`
`INNOLUX CORP. V. PATENT OF SEMICONDUCTOR ENERGY
`
`LABORATORY CO., LTD.
`
`|PR2013-00066
`
`
`
`SEMINAR M-3:
`
`ACTIVE-MATRIX LCDS
`
`Colin Prince
`
`ChiefEngineer, Displays
`Litton Systems Canada, Ltd., Etobicoke, Ontario, Canada
`
`m
`
`This seminar will describe the principles of matrix-addressing techniques and the evolution
`of AMLCDs which has resulted in the broad availability of personal information displays.
`Technology innovations which are expected to futher broaden the market for AMLCD—
`based products will also be described.
`
`
`
`SOCIETY FOH INFORMATION DISPLAY
`
`ISSN0887-915X/97/0000—M-3-$1.00 + .00 © 1997 SID
`
`
`
`NOTES
`
`
`
`AMLCDS
`
`C. Prince
`
`Litton Systems Canada Limited
`Etobicoke, Ontario
`Canada
`
`Introduction
`
`Over the last 25 years we have witnessed the emergence and evolution of a new class of
`high technology products, which we can generally describe as personal information displays.
`Very shortly the net revenue attributed to the display component of this product class will exceed
`20 billion dollars, yet at the onset of this evolution, the extent of this product was wrist watches
`utilizing Light Emitting Diodes, LEDs. As is the case today, the objective was to provide the
`maximum information content within the constraints of a portable product. To illustrate the
`magnitude of this evolution the LED watch display circa 1970 may have utilized some 28 display
`elements operating in a binary monochrome mode with a data update rate of one minute and a
`power consumption which prohibited continuous viewing for more than a few minutes. In
`comparison today’s portable display screens may provide more than one million full color
`elements with an update rate of 60 Hz, and it can be used continuously for a few hours.
`
`If one is allowed to use the product of these metrics to gauge the technology growth that
`has taken place, it results in the remarkable factor of 1010 over a 25 year span. The contribution
`of Active Matrix LCD to this evolution has been most pronounced over the last decade as
`illustrated in Figure 1 and 2.
`
`LCD Evolution
`
`The single technology which facilitated this evolution was the utilization of the twisted
`nematic, TN, configuration of liquid crystal, LC, materials. Initial exploitations of LC utilized
`the Dynamic Scattering mode of operation but was disadvantaged due to limited contrast,
`temperature dependency, and poor lifetime performance. It subsequently emerged that the TN
`mode would become the foundation for virtually all subsequent portable display products and
`continues to remain so. This is not to say that there have not been major advances and alternative
`configurations which have also emerged during this period of time and this topic will be
`addressed within this seminar.
`
`To begin with, LC technology offers the opportunity to mechanize a light valve and in
`itself is non—emissive.
`It simply and to a sufficient degree by the influence of an electric field is
`able to vary the transmission of light (Figure 3). If it were an ideal light valve it would provide
`
`M—3/3
`
`
`
`100% transmission in the on-state and zero transmission in the off—state and would thus provide
`infinite contrast.
`
`
`Contrast: To”
`TOPF
`
`(1)
`
`The TN mode is based on the properties of polarized light and the ability to rotate the
`plane of polarization as it propagates through the cell. The source of polarized light is created by
`an input polarizer which will have overall through—put of about 42% and since the output
`polarization state has to be analyzed in an output polarizer which will have a maximum
`transmission of about 85%, the Ton level is limited to 36%.
`In a practical display transparent
`electrodes of Indium Tin Oxide, ITO, are used to create the field effect thus reducing the
`transmission by at least 8%. This then limits the overall throughput, TON to about 30%.
`
`The off—state transmission is somewhat more difficult to precisely quantify since first
`order models indicate that for monochromatic light a genuine zero condition can be achieved. In
`practice, this does not occur due to a number of causes such as, non-ideal polarizer operation and
`LCD molecular alignment, fringe fields at the edge of the active area and leakage contributed by
`the presence of spacers within the structure. The consequence is that:
`
`Contrast (Normally Black — Parallel Polarizers) is typically
`
`6.3 _
`— 30
`0.01
`
`whereas Contrast (Normally White - Orthogonal Polarizers) is typically
`
`0.3
`-—— = 100
`0.003
`
`to more than £3— = 300 .
`0.001,
`
`2
`
`(
`
`)
`
`3
`
`(
`
`)
`
`(4)
`
`These values apply normal to the display and we are all aware that they are not sustained when
`viewed off-axis. They also rationalize why the vast majority of TN displays are configured in the
`Normally White mode. In the twisted state, the attainment of precisely 90° rotation requires that
`the cell spacing be in a specific relationship with respect to the birefringence of the LC fluid and
`in any event is wavelength dependent. This understanding was provided by Gooch and Tarry and
`explains the reason for the limited contrast of NB as well as the familiar purple coloration
`encountered in the off-state4 (Figure 4).
`
`‘ A similar situation prevails in the NW mode but in this case it corresponds to the on-state
`and a slight imbalance in the relative transmission of green with respect to blue/red is not
`noticeable (Figure 5).
`
`M—3/4
`
`
`
`Leakage in the non—transmissive state under off—normal viewing conditions is evident in
`both modes and particularly in the NW where birefringence is encountered as the light obliquely
`transverses the cell. This results in finite ellipticity in the emerging light and the inability of the
`output analyzer to absorb the orthogonal component (Figure 6).
`
`Matrix Addressing
`
`This rudimentary review of LC structures and properties has been provided so as to
`introduce the topic of matrix addressing. All the preceding applies to an LC cell when it is
`driven continuously with an a.c. waveform. This mode is known as “direct connection”, do. or
`shutter mode. Obviously such a system requires two connections per picture element where one
`connection can be common to all elements. Nevertheless such an arrangement becomes
`impractical for a large array with N elements necessitating N+1 connections.
`
`This introduces the concept of matrix display where N display elements can in theory be
`addressed by as few as 2” connections or circuits. For example a million (106) pixel matrix
`array can be driven with 2 x 103 circuits. Such a situation is entirely feasible and indeed is
`utilized on all familiar portable display products. However, the introduction of matrix addressing
`introduces criteria which cannot be entirely fulfilled and inevitably the use of matrix addressing
`result in some compromise in the optical performance with respect to that achieved in the dc.
`mode. In other words, we are conscious of the limitations of the LC in the dc. mode and these
`will only be exacerbated when matrix addressing is employed.
`
`If a matrix were constructed utilizing linear elements as may be approximated with, for
`example, incandescent lamps, it can be immediately appreciated that in attempting to illuminate a
`single lamp that adjacent lamps will also be illuminated and also that to achieve the rated (rms)
`power at a single lamp that during the select time it would have to be driven with a power level N
`times the continuous rating. Such impracticalities are resolved with the knowledge that matrix
`addressing is facilitated by the introduction of a non-linear element at each mode of the matrix
`and the higher the degree of non-linearity then, the better the discrimination and the larger the
`array that may be envisioned. P. Alt and P. Plesko provided the classic understanding of this
`topic and the simple expression
`
`vs
`
`VNs
`
`
`NV2 +1
`
`N 2 - l
`
`where VS is the select (on) state voltage, VNS is the non—select (off) state voltage and N is the
`number of multiplexed lines.5 Analysis of this function illustrates that the basic TN performance
`is compromised when employed in a passive matrix having more than 4 multiplexed lines. In fact
`there may have been instances when this limitation was not acknowledged and resulted in some
`negative customer reaction to LC displays in general.
`
`M—3/5
`
`
`
`it was necessary to either formulate an LC structure
`To overcome such limitations,
`having a higher degree of intrinsic non-linearity (Figure 7) or to deliberately introduce a
`nonlinear element within the matrix. In fact, both approaches would be pursued; one being a
`passive matrix having extended multiplexibility and the other being the evolution of the active
`matrix. Whereas the active matrix is best characterized by incremental progress, the passive
`matrix evolutibn benefited from a distinctive innovation by Scheffer and Nehring,
`the
`Supertwisted Birefringence Effect, SBE, which is now more familiarly known as Supertwisted
`Nematic, STN.7 They showed that when the twist angle is increased to beyond 270° that the LC
`transfer function has infinite slope thus enabling multiplexed ratios far in excess with that
`obtainable with the basic 90" TN structure (Figure 8). As in the case for the NB TN there was a
`high wavelength dependency associated with this configuration which resulted in a yellow on-
`state or a blue dark state (Figure 9). Given the significance of this structure it was quickly
`understood how this effect could be compensated. Initially this was accomplished by the use of a
`complimentary twisted non-driven cell. This technique was referred to as a double cell. Shortly
`thereafter,
`the non-active cell was replaced by a retardation film having the requisite
`complementary birefringent properties.
`
`It was these developments in the late 1980s which enabled the deployment of the first
`high information content portable display devices. The multiplexing capability in conjunction
`with achromatic operation enabled resolution at least up to the VGA level to become viable and
`despite limitations associated with switching speed, temperature dependence, viewing envelope
`and gray shade this technology has probably accounted for as much as 50% of the flat panel
`business revenue over the past 10 years. While STN enjoyed a distinct cost advantage relative to
`an active matrix LCD it maintained a majority share of the portable display market. However
`with the dramatic price erosion that has occurred within the AMLCD market within the last 18
`months, it is felt by many market analysts that this will now undermine the STN market position.
`However, such projections do not necessarily account for the improvements in STN performance
`that can be achieved with active addressing as originally described by Scheffer and which is the
`subject of other SID seminars.9
`
`For completeness it is also acknowledged that the multiplexibility can also be enhanced
`with the use of a bistable LC effect such as ferroelectric LC, but as yet there islimited evidence
`of their exploitation in current display products.
`
`Two Terminal Active Matrices
`
`With the appreciation that a highly non-linear characteristic was mandatory in order to
`achieve an adequate level of multiplexibility, it was intuitive that a simple diode function in
`series with the LC could provide the requisite features, and since the mid-19805 there have been
`a series of developments and products based on the use of diodes. These developments occurred
`over the same period of time as the evolution of the Thin Film Transistor, TFI‘, and again it
`would be intuitive that Thin Film Diode, TFD, would be a simpler process thus providing yield
`and cost benefits. Although TFD products are provided by a some suppliers, for example Philips
`and Seiko, the apparent cost advantages with respect to AMLCD has not materialized into a
`
`M—3/6
`
`
`
`corresponding market share. The reason for this is not entirely clear but it is worthy to review the
`technology evolution that has materialized with TFD.
`
`Diodes
`
`The immediate benefit that arises’ with a diode structure is that the bus structure is
`
`distributed on both of the LC plates, one plate providing the data line structure and the other
`supporting the select lines. This eliminates the possibility of cross-over shorts which is a
`dominant failure mode on a TPT substrate and it also offers a high aperture factor since the bus
`lines may be ITO therefore eliminating the need for non—active Opaque bus lines which are
`encountered in a TFI‘ matrix.
`
`Despite such simplicity, one has to be mindful of the constraints imposed by the LC, i.e.
`symmetrical a.c. drive with a minimal d.c. residual component. The earlier TFD displays reported
`by Yaniv were based on the use of a-Si:H PIN diodes.lo In this respect they had much in
`common with the TF1" process development of that era and benefited from the prior development
`that had materialized in the development of solar cells. In order to fulfill symmetrical drive
`requirements, it was necessary to use two diodes for each pixel and in consequence two select
`lines for each row of pixels. This is an example of a three terminal arrangement where all three
`terminals are driven. This airangement was subsequently simplified by Kuijk which is described
`as a diqde-diode—reset matrix where one of the terminals is returned to a common reference
`voltage.
`
`In the Yaniv circuit (Figure 10), the two select lines are driven in a complimentary
`manner so that when a row is selected the node of the two diodes establishes a low impedance
`virtual ground. The data voltage is thus established across the LC and for the balance of the
`frame time, the diodes are held under reverse bias conditions thus establishing a high impedance
`in series with the LC and hence holding the stored charge on the LC. On the subsequent frame
`the polarity of VD is alternated, and contingent upon the matching of the diode functions,
`symmetrical operation is obtained.
`
`The operation of the Kuijk circuit (Figure 11) is somewhat more complex in that it
`necessitates the use of a 5 level select circuit but enables the data voltage driver range to be
`reduced to LCON-LCopp which is typically as low as 3 volts.
`
`Despite the simple and elegant nature of these circuits none of them are now being
`utilized by the developers. The reasons for this are not self evident although it is clear that the
`requisite circuit function is dependent on the tracking of the diodes, not only at an individual
`pixel but also across the entire panel.
`
`As such temperature dependencies and spatial uniforrnities become an issue and as with
`any TFD it is impractical to insert a storage capacitor so as to minimize the effect of leakage as is
`inevitable during the held period of the refresh cycle and to minimize the effect of unwanted
`parasitic coupling effects. It is now anecdotal but in the formative years of TFI‘s there were
`strong arguments against incorporating a storage capacitor in a TF1“ matrix since it was seen as
`
`M—3/7
`
`
`
`detrimental to yield, a cause for cross—coupling and in any event a reflection of poor circuit
`design. Nowadays the benefits of storage capacitor are clearly recognized and it only remains
`debatable as to how large it should be made.
`
`Notwithstanding the apparent demise of PIN diodes, TFD products are available today
`and new developments continue to be reported but they are usually based on MIM (Metal—
`Insulator-Metal) devices which offer a two terminal circuit solution.8
`
`MIMs
`
`A MINI is a sequence of layers, metal—insulator—metal. Typical materials would be
`tantalum and ITO as the metals and tantalum pentoxide or silicon nitride as the insulator
`dielectric. Since the dielectric can exhibit sites of trapped electrons, under an electric field
`condition, a current IMIM flows according to the Poole-Frenkel effect:
`
`IMIM = 06 VMIMe
`
`mm _
`
`(6)
`
`In this expression, on is the conductance and B the gain factor. Examination of the Poole-Frenkel
`equation shows it to be symmetric thus exhibiting the necessary characteristic for the ac.
`charging of an-LC. The above equation can be rewritten in log form:
`
`10g IMIM = fileMlMlloga VMIM -
`
`(7)
`
`Detailed circuit parameters can be realized so that over a range of VMIM from 3 to 6 volts
`Ian can range from 10'13 to 10'7 and from O to 3 volts IMM is less than 10'13 amps (Figure 12,
`13). These values are highlighted since those of you familiar with the pixel charging
`requirements for a TFI‘ will recognize these values as being consistent with such familiarity. In
`practice it is found that dependent upon the selected material system there may be some polarity
`dependency. Such a problem is conveniently resolved by placing two MIMs “back—to—back” in
`series.
`
`The MINI forms an equivalent capacitor, Cmnvr, in series with LC capacitor, CLc, so that
`with a 3 level select (Vs+, 0, Vs') and a data voltage, VD, at the end of the select time, the voltage
`across the LC will be
`
`VLC = Vs " VD _ VMle -
`
`(8)
`
`In addition for effective charging and minimal cross talk during the non-select period it is
`necessary that
`
`CMIM << CLC
`
`and
`
`VD <'VM,Mm.
`
`(9)
`
`(10)
`
`M—3/8
`
`
`
`Alternative 4 level drive schemes are utilized to optimize for specific objectives of circuit
`features but all analyses acknowledge the necessity to minimize the value of CMIM with respect to
`CLC. Since Cu; is proportional to the area of a pixel i.e. inversely proportional to the square of
`resolution, it would be expected that this technique would be most appropriate to low/medium
`resolution displays. The validity of this deduction is broadly correct as is attested by a number of
`large area medium resolution products but there is at least one exception to this generalization.
`For example Citizen described a high resolution array based on MIM technology.13
`
`the non—linear
`the persistent critique of a two terminal structure is that
`However,
`component is in series with the LC and any variation in its threshold voltage will directly
`contaminate the pixel voltage.
`
`As will be shown,
`
`there is a threshold characteristic associated with three terminal
`
`devices, but in this case its influence and contamination of the pixel voltage is substantially
`reduced. This is probably the fundamental reason why industry elected to undertake the challenge
`and complexity incurred with the three terminal devices.
`
`Three Terminal AMLCDs
`
`It becomes abundantly evident that the ideal matrix element is a switch, so that the data
`voltage that is established on the columns of the matrix can be selectively applied to the pixel
`electrodes during the period of the appropriate line select time and then maintained on the pixel
`for the balance of the frame time (Figure 14).
`'
`
`With the realization of such a function it would then be possible to» achieve infinite
`addressability and relieve any constraints regarding resolution, refresh rate, size etc. It is this
`vision that has motivated the immense investment in AMLCDs. However it must be recalled than
`
`an active matrix is only a means for improving the addressability of a display media. It does not
`in any way enhance the fundamental attributes or limitations of the electro—optical effect as
`established under direct drive (non—multiplexed) conditions. In other words, any departure in the
`operation of the matrix from that obtained from an ideal switch will manifest as a degradation in
`the electro—optical effect with respect to that which is obtainable under direct drive operation.
`
`Two issues emerge from this introduction. One is the considerable effort that has been
`expended on understanding the imperfections of the switch function and the extent that they
`constrain the applicability and acceptability of an AMLCD product and the second is the
`continual investigation of alternate LC-based electro-optical effects so as to alleviate the familiar
`limitations of TN for universal display applications. This particularly refers to the angular
`dependencies of the TN structure. It will be rationalized in subsequent sections that the alternate
`LC structure may have been identified and thus in conjunction with the broad advances that have
`been accomplished in the fabrication of active matrix substrates, that a much broader market
`capture by AMLCD can be reliably anticipated.
`
`M—3/9
`
`
`
`The Electronic Switch
`
`As outlined above, the basic switching operation that is required is a sample—hold
`function, and as in conventional electronic circuit design this is accomplished with the use of a
`Field Effect Transistor, FET, where the parameters of the FET and its peripheral components are
`selected according to the timing conditions and the nature of the load impedance (Figure 15).
`This is precisely what is accomplished in an active matrix where the circuit component selection
`is restricted to that which can reasonably be accomplished with thin film technology. Specifically
`the use of a Thin Film Transistor, TFI‘, as opposed to that available from a silicon Metal Oxide
`Semiconductor Field Effect Transistor, MOSFET. The timing requirements are established
`according to the number of active lines, N, of the display and the load is that imposed by the LC
`which is dominantly that of a capacitor,,C1)c. If the refresh period is T, then the sample and hold
`times are respectively TIN and T(N—1)/N respectively, and if the corresponding equivalent
`resistances of the switch are RON and Roma, and since the pixel charging and discharging
`functions may be equated to a simple exponential function, then the accuracy of the circuit
`function is determined by the number of time constants available during the corresponding time
`periods.
`
`Thus the time constants are:
`
`RONCLC << TA;
`
`and
`
`ROFFCLC>> T(N‘%,
`
`(11)
`
`(12)
`
`and the accuracy to which data is coupled to the pixel is determined by the extent to which these
`conditions are achieved. As an example, for a 1% error, each of the above time constraints have
`to exceed their corresponding limits by yS and 100 respectively. This leads to the requirement
`’ that
`
`ROW/{0 2 SOON
`
`which in the case of a 1000 line display requires that
`
`R0% < 0.5x1o6N
`
`ON
`
`.
`
`(13)
`
`(14)
`
`From these cursory requirements that which is required is that the TFT provide a dynamic
`range in excess of 106 so as to ensure margin for other circuit leakage effects as well as
`temperature dependencies which will be encountered in the TFI‘ operation.
`
`This criteria is derived from the objective of achieving consistent gray shade performance
`and would be somewhat excessive for panels designed for strictly binary operation. It may be
`construed that the need to provide a gray shade accuracy of 1 part in a 100 may also appear to be
`excessive. However,
`it must be recalled that
`the human vision system has a logarithmic
`
`M—3/1 0
`
`
`
`sensitivity function and in addition, that the LC transfer function is non linear but has a distinct
`third order characteristic. The net result of combining these transfer functions is that a 1%
`increment in voltage can manifest as much as a 1 in 16 increment in equal perceived brightness.
`That is, it can be shown that in order to accurately portray a 16 level gray shade display
`necessitates a voltage precision of 1%.
`
`Beyond the dynamic range of the TF1” it is necessary to establish the absolute orders of
`magnitude. These are primarily determined by the value of Cu: whose value will be proportional
`to the area of the pixel, the characteristics of the LC fluid and cell spacing which result in a
`normalized capacitance value in the order of 10'5 pF per square micron (maximum). For a typical
`100 pixel/inch display (pixel pitch of 250 microns) the load capacitance, CDC, after adjustment
`for three elements per pixel and aperture factor, be in the order of 0.1 pF. According to the
`guidelines established previously and using the example of a 1000 line display refreshed at 60 Hz
`' leads to the requirement that
`
`l
`S —————————~———
`60x1000x0.1x10’12x5
`
`R
`0”
`
`< 3x106 ohms
`
`and
`
`R0” > 3x10120hms
`
`(15)
`
`(16)
`
`(17)
`
`and on the basis of a switching range of 10 volts, corresponds to the’requirement that the '] Fl
`provides
`
`ION greater than 3x106 amps and Ion- less than 6x10” amps.
`
`These requirements have been derived in a very simplistic manner but serve the purpose
`of evaluating the basic features which are demanded of the TFI‘ circuit operation.
`
`Semiconductor Characteristics
`
`In the pioneering era of AMLCD there was a high expectation that Cadmium Selenide
`would emerge as the preferred semiconductor for AMLCD TFI‘s.14 It has a high value of
`mobility which enables it to deliver high values of ION which would make it suitable for use in
`peripheral driver circuits, and was also capable of providing sufficiently low values of [opp so as
`to make it suitable for matrix devices (Figure 16). However, such expectations were eclipsed by
`the success which was achieved with the use of amorphous silicon. Despite its low value of
`mobility (usually less than 1) amorphous silicon is exceptionally well suited to the sample hold
`._ function required for matrix operation. It was the adequacy of this function in conjunction with
`the relative simplicity and consistency of the processing techniques which developed around
`amorphous silicon that transitioned AMLCD from a technical curiosity to its emergence as a
`commodity item as it is known today.
`
`Polysilicon is the other broadly utilized semiconductor material. It has a high value of
`mobility (typically greater than 100) thus making it suitable for peripheral driver circuits. Its off
`
`M-3/1 1
`
`
`
`current performance is not as consistently low (10'11 pA) as that of amorphous silicon but its
`major detriment is that it is most effectively processed at temperatures in excess of 600°C. Such
`temperatures are not compatible with low cost glass substrates but mandate the use of quartz.
`
`This limitation has caused the exploitation of polysilicon to be restricted to small area
`high resolution arrays such are used in viewfinders and projection devices. In these applications
`it is essential toprovide an on-substrate driver circuit since off—glass interconnection technology
`becomes problematic at less than 70 microns pitch (interconnect densities in excess of 300 per
`inch).
`
`With the appreciation that amorphous silicon is not generally capable of providing the
`driver function and the elegance of a full-function diSplay, it is likely that the majority of the
`current research work in TFT technologies has the objective of developing lower temperature
`polysilicon processes with the anticipation that it will displace amorphous silicon in some of its
`current market sectors.
`
`Parasitic Effects
`
`The preceding established the objective that the TF1" operation be nonintrusive to the
`optical operation of the LCD as to maintain its non-multiplexed performance. So far the dc.
`requirements of the TFI‘ to support this objective have been established but beyond this, there are
`ac. induced effects which are more compelling in their resolution.
`
`In this respect it is the charge coupling effects that are induced by the gate enable pulse
`that give rise to contamination of the pixel voltage and to the optical performance. In addition the
`load CLC is dependent on the orientation of the LC molecules where spammdicula, is typically 8.3
`whereas Emma is 3.1. Therefore CLC is dependent on the signal voltage.
`
`All these effects act so as to cause a dc. voltage to be induced across the LC and beyond
`the fact that this causes an error in the optical performance, the presence of a sustained d.c.
`component
`is detrimental
`to long term operation of the cell. From such experience,
`it
`is
`empirically established that the dc. component should be constrained to be less than 100
`millivolts (Figure 17 and 18).
`
`the magnitude of the induced effects would be considerably
`In all AMLCD circuits,
`greater than 100 mV, but is invariably reduced by the use of a storage capacitor, C5, so as to
`minimize the effect of the overlap capacitance, ng, and the signal dependent nature of CLC-
`Thus,
`
`C
`3"————-
`V, = AV
`8 CM + CLC + Cs
`
`(18)
`
`where ng is the gate-drain overlap capacitance and is a partial contributor to the overall parasitic
`capacitance, Cp.
`,
`
`M—3/1 2
`
`
`
`In theory Cs should be as large as possible but has to beconstrained by the limitations of
`the ION of the TFI‘ and the compromises that are encountered in process yield and aperture factor
`from a large value of CLC. It is therefore a generalized result to acknowledge that Cs is frequently
`sized so that Cs>3xCu3 (max.).
`
`The preceding equation provided the total magnitude of Val; and its nominal value can be
`reduced to zero by biasing the counter electrode or backplane voltage, Vcom, so that it balances
`the induced voltage. The design problem thus reduces to the variances in V“, that can be
`encountered due to data voltage, temperature, process variation, etc.
`
`Suzuki showed that with the use of tri—level gate pulse, it is possible that when the storage
`capacitor is returned to the prior gate line, a compensating pulse AVg’ can be induced such that if:
`
`(19)
`
`8
`
`AVE, =&
`AV
`Cs
`
`the dc. error can be made to be independent of CLC and temperature related effects and therefore
`only dependent on the ability to control the ratio of ng to Cs.16
`
`Obviously this is a powerful technique, but it does introduce a further variable, AV; , as
`
`well as the complexity associated with a tri-level gate driver. As will be elaborated there is a
`component of Cp which is polarity dependent, i.e. the gate induced charge coupling is different
`for positive pixel charging and negative pixel charging.
`
`In practice the parasitic coupling capacitance is not only derived from overlap capacitance
`formed by the gate and drain contacts but also the channel capacitance, C'ch, of the
`semiconductor.
`
`To appreciate the implications of this parameter it is convenient to refer to the equations
`which describe the operation of a single crystal MOSFET
`
`u=§Ug—KY
`
`for VdS 2V3; ‘ Vt
`
`u=§Mm—mm—fi]
`
`for Vd, < Vg, - V,
`
`W
`where . fl = ,uCox z— ,
`
`M—3/1 3
`
`am
`
`an
`
`(22)
`
`
`
`fl = mobiliU'[Cm% - sec]’
`
`C =unit ca acitance [F/ ]
`m:
`p
`cm2
`
`W = channel width,
`
`L = channel length,
`
`and
`
`V, = threshold voltage .
`
`From a fundamentalist. standpoint, it may be argued that they do not necessarily apply to
`an amorphous or polysilicon device, but it is sufficiently appropriate to appreciate the mechanism
`for asymmetric charge coupling, and furthermore,
`the role of TFI‘ threshold voltage and its
`influence in determining the long term stability and operational life of TFI‘ based displays.
`
`,
`
`From these equations, Singh provides a simple but adequate expression of the induced
`error voltage due to the channel capacitance which is in addition to that contributed by ng,
`equation (18 ):
`
`Voltage error =
`
`o.5c,,,(Vg, — v,) ,7
`.
`Cr + CLC + C311
`
`(23)
`
`It will be appreciated that this error voltage is asymmetric since Vgs is dependent on the
`polarity of VS.
`
`To some extent this effect can be compensated by appropriate modulation of the source
`voltage within the signal conditioning circuitry, or can be compensated with the use of a quad—
`level drive induced on the gate line.18
`
`Equation 23 also shows the influence of threshold voltage variation but more importantly
`introduces one of the implications of gate pulse distortion which occurs along-the gate line due to
`the distributed resistance and capacitative load (Figure 19).
`
`In addition to the above the distributed resistance of the storage capacitor return which
`may be the prior gate line, will result in an erroneous reference voltage being used at the end of
`the sample period.
`
`By the use of diligent design and process practice, which includes the minimization of Cch
`and ng as well as appropriate application of compensation techniques, these error sources can be
`reduced to an acceptable level.
`
`However, the issue of gate line resistance remains somewhat problematic, and to a large
`extent is the singular impediment to the ability to fabricate larger area AMLCDs. This topic has
`
`M—3/1 4
`
`
`
`is this
`been well presented in multiple references, for example lkeda19 and Howard.20 It
`dependence on a low resistivity gate line system that prompts the introduction of alternate
`metallurgies, such as copper as described by Colgan et al.21 In general, metals such as copper and
`aluminum, possess the requisite resistivity properties but are more reactive to subsequent
`processes than the more familiar metals such as chrome and tantalum. The process development
`is therefore to establish the passivation techniques that render these metals to be benign in the
`subsequent processing environment.
`
`However, if there was a legitimate broad demand for larger area AMLCD this could be
`accomplished using existing metallurgies in conjunction with redundant or double ended gate
`drive since this is equivalent to reducing the gate resistance by a factor x 4