BACKGROUND
1. Technical Field
This disclosure relates generally to displays.
2. Description of Related Art
Displays have become an important part of modern society. They are used in a wide range of devices such as TVs, smartphones, tablets, laptops, digital signage, and augmented reality and virtual reality devices. They are also used for many different applications, including entertainment, communication, education, and work. Displays can provide high-quality visual information, and they come in different sizes, resolutions, and formats to address different needs and preferences.
In particular there is high demand for small, bright and efficient displays, for example as can be used in small mobile devices. While there is also a high demand for large displays, such as those used in large size TVs and computer monitors, the design considerations for small displays are different than those for large displays. Hence there is a need for new and improved display technologies.
BRIEF DESCRIPTION OF THE DRAWINGS
Embodiments of the disclosure have other advantages and features which will be more readily apparent from the following detailed description and the appended claims, when taken in conjunction with the examples in the accompanying drawings, in which:
FIG. 1 is a block diagram of a multi-bit, memory-in-pixel-based, pulse width modulation (PWM) pixel driver.
FIGS. 2A and 2B are timing diagrams of operation of the PWM generator circuit of FIG. 1.
FIG. 3 is a circuit diagram of a four-bit implementation of a PWM generator circuit.
FIGS. 4A and 4B are timing diagrams of operation of the PWM generator circuit of FIG. 3.
FIG. 5A is a timing diagram that shows time reserved for programming memory cells.
FIG. 5B is a circuit diagram illustrating the addition of a reset switch to the PWM generator circuit of FIG. 3.
FIG. 6A is a circuit diagram of another four-bit implementation of a PWM generator circuit.
FIG. 6B are timing diagrams of operation of the PWM generator circuit of FIG. 6A.
FIG. 6C is a circuit diagram of another four-bit implementation of a PWM generator circuit plus local memory, using standard memory cells.
FIGS. 7A and 7B are a cross-sectional side view and an exploded perspective view of a display with an array of pixels.
FIG. 8 is a plan view of a backplane with an array of pixel driver circuits.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
The figures and the following description relate to preferred embodiments by way of illustration only. It should be noted that from the following discussion, alternative embodiments of the structures and methods disclosed herein will be readily recognized as viable alternatives that may be employed without departing from the principles of what is claimed.
A micro-LED display includes a light emission layer and a backplane layer. The light emission layer may be an array of LEDs. For example, the LEDs may use GaN quantum well active layers, which are fabricated on a GaN-on-sapphire or GaN-on-Si substrate. Other materials systems may also be used. Organic LEDs (OLEDs) may also be used. The backplane may be fabricated on a Si CMOS wafer. The two wafers may then be bonded together so that CMOS circuitry provides drive current to the LEDs. Micro-LED displays may also be constructed using techniques other than wafer-to-wafer bonding.
Generally the term “micro-LED” means that the size of a display pixel is very small. The pitch, or distance from the edge of one pixel to the corresponding edge of its nearest neighbor is also very small. For example, micro-LEDs may be as small as about 4 μm or less, and also laid out with a pitch of about 4 μm or less. When frontplane and backplane wafers are combined to form pixels, the area of the CMOS driver circuitry preferably does not exceed the area occupied by the LEDs. The CMOS driver circuitry may be positioned below the LEDs.
Within the backplane, pixel driver circuits supply and control current to LEDs of individual display pixels to vary their brightness. The current may be varied via pulse amplitude modulation (PAM), pulse width modulation (PWM) or a combination of the two techniques. In some implementations PAM is used to control the brightness of the display as a whole, while PWM is used to control the brightness of individual pixels.
PWM may be implemented in CMOS digital circuits in a small area. CMOS technology nodes are now so small that each display pixel (or cluster of pixels) may have its own local memory (such as SRAM), and that enables each pixel to have its own PWM generator circuit. However, the PWM generator circuit cannot take up too much area, especially when the pixel size is about 3 or 4 μm or less. The available area may be further reduced if separate generator circuits are needed for different subpixels (e.g., red, green and blue subpixels) within each pixel.
As described herein, a PWM generator circuit may be made more compact (i.e., occupy less area) by taking advantage of certain properties of the pulse width modulation for this particular application. In PWM, the temporal width of the drive signal is modulated to be longer or shorter, depending on the desired brightness for the LED. This can be implemented by using a set of N clock pulses CKn, n=1 to N, of different widths and a set of corresponding N control bits Bn for each pixel or subpixel that determine which clock pulses are activated for that pixel. The total pulse width is the sum of the pulse widths of the activated clock pulses.
The PWM generator circuit for each pixel may be designed as a modified multiplexer. In a conventional multiplexer, control bits are the select inputs for the multiplexer. Here, however, control bits Bn are the signal inputs to the multiplexer and clock signals CKn are the select inputs to the multiplexer. The multiplexer may be simplified because the clock is designed such that not more than one clock signal CKn is asserted at any time. The modified circuit does not need the ability to resolve contentions among select inputs as would be the case in a conventional multiplexer. The PWM generator may be implemented using transistors whose source or drain is connected to the control bit Bn and gate is connected to the clock signal CKn. This design reduces the area occupied by the circuit.
FIG. 1 is a block diagram of a multi-bit, memory-in-pixel-based, PWM pixel driver. In the example of FIG. 1, CKn are non-overlapping clock signals CKn for n=1 to N. The memory 110 stores the N corresponding control bits Bn, each of which determines whether the corresponding clock signal is activated. The memory 110 may be SRAM, or other kinds of memory such as magnetic RAM, resistive RAM, or dynamic RAM. If control bit Bn is asserted, then the corresponding pulse CKn is activated and adds to the aggregate pulse width of the drive signal DRV. The PWM generator circuit 120 receives the control bits Bn and clock signals CKn, and generates the pulse width modulated drive signal DRV. The drive signal DRV controls the LED current, ILED, and therefore the brightness of the LED. Of course, the complementary circuit may also be implemented where the drive signal that controls the LED current is DRV_bar, where the_bar indicates the complement.
From a Boolean operation point of view, the generator circuit 120 performs a logical AND operation between each clock signal CKn and its corresponding control bit Bn to produce individual component pulses CPn, followed by a logical OR operation between these component pulses CPn:
CPn=CKnAND BN DRV=CP1 OR CP2 OR . . . CPN (1)
FIGS. 2A and 2B are timing diagrams of operation of the PWM generator circuit for bit sequences of Bn=1111 and 1010, respectively, where n=1 is the most significant bit. The corresponding clock signals have pulse widths that increase by powers of 2. If clock signal CK4 has a pulse width of T, then CK3 has a pulse width of 2T, CK2 has a pulse width of 4T, etc. The clock signals are binary weighted. Each clock pulse is twice as long as the clock pulse selected by the next most significant bit. The pulse widths of the clock signals CKn increase by powers of 2 for different values of n. However, binary weighting is not required. Other weightings may be used, including equal width pulses. In this example, the clock pulses are adjacent to each other but non-overlapping, and the time window occupied by the clock signals is referred to as the pulse window 210.
FIG. 2A shows a timing diagram for Bn=1111. The top four timing diagrams show the clock signals CKn as previously described. Because all four bits Bn are asserted, all four clock pulses CK1-4 are active resulting in the component pulses CP1-4 shown in the timing diagrams below the clock signals. The resulting drive signal DRV is the concatenation of these component pulses CP1-4, as shown in the bottom timing diagram. In FIG. 2A, the DRV signal has an aggregate pulse width of 15T.
FIG. 2B shows a timing diagram for Bn=1010. The timing diagrams for the clock signals CKn are omitted since they are the same as in FIG. 2A, but their pulse widths are indicated by the dashed vertical lines. In this figure, only bits B1 and B3 are asserted, so only component pulses CP1 and CP3 are activated, and the other two clock pulses are not. The resulting drive signal DRV has an aggregate pulse width of 8T+2T=10T. T is selected to be small enough that the gaps between pulses are not observable by humans.
FIG. 3 is a circuit diagram of a four-bit example implementation of a PWM generator circuit that implements the timing diagrams shown in FIG. 2. The generator circuit in FIG. 3 has N input nodes that are coupled to receive the N control bits Bn from the local memory. The PWM generator circuit also receives the clock signals CKn and generates the drive signal DRV from the incoming signals Bn and CKn. In this design, the generator circuit has N input arms, one between each of the N input nodes Bn and the output node DRV. Each arm n contains a switch that couples the corresponding input node Bn to the output node DRV, as controlled by the clock signal CKn. In this particular implementation, the switches are implemented as two-transistor pass gates PGn. Each pass gate includes two opposite polarity transistors (NMOS and PMOS), which are controlled by the clock signal CKn and its complement respectively. For both transistors, the source and drain are connected between one of the input nodes and the output node. The gates are controlled by signals based on the clock signal CKn. When the clock signal CKn is asserted, both transistors turn on and signal Bn is passed to the output node DRV.
The outputs of the pass gates PGn are the component pulses CPn, which are combined to form the output signal DRV, which drives the LED. Because only one clock pulse CKn will be asserted at any time, this means that only one pass gate PGn will be open at any time, which means that only one component pulse CPn will be active at any time. As a result, the component pulses CPn may be combined simply by connecting all of them to the output node DRV. No additional circuitry is needed.
The PWM generator circuit functions as a multiplexer implemented using pass gates and without protection against contention among the select signals CKn. A conventional multiplexer would be implemented using Boolean logic gates, but the simpler pass gates are used instead in FIG. 3, which reduces the transistor count and the area of the circuit. In addition, a conventional multiplexer would include circuitry that handles conflict situations when more than one clock signal CKn is asserted. The generator circuit of FIG. 3 does not contain this protection, because it is unnecessary since the clock signals CKn are not overlapping by design. This further reduces the transistor count and area.
An N-bit PWM generator circuit that uses logic gates to implement Eqn. 1 would require at least 6N transistors: 4 transistors for a NAND gate to construct CPn and 2 transistors for the OR operation, per bit. It could be even more, since the OR operation may require multiple gates and stages if N is large. For comparison, the circuit in FIG. 3 uses only 2N transistors. In addition, the design in FIG. 3 may be extended straightforwardly for larger values of N by adding additional input arms while keeping the transistor count to 2N.
FIGS. 4A and 4B are timing diagrams of operation of the PWM generator circuit of FIG. 3, for bit sequences of Bn=1111 and 1010 respectively. These are analogous to the timing diagrams of FIGS. 2A and 2B. The pulse widths of the clock signals CKn are indicated by the dashed vertical lines. As indicated by the notation, the activation of clock pulse CK1 opens the corresponding pass gate PG1 while the other pass gates remain closed (because the other clock pulses are not asserted). When pass gate PG1 is open, it passes the value of B1=1, producing the component pulse CP1. The control bit B1 is stored in SRAM. Opening the pass gate PG1 connects the SRAM memory cell to the drive output DRV. The same is true for the other clock signals CKn and their corresponding pass gates PGn and component pulses CPn. The resulting DRV signal is shown at the bottom of FIG. 4A.
FIG. 4B operates the same way as FIG. 4A, but for a different bit sequence of Bn=1010. When clock pulse CK2 is activated, pass gate PG2 opens and passes the control bit B2=0. The same is true for bit B4=0. This results in the drive signal DRV with shorter aggregate pulse width shown at the bottom of FIG. 4B.
FIG. 5A is an example timing diagram that shows time reserved for programming the memory cells during every frame. The duration of the clock signals CKn defines a pulse window 510 when the clock signals may be activated. Dead time 520 between these windows may be used to program the SRAM memory with bit sequences Bn. During this programming window 520, the clock signals CKn are not activated. The DRV output is undriven, leaving it in its most recent state. If this is not desirable, it may be corrected by the addition of a reset mechanism, as illustrated in FIG. 5B.
FIG. 5B is a circuit diagram illustrating the addition of a reset switch 530 to the PWM generator circuit of FIG. 3. In FIG. 5B, the reset switch 530 is a transistor driven by the RST signal. The RST signal is asserted during the programming window 520. This pulls DRV to zero, or logical low, when no pass gates PGn are turned on. Even with the addition of the RST control, the N-bit PWM generator circuit can be implemented using as few as (2N+1) transistors.
The examples of FIGS. 3-5 use pass gates as switches. Pass gates use both NMOS and PMOS transistors and also use both polarities of the CK signals to operate properly. FIG. 6A is a circuit diagram of a four-bit example implementation of a PWM generator circuit that does not use both polarities of transistors and clock signals.
In FIG. 6A, the switches in each input arm are implemented as a single transistor Tn controlled by the corresponding clock signal CKn. The source and drain of each transistor are connected between one the input nodes and the output node DRV. The gate of each transistor is connected to CKn_bar. The generator circuit also includes a RST transistor 630. FIG. 6B is the corresponding timing diagram, which introduces dead time 635 between the clock pulses CKn. The reset signal RST is asserted during these dead times. In this version, the N-bit PWM generator circuit uses as few as (N+1) transistors for N-bit PWM current control. The clock signals CKn are inverted. Each of the switches is implemented with a single MOS transistor Tn (PMOS in this example) and the circuit uses only a single polarity of the clock signals.
PMOS transistors are generally capable of pulling a signal line high, but not low. The extra reset pulses in FIG. 6B during the dead times 635 ensure that DRV is pulled low before a rising edge of the next clock signal CKn. The use of the reset device and reset pulses in the dead time, allows the use of single-transistor gates with single-polarity input. Slightly increasing the complexity of the reset signal RST allows the elimination of almost half of the transistors shown in FIG. 5B. The decrease from (2N+1) to (N+1) transistors saves valuable area on a semiconductor die, allowing the circuit (and the associated memory) to fit under (and within the same area as) a very small pixel.
The complementary circuit may also be implemented, using NMOS transistors instead of PMOS transistors. In that case, the gates are connected to CKn and the reset transistor is connected to the other supply node Vdd rather than ground.
FIG. 6C shows an example using NMOS transistors and using a standard memory cell from a cell library. In this example, the standard memory cell is an 8T (eight transistor) SRAM cell. These cells have been optimized and are provided by the fab. In FIG. 6C, the extent of the memory cells 640 is indicated by the dotted lines. The top memory cell is annotated and described below. The other memory cells are analogous. Comparing to FIG. 6A, the memory cell 640 includes both the local memory 642 storing bit B1 (which is not shown in FIG. 6A) and the switch/transistor Ti.
The memory cell 640 from the cell library includes ports or nodes as indicated by the solid circles. din (data in) and wrt (write) and their complements are used to program bit B1 into the SRAM cell 642, which consists of four cross-coupled transistors. These may be used during a programming window (such as window 520 in FIG. 5A) to update the value of the stored bit B1. rd (read) is a read control node that is used to control read out of the stored data. When rd is asserted, bit B1 is available at the read port dout. In FIG. 6C, the read control node rd is connected to the clock signal CK1. To complete the circuitry, the outputs dout of the four memory cells 640 are connected to the drive node DRV, and a reset transistor is also added.
The above examples use transistors to directly connect bits of local memory to drive nodes for LEDs. They occupy less area than other implementations because, for example, they do not contain any Boolean logic gates. In addition, they assume that only one clock signal CKn is asserted at any time and do not contain circuitry capable of resolving conflicts if clock signals are concurrently asserted.
FIGS. 1-6 show PWM generator circuits for individual pixels. A display includes an array of pixels. FIGS. 7A and 7B are a cross-sectional side view and an exploded perspective view of a display with an array of pixels 750. For clarity, FIG. 7B shows the footprints of various components, but without any thickness. The top plane is the LED frontplane, and the bottom plane is the CMOS driver backplane.
FIG. 7 shows a 2×2 array of color pixels, as indicated by the dashed lines. In this example, the display is a color display and each color pixel includes a 2×2 array of micro-LEDs 760 that provide red, green and blue subpixels. Each subpixel includes one or more separately controllable micro-LEDs. This example uses a Bayer pattern of one red (R), one blue (B), and two green (G) subpixels per color pixel. This could be implemented by using different color micro-LEDs, same color micro-LEDs with color conversion materials, or broadband micro-LEDs with color filters. Other patterns may also be used, such as RRGB if the red micro-LEDs are less efficient.
The pixel driver 700 for each pixel includes three of the PWM generator circuits 720 described above and also the corresponding local memory 710. One PWM generator circuit drives the LED for the red subpixel, one drives the LED for the blue subpixel and one drives the two LEDs for the green subpixel. The backplane area for each pixel is divided into quadrants, with the three PWM generator circuits 720 located in three of the quadrants and the local memory 710 for all three generator circuits located in the fourth quadrant. With this arrangement, the pixel driver circuitry does not occupy more area than the micro-LEDs. In this particular example, the PWM generator circuits 720 and local memory 710 for each pixel are positioned under the micro-LEDs for that pixel.
The arrangement shown in FIG. 7 is just one example. Other arrangements will be apparent. For example, the micro-LEDs may be arranged as stripes, in hexagonal arrays or other geometries. There may also be different combinations of different colors or other different types of LEDs. Different arrangements of micro-LEDs to color subpixels may also be used. The pixel generator circuits also may be arranged as stripes or in other geometries. The circuits for different subpixels and pixels may not be physically segregated as shown in FIG. 7. The different instances of the circuits may also be rotated or flipped in orientation. These variations may be used to facilitate routing of circuit interconnects or sharing of common components. For example, the backplane circuitry for each individual pixel may not be the same shape as the area occupied by the light emission layer. The arrangement of backplane and light emission layer may use multi-pixel building blocks, for example repeating 2×2 or larger arrays of pixels.
FIG. 8 is a plan view of a backplane with an array of pixel driver circuits. The pixel driver circuits 800 receive clock signals CKn and reset signal RST as inputs. They also contain local memories which are programmable to store the control bits Bn for each pixel.
In FIG. 8, the pixel drivers 800 are shown as squares and FIG. 8 shows an 8×8 array of these squares. The pixel drivers 800 may all receive the same clock signals CKn and reset signal RST. Accordingly, the backplane includes a clock source 842 and a RST source 844 that are located outside the array area 802 occupied by the pixel drivers. These sources generate the clock signals CKn and reset signal RST, which are then distributed to the pixel drivers by some signal distribution network. In FIG. 8, the distribution tree has a main vertical branch that runs along every fourth column and then smaller horizontal branches that service the two pixel drivers to either side of the main branch.
Note that the clock signals CKn are not the clock signals that are used to synchronize the operation of digital logic circuits. Rather, these clock signals CKn are used to determine the aggregate pulse width of the driver signals for LEDs. As such, many of the timing requirements on synchronous clock signals do not apply to these PWM clock signals. For example, the pulse widths of clock signals CKn may be on the order of the frame rate, which is much slower and provides a much larger timing tolerance. In addition, because the different pixels do not have to operate synchronously with each other, the clock signals CKn may arrive at different times at different pixel drivers 800. Thus, the designer has more freedom in designing the clock and reset distribution networks.
The pixel drivers 800 also include local memories that store the bit sequences Bn. These may be programmed using row and column addressing, with the memory controller also located outside the array area 802 for the pixel drivers. These local memories may be shared between pixel drivers. For example, there may be a local memory that stores the control bits for a group of pixel drivers.
Although the detailed description contains many specifics, these should not be construed as limiting the scope of the invention but merely as illustrating different examples. It should be appreciated that the scope of the disclosure includes other embodiments not discussed in detail above. Various other modifications, changes and variations which will be apparent to those skilled in the art may be made in the arrangement, operation and details of the method and apparatus disclosed herein without departing from the spirit and scope as defined in the appended claims. Therefore, the scope of the invention should be determined by the appended claims and their legal equivalents.
Patent Application
20240249665
