp period remains fixed for all time. The counter overflows after 512 counts to begin a new vertical ramp period. The standard line sequential pattern is easily implemented at any scan rate by locking out the input register strobe and counting horizontal synchronization pulses in place of master clock pulses as shown in Figure 3»^«1« 3.5 Scan Program Control The sole purpose of the scan program control is to implement the interlace sequence programmed by the system operator. Basic to this scheme is the use of two 6-bit up-down counters in the derivation of the required digital number at the beginning of each field. Either one or the other or an alternate use of both counters may be employed in the scan program. The essential elements of the scan program control are shown in Figure 3-5»l« The AND-OR- IJWERT gates shunt the outputs of the appropriate counter to the input register of the vertical ramp generator. The state of the jk flip-flop determines which counter is being employed. In the event of an alternate use of both counters, the flip-flop is clocked once each field, and the counters are alternately switched to the common output leading to the vertical ramp generator. In this manner, a starting line sequence such as the following may be realized: 1, n, 2, n-1, 3? n-2, • • •> n /2, n/2+1. In this particular example, counter number 1 is set to count up from line number 1, counter number 2 is set to count down from line number n. Counter 1 is switched into the circuit during the first field, counter 2 is switched in during the second field, then back to counter 1 h Ill I s s CP 27 ,1 3.* II > = e 1- ■'I— Oy^ o — o i- W 28 and so on. Each counter is clocked once every other field in this alternate program. If only one counter is used to implement a starting line sequence of 1, 2, 3> • • .» n, for example, the jk flip-flop remains in the state to which it was initially preset or cleared (corresponding to the use of counter 1 or counter 2, respectively) and the counter being used is clocked once every field. The initial contents of each counter are set via the starting line switches on the control panel. The clock steering circuit shunts the field clock signal to the appropriate counter, depending upon the state of the jk flip-flop. During single counter operation the preset clear steering circuit sets the state of the flip-flop, thus selecting one of the two counters and steering the field clock accordingly. 3.6 Variable Scan Rate Television Camera The variable scan rate television camera was designed and built to operate at any one of 6h possible scan rates, the maximum rate being that of the conventional television system; namely, 15,750 horizontal television lines per second. Slower rates were derived by dividing the maximum rate by the integers 2, 3> • • •, 64. The deflection waveforms and blanking signal corresponding to these different rates were supplied to the camera by the main frame. The camera itself consists of a power supply, four deflection amplifiers, a preamplifier, amplifier, gain control circuitry, protection circuitry, and the all-electrostatic vidicon tube. A block diagram of the camera system is shown in Figure 3-6.1. The central problem encountered in the design of the camera was to maintain a constant signal level for all possible scanning rates. Without the use of a variable speed mechanical shutter, it was questionable as to whether or not the signal amplitude, extracted at the vidicon target, would remain 29 c {* CM- O- en c tn «n 4f r .£ c 1 p •J > c 1 > V 2 J — c c c ■~* 1 V / a. a. £ t I ~~ . 1 Ml J / ^ / k r j A a) o 13 o 0) rH •H O •H o O O H PQ H 0) bO ■H 30 the same for the various scan rates. The following section is devoted to an analysis of this problem in terms of the dynamic characteristics of the vidicon tube. 3-6.1 Signal Level as a Function of Scan Rate A brief review of the dynamic operation of the vidicon will establish the basis for the following analysis. The vidicon target consists of a photoconductive deposition backed by a transparent conductive coating which is electrically connected to the target flange. This results in an effective capacitance from one side of the target surface to the other, with the photoconductor acting as a dielectric. Figure 3 • 6. 1.1 shows the physical aspects of the target along with the external circuitry and the equivalent circuit of each small target element. Incident light focused on the target causes a certain leakage or conduction to occur through the photoconductive material at each small location, the exact amount depending on the intensity of the incident light at that location. The resistance R of each element, a function of the light intensity at that point, accounts for this leakage. The signal voltage, v (t), is extracted across o resistance R as the beam impinges on the target surface, recharging the o capacitance C of each element that it strikes. The beam side of each element is charged down to cathode potential, or zero volts, while the conducting side is returned to the target voltage, V . Resistance R^ accounts for the so-called dark current which adds to the signal current, producing the signal voltage. The dark current is generally small compared to the signal current if the target is constantly illuminated. 31 v g (t) Small target element Photoconductive material Target flange Electron beam Incident light Transparent conductive coating Figure 3.6.1.1 The Vidicon Target and Associated Circuitry 32 The target voltage, V , is typically 30 volts. The incident light causes the capacitance C of each element to be discharged from 30 volts in the manner of an exponential decay. At reduced scan rates, the incident light integrates over a longer period of time, resulting in a larger voltage decay. When the light is permitted to integrate in this manner, the frame rate must not be so low that the voltage across the target decays by more than several volts , otherwise the resulting electric fields will become sufficiently large to distort and defocus the electron beam. Since this decay may not exceed roughly 10 percent, the discharge occurs over a nearly linear portion of the exponential decay. The extent to which a linear approximation is valid is termed "linear storage". The salient implication here is that to a point, the voltage decay is nearly a linear function of frame time, assuming constant illumination. The recharging process occurs as the beam strikes each element of the target surface. Charging occurs until the surface is brought down to cathode potential. If it can be shown that the voltage increment by which C is recharged each frame is a linear function of frame time, then this recharge will exactly compensate the linear discharge, regardless of the frame time or scan rate. It then follows that the signal level cannot be a function of scan rate, assuming constant illumination and linear storage. Evidently, this result is well known to those experienced in the field of television pickup tubes, although to the best of this author's knowledge, no detailed explanation 13 1^+ has been set forth. ' The following simple analysis will show, to a good approximation, that the increment by which C is recharged each frame is propor- tional to the frame time. 33 Figure 3 • 6. 1.2 shows the complete circuit for any one element of the target. The current source and the diode represent the effect of the impinging electron beam. Assume that the function if (| ) represents the spatial current distribution of the spot in the £ direction through the center of the spot. As a function of time, therefore, the current impinging on each element may be represented by the waveform shown in Figure 3.6.1.3. (This figure neglects distortion of the beam as a function of velocity in the direction of motion.) Here again x is the scan rate factor. At time t = nxT + xt , a frame time of xT seconds, diode D becomes forward biased, and the video signal v (nxT + xt ) : i(nxT + xt )R is maximum. IL and R are very large, and may be neglected during the charging process. The signal voltage developed across R is generally small compared to the voltage developed across C, in which case v~(t) may also be neglected. Under these conditions, the voltage increment by which C is charged during each charging interval is v c (nxl + xt ) =^/ Xt O f (£)dt By a simple change of the variable of integration, M xt °f(i)dt = x±2r to f(t.)dt. c ,/ Q x c ./ Q Or, v (nxT + xt ) = xv (nT + t ). This is the desired proportionality relationship which implies that under the stated assumptions, the signal level is independent of scan rate, with constant illumination. If this relationship were anything but linear, the diode D would become forward biased at different phase angles of the i(t) curve, depending on the scan rate, resulting in & 3k v| >"S *D i(t) CO Incident light Figure 3.6.1.2 Complete Circuit for Each Element of the Target Surface 35 -p 3 o o -p -p a -p CO 'd C o co o S QJ •H CO EH EH <+H X O <+H o o •H -P o Ph CD cd § CO -t Diode D forward biased at time t = t. »->- t Figure 3.6.1.U Waveforms Illustrating Derivation of Signal Amplitude 38 -p E -PlEH w o o o o -p. •H T til o The voltage by which v is increased at time t during the recharging period is v c (t) . i /* i(t)dt = ^ /* [1 - =o s( M)]dt . ^[t - f sin(M)]. At time t , "when diode D becomes foward biased, the following relation must obtain: V Q + v (t ) + Kt )R g = V T Therefore, -xT v R L C 1 r+ xT . , kt n V^, , kt 0x n v Rearranging, -xT V T (1 - e ) = _[t - T sxn(— )] + — [1 - cos(— )] The solution of this transcendental equation determines time t , which in turn may be used to find the signal amplitude i R kt max -xT R L° Notice that if e is replaced by its linear approximation, and the last term on the right is ignored, the equation becomes 2kV T tt . , k \ i Q R L xT v xT ' Uo the solution of which gives t as a linear function of x. This implies that v is independent of x, the result first predicted assuming linear storage max and neglecting the signal voltage during the recharging process. The numerical results of this analysis are displayed on the graph of Figure 3.6.1.6. Typical values were chosen for all the circuit elements. Values of t were calculated for each integral value of x from 1 to 10. Each value of v was then calculated and normalized. Also shown in Figure 3-6.1.6 max are the linear storage approximation and the experimentally measured values of v s •* max *The stray capacitance between the target and all other electrodes in the vidicon has been omitted in the entire preceding analysis. This capacitance is by no means negligible, amounting to something of the order of 5 or 6 picofarad. This, in conjunction with a resistance R = 50 kilohms, for example, results in a time constant of 250 or 300 nanoseconds. Although this has a drastic effect on the high frequency response of the television camera, ommission of the stray capacitance in no way invalidates the preceding ■ . results. The stray capacitance tends to average the voltage across R due to the current contributions from all the individual elements. This and the fact that the target surface is, in reality, a continuum accounts for the smooth nature of the video signal as the beam traverses areas of the target which are uniformly illuminated. f Fh >) a H CD H G cO •H -p H 0) si S •p •H •H u s CD P. 4) 0) X xs X5 0) 2 d -p +5 Eh T? •H •H O CD H H •H ^i & Pj -p 3 a g -0 03 0} to s 03 •H rH H X S 03 CO o C fl Jh cu M M ft T3 •H •H Ph 3 « CO a5 -p •H H H 0) H cO c8 M P. o o nJ a •H •H fn cO -P -P O 0) 0) 4-3 H !h Fh w cO O o d CO -p •H O u •H O o CJ •H T) ■H > O -P d 0) H a3 > •H g. O bO c3 •H O -d 0) •H «h ■H H •H CO en •H kk the controlling mechanism and the input resistance R , an important consideration at such low signal currents. This circuit is shown in Figure 3.6.2.2. A feedback signal is returned to the gain control circuit from the preamplifier in order to bootstrap the gate and null the effect of the gate-to-drain capacitance. A field effect transistor is used at the input of the preamplifier for its high input impedence characteristics. This prevents attenuation of the already small signal level. Figure 3 '6. 2. 3 shows the preamplifier circuit. Here Q, , used as an emitter follower, bootstraps the source and drain through C, and C , respectively, thus eliminating the effect of the gate-to-source capacitances. In a similar manner, the collector of Q, is bootstrapped through CL . The input capacitance to the preamplifier circuit was measured at approximately 0.8 picofarad. The input resistance is well over 10 megohms. A combination of series and shunt peaking is employed in stages Q, and Q^. In order to enhance television picture detail, conventional television preamplifiers employ a peaking stage to boost the high frequency response. This technique was employed in this preamplifier design, though not to the extent that it nulled the noise reduction effect described above. Here an additional complexity arose due to the variable gain feature of the preamplifier. As the upper frequency limit falls off with increasing R , as discussed above, the b peak must shift accordingly. Thus a different pole- zero combination is switched into the circuit for each of the gain settings. This is shown in the emitter circuit of Q,^ in Figure 3-6. 2. 3- The networks switch simultaneously with the different values of R . b The voltage gain of the preamplifier is approximately 40 decibels. ^ -P •H 3 O u •H O H O ■s o o id •H CM OJ MD en CD •H k6 u v •H H & O (D -3 •H > m OJ 0) •ri 4 7 3.6.3 Range Discriminator Circuit Each scan rate -was catagorized into one of nine different ranges, each range corresponding to one of nine different preamplifier gain and filter values. An analytic approach was extremely useful as a basis for selecting the different ranges. This derivation is outlined in the appendix along with a typical example. It was immediately obvious that an analog, rather than digital approach was far better suited to the classification of 6h different scan rates among nine ranges. Decoding 6k different possibilities of a 7-bit binary number digitally would have been a formidable task, indeed. Instead, an analog conversion was performed, and the resulting analog number was classified in the manner of Figure 3-6.3.1. Each pair of differential amplifiers were connected in such a way that each responded if the analog voltage fell within its window. The numbers corresponding to the first two windows, being very simple binary numbers, were coded digitally while the analog technique was applied to the remaining ranges . The digital output of each pair of amplifiers controls a simple transistor switch. This, in turn, activates one of nine different LC filters in the preamplifier and generates one of nine possible values of current in the photocell- lamp gain control facility. An attractive feature of this scheme is that the integral nature of the scan rates allows the windows to be chosen such that no overlap occurs, and the scan rate numbers will always fall within one and only one window. ifiO — i+8 A o -p a a e •H O CO •H Q CD a •H cci en fa b. Q U9 3.6.4 Vidicon Protection in the Event of Scan Failure The problem of vidicon protection was compounded by the variable scan rate nature of the system. In a single scan rate system, slope detection applied to the horizontal voltage ramp is a simple and useful approach. In such a scheme, when the voltage amplitude corresponding to the first derivative of the ramp falls below a predetermined value, a blanking voltage is applied to the vidicon to prevent burning the target. This technique, with several simple modifications, was applied to the case of the variable rate horizontal scan. The deflection failure detector shown in Figure 3.6.4.1 is composed of two differential amplifiers together with their related components. The .001 microfarad capacitor and the 10 kilohm resistor tied to ground constitute the differentiating network. The differentiated voltage is compared with a reference via the first differential amplifier. When the threshold is exceeded, an output pulse is applied to the logic elements and the following integrating network. These elements prevent the circuit from reacting during horizontal retrace, or that portion of the ramp with negative slope. The protection circuit may blank the cathode of the vidicon only when the blanking input is a logical "1". The derivative of the horizontal ramp appeared to be an ideal source for extracting the blanking signal directly. Unfortunately, however, the time delay due to the differentiating network could not be tolerated, hence the blanking signal was derived at the horizontal ramp generator. The stair-step nature of the vertical ramp dictated a different scheme for protection of the vidicon. The nine binary inputs to the D/A converter, which produce the vertical deflection waveform, are monitored collectively. Each input is applied to a differentiating network, and the combined output is compared with the horizontal blanking signal. If at least one of the binary inputs changes state, in either direction, a flip-flop will 51 be reset. If no change occurs, indicating that the vertical waveform has not changed between two successive scans, the flip-flop will not be reset, and the cathode of the vidicon will remain blanked. Figure 3.6.U.2 shows the complete circuit. This scheme is obviously not infallible, for the fault could lie in the D/A converter. A trade-off between complexity and probable failure rate resulted in the choice of this scheme. 3.6.5 Deflection Amplifiers Due to the shifting nature of the television raster, direct-coupled deflection amplifiers were required to drive the plates of the electrostatic vidicon tube. The specifications called for 30 volts peak-to-peak deflection voltage per plate with an average voltage of 225 volts. An inverting and a non-inverting amplifier circuit were designed using operational amplifiers and high voltage transistors as shown in Figure 3-6.5.1. The signal outputs were returned through feedback resistors to the inputs. One such pair of amplifiers is used to drive the horizontal plates, and an identical pair to drive the vertical plates. The linearity and response of the amplifiers are excellent over the entire range of scan rates . 52 O -P O CD -P CD o H •H cd £ O •H •P O CD Q ■d o •H -P (1) > CM VD OO a; •H + 300V 53 -25V Figure 3.6.5.1 Deflection Amplifiers 5h 4. CONCLUSION The success of the VISTA system was limited to a large extent by the variable persistence oscilloscope. Although the persistence control performed admirably over the useful range of scan rates, the device was incapable of displaying shades of gray between the black and white extremes. The tube is essentially a bistable storage tube, and as such is useful for displaying printed matter only. Any attempt at displaying a multishade picture resulted in an enhancement of those levels above the bistable brightness threshold, and a fading away of those levels below the threshold. Although the manufacturer claimed that four gray levels could be resolved, these levels were indisting- uishable in the presentation of typical television pictures. In spite of the inability to display gray levels, however, the device worked quite well. The presentation of black and white pictures or printed matter clearly demonstrated the usefulness and success of the variable persistence technique. An optimum adjustment exists beyond which the picture will begin filling in (this phenomenon is known as picture "bloom"), and below which the bar drift effect will be noticeable. Of course at high persistence levels the inability to transmit picture motion becomes more pronounced due to the longer decay time. This effect predominates over the inability to transmit picture motion due to the reduced scan rate. Presumably, however, insufficient motion occurs at such reduced rates to warrant a faster decay rate. The bandwidth compression ratio corresponding to the least acceptable picture quality was in the vicinity of eight-to-one. Again, the persistence feature was the limiting factor. The degree of persistence required to null the bar drift effect beyond this ratio caused the picture to fill in, or bloom. It was found that a single setting of the persistence control sufficed for all scan rates up to the eight-to-one reduction. 55 Strangely enough, the television camera was not the limiting component as had originally been anticipated. In fact, satisfactory video signals were derived over the entire range of sixty-four different scan rates.' Even more surprising was the fact that no gain control was required over this extreme range. In view of the detailed analysis in section 3*6, the author concludes that the estimate of the capacitance value per element was incorrect in spite of the reliable source of information. The reason for this conclusion was the persistence characteristics, or more precisely the signal decay charac- teristics associated with the vidicon upon removal of the light excitation. Evidently the elemental capacitances are not fully recharged as the beam strikes each element. This, of course, is one cause of picture lag as mentioned previously. In this case the signal level is a function only of the beam current, and hence is not a function of frame rate assuming constant illumination. The capacitance associated with each element must be larger than anticipated to substantiate this result. This would also lead to a more linear storage since the exponent of the discharging relationship varies as 1/C. This lends further credence to the above argument. A second observation is consistent with the incomplete recharging of the elemental capacitances. An acceptable picture was derived after scanning only one complete frame at any one of the different scan rates. Thus it is sufficient to scan a single frame in each of the slow and fast modes, alternately. This implies that the information change may be monitored more often since it is not necessary to allocate several frames for steady state operation of the vidicon in the slow scan mode. The most acceptable interlace pattern was an alternate up and down shift of the raster. This pattern resulted in the maximum amount of raster motion, thus reducing the bar drift effect most successfully. The exact 56 starting line sequence appeared not to matter as long as the raster shifted up and down alternately in large steps. The following two starting line pro- gressions were equally effective and gave the best quality pictures: 1, n, 2, n-1, . . ., n/2, n/2+1 and n/2+1, n/2, n/2+2, n/2-1, . . ., n-1. The use of two counters in the alternate mode was noticeably better than the use of either counter alone. A comparison between the n-to-one interlace scheme and the line sequential scheme left no doubt as to the effectiveness of the former. At any one of the reduced scan rates, the flicker problem rendered the line sequential pattern completely unacceptable as anyone might expect. Ideally, a completely random interlace scan would probably be the most desirable type of interlace pattern, though the dynamic characteristics of the vidicon preclude such an application. A pseudo-random type dot scan has been dealt with quite 10 extensively, out the complexity of the synchronization signal requires a rather elaborate decoding process. Here the variable interlace scheme enjoys quite an advantage, though it does not approach the degree of randomness associated with the pseudo-random dot scan. The required synchronization signals add little complexity to the commercial television scheme since the variable scan is a simple extension of the standard two-to-one interlaced scan. The usefulness of the variable interlaced scheme depended to some extent on the nature of the picture being displayed, as well as on the scan rate. The display of regular geometrical figures resulted in objectionable moire patterns due to the regular, repetitive nature of the interlaced sequence. Picture detail in the vertical direction became obscured as some horizontal lines faded while others were being scanned. Combined with the variable persistence feature, however, a high quality picture was obtained at compressions up to eight-to-one with no deterioration due to moire patterns or loss of detail. It was concluded, then, that both the alternating up and down raster interlace b7 and the variable persistence feature were essential in obtaining high quality pictures at reduced scan rates. At the time of this writing, a rather different approach to the variable persistence display is being pursued which will adequately present a multishade television picture. This idea was set forth by Sylvania, Division of General Telephone and Electronics Company, and is an extension of the single gun color tube principle. In this scheme, two different phosphor layers are deposited on the CRT screen, and a variation in the accelerating potential alters the degree of penetration of the electron beam in the phosphor coatings. This provides a continuous color variation from one color phosphor to the other. In the VISTA system application, the phosphor coatings would be of the same color, but possess different decay characteristics. Thus the persistence could be controlled continuously between the two extremes. The detection of information changes between successive television frames and the automatic selection of a bandwidth compression ratio was a successful venture, though slightly more complex than was initially anticipated. An unexpected problem was the large measure of cross talk between channels of the video disc recorder during the READ and WRITE operations. The cross talk was so excessive, in fact, that the sequential video switching scheme between channels had to be altered. In the final design, the READ and WRITE operations were sequenced such that the two operations were never performed simultaneously. This entailed the use of an additional frame period at the end of each fast scan mode. This also permitted the use of two tracks instead of three with a vacant track between the two to further isolate the channels during the erase operation. These measures successfully inhibited any noticeable cross talk. 58 Due to the nature of the magnetic disc and the desirability of storing large shaded picture areas instead of fine details alone, the disc was used in a "chopped" video mode. The result was an amplitude modulated 3MHz signal which had to be demodulated before a comparison beteen video signals could be performed. A demodulator was thus designed and built for this purpose which consisted of a full-wave rectifier and peak- follower circuit , An accurate control of the video signal levels and amplitudes was required for a valid comparison between successive frames. Any unbalance was integrated by the compare and hold circuit and resulted in a false indication of the information change. Many signal level restoration and gain control circuits were thus required, particularly since the WRITE signal amplitudes required by the disc differed for the two channels, and the video signals read from the disc also differed. In general, the system functioned rather well in the automatic mode. Any object waved in front of the camera, for example, resulted in the selection of a larger bandwidth for the video signal, the exact selection depending on the motion and size of the object. 59 APPENDIX 6o A.l Disc Video Control The mechanism by which the video signal emerging from the camera is simultaneously stored and compared with the previously stored frame is the disc video control. The video signals must be shunted between three tracks of the disc recorder in order to read, write and erase simultaneously during each fast scan mode of operation. A newly erased track accomodates the video signal during the next fast scan, at which time a different track will be erased. Similarly, a different track will be written upon. Figure A. 1.1 illustrates the disc video control which comprises a modulo 3 counter, a decode section and a series of video gates. The modulo 3 counter changes state each time the system reverts to the fast scan mode. During this time the decode AND elements are activated. Depending on the particular count, one each of the erase, read and write gates will open to permit passage of the corresponding signals. Each time the count changes, a different set of three gates will open, thus permuting the read, write and erase signals among the three tracks. A. 2 Analog-to-Digital Converter The analog delta density signal is converted to a digital signal for use by the horizontal ramp generator and other subsystems. A six-bit conversion is required for selection of one out of 6k possible scan rates. Figure A. 2.1 is a simplified block diagram of the A/D converter. A trigger signal activates the clock for six complete periods which enables the ring counter to complete one full cycle. As each D-type flip-flop is strobed, beginning with the most significant bit, the resulting analog voltage at the output of the D/A converter is compared with the analog delta density input signal. If the analog input exceeds the D/A output, the D-type flip-flop remains in the "1" state. If the D/A output exceeds the analog input a logical 61 O W •H O H t-H bO •H 62 4 CD "J L A 1 A V N OJ fM i/> N 63 "1" appears at the output of the differential comparator which clears the Q, output of the D-type flip-flop to a logical "0". A similar comparison is made for each of the less significant bits, in descending order, during each successive clock period. When all six D-type flip-flops have been so addressed, the digital conversion appears on the output lines shown in Figure A. 2.1. A. 3 Partitioning the Scan Rates into n Ranges Having calculated the desired gain for each of the 6h possible scan rates, an algorithm was developed for partitioning these scan rates into n ranges, n < 64, to economize and simplify the circuitry. Obviously this results in some error, since only one scan rate in each of the n ranges will correspond exactly to the gain value chosen for all the scan rates within that range. Clearly, the greater n is, the closer the scan rates 1/xT will come to their corresponding gains G(x) (recall that T is the fastest frame time, x is the factor by which the fastest scanning speed is reduced), and the smaller will be the error. For the purposes of the derivation, Figure A. 3*1 presents a hypo- thetical gain versus scan rate factor curve illustrating the partitioning procedure. For any x ., j = 1, 2, . . . , n, the band corresponding to the J coordinate distance 2Ax . will be J G(x. + Ax.) - G(x. -Ax.) * 2 |£ | ox . J 3 3 J ox ' j X = X . Summing these increments, G -G. =2^1 Ax. +2^|Ax+...+2^|Ax =2 Z^Iax. max nun 5x ' 1 5x ' 2 5x ' n 5x ' j X=X n X=X X=X ' x=x 12 n J 6h G max SG slope = — * 5x x=x • • Ax, Ax 1 Ax 2 Ax, Ax Ax Ax Ax x. Scan rate factor x x n Figure A. 3.1 Partitioning Procedure Applied to a Hypothetical Gain Curve 65 A convenient criterion to apply to the partitioning procedure is that the maximum per cent error in gain not exceed some value 5 for every one of the n ranges . Thus , 1 5G | A 1 SG i A 1 6G n 5 ~ G(x n ) 5x I ^1 - GlxTT 6x I ^2 " gJTJ 6x ' **■ X ^X -, X = X ,>. X = X 12 n Therefore, n G - G . = 2S.E,G(x.) max mm 3=1.' J A useful recursion relation is the following: G(x.) + 5G(x.) = G(x. _) - SG(x. ) J J J+l j+1 Therefore, G ^)=Hf G ^ + i) 1 - Pi Letting 7 = —, G(x.) = ?G(x. 1 ). An additional relationship now follows G(x.. k ) . A(x.) Substituting this expression, the series becomes 66 G - G . = 26 (G n + G + . . . + G ) max mm 12 n = 2&(7 n_1 G + 7 n " 2 G + . . . + yG + G ) v/ n / n ' n n __ / n-1 n-2 . v_ = 25(7 +7 + . . . + j + 1)G where G 1 = G(x ), G = G(x ), . . ., G = G(x ). The sum of this geometric series is -1 ^ -1 n / n-1 n-2 _ . 1 - 7 (7 +7 +...+7+1)= 1 _ I Therefore, 1 - y n G - G . =25 i 2U G max: man 1 - 7 n G may be expressed in terms of G : n * max G max n 1+5 Substituting this expression, mm 25 1-7 G "1+51-7 max ' 25 Since - — = 1-7, the expression above reduces to G . n mm A 1 ' G = ° max 67 G . is unity, so the final expression is nun J ' ^ 7 G - w max Thus it suffices to choose the desired number of ranges n from whence follows the per cent error 5 and the range boundaries . The following example will illustrate the usefulness of these results. Consider the linear gain function G = x. Ten ranges represent a reasonable yet not extravagant number of circuits. For n = 10 and G = 6k, „ max 7 = 0.66. Therefore, 5 = 25.k per cent maximum error. G-. = ■= = 51 • 0. J_0 J. + o G = 7G, n = 33*7? etc. Using the G(x) relationship, the scan rates within each range may now be calculated. A.k Steady-State Error Correction Applied to Horizontal Ramp Generator An inherent property of the horizontal ramp generator, as pointed out in section 3.2, was its ability to correct for sudden or spurious dis- placements of the ramp voltage. A second type of error encountered is a steady state voltage error in the ramp amplitude. This may arise due to temperature effects on the constant current charging or discharging sources, as variations in the capacitance values, leakages or any other such slowly changing mal- function. A rather attractive scheme was developed for eliminating any voltage error within several ramp periods . Before departing on another analytic adventure, however, it should be pointed out that this error correction scheme was never actually implemented. Although a small error was detected, it was not sufficiently large to necessitate the use of an elaborate correction circuit, Nevertheless, the successful theoretical aspects of the scheme warranted a 68 description of the circuit and its effect upon the ramp voltage in the event of a hypothetical error. The outstanding characteristics of the scheme is its ability to cope with and correct an error almost instantaneously. The basic idea is the detection of the voltage error by comparison of the peak ramp voltage with a reference, and the subsequent generation of a charging error current 5I n . 5I n is a constant current which is added to the ramp charging current. At the conclusion of every horizontal line the ramp voltage is sampled in this manner, and the error currents generated are (8l n )-,, (SI n ) p j • • . , at the end of the first period, second period, and so on respectively. (&I~) is proportional to the voltage error at the end of the J jth ramp period. Finally, as the error converges to zero the total charging current converges to I + (&I ) + (si n ) p + . . . = I + 8I n . Figure A. 4.1 shows the bare essentials for implementing this correction scheme. The new error current for each cycle is derived during the previous retrace period and generated commencing with the beginning of the positive going ramp. The analog adder shown in Figure A. 4.1 insures that the new error current adds to the sum of all the previous error currents. The sample and hold circuit shown in the upper part of the figure samples and stores this sum for addition at the end of the next cycle. The nomenclature of Figure 3-2.3 will again be used in this analysis, where SV is the error at the end of the first ramp period, 5V the error at the end of the second ramp period, and so on. One complete ramp period here shall be defined as the elapsed time between voltage peaks of the ramp. Recalling that I is the charging current source and I ' the discharging current source, the retrace time is 69 I WV — • Wv 1 TO c(v - &v . ) n-1 I ' - I - (F5V, + F5V + . . . + FSV _) 12 n-2 where F5V . is the error current generated during the Jth cycle and held J thereafter. F is the feedback factor expressed in ampere s /volt . The duration of the positive going portion of the ramp is given by C(V - 5V ) n 7 I + (F6V n + F&V + . . . + F6V , ) 1 2 n-1 7 The following relation is thus obtained: C(V - 5V _) n-1 T = I Q ' - I Q - (F6V 1 + F6V 2 + . . . + F5V n _ 2 ) + C(V - 5V ) n I + (F6V n + FSV + . . . + FSV "J 1 2 n-1 7 With no feedback applied it follows that C(V - BV 1 ) C(V - 5V X ) x o I o' - J o which simplifies to v i - — VC I I ' = T Substituting this in the equation above and rearranging leads to the recursion relationship 71 5 v n [I + T^ ( — + ~v~ + * ' • + ~T~ )](1 ■ ~T~ ) ~^ = V, FV, BV 1 + 8 V 5V n-2 , "i^~ ' 2 " i^HT + ~v~ f * * * + ~T~ } 5V 1 - -^ w 5V n & V o & v i V r-. FV, 1,2 ■ n-L-i . x o I ' x Assuming some initial per cent error, this relationship may be employed to determine the amount of feedback required for the most rapid convergence to zero of the error voltage. The plots sho-wn in Figure A.U.2 indicate that the most rapid convergence occurs for an initial feedback current F5V which is ten per cent of the ramp charging current, assuming an initial voltage error of ten per cent. 72 7 -■ 6 -- Initial error - 10% ■ 9% of charging current fedback 11$ of charging current fedback Number of cycles after application of feedback Figure A. k. 2 Per cent Error Versus Number of Cycles Illustrating Convergence of Ramp Voltage Error 73 LIST OF REFERENCES 1. Seyler, A. J. "Visual Communication and Psycho-Physics of Vision", Proceedings IRE , Australia, Vol 23, May 1962, pp. 291-30*+. 2. Gouriet, G. G. "Bandwidth Compression of a Television Signal", Proceedings IEE (British), Vol. 10U, Part B, May 1957, PP- 265-272. 3. Nishikawa, S., Massa, R. J., Mott-Smith, J. C. "Area Properties of Television Pictures, IEEE Transactions on Information Theory , Vol. IT-11, July 1965, PP. 3^8-353. k. Gouriet, G. G. "Bandwidth Reduction in Relation to Television", British Communication and Electronics , Vol. 3, 1956, pp. 424-4-29. 5. Winkler, M. R. "Saving Television Bandwidth", Wireless World , Vol 59, April 1953, PP. 158-162. 6. Cherry, E. C, Gouriet, G. G. "Some Possibilities for the Compression of Television Signals by Recoding", Proceedings IEE (British), Vol. 100, Part 3, 1953, PP. 9-18. 7. Cherry, E. C. "The Bandwidth Compression of Television and Facsimile", Television Society Journal , Vol. 10, February 1962, pp. 40-49- 8. Cherry, E., et al. "An Experimental Study of the Possible Bandwidth Compression of Visual Image Signals," Proceedings of IEEE (British), Vol. 51, November, 1963, PP- 1507-1517- 9. Geddes, W. K. E., Newell, G. F. "Tests of Three Systems of Bandwidth Compression of TV Signals", Proceedings IEE (British), Vol. 109, Part 5, July 1962, pp. 311-32*1. 10. Deutsch, S. "Narrow Band TV Uses Pseudo -Random Scan", Electronics , April 27, 1962, pp. ^9-51. 11. Winkler, M. R. "Saving Television Bandwidth", Wireless World , Vol. 59, April 1953, PP. 158-162. 12. Bell, D. A., Howson, E. A. "Reduction of Television Bandwidth by Frequency Interlace", Journal IRE (British), Vol. 20, February I960, pp. 127-136. 13. Johnson, R. E. "Vidicon Performance Characteristics at Slow Scan Rates", RCA Review , Vol. XXVII, No. 1, March I966, pg. 66. 14. Shelton, C. T. and Stewart, H. W. "Pickup Tube Performance With Slow Scanning Rates", Journal of the Society of Motion Picture and Television Engineers , Vol. 67, No. 7, July 1958, pg. 448. 7^ 15. Weimer, P. K. "Television Camera Tubes: A Research Review", Advances in Electronics and Electron Physics, Vol XIII , Academic Press, New York, I960, pp. 400-^06. 16. Fink, D. G. Television Engineering , McGraw-Hill Book Company, Inc., New York, 1952, pg. 15k. 75 VITA David Charles Rollenhagen was born in New Jersey on March 27, 19^2. He attended Brown University where he received his Sc.B. degree in June, 196^, graduating cum laude with high honors in electrical engineering. He received his Sc.M. degree at Brown in June, 1965 • He has worked as a research assistant in the Department of Computer Science since September, I965. He is a member of Tau Beta Pi and an associate member of Sigma Xi. ico* "ffT *» ««• Form AEC -427 (6/68) AECM 3201 U.S. ATOMIC ENERGY COMMISSION UNIVERSITY-TYPE CONTRACTOR'S RECOMMENDATION FOR DISPOSITION OF SCIENTt/IC AND TECHNICAL DOCUMENT i S*e Inttrvctfom on rt< Slthi 1. AEC REPORT NO. coo- 1469- 01^+7 2. TITLE The Vista System for Compression of Television Signal Bandwidths 3. TYPE OF DOCUMENT (Check one): Pm a. Scientific and technical report l~l b. Conference paper not to be published in a journal: Title of conference Date of conference Exact location of conference. Sponsoring organization □ c. Other (Specify) 4. RECOMMENDED ANNOUNCEMENT AND DISTRIBUTION (Check one): Km a. AEC's normal announcement and distribution procedures may be followed. I I b. Make available only within AEC and to AEC contractors and other US. Government agencies and their contractors. I I c. Make no announcement or distribution. 5. REASON FOR RECOMMENDED RESTRICTIONS: 6. SUBMITTED BY: NAME AND POSITION (Please print or type) David Charles Rollenhagen Research Assistant Organization Department of Computer Science University of Illinois Urbana. Illinois 6l8oi Signature Date September 22, I969 FOR AEC USE ONLY AEC CONTRACT ADMINISTRATOR'S COMMENTS, IF ANY, ON ABOVE ANNOUNCEMENT AND DISTRIBUTION RECOMMENDATION: 8. PATENT CLEARANCE: □ a. AEC patent clearance has been granted by responsible AEC patent group. I I b. Report has been sent to responsible AEC patent group for clearance. I~l c. Patent clearance not required. :\