IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 8, AUGUST 1986 1107

Multiple-Beam Cathode Ray Tube Design Overview VERNON D. BECK AND BRUCE P. PIGGIN

Abstract-A CRT that simultaneously writes 16 adjacent raster lines is described. The electron beams that write the lines are produced from very small electron guns fabricated on a single sapphire chip using thin- film techniques. Subsequently a single lens is used to focus the 16 beams on the screen, and a conventional magnetic deflection yoke is used to deflect the beams to form the raster. Because the displayed data is han- dled in parallel, higher content flicker-free images can be displayed. The dynamic corrections required to achieve display contents above 10 Mpixel are discussed.

T I. INTRODUCTION

HIS PAPER describes the electron optics and dy- namic corrections for a high-resolution multiple-beam

cathode ray tube (MBCRT). A companion paper de- scribes the technology used to make thin film sapphire chips that comprise the electron guns used in this MBCRT ;: 11. The chip technology makes it possible to make small, closely spaced, and closely matched electron guns. The w e of such electron guns permits better MBCRT electron Ioptics.

The introduction of parallel electron beams into the ZRT can permit higher levels of performance. This can 4:ome about in two ways: alleviation of the requirements ]:laced on the electronic drive circuitry or avoiding elec- 1:‘on optical limits to the luminance of the CRT.

A CRT raster display must repetitively display an im- tzge frame at a minimum frequency so that the user of the l:lisplay will not perceive flicker. This requirement can be uanslated into a minimum horizontal deflection frequency ij:ld a minimum video bit rate for a given raster format. The introduction of parallel electron beams allows the hc~rizontal deflection frequency to be reduced by a factor tl I at is the number of parallel beams used without reducing the frame rate. Although the introduction of parallel elec- tion beams does not reduce the video bit rate, it does make ilt easier to achieve high bit rates because the video data c m be handled in parallel and does not need to be fully scrialized.

The luminance of a raster-mode CRT display is pro- pl:rtional to the total current available to write the image 011 the screen. The maximum current in an electron beam 01 a given size is limited by space charge and cathode loitding, which is the maximum current density that can

Manuscript received October 23, 1985; revised May 2, 1986. ’/. D. Beck is with the IBM Thomas J . Watson Research Center, York-

t0ri.n Heights, NY 10598. 13. P. Piggin was with the IBM Thomas J . Watson Research Center,

‘Ycrktown Heights, NY 10598. He is now with Birchlake Consultants, 1V11lth Common, Sherfield English, Romsey, Hampshire, United Kingdom, !3O,i OJT.

lEEE Log Number 8609541.

be drawn from the cathode [2]. A larger total electron- beam current can be used if the current is divided among several electron beams.

This situation has been recognized for many years and several different designs have been proposed for MBCRT’s [3], 141. One key feature of the MBCRT de- sign set forth in this paper is the integrated thermionic chip [ 11, [SI that is used to produce the individual electron beams. The electron optics of multiple-beam CRT’s are more complex than those used in single-beam CRT’s. A larger set of dynamic corrections must be used to ensure that the electron beams remain in focus and maintain the same spatial relationship with respect to one another when they are deflected.

Although the benefits of multiple beams could be used in many ways, the work reported in this paper is directed at application of multiple beams in CRT’s for high-reso- lution displays.

11. ELECTRON OPTICAL DESIGN A . Overview

The overall design of the MBCRT is shown in Fig. 1. Individual cathodes create the electron beams that are all accelerated to approximately 6 kV in a common field. A nonuniform accelerating field was designed that causes the central ray of each of the electron beams to pass through a common point in the center of the main focus lens. This situation reduces off-axis aberrations and prevents size changes in the pattern of beams on the screen when the strength of the focus lens is changed.

A bi-potential lens focuses the electron beams on the screen and raises their energy to that of the anode, 16 kV. The focus lens and the accelerator are designed to use the maximum diameter possible in the neck to minimize ab- errations.

A stator-type cosine wound magnetic dleflection yoke is used to deflect the electron beams. As the electron beams are deflected, the yoke introduces aberrations that defocus the individual beams and alter the relative positions of the beams with respect to one another. The former effect is common to all CRT’s while the latter is unique to MBCRT’s. In order to evaluate the performance limits of the MBCRT, it is necessary to understand the electron optical aberrations of the MBCRT and how they can be minimized or dynamically corrected.

B. Chip/Gun Design The electron guns used in the MBCRT described in this

paper are built using the integrated thermionic technology

0018-9383/86/0800-1107$01 .OO 0 1986 IEEE

1108 IEEE 'TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 8, AUGUST 1986

Multiple Beam Cathode Ray Tube

SHEAR) LENS\

ROTATION' STIGMATOR magnetostatic \ I

Fig. 1. Schematic layout of MBCRT showing 2 of 16 beams and element!; used for dynamic corrections. The beams originate from microcathooes built on a sapphire chip. A short nonuniform field common to all the beams is used to accelerate the beams produced by the chip to about I5 kV. The beams then pass through a common point in the center of a si- potential focus lens, diverge, and strike the phosphor screen producing an array of spots. A deflection yoke is used to scan the array over [he screen in a raster producing swathes of adjacent scan lines. Two pairs of orthogonal quadrapoles are used to correct astigmatism and array shear. A solenoid is used to adjust the rotation of the array of beams. The exact location of the beam intersection point, and hence the size of the array on the screen, is controlled by an electrode in the accelerator.

developed at Los Alamos Scientific Laboratory [ O ] . Briefly, the chip is built on a sapphire substrate that has a thin film of refractory metal deposited on both surfaces [7]. On the back surface of the chip the metal is etched into a serpentine resistive heater which is used to heat the chip to the operating temperature of the conventional ox- ide cathode, about 750°C. On the front surface, moats are etched through the thin film of metal to electrically isolate individual cathodes from a common first grid. The fr1:1nt surface of the chip is shown in Fig. 2 . Individual cathode pads are connected to bond pads on the perimeter of the chip, by thin exposed conductors which penetrate the first grid. The conventional triple carbonate oxide cathode rna- terial is deposited on the cathode pads using photolith'o- graphic techniques. The total cathode area used in .o:his MBCRT is comparable to the area of the GI aperture used in a conventional CRT with similar screen size but with a much larger spot size that allows the display of c!mly about 0.5 Mpel. Because the cathode area is the same in the two cases, both CRT's can obtain equal beam currents at equal cathode loadings and therefore produce images of equal luminance.

The completed chip is wire bonded in the center of a square hole in a ceramic ring. The 21 bond wires of 75- pm (0.003-in) diameter wire provide both the electlrical connection to the chip and the mechanical support fo:. the chip. Approximately 3 W of power must be dissipated. in the heater in order to maintain the chip at operating tem- perature. About half this power is lost by radiation from the chip and half is lost by conduction down the bond wires. Leads on the ceramic ring run from the edge of the square hole to the outside perimeter where they are bonded to the stem pins which are the external electrical corznec- tions of the CRT.

Fig. 2. Top surface of the integrated thermionic chip. The chip is built on a sapphire substrate, the top of which is entirely metalized except in moats which separate the first grid from the cathodes and leads and are shown in black in the figure. The 4 X 4 array of squares in the center of the chip are the cathodes which have a pitch of ,010 in. They are pho- tolithographically coated with the carbonate mixture conventionally used to make cathodes. Bond pads are on the edges of the chip. A resistive heater is on the backside of the chip. The square array is canted at an angle of 14" when scanned so that 16 equally spaced lines will be traced out on the screen.

One obvious feature of the chip is the arrangement of the cathodes in a 4 X 4 square array. The array is tilted at an angle of 14" when it is scanned so that 16 equally spaced lines are traced out on the screen. The top of Fig. 5 (given later) shows how the array produces parallel scan lines. By arranging the cathodes into a solid array of roughly equal width and height, the separation between cathodes is maximized without having the cathodes far from the axis of the system. The large separation allows more space to build individual guns, allows large cathode areas to permit large beam currents, and minimizes the interaction between guns and between beams. This ar- rangement requires that the video data for each beam be appropriately timed to compensate for the fact that all the beams are not in a single vertical column. The video data for different beams can be deskewed with shift registers of different lengths. This occurs for the same reason that the 16 beams trace out 16 lines equally spaced in the ver- tical direction; each of the beams falls on a different hor- izontal pixel cycle. Because both the width and height of the array of beams are comparable, one should use appro- priate electron optical techniques. The design of the color in-line self converging yoke is an example of a design that would be inappropriate for this MBCRT because it relys on the array height being much less than the width.

Although the mechanical structure of the chip is simple, its electron optics is not simple. The traditional triode electron gun used in the CRT [2] employs a planar first

BECK AND PIGGIN: MULTIPLE-BEAM CATHODE RAY TUBE 1109

grid with a circular aperture which is spaced above the plane of the cathode. Reducing the space between the cathode and first grid increases the cutoff voltage. Despite the fact that this space has been reduced to zero in this MBCRT gun design by making G, and the cathode co- planar, the cutoff voltage of the individual guns is only about 8 V because of the small size. A second grid is also used in the traditional triode. In the case of the MBCRT gun, the 6-kV accelerator anode is used as the second grid. It is located far from the chip and has a large aperture which is shared by all 16 beams. This aperture does not intercept any of the beam and does not require accurate alignment. The elimination of a series of accurately aligned apertures for each beam is an important aspect of this MBCRT design.

This MBCRT design uses cathode drive and a common first grid, which eliminates the necessity for a second grid spaced close to the first grid. If first grids were driven instead of the smaller area cathodes, much 1arger.pertur- bations would occur in the field above the chip, especially above adjacent cathodes. These perturbations could be shielded by placing a common second grid slightly above the first grid, but this second grid would require apertures which were accurately aligned with those in the first grid. Thus, the common first grid electrostatically isolates the individual guns.

The cathode leads also perturb the field above the chip and have been made as narrow as practical to minimize this effect. To minimize the effect of a lead on the cathode it controls, square cathodes are used with the lead at- tached to the corner and exiting radially. The reasoning behind this design choice is described below. The ideal jesign would have cylindrical symmetry. Given that one were to have the cathode lead pass through GI , one would want to compensate for the asymmetry introduced by the ead. The asymmetry can be cancelled by adding a short ,second lead or stub diametrically opposite the actual lead. 'When the stub is added, the grid aperture develops an el- , lpticity, which introduces astigmatism. Adding another pair of stubs at right angles can be used to cancel this dlipticity and its astigmatism. In this way a cross is gen- crated, the center of which can be approximated with a square. The corners of the square act as the stubs.

Three-dimensional modeling of the electron optics ,above the chip has been done to evaluate possible inter- rctions between beams or guns. The modeling predicted Ithat the interaction between gun drives would cause dis- placements of the spots that are much smaller than the #;pot size. The fields in the vicinity of a gun were modeled : ~ y dividing the grid and cathode regions of the chips into 1:ectangular patches. The charge on each of the patches ,vas then calculated using an integral equation technique. IPields from different patches were then superposed to ap- ,:'roximate the field above the chip. Electrons were then I raced through this field using the same technique de- !;xibed later in the section on the deflection yoke. No in- I 1:ractions have been experimentally detected in attempts that were made to observe them.

. .

The actual size of the crossover produced in the gun is well predicted by gun models, and space charge does not significantly affect the individual beams. Because the in- dividual beams pass through one another in the focus lens, the space-charge force between beams has opposite signs on the two sides of the lens. This situation cancels most of the net deflection at the screen caused by the space- charge repulsion between beams.

Since the geometry of the grid-cathlode region is deter- mined lithographically and not subject to substantial ther- mally induced variations, the cutoff voltage is very stable and consistent among all the guns on a chip. Residual variations in the cutoff voltage of the guns on a chip are due to variations in the thickness of the oxide coating on each cathode pad. Because the eye is very sensitive to periodic variations in intensity, one should provide a means to trim the drive to each cathode so that all beam currents are equal.

C. Accelerator Customarily, what is referred to as the accelerator in

this paper is the GI - G2 region of a conventional CRT triode gun. The distinction is made in this case because the gun uses an unusually high G2 potential and because the accelerator field is used to cause the individual beams to pass through one another in the center of the focus lens. The primary objective of the accelerator design was min- imizing the length of the CRT, so the functions of increas- ing the energy of the beams and focusing the central rays of the individual beams to a common point were com- bined in the accelerator.

The design of the accelerator is closely related to the design of the so called Butler gun used in the scanning transmission electron microscope [8]. The simplest ac- celerating field is a uniform electric field. A uniform field can be established easily between two flat parallel con- ducting plates. Unfortunately, there are no materials that are totally transparent to electrons, so it is necessary to put a hole into each plate to allow the electron beam to enter and leave the accelerating field. The hole in the plate alters the field and large gradients arise near the edge of the hole and introduce large electron optical aberrations. The problem introduced by the holes can be reduced by choosing a field that goes to zero at the beginning and end of the accelerator. A cylindrically symmetric field can be constructed in which the electric field on axis is of the form z 2 - a2. This field goes to zero at z = f a . The equipotential surfaces for this field can be constructed by first integrating the field on axis to obtain the potential on axis and then using the following kernel which gives the potential at all points from the potential. on axis.

. r27r

The equipotential surfaces are shown in Fig. 3 . At the points where the field on axis vanishes, the equipotential surfaces become cones with half angles of 54"44'. If small

1110 IEEE 7!T.ANSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 8, AUGUST 1986

equipotentials

entrance L exit entrance exit

Fig. 3 . Equipotential surfaces for a cylindrically symmetric field which has the form z 2 - a' on axis. The region of this field between the two heavy equipotentials is especially useful as a lens because the field goes to zero at the entrance and exit apertures. Holes placed in the electrodes at these points cause very little perturbation to the field.

holes are put into conductors that conform to the entrance and exit equipotential surfaces, the field in the accelera1;- ing space will not be significantly altered, because it will not bulge out the entrance and exit holes. The reason th:ts field introduces such little aberration is because the field and its derivatives along the axis remain small.

In order to extend this design to the accelerator require,d for the MBCRT, it is necessary to make some assump- tions about the field. First, assume that the field on the axis is represented as a polynomial. This allows the in-. tegration (1) to be carried out easily. In order to simplify the manufacture of the sources, next assume that all the sources lie in a plane perpendicular to the axis at z = 0 ;at zero potential. If the field is symmetric about z = 0, thtm this condition will be satisfied. The polynomial expres- sion for the field can easily be made symmetric about z =: 0 by excluding all odd power terms. The next condition is that the electric field on axis vanish at the exit aperture. The last condition is that electrons that leave the sourjce plane at different off-axis distances pass through a corn- mon point on the axis.

The simplest polynomial that satisfies the first three conditions is z 2 - u2, which has the sources at z = 0 and exit at z = a. Unfortunately, this field causes electrons leaving the accelerator to diverge. The next simplest field is of fourth order in z . In this case the field has a fre'e parameter, namely the ratio of the field at the source plane to the maximum field within the accelerator. Adjustmmt of this parameter allows the fourth condition to be m:t. Fig. 4 shows the equipotentials and electron trajectories for such an accelerating field. The accelerator field irk- tially increases in magnitude, causing the beams to be '&-

retted toward the axis. Near the exit of the accelerator, the field decreases to zero and causes the beams to 'ldi- verge, but the initial convergence dominates. The effect of the converging and diverging is to cause the beams to appear in the exit aperture as though they came from a point 30 percent closer to the axis than the actual source point.

The equipotential surfaces which are the entrance and exit electrodes extend to infinite radius and cannot be ac- commodated in any physically realizable CRT. It is pos- sible to construct electrodes which generate a field on a.xis

cathode I exit

trajectories electron

cathode

Fig. 4. Equipotential surfaces and electron trajectories for an accelerator field that is a fourth-order polynomial. On the right the axial field and potential are shown. The two heavy equipotentials serve as the elec- trodes. The electrode at zero potential has an unusual shape which is flat in the center. The point at which the trajectories intersect is dependent on the ratio of the axial field at the source plane to the maximum axial field within the accelerator.

which closely approximates the desired polynomial. Fi- nite-element modeling was used to compute the fields pro- duced by physically realizable electrodes. Several design iterations were made to obtain suitable accelerator elec- trodes.

The first accelerator design used electrode shapes that closely matched the equipotential surfaces generated from the polynomial field. It was found to be necessary to have an electrical adjustment of the beam intersection point and this was achieved in the second iteration by splitting the electrode at the source potential into an inner and outer part. Variation of the potential on the outer part allowed the balance between the converging and diverging sec- tions to be altered to shift the beam intersection point. A small kink can be introduced in the field where the outer part meets the inner part containing the chip. This kink does not substantially increase the aberrations of the ac- celerator. The size of the bundle of beams on the screen can be trimmed by about 10 percent by varying the poten- tial on the outer part by about 100 V.

This design had another deficiency. It allowed line of sight between the beam and the glass rails and neck. These insulators could and did accumulate surface charge which altered the accelerator field and degraded the optics. This problem was solved in subsequent designs by shielding insulators from the electron optically active part of the accelerator with conductors at controlled potentials. Elec- tron optical models of the accelerator predict that curva- ture of field will contribute about 14 pm to the spot size when the focus is optimized so that spots in the array nearest and farthest from the axis are equally defocused by curvature of field.

In general, long fields with low gradients have rela- tively smaller aberration. Fields of this type are most eas- ily created with large diameter electrodes and the accel- erator design uses the full diameter of the 36-mm neck. An even longer accelerator could have been made if a nonlinear graded resistive film were used to establish the potential on the the outer radial boundary of the acceler- ator. This was not done because of the difficulty of mak- ing the grading resistive film.

'BEi1::K AND PIGGIN: MULTIPLE-BEAM CATHODE RAY TUBE 1111

D. Focus Lens Initially, magnetic focus was used in order to avoid

components within the CRT envelope and minimize ab- er..ations from the focus lens. It was possible to make dy- na~nic focus corrections on a 2-Mpixel raster at a 60-Hz ffrtllme rate and a 4-kHz horizontal rate. It was necessary ta :ompensate for the rotation introduced by the magnetic !k!s by including a second lens coil, which generated a filrLId of the opposite polarity. When the raster size was irlt8;reased to 9.3 Mpixel, the drive power requirements for m~~.gnetic dynamic focus and rotation compensation be- C ; ~ I I K . excessive and a change was made to electrostatic focus. Since the magnification between the crossover and thf: screen is rather low, the voltage swing required for d'ynamic focus is larger than usual. Dynamic focus drive rquirements were computed for different lens designs, a . d a bi-potential lens was chosen because of its drive requirement.

The focus potential is taken out through the side of the nwk. This is necessary because of the large number of pins in the stem to accommodate the 16 individual cath- oc jes. Nominally the focus voltage is 6 kV and a horizon- tal parabola of 700 V and a vertical parabola of 400 V are used for dynamic focus. These large swings are generated wing transformers. Because the low voltage end of the fi:)zus lens is electrically connected to the accelerator, dy- mmic focus corrections appear in the accelerator. These wriations do not affect the electron trajectories in the ac- cl:lerator, but they do affect the cutoff voltage of the in- d lvidual guns. This can be compensated for by feeding a fj'action of the dynamic focus voltage to the video drive a .nplifiers. An alternate solution to this problem is to sup- ~ t ' l y a fixed dc potential to the accelerator electrode through al second neck feedthrough. If the dc potentials of the ac- c : lerator and focus electrodes are the same, the variable lcnsing action that occurs between the accelerator and fo- c'llIs sections due to dynamic focus is negligible.

A shrinking raster spot size of about 100 wm (0.004 in) ir; obtained in the center of the tube. The principal con- tl~bution to the spot size is the size of the crossover in the y ln . Only a few micrometers of spherical aberration arise L;om the focus lens. A' Dejection Yoke

A 16-slot 90" stator-type deflection yoke with a 40-mm- diameter bore is used in this MBCRT. It is excited with (::)sine distributed windings in the horizontal and vertical directions. Presently third order deflection distortion is corrected using an electromagnetic octopole to cancel pin- cushion and a third order S-correction to cancel nonlin- c:rrity.

The yoke is the dominant source of aberrations when 1 ~ , e array of beams is deflected. An electron optical model (: F the yoke was constructed in order to study the aberra- tions of the yoke. This model traces electron trajectories L y integrating Newton's equation with a relativistic cor- xction. The force on the electron is given by the cross product of the velocity of the electron with the magnetic :lidd. The magnetic field of the yoke is obtained by sum-

ming the fields from about 1000 straight line segments of current. The magnetization of the ferrite used within the yoke is represented by surface and bulk magnetization currents that are also represented as stlraight line seg- ments. The trajectory integration is done with an embed- ded fifth- and sixth-order Runge-Kutta integration routine (Fehlberg form) which monitors the truncation error to automatically adjust the step size to achieve a specified accuracy. Typically an accuracy of lo-' or better is used, so computed electron landing positions on the screen should be accurate to better than 0.1 pm. Aberrations are found by tracing several rays in a bundle and evaluating the differences in landing position and slope at the screen.

The usual deflection aberrations of curvature of field and isotropic astigmatism were found using the model. Curvature of field of the deflection yoke is corrected by changing the strength of the focus lens. Isotropic astig- matism is corrected using two orthogonal magnetic quad- rapoles that are located in the same plane as the focus lens. It is important that the central rays of the individual beams pass through the center of these elements so that dynamic corrections for focus do not introduce changes in the relative positions of the beams on the screen.

The deflection yoke also introduces deflection distor- tion exactly as in the single-beam CRT. The techniques used in the design of single-beam CRT yokes can be used to correct the third-order mixed terms (pincushion) by in- troducing third harmonic components to the yoke winding distribution or by adding a static octopole on the screen side of the yoke.

In the MBCRT an additional type of aberration appears. It is the distortion of the array of beams that occurs when it is deflected. It was found that the dorninant distortion of the array is linear. A general linear distortion of a two- dimensional object, like the array on the screen, can be factored into 4 components: a change of size, a rotation, and two orthogonal components of shear. It is fortunate that electron optical elements exist to correct each of these. The size of the array on the screen can be changed using zoom optics. In practice, since the size change is small and amounts of about 7 percent between the center of the screen and the comer, the size change can be af- fected by moving the beam intersection point slightly ahead or behind the focus lens. The rotation can be cor- rected by using a solenoidal magnetic field. In this case the maximum correction required is about 3" so the fo- cusing action to the solenoid is very weak. The two com- ponents of shear can be corrected by using a pair of or- thogonal quadrapoles located where the central rays of the beams are separated. The shear correctors will introduce astigmatism into the individual beams, but this can be cancelled using the pair of quadrapoles used to dynami- cally correct astigmatism of field.

111. U S E OF THE MBCRT A. Application Constraints

The MBCRT described above has several constraints that affect its use in displays. First, the size of the array

1112 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 8, AUGUST 1986

of beams on the screen is essentially fixed by the spacing of the guns on the chip and the ratio of the distances From the chip to the lens and the lens to the screen. Because these parameters are determined when the tube is mlanu- factured, a given MBCRT is suitable for only one pixel to pixel spacing. Since the screen size is limited, the ras- ter format is fixed.

The MBCRT technology described thus far is a mono- chrome technology and extension to color is difficu1.t. It is probably inappropriate to use conventional shadow mask technology in combination with this MBCRT tech- nology because it will require beams be far from the axis of the lens and yoke. Beam index technology may be use- able, but developing a method to determine the exact beam position on the screen will be difficult. A variant of beam index, which can be referred to as beam servo, is also feasible. In this case the beams are scanned along the phosphor stripes, rather than across them. This situ2:t;ion avoids the higher video bandwidth typically required for beam index and alters the interrelationship between guard bands, phosphor stripes, color purity, video bandwidth, and luminance. The lower horizontal deflection rate of an MBCRT lowers the servo bandwidth and the large ]nul- tipixel vertical space between scans allows a large inlitial lock-in range. Again, the usefulness of such a schene is dependent on a method of determining the exact position of the beams on the screen. Cathode drive causes small shifts in the landing positions of beams deflected into the comer. The motion caused by drive is less than 70 pm in the comer of a 19-V MBCRT.

Field sequential color could be used and MBCRT twh- nology is useful in allowing the higher video bandwidth and beam current that is required. Field sequential color could be done using either a penetration phosphor OF a liquid-crystal color shutter.

B. Drive Electronics The MBCRT technology described above alters the re-

quirements placed on the drive electronics. As a reference in describing the drive requirements, consider a '3.3- Mpixel raster (2640 V X 3520 H) running at a frame rate of 75 Hz. Any type of CRT display with a content this high requires exceptionally high levels of stability.

It would .be difficult to construct a video amplifier that would drive a single-beam CRT at the rate required above. Conventional CRT drive circuitry presently can reach video bandwidths of about 250 MHz and line rates of about 130 kHz. [9] The amplifier would receive a 11c:w pixel approximately once every nanosecond and have to swing 30 to 100 V to turn the beam on or off. In the c'ase of the MBCRT, 16 video amplifiers would be used in par- allel. Each amplifier would receive new pixel data every 18 ns and have to swing only about 8 V to turn the bt:am on or off. It is likely that integrated circuits could supply the required video drive for the MBCRT. The video (lata for this MBCRT must be deskewed, and this can be d m e easily with shift registers of different lengths.

The luminances of each of the individual beams used in an MBCRT must be well matched to avoid stripe:.; in

regions of the screen where all pixels are on. This is a difficult requirement to meet because the human eye is very sensitive to periodic intensity variations. The verti- cal spatial frequency of an individual beam is 5 cycles per degree for a 19-V MBCRT viewed from a distance of 50 cm. The human eye has its maximum sensitivity to sinu- soidal intensity variations at a spatial frequency of 4 cycles per degree, and at this frequency the human eye can detect intensity variations of only 0.2 percent [ 101, Because the carbonate crystals that comprise the cathode have a size which is about 1 pm, variations in the thickness of the cathode coating will be in the range of 1 pm. A 1-pm variation in the thickness of the cathode coating shifts the cutoff voltage of the guns by about 5 percent. Under these circumstances, it is probably necessary to provide indi- vidual gun bias adjustments to avoid stripes in regions of the screen where all pixels are on.

The horizontal deflection circuitry must operate at 13 kHz for the MBCRT whereas it would have to operate at 210 kHz for the single-beam CRT. Obtaining horizontal deflection at 210 kHz is difficult. The video deskewing operation used in the MBCRT requires a constant hori- zontal writing speed for a fixed pixel frequency. Varia- tions in the writing speed that arise from the electron op- tics of the yoke are included in this requirement. Because some of the video data is held for as long as 15 pixel times, the writing speed must be constant to within 1 part in 60, if the video data is to be written at the correct lo- cation on the screen to within 1/4 of the pixel to pixel spacing.

The vertical deflection circuitry has essentially the same requirements for the MBCRT and the single-beam CRT. In the case of an MBCRT, a staircase, rather than a saw- tooth, would be preferred for vertical deflection because of the larger vertical jumps that occur between horizontal scans.

The requirements for generation, drive, and setup of dynamic corrections must be considered. Compared to the single-beam CRT, this MBCRT requires four additional channels of dynamic correction for the .distortion within the array of beams that arises from deflection. The feasi- bility of the dynamic corrections depends on their fre- quency, power, and accuracy requirements. The MBCRT has a significant advantage because the required band- width for the dynamic correction signals is much lower.

There are several methods for generating dynamic cor- rection signals. The method used for the work reported here is based on digital techniques very similar to those used for dynamic convergence in color CRT displays [ 111. In this scheme the screen is divided into a number of zones; for example,' 28 vertical zones by 18 horizontal zones. The value of each dynamic correction in each zone is stored in a memory that is read out during the scan. The output is converted to an analog voltage and used to drive a given correction element. The digital correction gener- ation scheme is desirable because it allows great flexibil- ity in the waveforms possible and avoids interactions be- tween zones during adjustment.

The drive required for each of the dynamic correction

Bli[:K AND PIGGIN: MULTIPLE-BEAM CATHODE RAY TUBE 1113

e1l:ments will be discussed. Dynamic focus is necessary ICI, cancel the curvature of field that arises from the yoke. 'Tlle dynamic voltage swing required for this MBCRT is la rge because of the low magnification used in the focus le -s. Since this correction is electrostatic and no dc cur- re:rit is drawn to the focus electrode, only the 22-pF ca- p;xitance of the focus electrode needs to be driven. The focus voltage needs to be accurate to about f 30 V, and this implies that the dynamic corrections, which are about 11l;lOO V, need to be accurate to about f 3 percent.

Dynamic stigmation is also used in this MBCRT to a(: nieve high resolution. This is done using two orthogo- n;.l magnetic quadrapoles located in the plane of the focus letls. This location is used for the stigmators so that they will not alter the geometry of the array. Approximately 1 VA of drive power needs to be supplied to each of the quadrapoles for magnetic stigmation. The signals needed f(: r dynamic stigmation require accuracy of about k 3 per- ci:nt.

The drive requirements for the array geometry correc- tims are based on measured values of size change, rota- til :)n, and shear introduced by deflection. The accuracy re- q'Jirements for the drive signals are deduced from the rt,;luirement that error between the actual and correct pixel 1;i.Iding point be less than 1/4 of the pixel to pixel spac- ing. This is an arbitrary requirement that seems to be ad- e1:uate to avoid serious degradation of the image quality. Elecause there are four orthogonal errors, the individual e 'Tors should not exceed 1/8 of the pixel to pixel spacing.

The dynamic correction for change of size of the array ic, done electrostatically. The voltage supplied to the ex- tmltN:rnal portion of the low-voltage electrode of the accel- erator typically varies from 100 to 200 V and needs only to drive the capacitance of the electrode. The variation in the size of the array between the center and corner of the s8:reen is about 7 percent. This variation will cause the c:l:)rner pixel of the array to land (0.07) X 8.75 (length of t l ~ e array half diagonal in terms of the pixel to pixel spac- iilg) units from the correct location. If the contribution of tlle change of size is to be less than 0.125 units, then the ~ ~ ~ z e correction must be accurate to 1 in 4.9, which should t j c easily achieved.

The required accuracy for the rotation of the array is .. . 0.4". The maximum rotation introduced by deflection is'; 3 " , and one may add to that rotational misalignments (:-' +2", which may occur during manufacture when the :item is sealed into the neck. This yields an accuracy for 1 ..e dynamic rotation correction of 1 in 12.5, which again il 5 quite modest. The dynamic drive power for the rotation correction solenoid is 1 VA.

When the array is deflected into the corner, array shear uses the corner pixel in the array to land 3 units from >.he correct location. In order that shear not contribute -nore than 0.125 units error, shear must be corrected to 1 .lpart in 24. Shear is corrected using two orthogonal quad- Itapoles located between the accelerator and the focus lens 'where the individual beams are spatially separated. The i'1ear correctors introduce astigmatism into the individual heams that must be cancelled with the quadrapoles used

for stigmation. The drive power requirements for the shear is fairly large: one of the quadrapoles requires 19 VA and the other requires 6 VA.

The elements introduced to carry out the dynamic cor- rections are also useful in eliminating on-axis aberrations that arise from misalignments introduced during assem- bly. Rotational misalignment of the chip has already been discussed. Translation misalignments of the chip in the accelerator or of the electrodes in the accelerator of lens can occur. In lowest order these misalignments may cause the electron optical axis of the system to be bent and this can be canceled with external magnetostatic dipoles. The next higher order effect of these misalignments is to intro- duce astigmatism and shear, which can be cancelled with the quadrapoles.

C. MBCRT Setup Procedure A description of the method currently wed to set up the

MBCRT is helpful in understanding how the tube func- tions and the limitations it has. The first step is to focus the beams statically in the center of the screen. This can be done with the scan turned off if the chip temperature is reduced.

The focus lens and stigmators are adjusted for mini- mum spot size, the rotation solenoid is adjusted so the array is properly canted, and the shear quadrapoles are adjusted so the array is square. The potential of the sec- ond grid is then adjusted so that variations in the strength of the focus lens do not affect the size of the array. Lateral motions of the array arise from lateral misalignment of the focus lens which can be corrected magnetostatically. These adjustments ensure that all the individual beams are centered in the focus lens. This procedure establishes the size of the array at the center of the screen and therefore the pixel to pixel spacing.

The next operation sets up the raster to match the pixel to pixel spacing at the center of the screen. When the ras- ter is applied it may be necessary to trim the rotation so- lenoid so that all 16 lines in a swathe are equally spaced. The vertical ramp (or stair step) amplitude is adjusted so the swathe is moved down exactly 16 line spaces at the end of each horizontal scan in the centler of the screen. The horizontal writing speed must be matched to the pixel clock so that in the center of the screen columns of pixels that are vertically above one another are straight. Pin- cushion and S-correction are next adjusted so the swathe spacing and writing speed are constant over the entire screen. Currently, only the third-order corrections are being done, but arbitrary deflection distortion corrections could be made.

The remainder of the setup procedure is used to ensure that the array is in focus and fits in the raster over the entire screen, The actual adjustment is easier to do if lin- ear combinations of the four array geometry elements are used instead of the actual elements themselves. Fig. 5 shows the affect of the individual elements and a set of linear combinations which affect the x 01- y discontinuities which occur with a vertical periodicity of either 4 or 16 lines. Using this set of linear combinations, a user would

1114 I3EE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-33, NO. 8, AUGUST 1986

No Errors

MBCRT Array Distortions

Rotation Error '?;, -. ---

Fig. 6 . Photographs of various sections of a 6-Mpixel MBCRT screen. These photographs show 3-mm-high sections of the screen taken off a video monitor driven by a camera with a microscope objective. In the

Shear, Error center of the screen the individual spot are well focused and individual .e -- scan lines are clearly visible. Video dwell time, rather than astigmatism,

is the source of the length of the spots. The geometry of the array is ' 3 s correct and well matched to the raster. This is apparent from the absence

Shear, Error :! ------e-..----.

a ' b * =if== * ' = E = * * _ I f _

b 9 -== of extraneous discontinuities in the vertical and diagonal lines. Behind

9 .*- e --- the photograph of the screen center is a photograph of a square array of

e --- i .

0 *

a * 6

e ' a

0 pixels with a side 8 pixels long. If all the focus and geometry corrections are left fixed and the array deflected to the upper left corner of the screen, the spots are strongly defocused as shown in the corner background pho- tograph. The photograph overlaying it shows that the spots can be well focused and the array distortion corrected. Two other photographs show

Magnif icat ion + Shear, Rotation + Shear, different regions at the edge of the screen. A defect in the bottom cathode of the swathe causes the streak in the one of the 16 scan lines. At the

__ bottom two additional photographs are shown at the same scale for ref-

on a conventional 25 X 80 character video display terminal. At the right,

----e-- -- -------*..--.. e.

__ . . * __ erence. At the left, a photograph is shown of 7 X 9 characters displayed

* * -_ some 8-pixel-high MBCRT characters are displayed. * * * * * .___._._ 0 ._.__.

. * that sample the entire screen, correction values for all the remaining zones would be calculated by interpolation.

Rotation - Shear, Magnif icat ion - Shear,

---.-.-e-.---.. _--- ~

Fig. 5 . Linear distortions in the image of the square array as it appears on the MBCRT screen. The dominant distortion in the geometry of the square array introduced by deflection is linear and can be corrected using four electron optical elements. The distortions introduced by these eie- ments are shown in the upper half of the figure. Next to each pictule of a distorted array is a picture of what a vertical line would look like under this distortion. In the lower half of the figure, four distortion patterns, are shown which are linear combinations of the distortions introduced by the individual electron optical elements. These distortions have an e m l y understood effect on the displayed image and comprise a preferred sct of adjustments for correction of array distortion.

focus the beams in a zone using the lens and stigmatcars. Then the geometry of the array would be adjusted so that there were no discontinuities in the x or y directions at every fourth line. Finally x or y discontinuities which oc- cur every 16 lines between swathes would be adjusted out. After this procedure is carried out at a number of zones

D. MBCRT Laboratory Performance Thus far only one chip design has been used to make

MBCRT's in two different sizes. Initially, 70" 15-V CRT's were made with P39 phosphor. These CRT's were magnetically focused and had a geometry that gave an ar- ray size suitable for displaying a 2-Mpixel raster. These tubes had a length of 20 in. Subsequently, 90" 19-V CRT's were made with P45 phosphor. These CRT's were electrostatically focused and had a geometry suitable for displaying a 6-Mpixel raster. These tubes had a length of 19 in. Currently, plans are underway to lengthen these tubes by 1 in to reduce the array size to allow 9.3-Mpixel rasters to be displayed.

Both sizes of MBCRT gave spots that were significantly smaller than the pixel to pixel spacing so that individual raster lines were visible. Fig. 6 shows photographs of sec- tions of the screen of the 19-V MBCRT.

IV. SUMMARY A design for a multiple-beam cathode ray tube has been

presented that uses a thin-film cathode chip to generate 16 electron beams. These beams can be deflected over a CRT

BECK AND PIGGIN: MULTIPLE-BEAM CATHODE RAY TUBE 1115

screen to produce high content images. This MBCRT re- lieves the high bandwidth requirements for the video am- plifier and horizontal deflection circuit. Dynamic correc- tions prevent spot size growth and array distortion when the array of beams is deflected.

REFERENCES [ l ] B. P. Piggin, E. Segredo, J . F. O’Hanlon, J . L . Staples, and A. V.

Brown, “Integrated electron sources for multibeam CRT’s,” IEEE Trans. Electron Devices, vol. ED-33, no. 8, pp. 1116-1122 Aug. 1986.

[2] H. Moss, “Narrow angle electron guns and cathode ray tubes,” supp. 3 to Advances in Electronics and Electron Physics, L. Marton, Ed. New York: Academic, 1968.

[31 D. L. Say, “A multibeam CRT,” Information Display, vol. 11, pp. 29-34, May 1970.

[4] D. S. Hills, “Multi-beam cathode-ray-tube displays,” in SID Dig. Tech. Papers, pp. 74-75, 1972.

[5] S. W. Depp and B. P. Piggin, “Multiple electron beam cathode ray tube,” U.S. Patent 4 361 781, Nov. 30, 1982.

[6] B. McCormick, D. Wilde, S. W. Depp, D. Hamilton, and W. Ker- win, “Development of integrated thermionic circuits for high tem- perature applications,” IEEE Trans. Ind. Electron., vol. IE-29, May 1982. J . Souk and J . F. O’Hanlon, “Characterization of electron-beam de- posited tungsten films on sapphire and silicon,” presented at the Int. Conf. on Metallurgical Coatings, Los Angeles, 1985. J . W. Butler, “Digital computer techniques in electron microscopy,” presented at the 6th Int. Congress for Electron Microscopy, Kyoto, 1966. C. Infante, D. Denham, and B. McKibben, “A 230 MHz bandwidth high-resolution monitor,” in SID Dig. Tech. Papers, p. 124, 1983. F. L. Van Nes and M. A . Bouman, “Spatial modulation transfer in the human eye,” JOSA, vol. 57, p. 401, 1967. J . S. Beeteson, K. T. Jarzebowski, andB. R. Sowter, “Digital system for convergence of three-beam high-resolution color data displays,” IBM J. Res. Develop., vol. 24, pp. 598-61 1, Sept. 1980.

Vernon D. Beck received the Ph.D. degree in physics from the University of Chicago in 1977. His dissertation work dealt with the correction of electron optical defects in the electron micro- scope.

While an undergraduate student at the Univer- sity of Chicago, he received the Telegdi prize for the highest score on the Ph.D. qualifying exami- nation in physics. He was the recipient of a Hertz fellowship at the University of Chicago from 1971 to 1975. After graduation, he joined the research

staff of the IBM Thomas J. Watson Research Center in Yorktown Heights, NY, where he has continued his study of electron optics.

* Bruce P. Piggin was born in London, England, in 1928. He received the B.Sc. degree from Lon- don University in 1953 in mathematics and chem- istry and the Ph.D. degree from the University of Southampton in 1967 for work on “very fast pulse studies of electrochemical systems.”

He initially worked for the United Kingdom Government at the Royal Aircraft Establishment Farnborough on computer design. In 1954, he joined Plessey Ltd. as Chief Chemist of their sub- sidiary Leigh Electronic Developments Ltd.

working on the design, development, and automatiNc manufacture of super- rugged high-performance minature electron tubes. In 1960, he joined IBM U.K. Ltd. working on the early manufacturing method problems of the IBM 360 systems and spent a lot of time in the United States on assign- ments. In 1968, he was appointed IBM U.K. Manufacturing Research Manager where he led a team working on many aspects of computer de- velopment and manufacture from silicon devices, automated E-beam test- ing, computer language development through to chemical processing, and surface chemistry problems. In 1977, he joined the IBM Thomas J . Watson Laboratory at Yorktown Heights, NY, on assignment and did a lot of the early work on the Multibeam CRT. In 1981, he returned to IBM U.K. and took early retirement. He now runs Birchlake, a consultancy specializing in device design, development, and processing, and maintains a particular interest in the problems of information display.