Power Efficient Sustain Drivers and Address Drivers for Plasma Panel

This invention relates to plasma panels and to improvements in address driver circuits and sustain driver circuits for plasma display panels, particularly for independent sustain and address plasma display panels.

Plasma display panels, or gas discharge panels, are well known in the art and, in general, comprise a structure including a pair of substrates respectively supporting thereon column and row electrodes each coated with a dielectric layer such as a glass material and disposed in parallel spaced relation to define a gap therebetween in which an ionized gas is sealed. Moreover, the substrates are arranged such that the electrodes are disposed in orthogonal relation to one another thereby defining points of intersection which in turn define discharge cells at which selective discharges may be established to provide a desired storage or display function. It is also known to operate such panels with AC voltages and particularly to provide a write voltage which exceeds the firing voltage at a given discharge point, as defined by a selected column and row electrode, thereby to produce a discharge at a selected cell. The discharge at the selected cell can be continuously "sustained" by applying an alternating sustain voltage (which, by itself is insufficient to initiate a discharge). This technique relies upon the wall charges which are generated on the dielectric layers of the substrates which, in conjunction with the sustain voltage, operate to maintain discharges.

Details of the structure and operation of such gas discharge panels or plasma displays are set forth in U.S. Pat. No. 3,559,190 issued Jan. 26, 1971 to Donald L. Bitzer, et al.

In the past two decades, AC plasma displays have found widespread use due to their excellent optical qualities and flat panel characteristics. These qualities have made plasma displays a leader in the flat-panel display market. However, plasma panels have gained only a small portion of their potential market because of competition from lower costs CRT products.

The expense of the display electronics, not the display itself, is the most significant cost factor in plasma displays. Because of the matrix addressing schemes used, a separate voltage driver is required for each display electrode. Therefore, a typical 512.times.512 pixel display requires a total of 1024 electronic drivers and connections which add considerable bulk and cost to the final product.

In a co-pending U.S. patent application Ser. No. 787,541 filed Oct. 15, 1985, and assigned to the same assignee as herein, there is described an Independent Sustain and Address (ISA) plasma panel. Also, see the publication L. F. Weber and R. C. Younce, "Independent Sustain And Address Technique For The AC Plasma Display", 1986 Society For Information Display International Symposium Conference Record, pp. 220-223, San Diego, May, 1986. The ISA plasma panel technique includes the addition of an independent address electrode between the sustain electrodes. These address electrodes are then connected to the address drivers. The sustain electrodes can be bused together and connected directly to the sustainers.

The ISA plasma panel offers two significant advantages. First, since the address electrodes do not have to deliver the large sustain current to the discharging pixels, the address drivers have low current requirements. This allows lower cost drivers to be used. The second advantage is that only half the number of address drivers are needed since one address electrode can serve the sustain electrode on either side.

Despite the significant advantages afforded by the ISA panel, it is still desired to reduce as much as possible the manufacturing cost of such panels. However, while the ISA panel has enabled a reduction of the address drivers of a typical 512.times.512 pixel display from 1024 electronic address drivers to only 512 drivers, this is still a significant number of required electronic components. In fact, the plasma panel cost is dominated by the cost of the associated required electronic circuits such as the addressing river circuits and sustain driver circuits. In addition, it is desired to reduce the amount of energy normally lost in charging and discharging the capacitance of the plasma panel.

It is therefore desired to reduce the cost of plasma panel reduction by reducing the cost by the associated electronics.

It is also desired to reduce the operational cost of plasma panels.

In accordance with one aspect of the present invention, an improved address driver circuit is provided for the ISA plasma panel. The new driver circuit utilizes open-drain (N-channel or P-channel) MOSFET output structure which can be made at a lower cost compared to the normally used totem-pole drivers. A unique feature of the present invention resides in a technique used to apply the proper positive and negative pulses to the ISA plasma display panel by using identical, low cost N-channel open-drain MOSFET devices. Thus, in contrast with prior plasma panel address driver circuits that must be able to pull high (i.e., drive the plasma panel with a positive pulse) and pull low (i.e., drive the plasma panel with a negative pulse) the unique feature of the present invention enables the N-channel open-drain MOSFET devices only to be designed to pull low.

In accordance with another aspect of the present invention, a power efficient sustainer circuit has been developed for use with flat panels having substantial inherent panel capacitance due to the panel electrodes, such as plasma display panels, electroluminescent panels, liquid crystal displays, etc. The new sustain driver circuit uses inductors in charging and discharging the panel capacitance so as to recover 90% of the energy normally lost in driving the panel capacitance. Accordingly, a plasma panel incorporating a power efficient sustain driver circuit according to the present invention can operate with only 10% of the energy normally required with prior art plasma panel sustaining circuits.

FIGS. 1a, 1b, 1c are schematic representations of switch devices useful in explaining an address circuit driver;

FIG. 2 is a plan view of a plasma panel with open-drain address drivers and sustain drivers in accordance with one aspect of the invention;

FIG. 3 are waveform diagrams useful in understanding the operation of FIG. 2;

FIG. 4 are waveform diagrams showing an expanded view of the section of FIG. 3 labeled 4--4;

FIG. 5 is a schematic circuit diagram showing an ideal model of a new sustain driver according to the invention;

FIG. 6 are waveform diagrams useful in understanding the operation of FIG. 5;

FIG. 7 is a schematic circuit diagram showing a practical circuit model of a new sustain driver according to the invention;

FIG. 8 are waveform diagrams useful in understanding the operation of FIGS. 7 and 9;

FIG. 9 and 9a are schematic circuit diagrams showing a constructed embodiment of a new sustain driver according to the invention;

FIG. 10 is a schematic circuit diagram of a new sustain driver in an integrated circuit design;

FIG. 11 is a schematic circuit diagram of an XAP address pulse driver incorporating energy recovery techniques according to the invention;

FIG. 12 are waveform diagrams useful in understanding the operation of FIG. 11;

FIG. 13 is a schematic circuit diagram of YAP address pulse driver incorporating energy recovery techniques according to the invention; and

FIG. 14 are waveform diagrams useful in understanding the operation of FIG. 13.

The present invention will be described in connection with an ISA plasma panel to which has been incorporated a new and improved address driver circuit in accordance with one aspect of this invention, and a new power efficient sustain driver circuit in accordance with another aspect of the present invention. For convenience of description, the first aspect of this invention, i.e., the new and improved address driver circuit will be described followed by the description of the power efficient sustain driver circuit.

A major advance of this invention is the simplification of the address circuit drivers. These drivers only need to be designed to pull low. This contrasts with the normal plasma panel circuits that must be able to pull high and pull low. The pull low type driver can be fabricated at considerably lower cost. FIG. 1 shows the basic type of address circuit driver that can be used in this invention. FIG. 1a shows a simple switch in parallel with a diode. The switch is used to apply selective address pulses to the plasma panel depending on the state (open or closed) of the switch. With today's solid state switching technology, this switch usually takes two forms: the MOS Field Effect Transistor (MOSFET), shown in FIG. 1b and the Bipolar transistor shown in FIG. 1c. Usually there is an inherent parallel diode associated with these transistors so that the diode in parallel with the switch in FIG. 1a should be understood as being included in the circuit model. The examples presented here are for N-channel MOSFETs and npn Bipolar transistors because these are usually the best devices for integration. However, devices of the opposite polarity could be used with the appropriate adjustment in the waveforms and circuits.

FIG. 2 shows a circuit diagram for applying the concepts of this invention to drive the address electrodes in an ISA plasma panel i.e., a plasma display panel having independent sustain and address electrodes as previously described.

This example uses the N-channel MOSFET devices shown in FIG. 1b, but of course other suitable switches could be used. The basic concept is to connect the drain electrode of each MOSFET to each address electrode of the ISA plasma panel and to then connect all of the sources of the MOSFETs on a given display axis to a common bus. When such MOSFET transistors are integrated, it is very easy to fabricate arrays of these transistors when they have all of the sources connected to a common bus. This arrangement is commonly referred to as the open drain configuration. Note that both the X axis and Y the axis address electrodes in FIG. 2 use N-channel MOSFETs in the open drain configuration. This has the advantage that the same electrical parts can be used for both the X and the Y axis. This allows lowering of circuit costs because normally two distinct parts must be designed, fabricated and stocked. In addition, a single part will be made at twice the volume of that of the systems that require two parts and therefore the higher volume of the single part will result in lower costs. Two parts are normally required because the X and Y axes require different polarity address pulses. In the example shown here, the X axis requires a positive pulse and the Y axis requires negative pulses. A novel feature of this invention is the technique used to apply the proper positive and negative pulses to the ISA plasma display panel address electrodes by using identical low cost N-channel open drain MOSFET devices.

FIG. 3 shows the waveforms used to drive the ISA panel. This shows a portion of the video scan of the panel for addressing the eight rows of pixels shown in FIG. 2 in a top to bottom sequence. Other scanning techniques may be used rather than the video scan example illustrated here. Each row of pixels requires two of the 20 microseconds addressing cycles. The top four waveforms show the signals applied by the four sustainers. The phasing of these waveforms selects which of the four pixels surrounding each address cell in FIG. 2 can be addressed during a given addressing cycle. The fundamental periodicity of this phasing is every eight addressing cycles because of the sustain electrode connection technique used in FIG. 2.

Below the sustain waveforms are the signals associated with the address electrodes. The waveforms labeled XAP and YAP are supplied from address pulse generators that are connected to the common bus of the address driver transistors as shown in FIG. 2. These address pulsers generate the special waveforms needed for the address drivers to apply the proper signals to the address electrodes. The XA waveform shows the selective erase signals on the X address electrodes. A high XA level will erase a selected pixel and a low level leaves the pixel on. The YA waveforms for four adjacent Y address electrodes are shown at the bottom of FIG. 3.

We will now investigate the details of how the FIG. 2 circuit operates. The Y axis will be examined first since its operation is the simplest. The linear array of open drain transistors have all of their source electrodes connected to a common bus. This bus is connected to a pulse generator called the Y address pulser and labeled YAP. The purpose of this generator is to supply the energy for the address pulses and to determine the shape of the waveforms applied to the selected Y address electrodes. Notice that, as shown in FIG. 3, this generator supplies double amplitude negative pulses. For instance, during the address period, a negative pulse needs to be applied to the selected Y address electrodes. During this period, a negative pulse is generated by YAP and this pulse is applied to the source electrode of all of the Y address transistors. Any transistors that are off do not conduct and their associated plasma panel address electrodes remain at virtually the same potential as they were before the generation of the negative pulse. Any transistors that are turned on will conduct and their associated plasma panel address electrodes will be pulsed negative to cause an address operation in the plasma panel. Any number of Y address electrodes could be selectively pulsed negative with this technique, however, in video mode, the Y axis address electrodes are usually pulsed one at a time in a sequential manner that causes the image to be scanned.

Since the address electrodes of an ISA plasma panel can be reasonably modeled as a simple capacitance, the current through the transistors flows predominently during the transitions of the YAP generator. During the negative transition of the YAP generator the conduction current must flow predominantly through the transistor. However, during the positive going transition of the negative address pulse (as it returns to the initial level before the application of the negative pulse), the current can flow through both the MOSFET transistor and also through the body diode that is associated with the transistor. This body diode will of course conduct whether the transistor is in the on or off states. This will allow all of the Y address electrodes to be pulled to the same high level when the YAP generator is at its high level.

We will now discuss the operation of the X axis circuits shown in FIG. 2. This circuit differs from that of the Y axis because the X axis must be capable of applying a positive pulse as opposed to the negative pulse of the Y axis. Note that just like for the Y axis, the array of N-channel open drain MOSFET transistors has all source electrodes connected to a common bus and this bus is connected to the X address pulse generator labeled as XAP. This XAP generator operates quite differently from the YAP generator because of the opposite polarity of the output pulse. The shape of the XAP waveform is two short pulses (see FIG. 3 and the expanded view of FIG. 4) used to generate a single longer pulse on the plasma panel address electrodes. The first XAP pulse corresponds to the leading edge of the address electrode pulse and the second XAP pulse corresponds to the trailing edge of the address electrode pulse.

Now we examine the first XAP pulse. It is assumed that all of the address electrodes are initially at the same potential as the XAP generator just before the application of the first pulse. As the XAP generator rises, current flows through all of the body diodes of the MOSFET transistors. This pulls up all of the X address electrodes to a level that is just one diode drop lower than the XAP generator. This action continues until the XAP generator reaches its first peak. Note that all X address electrodes ar pulsed positive at this time regardless of whether they are selected or not.

The selection operation does not occur until the falling edge of the first XAP pulse. During this time, if a positive pulse is to remain on any selected X address electrodes, then the associated MOSFET transistor is turned off. The transistors that are left on will pull their address electrodes down as the first pulse of the XAP generator falls. This action continues until the XAP generator stops falling at the end of the first pulse. At this time, all of the selected address electrodes are at a high voltage level and the unselected address electrodes are at a low level. This situation can continue for a long period until the second XAP pulse. The selected address electrodes are held high by the capacitance of the plasma panel address electrodes to the sustain electrodes. The unselected address electrodes are held at the low voltage of the XAP generator by the MOSFET transistors that are turned on.

The selection pulse can be terminated by turning all of the transistors on while the XAP generator is at the low level. This works but with some undesirable characteristics. First of all, when the selected transistors are turned on, they quickly discharge the voltage of the address electrode. The discharge rate is frequently so fast that a large amount of displacement current flows through the transistors and the plasma panel capacitance. This displacement current can cause a number of problems. First, this current frequently grows and decays at a very fast rate so that large amounts of electrical noise is generated. This noise tends to create problems for other circuits in the system and can easily mis-trigger many of the logic gates that ar used to control th operations of the plasma panel. A second problem of this large current is the large energy dissipation that occurs in the transistor to discharge the capacitance. This energy dissipation can be enough to burn out the transistors in some cases. It also makes the transistors hot and requires special heat sinking requirements. In addition, the energy lost in heating these transistors cannot be recovered and it increases the power requirements of the power supply and the power consumption of the plasma display system.

All of these problems can be significantly reduced with the following switching technique. Shortly before the X address pulse needs to fall, the XAP generator begins the rise of its second pulse. Recall that the first XAP pulse was used to initiate the address pulse. During the rise of the second pulse, current flows through the body diodes of the MOSFETs associated with the unselected X address electrodes. If the MOSFETs of the unselected transistors are still on, there will also be some conduction through these MOSFETs. This current charges up the unselected address electrodes and causes their voltage to rise. This charging continues until the second X pulse reaches its peak. At this peak, all of the X axis MOSFETs should be turned on. As the second XAP pulse begins to fall, a current flows through all of the X MOSFETs which discharges all of the address electrodes. This action continues until the second X pulse completes its fall to its lowest level. At this point, all of the address electrodes should be at this low XAP voltage. This is the final stage of the addressing operation and all of the X address electrodes will be held at this low voltage level until the next addressing operation.

The write before erase addressing proceeds with the following sequence. FIG. 3 shows that a write pulse is first applied to the YAn+1 electrode which turns on all of the pixels in the two rows on either side of YAn+1. After the completion of this write pulse, four erase pulses are used to selectively erase the pixels in the two rows on either side of YAn. The image is introduced in the panel through a selective erase by controlling the voltage of the XA address electrodes during the erase operation. The sequence continues by writing the two rows on either side of YAn+2 and then selectively erasing the two rows next to YAn+1. This staggering of the write and erase operation improves panel voltage margins by allowing the written cells to stabilize for at least four cycles before the selective erase operation occurs. Note that the addition of the write operation to the addressing sequence does not require any additional time beyond that already needed for the sustain and selective erase operations. This allows higher update rates.

A key factor that allows the use of low-cost open-drain address drivers is the design of the address pulser waveforms. FIG. 3 shows that the YA address electrodes require selectively applied negative pulses and the XA address electrodes require selectively applied positive pulses. The design of the X and Y address pulser waveforms allows these two polarities with the same N-channel IC design.

In summary of the YA operation first, note that the YAP signal applied to the sources of all of the Y address transistors closely follows the selected YA address electrodes signals. At a given time a selected YA electrode transistor is turned on and all of the other YA transistors are kept off. Thus the negative pulse generated by YAP is transferred to the selected YA address electrode.

A summary of the operation of the XA address electrodes is more complicated. This is shown in the FIG. 4 expanded view of the FIG. 3 waveforms. Note that the XAP waveform shows two short pulses for each XA erase pulse. These pulses define the leading and trailing edges of the XA erase pulse. They have a sine wave shape since in a constructed embodiment of the invention they are generated with an energy recovery circuit similar to the sustain drive circuit described hereinafter. The rise of the first XAP pulse pulls all of the XA address electrodes high through the body diode and conduction channel of the MOSFET address drivers. At the peak of the first XAP pulse the selected MOSFETs are turned off if the selected pixel is to be erased. The MOSFETs that are left conducting will pull their XA address electrodes low as the first XAP pulse falls low. The selected MOSFETs that are not conducting will remain high by means of the capacitance of the address electrode to the sustain electrodes. This high level on the address electrode causes erasure of the pixel.

The rise of the second XAP pulse pulls all of the non-selected XA address electrodes to the same high level as the selected XA address electrodes. At the peak of the second XAP pulse, all of the X axis address drivers are turned on so that the fall of the second XAP pulse will pull all of the address electrodes to the initial low level.

The above XA address technique successfully places positive pulses on the selected XA address electrodes, however, it is also places two short positive pulses on the non-selected XA address electrodes that correspond to the pulse of XAP. To prevent these two short pulses from causing mis-addressing of the non-selected pixels, the YAP pulse is properly phased as shown in FIG. 4. The YAP pulse falls after the fall of the first XAP pulse and YAP rises before the rise of the second XAP pulse. This prevents the non-selected XA pulses from adding to the selected YA pulse to cause a mis-addressing discharge.

One concern is that when the column drivers are in a high impedance state, the pulses applied to a neighboring electrode in the low impedance state will capacitively couple to the high impedance electrode and cause it to receive the wrong voltage amplitude. This is not a significant problem for two reasons. First, note that in FIG. 2, the address electrodes are shielded from each other by the sustain electrodes. This makes the variations in pulse amplitude, due to address line-to-line coupling, less than 10% of the address pulse amplitude as shown in FIG. 4. The second point is that this 10% variation is not a significant problem because of the excellent address margins of the ISA design.

Standard voltage pulse generators can be used as the XAP and YAP address pulse generators supplying th corresponding waveforms of FIG. 3. Alternatively, the energy recovery technique described hereinafter with respect to the power efficient sustain driver circuit ca be used for the XAP and YAP address pulse generators.

The plasma panel requires a high voltage driver circuit called a sustainer, or sustain driver circuit, which drives all the pixels and dissipates considerable power. As an example, four sustainer drivers XSA, XSB, YSA, YSB are shown in FIG. 2 with the ISA panel.

The following describes a new high-efficiency sustainer that eliminates most of the power dissipation resulting from driving the plasma panel with a conventional sustainer. With this new sustainer, considerable savings can be realized in the overall cost of the plasma panel. The new sustainer can be applied to standard plasma panels, or the new ISA plasma panel, as well as to other types of display panels requiring a high voltage driver, such as electroluminescent or liquid crystal panels having inherent panel capacitance.

When the plasma panel is used as a display, frequent discharges are made to occur by alternatively charging each side of the panel to a critical voltage, which causes repeated gas discharges to occur. This alternating voltage is called the sustain voltage. If a pixel has been driven "ON" by an address driver, the sustainer will maintain the "ON" state of that pixel by repeatedly discharging that pixel cell. If a pixel has been driven "OFF" by an address driver, the voltage across the cell is never high enough to cause a discharge, and the cell remains "OFF".

The sustainer must drive all of the pixels at once; consequently, the capacitance as seen by the sustainer is typically very large. In a 512.times.512 panel, the total capacitance of all the pixel cells in the panel, Cp, could be as much as 5 nF.

Conventional sustainers drive the panel directly, and thus 1/2CpV.sub.s.sup.2 is dissipated in the sustainer when the panel is subsequently discharged to ground. In a complete sustain cycle, each side of the panel is charged to V.sub.s and subsequently discharged to ground. Therefore, a total of 2CpV.sub.s.sup.2 is dissipated in a complete sustain cycle. The power dissipation in the sustainer is then 2CpV.sub.s.sup.2 f, where f is the sustain cycle frequency. For Cp=5nF, V.sub.s =100V, and f=50 kHz, the power dissipation in the sustainer, resulting from driving the capacitance of the panel, is 5 W.

If an inductor is placed in series with the panel, then Cp can be charged and discharged through the inductor. Ideally, this would result in zero power dissipation since the inductor would store all of the energy otherwise lost in the output resistance of the sustainer and transfer it to or from Cp. However, switching devices are needed to control the flow of energy to and from the inductor, as Cp is charged and discharged. The "ON" resistance, output capacitance, and switching transition time are characteristics of these switching devices that can result in significant energy loss. The amount of energy that is actually lost due to these characteristics, and hence the efficiency, is determined largely by how well the circuit is designed to minimize these losses.

In addition to charging and discharging Cp, the sustainer must also supply the large gas discharge current for the plasma panel. This current, I, is proportional to the number of pixels that are "ON". The resulting instantaneous power dissipation is I.sub.2 R, where R is the output resistance of the sustainer. Thus, the power dissipation due to the discharge current is proportional to I.sub.2, or the square of the number of pixels that are "ON".

There are two ways to minimize this dissipation. One is to minimize the output resistance of the sustainer by using very low resistance output drivers, and the other is to minimize the number of pixels that are "ON" at any time.

This invention provides a new sustainer circuit that will recover the energy otherwise lost in charging and discharging the panel capacitance, Cp. The efficiency with which the sustainer recovers this energy is here defined the "recovery" efficiency. When Cp is charged to V.sub.s and then discharged to zero, the energy that flows into and out of Cp is CpV.sub.s.sup.2 ; therefore, the recovery efficiency is defined by ##EQU1## where E.sub.lost is the energy lost in charging and discharging Cp.

Notice that the recover efficiency is not the same as the conventional power efficiency, defined in terms of the power delivered to a load, since no power is delivered to the capacitor, Cp; it is simply charged and then discharged. The recovery efficiency is a measure of the energy loss in the sustainer.

A circuit proposed for driving electroluminescent (EL) panels, published in M. L. Higgins, "A Low-power Drive Scheme for AC TFEL Displays", SID International Symposium Digest of Technical Papers, Vol. 16, pp. 226-228, 1985, was tested in the laboratory, but was abandoned since it was not capable of better than 80% energy recovery, and it has undesirable design complexities. A new, very efficient sustain driver was then developed which eliminates the problems inherent in the prior proposed circuit.

First, a circuit model of the new sustain driver circuit will be analyzed to determine the expected recovery efficiency. The reasons why greater than 90% recover efficiency is possible with this new sustain driver will be explained, and several design guidelines will be given. Next, a constructed prototype of the new sustain driver will be discussed.

An ideal sustain driver circuit will be presented first to show the basic operation of the new sustain driver, given ideal components. As would be expected, given ideal components, this circuit has 100% recovery efficiency in charging and discharging a capacitive load. The schematic of the ideal sustain driver circuit is shown in FIG. 5, and in FIG. 6 are shown the output voltage and inductor current waveform expected for this circuit as the four switches are opened and closed through the four switching states. The operation during these four switching states is explained in detail below, where it is assumed that prior to State 1, Vss is at Vcc/2 (where Vcc is the sustain power supply voltage), Vp is at zero, S1 and S3 are open, and S2 and S4 are closed. The reason that V.sub.s s is at Vcc/2 will be explained, below, after the switching operation is explained:

State 1. To start, S1 closes, S2 opens, and S4 opens. With S1 closed, L and Cp form a series resonant circuit, which has a forcing voltage of Vss=Vcc/2. Vp then rises to Vcc, at which point I.sub.L is zero, and D1 becomes reverse biased. Alternatively, diode D1 could be eliminated and S1 opened when Vp rises to Vcc (at the point where I.sub.L is zero).

State 2. S3 is closed to clamp Vp at Vcc and to provide a discharge current path for any "ON" pixels.

State 3. S2 closes, S1 opens, and S3 opens. With S2 closed L and Cp again form a series resonant circuit, which has a forcing voltage of Vss=Vcc/2. Vp then falls to ground, at which point I.sub.L is zero, and D2 becomes reverse biased. Alternatively, diode D2 could be eliminated and S2 opened when Vp falls to zero (at the point where I.sub.L is zero).

State 4. S4 is closed to clamp Vp at ground while an identical driver on the opposite side of the panel drives the opposite side to Vcc and a discharge current then flows in S4 if any pixels are "ON".

It was assumed above that Vss remained stable at Vcc/2 during the above charging and discharging of Cp. The reasons for this can be seen as follows. If Vss were less than Vcc/2, then on the rise of Vp, when S1 is closed, the forcing voltage would be less than Vcc/2. Subsequently, on the fall of Vp, when S2 is closed, the forcing voltage would be greater than Vcc/2. Therefore, on average, current would flow into Css. Conversely, if Vss were greater than Vcc/2, then on average, current would flow out of Css. Thus, the stable voltage at which the net current into Css is zero is Vcc/2. In fact, on power up, as Vcc rises, if the driver is continuously switched through the four states explained above, then Vss will rise with Vcc at Vcc/2.

If this were not the case, a regulated power supply would be needed to supply the voltage Vss. This would increase the overall cost of the sustain circuitry and could make this design less desirable.

The energy losses due to the capacitances and resistances inherent in the real devices, i.e., the switching devices, the diodes, and the inductor, can be determined by analysis of a practical circuit model shown in FIG. 7. The switching devices are modeled by an ideal switch, an output capacitor, and a series "ON" resistor. The diodes (except Dc1 and Dc2) are modeled by an ideal diode, a parallel capacitor, and a series resistor, and the inductor is modeled by an ideal inductor and a series resistor.

Dc1 and Dc2 are ideal diodes. They are included to prevent V1 from dropping below ground and V2 from rising above Vcc. As will be shown below, if Dc1 and Dc2 were not included, then the voltages across C1, Cd2, C2, and Cd2 would be higher than otherwise, which would lead to additional energy losses.

The switching sequence of this circuit is the same as that of the ideal model shown in FIG. 5. FIG. 8 shows the voltage levels for Vp, V1, VL, and V2 and the current levels for I.sub.L, I1, and I2 during the four switching states. Again, it is assumed that Vss is stable at Vcc/2.

The recovery efficiency in the practical circuit model of FIG. 7 can be determined below with reference to FIG. 8. For example, the energy losses due to the capacitance of the switching devices (C1 and C2) and the diodes (Cd1 and Cd2) can be determined; then, the energy losses due to the resistances of the switching devices (R1 and R2), the diodes (Rd1 and Rd2), and the inductor (R.sub.L) can be determined; and finally, the energy loss due to the finite switching time of the switching devices can be determined. In each case, reference ca be made to the four switching states, shown in FIG. 8.

To find the power dissipation resulting from the capacitances of the switching devices and the diodes, an account is made of all the 1/2CV.sup.2 loss. It is assumed that, initially, S1 and S3 are open, S2 and S4 are closed, V.sub.L is at ground, and Vss is at Vcc/2.

State 1. To start, S1 closes and S4 opens. V1 and V.sub.L then rise to Vss, and the voltages across Cd2 (V2-V.sub.L) and across C1 (Vss-V1) both fall from Vss to zero. Thus, C1Vss.sup.2 /2 is dissipated in R1 and Cd2Vss.sup.2 /2 is dissipated in R1, Rd1, and R2. S2 then opens. With S1 closed, the series combination of R1, Rd1, L, and Cp is a series RLC circuit with a forcing voltage of Vss=Vcc/2. The waveforms are shown in FIG. 8. As I.sub.L falls to and crosses zero, then D1 becomes cut off and V.sub.L begins to rise.

State 2. S3 is closed to clamp Vp at Vcc. (Notice that before S3 closes Vp has not completely risen to Vcc, due to the damping that was caused by R1, Rd1, and RL. Thus, when S3 is closed, Vp is pulled p to Vcc through S3, and a small amount of overshoot cold occur if there were stray inductances present in the real circuit. This overshoot is shown in the waveform for Vp in FIG. 8). I.sub.L then becomes negative as C2 and Cd1 (V.sub.L -V1) both rise from zero to Vss, at which point Dc2 becomes forward biased and I2 begins to flow. The energy in the inductor, when I2 begins to flow, is then 1/2(C2+Cd1)Vss.sup.2. This energy is dissipated in R.sub.L, Rd2, and R3 as I2 falls to zero.

State 3. After the discharge current for any "ON" pixel cells has been supplied, then S2 closes and S3 opens. V2 and V.sub.L then fall to Vss, and the voltages across Cd1 (V.sub.L -V1) and across C2 (V2-Vss) both fall from Vss to zero. Thus, C2Vss.sup.2 /2 is dissipated in R2 and Cd1Vss.sup.2 /2 is dissipated in R2, Rd2, and R1. S1 then opens. With S2 closed, the series combination of R2, Rd2, RL, L, and Cp is a series RLC circuit with a forcing voltage of Vss=Vcc/2. The waveforms are shown in FIG. 8. As I.sub.L rises to and crosses zero, then D2 becomes cutoff and VL begins to fall.

State 4. S4 is closed to clamp Vp at ground. (Notice that before S4 closes, Vp has not completely fallen to ground, due to the damping that was caused by R2, Rd2, and R.sub.L. Thus, when S4 is closed, Vp is pulled down to ground through S4, and a small amount of undershoot could occur if there were stray inductances present in the real circuit. This undershoot is shown in the waveform for Vp in FIG. 8.) I.sub.L then becomes positive as CC1 and Cd2 are charged from the inductor. The voltages across C1 (Vss-V1) and across Cd2 (V2-VL) both rise from zero to Vss, at which point Dc1 becomes forward biased and I1 begins to flow. The energy in the inductor when I1 begins to flow is then 1/2(C1+Cd2)Vss.sup.2. This energy is dissipated in R.sub.L, Rd1, and R4 as I1 falls to zero.

Thus, it can be determined that the practical circuit model of FIG. 7 results in a power loss of (f)E.sub.lost =0.17 W, where the sustain frequency is equal to f =50 kHz. By comparison, if there were no energy recovery, then the normal loss from charging and discharging Cp would be (f)CpVcc.sup.2 =2.5 W. The recovery efficiency (as previously defined) of the circuit of FIG. 7 is

where Cp=5 nF and Vcc=100 V.

In summary, the practical circuit model of FIG. 7 predicts the the new sustain driver will be capable of 93% recovery, assuming that the Q of the inductor is at least 80 and that the optimum tradeoff between switch output capacitance and "ON" resistance is realized.

The schematic of a constructed prototype sustain driver circuit is shown in FIG. 9, and a complete parts list is given in Table 1.

It was found that the waveforms of the constructed circuit of FIG. 9 correspond almost exactly with the waveforms of FIG. predicted from the circuit model of FIG. 7.

Switches S1, S2, S3, and S4 in FIG. 7 were previously described as being switched at the appropriate times to control the flow of current to and from Cp. In the prototype circuit of FIG. 9, the power MOSFETs (T1, T2, T3, T4) replace the ideal switches of FIG. 7 and must be switched at the appropriate times by real drivers to control the flow of current to and from Cp. Switching T1 and T2 at the appropriate times requires only that they are switched on the transition of Vi. Thus, only a single driver (Driver 1) is required. Switching T3 and T4 presents a more difficult problem, however, since in addition to being switched on the transition of Vi, they must also be switched whenever the inductor current crosses zero. This could have required that T3 and T4 be controlled with additional inputs to the FIG. 9 circuit if it were not the case that V1 and V2 make voltage transitions whenever Vi makes a transition and shortly after the inductor current crosses zero. Thus, the switching of T3 and T4 is accomplished by using the transitions of V1 and V2 to switch the Drivers (2 and 3) in FIG. 9 at the appropriate times and no additional inputs are required.

Switching the MOSFETs can be seen with reference to FIG. 9 and the following description. When Vi rises, the output of Driver 1 is switched "LOW" and the gates of T1 and T2 are driven "LOW" through the coupling capacitors, C.sub.g1 and C.sub.g2. Thus, T1 is switched "ON", T2 is switched "OFF", and current begins to flow in the inductor to charge Cp. Also, D3 becomes forward biased and D4 is reverse biased. This causes Driver 2 to quickly switch "LOW", thus driving T4 "OFF", while Driver 3 is delayed from switching "LOW" until after Vp has risen. (As will be explained later, R1 and R2 are needed only during initial startup when Vcc power is first applied and before Vss has risen high enough for Drivers 2 and 3 to be switched from the changes in voltage of V1 and V2.)

Referring back to the end of State 1 in FIG. 8 it can be seen that V2, in FIG. 9 will begin to rise from Vss to Vcc shortly after the inductor current into Cp has fallen to zero, at which time, T3 must be switched "ON" to clamp Vp at Vcc. In FIG. 9, when V2 rises, then the input of Driver 3 also rises, due to the current through the coupling capacitor C4. The output of Driver 3 then switches "LOW", and the gate of T3 is driven "LOW" throughout he coupling capacitor, C.sub.g3. Thus, T3 is switched "ON" and clamps Vp to Vcc.

Later, when Vi falls, the output of Driver 1 is switched "HIGH" and the gates of T1 and T2 are driven "HIGH" through the capacitors, C.sub.g1 and C.sub.g2. Thus, T1 is switched "OFF", T2 is switched "ON", and current begins to flow in the inductor to discharge Cp. Also, D4 becomes forward biased and D3 becomes reverse biased. This causes Driver 3 to quickly switch "HIGH", thus driving T3 "OFF", while Driver 2 is delayed from switching "HIGH" until after V.sub.p has fallen.

When V1 begins to fall from Vss to ground, shortly after the inductor current flowing out of Cp has fallen to Zero (as at the end of State 3 in FIG. 8), then the input of Driver 2 falls because of the coupling capacitor C3. The output of Driver 2 then switches "HIGH", and the gate of T4 is driven "HIGH". Thus, T4 is switched "ON" and clamps Vp to ground.

Notice that an external timing circuit is not needed to determine when to switch T3 and T4 because the switching occurs shortly after the inductor current crosses zero, independent of the rise or fall time of Vp. This leads to simple circuitry that is independent of variations in the inductance (L) or the panel capacitance (cp) and is a significant advantage over prior proposed sustain drivers. It also makes it possible to drive the circuit with only one input, so that if the input becomes stuck ("HIGH" or "LOW"), T3 and T4 cannot both be "ON" simultaneously, which would result in the destruction of one or both of the devices.

Another advantage that this circuit has over prior proposed circuits is that T1, D1, T2 and D2 need only be 1/2 Vcc rather than the full Vcc voltage of prior circuits. Lower voltage switching devices, requiring lower breakdown voltages, are typically less costly to fabricate. This results in a lower parts cost for a discrete sustainer and lower integration costs for an integrated sustainer.

The resistors, R1 and R2 are provided for the case in which Vss is at a very low voltage, such as during initial power up of Vcc. In this case, the voltages V1 and V2 do not change enough to cause the Drivers 2 and 3 to switch. The resistors will cause the Drivers 2 and 3 to switch, after a delay time, which is determined by the value of the resistors and the input capacitance of the Drivers.

The reason it is necessary to switch the Drivers 2 and 3 during initial power up when Vss is very low is as follows. In order for Vss to rise, it is first necessary to T3 to switch "ON" and bring Vp up to Vcc. Then, when T2 turns "ON", a current will flow from Cp to Css. If T4 is later switched "ON", thus clamping Vp to ground, then when T1 turns "ON", the current that flows out of Css will prevent Vss from rising above Vcc/2, and Vss will begin to stabilize at Vcc/2 after several cycles of charging and discharging Cp. Thus Vss will not achieve the proper voltage unless T3 and T4 are switched "ON" by the action of R1 and R2 during power up.

The resistor, R3, is provided to discharge the source to gate capacitance of T3 when the supply voltage, Vcc, suddenly rises during power up. Without R3, the source to gate voltage of T3 would rise above threshold, as Vcc rises, and remain there, with T3 "ON", after Vcc has risen. Then, if T4 were switched "ON", a substantial current would flow through T3 and t4 and possibly destroy one or both of the devices.

In an experimental setup for measuring the efficiency of the prototype circuit in FIG. 9, the supply voltage (Vcc) and the supply current were accurately measured while the circuit was driving a 5 nF capacitor load (Cp). The load was driven at a frequency of f=50 kHz, with the supply voltage at 100 V. Thus, the normal power dissipation expected in this case was ##EQU2##

The measured supply current for the FIG. 9 circuit was 2.0 mA, so the actual power drawn from the supply and dissipated in the driver was 0.2 W. Thus, this circuit recovered all but 0.2 W of the normally lost power. The previously defined recovery efficiency is therefore 92%.

By comparison, the recovery efficiency predicted by analysis of the circuit model of FIG. 7 is 93%. This is an indication that the most significant sources of power loss in the real circuit of FIG. 9 have been accurately accounted for in the model of FIG. 7, and the model is a valid representation of the real circuit.

The sustain driver of FIG. 9 can be used on each side of an ISA plasma panel. As an example, each of the sustain drivers XSA, XSB, YSA, YSB, in FIG. 2 could be a sustain driver of FIG. 9, and could be used with the open-drain address drivers previously described in connection with FIGS. 1-4.

After testing two sustain drivers (each as shown in FIG. 9 with capacitor loads, one sustain driver was connected to each side of a 512.times.512 ac plasma display panel. It was found that these sustain drivers could drive the panel with 90% recovery efficiency when no pixels were "ON", and that with all of the pixels "ON", the dissipation was still low enough that heat sinks were not necessary. With all of the pixels "ON", the power dissipation in T1 and T2 did not change, but the power dissipation in T3 and T4 increased due to the I.sup.2 R losses resulting from the flow of discharge current. This power dissipation can be lowered by using lower "ON" resistance devices for T3 and T4.

In testing the prototype sustain driver circuit of FIG. 9, it was found that this circuit continued to charge and discharge the panel at the sustain frequency with high recovery efficiency, regardless of large variations in the panel capacitance or in the inductance of the coil. This is a distinct advantage over prior proposed sustain driver circuits.

It may be possible to substitute bipolar power transistors for the power MOSFETs, T1 and T2 in FIG. 9 in a suitably designed circuit. Also, since the power dissipation and, hence, the cooling requirements have been significantly reduced in the sustain driver circuit of FIG. 9, if all of the sustainer electrodes can be economically integrated onto a single silicon chip, then the complete sustainer can be packaged into a single case with one heat sink.

With reference to FIG. 10, there is illustrated an integrated, power efficient sustain driver circuit according to the invention that does not require resistors or capacitors. In the circuit of FIG. 10, T1 and T2 are driven directly by the Level Shifter, T3 is driven directly from the CMOS Driver Dr1, and T4 is driven directly from the CMOS driver Dr2. If Css1, Css2 and the inductor are excluded from integration, then the integrated circuit is made up entirely of active components. Thus, the silicon area required is minimized.

The operation of this circuit is basically the same as the circuit of FIG. 9. As before, T1 and T2 charge and discharge Cp via L, and T3 and T4 clamp Vp at Vcc and ground, respectively. The difference is in the gate drive circuits Dr1, Dr2, and the Level Shifter, and in the addition of Css1.

Css1 and Css2 form a voltage divider where Css1=Css2. Thus, at power up, when Vcc begins to rise, Vss will rise at Vcc/2. Later, when Vss has risen above the threshold level of the MOSFETs, then Vss will be held at Vcc/2.

The Level Shifter is a set-reset latch, with its output at either Vcc or ground. When Vi switches "HIGH", the output of the Level Shifter drops to ground and forces -Vss across the gate to source of both T1 and T2. This turns T1 "ON" and T2 "OFF". The input to Dr2 is then forced to Vss, the output of Dr2 drops to ground, and T4 is turned "OFF". Later, when I.sub.L falls to zero and then reverses, the input to Dr1 rises from Vss to Vcc, the gate of T3 is then pulled down by Dr1 to Vss, and T3 turns "ON". Thus, Vp is driven to Vcc when Vi switches "HIGH".

When Vi switches "LOW", the output of the Level Shifter rises to Vcc and forces Vss across the gate to source of both T1 and T2. This turns T1 "OFF" and T2 "ON". The input to Dr1 is then forced to Vss, the output of Dr1 rises to Vcc and T3 is turned "OFF". Later when I.sub.L falls to zero and then reverses, the input to Dr2 falls from Vss to ground. The gate of T4 is then driven up by Dr2 to Vss, and T4 turns "ON".

The XAP and YAP address pulse generators may also be designed with the energy recovery technique previously described in connection with the sustain driver circuit. As an example, reference may be made to FIGS. 11-14. FIG. 11 illustrates an XAP address pulse generator connected to the panel electrodes at the output terminal. FIG. 12 illustrates the output voltage and inductor current waveforms (similar to FIGS. 5 and 6 with respect to the sustain driver) as switches S1 and S4 are opened and closed through the switching states. The output voltage waveform in FIG. 12 is a positive double pulse conforming to the desired XAP waveforms of FIGS. 3 and 4. Notice that switch S2 of FIG. 5 has been eliminated in the XAP generator of FIG. 11 since diode D3 diode D2 and S2 in FIGS. 5 and 6.

FIG. 13 illustrates YAP generator and FIG. 14 illustrates the corresponding waveforms in the switching states. Capacitor C.sub.D and the output capacitance connected to the output terminal function as a voltage divider of voltage Vcc supplied to the circuit. When a Write Pulse is required (See FIG. 14), switch S5 is closed to short capacitor C.sub.D to provide the full amplitude Write Pulse to the panel. If an Erase Pulse is required, switch S5 is opened to provide the reduced amplitude Erase Pulse to the panel.

If desired, an ISA panel can be provided with N-channel MOSFET address drivers on one axis and P-channel MOSFET address drivers on the other axis, using techniques similar to the YAP and XAP address driver circuit techniques previously described. For example, a YAP address pulse generator with an N-channel MOSFET driver could be used with negative pulse similar to the negative pulses of the YAP pulses in FIG. 3. For the XAP address pulse generator a P-channel MOSFET driver could be used with a positive going single pulse having a pulse width equal to the width between the two double XAP pulses shown in the expanded view of FIG. 4.

The foregoing detailed description has ben given for clearness of understanding only, and no unnecessary limitations should be understood therefrom, as modifications will be obvious to those skilled in the art.