Driver Architectures of Current Displays and their Limitations

Introduction

The term “Active Matrix” is fifty years old this year. The term was coined by T. Peter Brody in 1975 (The 60th Anniversary of TFTs and the Evolution of High‐Resolution Displays) who was part of the team developing flat panel technology at Westinghouse Research Labs. Active matrix displays based on thin-film transistor (TFT) arrays now dominate the display market. Variants of TFT based displays account for all displays used in consumer electronics and as monitors in professional applications. 

Active Matrix displays are defined by having the state of the pixel maintained locally, at the pixel. This is a core capability of modern flat panel displays and it has contributed to the steady improvement in the density and quality of consumer displays. And for our purposes all of these displays are now built on top of a TFT backplane. This ubiquity is driven by the complexity of display architecture when produced at scale and by the industrial maturity of the TFT display stack and the level of development time for the integrated components at the display controller and timing controller level. The monolithic application specific integrate d circuits (ASIC) require long development timelines and all of these components are used in very high volumes. For these reasons successive waves of display technology have adopted the existing TFT stack rather than creating novel driver architectures that may have been more optimal for those displays. Adapting to TFT is simply more efficient and more cost effective than building something new at scale. 

There are alternative architectures that have been proposed for active matrix displays including Active Addressing, Dual Scan Super-Twisted Nematic, and other driver topologies and research is still ongoing. There are also Complementary Metal Oxide Semiconductor (CMOS) based backplanes that are used for smaller displays such as the imagers for head mounted displays. CMOS based active matrix displays tend to be small as the process (and cost and yields) has not been scaled to support larger displays. 

The subject of TFT backplanes is interesting because of the specific requirements of the next generation microLED displays. Existing passive matrix LED displays are driven by current and controlled via pulse width modulation (PWM) pixels. PWM is also used in LED lighting and in display backlights for liquid crystal displays, and it is notably absent from the pixel driving mechanisms of large format active matrix displays such as LCD and OLED panels. This document examines the material limitations, technical challenges, and practical considerations that have prevented PWM implementation in existing active matrix display technologies.

Material Limitations in Display Technologies

Liquid Crystal Response Time
Liquid crystal materials exhibit inherently slow response times, typically ranging from 1-10 milliseconds in modern displays. PWM typically requires switching speeds several orders of magnitude faster than the content refresh rate to provide adequate grayscale depth. At a 60Hz display refresh rate with 8-bit color depth (256 levels), PWM would require liquid crystal switching capabilities in the microsecond range, far beyond the physical capabilities of even the fastest LC materials available today.

Liquid crystal displays create an image by controlling a light source passing through an array of openings. The liquid crystal itself is a type of light valve with each pixel in the array moving from a 0 degree state to a 90 degree state when it needs to turn on (note that these degrees are just relative to the display in one orientation). The polarization of the light source into the display is set by a polarizer at the back of the display and this limits the output to just the pixels where the polarization aligns with the front of the display. But it takes the liquid crystal time to change between the 0 degree and 90 degree states. 

The relatively slow response times of liquid crystal have been responsible for several branches of the active matrix display architecture that support dithering, and other methods of compensating the relatively slow switching speed of liquid crystal. This has also led to GPU based compensation systems such as variable refresh rates that compensate for the slow switching speeds of liquid crystal. 

OLED Material Degradation
OLED materials suffer from accelerated aging when subjected to frequent switching cycles. PWM implementation would dramatically increase the number of on-off cycles compared to constant-current approaches:

• At 8-bit resolution (256 brightness levels), PWM would require up to 255 on-off transitions per refresh cycle
• For a 60Hz display, this results in over 15,000 switching cycles per second

Technical Challenges of High-Frequency PWM in Large Arrays

Power Consumption
PWM’s rapid switching inherently consumes more power than constant-voltage or constant-current approaches in LCD and OLED:

• Each transition incurs capacitive charging and discharging losses in the TFT gates and data lines
• In a Full HD (1920×1080) display, PWM would require charging and discharging over 6 million capacitive nodes (subpixels) thousands of times per second
• PWM’s rapid switching would likely increase power consumption compared to constant-voltage approaches due to the energy required to repeatedly charge and discharge the parasitic capacitances in the TFT backplane and signal lines. The exact magnitude of this increase would depend on multiple factors including switching frequency, display resolution, and specific TFT technology used.

Signal Integrity and Crosstalk
High-frequency PWM signals face significant integrity challenges when distributed across large matrices:

• Display panels have high parasitic capacitance between adjacent signal lines
• As PWM frequency increases, capacitive coupling between adjacent data lines increases proportionally
• The thin-film structure creates significant distributed RC delays across the panel
• These effects combine to create signal crosstalk, ghosting, and timing inconsistencies across the display
• You should not try to make high resolution active matrix displays at home in your kitchen

Perceptual and Quality Issues

Perceivable Flicker
PWM can introduce perceivable flicker, particularly at lower frequencies:

• Achieving flicker-free operation would require PWM frequencies in the kilohertz range which is well beyond the range of LCD panels.

Current Solutions and Alternatives
Instead of PWM, modern active matrix displays use several alternative approaches:

1. Analog voltage control: LCDs primarily use variable voltage levels to control the twist of liquid crystals, providing different levels of light transmission
2. Current control: AMOLEDs use variable current levels to control the brightness of each OLED pixel
3. Frame Rate Control (FRC): Additional gray levels are created by alternating between two brightness levels over multiple frames (a sort of temporal dithering)
4. Spatial dithering: Adjacent pixels use slightly different brightness levels to create the perception of intermediate levels

LED Display Architecture

A LED screen made in 2025 will use a driver from a company such as Macroblock, Chip-one, or XMplus. These drivers manage a small area of pixels that is part of a single module in a larger LED display. The driver will incorporate a number of functions and components including a serial interface, an oscillator, a clock, counters for a PWM generator, and a number of Metal Oxide Semiconductor Field Effect Transistor (MOSFET) in order to  be capable of switching an X, Y grid of pixels. The driver scans through this array on one axis with the MOSFETS controlling the other axis. The PWN generated in the IC is distributed and available to the LED only when the driver is addressing that LED. There is no local source at the pixel and this is a major difference between a typical LED screen and a typical active matrix display. In an active matrix display the state of the pixel is controlled locally by the TFT whereas in an passive matrix LED display access to the PWM circuit is turned off when a pixel is not being directly addressed. 

In this sense while both systems scan across a grid there are a number of important differences. In a passive matrix LED display the drivers scan in small rectangles (a driver group) within the larger finished module, which in turn is part of a larger display. These groups are often single color so there will be a red, green, and blue driver IC. The drivers in a large LED display may scan in sync through use of a timing source or they may scan randomly. For this reason the number of segments of LED that are scanned by each driver is of great interest to a high end user. This is the Scan Multiplexer number and the higher this number the lower the percentage of time any one segment of the LED display will be illuminated. This means that each LED must be driven at a brightness level for each pixel where it will be at the appropriate brightness for that moment in a video. 

In an active matrix display a pixel may hold local memory allowing the pixel ON time for that moment of video to be divided across the video frame in such a way that the ratio of on time to off time can be managed in real time in a manner that is optimal for a specific piece of video. In such a display the timing controller and the display controller will generally refresh the whole display by scanning through the rows. The Gate ICs at the side of the display will activate all the TFTs in a row while the Source ICs along the bottom or top of the display (in landscape) update the information in the active pixel in each column. The Gate ICs and the Source ICs are synchronized so that the scan of the display smoothly updates the data held locally in the TFTs that are driving each pixel. 

Conclusion

We are still early in the efforts to adapt existing TFT production designs to the requirements of new microLED displays. While PWM offers elegant solutions for many electronic control applications, the material properties of display elements, physical limitations of TFT backplanes, and practical engineering constraints of large matrix arrays make it unsuitable for pixel-level control in active matrix displays. The industry has instead developed specialized analog driving schemes optimized for the unique characteristics of each display technology. Unless there are fundamental breakthroughs in materials science or novel architectures for display driving, PWM will likely remain confined to backlight control rather than becoming part of the pixel-level driving mechanism in active matrix displays.

Technical Notes
While pure PWM isn’t used for pixel-level control in TFT displays, some pulse-based techniques are employed in specific implementations:

• Pulse Amplitude Modulation (PAM):
  • Used in some specialized LCD driving schemes where voltage pulses of varying amplitudes are applied during the addressing phase
  • Often used in automotive displays and industrial panels that need to operate in extreme environments
  • Helps reduce power consumption while maintaining acceptable response times
• Modified Pulse Forms:
  • “Dynamic Capacitance Compensation” techniques use shaped voltage pulses to overcome the capacitive effects in LC cells
  • These aren’t traditional PWM but use carefully timed and shaped pulses to improve response time
• Hybrid Approaches:
  • Some high-end displays use “Impulse Driving” methods that combine short voltage pulses with hold periods
  • This reduces motion blur by creating a display response more similar to CRTs
  • Sony’s “Motionflow Impulse” and similar technologies use this approach
• Row/Column Addressing Pulses:
  • The fundamental addressing of TFT arrays does use pulses to select rows and columns
  • These are more digital select/enable signals rather than modulation techniques
• Temporal Dithering Techniques:
  • Frame Rate Control (FRC) uses a form of pulse position modulation spread across multiple frames
  • This creates intermediate brightness levels on displays with limited native bit depth

What’s notably absent is pure pulse width modulation at the pixel level, primarily due to the material and technical limitations I described in the above document. The techniques that do exist tend to be hybrid approaches or specialized implementations that work around the fundamental limitations of LC materials and TFT arrays.​​​​​​​​​​​​​​​​ PWM is not used as part of the pixel driver in any of these approaches but rather as a compensation for some component in the pixel driving architecture.