Introduction: Why “Inspection” is Not Enough; “Compensation” is Also Needed#
In the LCD era, the ultimate goal of Mura inspection was to judge a panel as pass or fail—defective products entered rework or scrap processes. However, as the display industry entered the OLED and MicroLED eras, this “inspect and judge” logic encountered a fundamental challenge.
In OLED panels, each pixel is independently self-luminescent. Tiny fluctuations in organic material evaporation, TFT backplane performance, and encapsulation processes inevitably lead to luminance and chromaticity differences between pixels. For MicroLEDs, the precision limits of mass transfer, inherent performance dispersion of micron-scale chips, and splicing gaps make it physically almost impossible to completely eliminate pixel-level inconsistencies. If a “inspect and scrap” strategy were relied upon alone, yield and cost would not meet commercialization requirements.
The emergence of Demura (literally “removing Mura”) technology marks a significant evolution in display inspection functionality: a shift from “inspection and judgment” (Pass/Fail) to “measurement and correction” (Measure and Correct). The inspection system is no longer just a gatekeeper for quality but has become an active participant in improving yield and enhancing product performance.
Fundamental Causes of Luminance Non-Uniformity in OLED and MicroLED#
Sources of Non-Uniformity in OLED#
The causes of Mura in OLEDs are fundamentally different from those in LCDs, concentrating on pixel-level inconsistencies and manufacturing process uniformity control:
Differences in Pixel Emitting Units. Each subpixel in an OLED emits light independently. Minor deviations in processes such as organic material evaporation and thin-film encapsulation lead to differences in actual luminance and chromaticity when different subpixels receive the same electrical signal. This pixel-to-pixel inconsistency is the primary source of OLED Mura.
Non-uniform Material Deposition. Controlling the deposition thickness of organic light-emitting material layers and charge transport layers is critical. Deformations or alignment deviations in the Fine Metal Mask (FMM) used in evaporation cause regional fluctuations in material thickness, forming line or area Mura.
TFT Backplane Non-uniformity. Performance differences in the LTPS or Oxide TFT backplanes driving OLED pixels affect the precise control of current to each pixel, causing spatial distribution deviations in luminance.
Non-uniform Aging. Light-emitting materials of different colors in OLEDs age at different rates (blue material decays fastest). Long-term use further exacerbates performance differences between pixels, manifesting as color temperature drift and worsening uniformity.
Sources of Non-Uniformity in MicroLED#
The uniformity challenges for MicroLED are even more severe:
Inherent Chip Differences. Even at the epitaxial growth and chip manufacturing stages, characteristics like luminance, wavelength, and voltage of individual MicroLED chips exhibit inherent dispersion. Due to the extremely small chip size (potentially less than 100 microns), traditional LED binning methods are difficult to implement effectively.
Mass Transfer Precision. Accurately transferring millions to tens of millions of micron-scale chips from wafers to driving backplanes requires precision within ±1.5 microns or better. Any position offset, rotation, or damage results in pixel defects or luminance anomalies.
Splicing Mura. Large-size MicroLED displays are usually spliced from multiple modules. Alignment precision between modules and control of luminance and color consistency for edge pixels constitute independent uniformity challenges.
Basic Principles of Demura#
The core idea of Demura is straightforward: rather than physically repairing defects, it involves accurately measuring the actual output of each pixel, calculating compensation coefficients, and adjusting driving electrical signals so that the final light output of each pixel reaches the target value.
Specifically, Demura follows two technical paths:
Electrical Compensation#
Compensation is performed by obtaining the actual I/V characteristics of the driving TFT or OLED and identifying the difference between actual and target values. This method requires coordination at the levels of panel design, IC architecture, driving circuitry, and algorithms, making its implementation complex.
Optical Compensation (External Demura)#
This is currently the more widely adopted method. Its principle is more direct: using high-resolution imaging equipment to record the actual luminance of each pixel and calculating compensation values based on the measurement results. Optical compensation does not rely on an understanding of internal OLED/TFT characteristics; instead, it targets the final optical output for correction, giving it better universality.
The essence of a Demura algorithm can be simplified as: dimming pixels that are too bright and brightening pixels that are too dim (or eliminating color shifts) so that the luminance and color of various regions on the panel tend toward consistency under pure-color frames.
Process of Generating Correction Maps using Imaging Colorimeters#
The core step of optical Demura is using an imaging colorimeter to generate an accurate Correction Map. This process involves several key steps:
Acquisition Phase#
The AMOLED panel is lit to sequentially display frames of different gray levels. A typical acquisition scheme covers low, medium, and high gray-level ranges—for example, capturing R, G, and B images at gray levels 32, 64, 96, 160, 192, and 224, totaling 18 images. The purpose of this multi-gray-level acquisition strategy is to obtain the Gamma characteristics of pixels at different brightness levels, as the degree of pixel non-uniformity typically varies with brightness level.
The technical requirements for the imaging equipment are extremely strict:
- Spatial Resolution: Must be able to resolve individual subpixels; microscope lenses may be needed for high-resolution panels.
- Frame Rate: Production line applications require 15 fps or higher.
- SNR: Sufficient Signal-to-Noise Ratio to distinguish faint pixel-level differences from random noise.
- Chromaticity Accuracy: Accurate extraction of chromaticity information for each subpixel.
Data Processing Phase#
The captured images undergo multiple processing steps:
Flat-Field Correction (FFC). Eliminating non-uniformity inherent to the imaging equipment itself. Without FFC, the vignetting effect of the camera lens (luminance decay at the edges) would be mistaken for panel Mura, leading to incorrect compensation data.
Moire Removal. Frequency interference between the camera sensor pixel array and the display pixel array generates Moire patterns, which must be eliminated via algorithms.
Distortion Correction. Geometric distortion introduced by the optical system must be corrected using calibration parameters to ensure precise correspondence between image pixels and panel pixels.
Pixel Mapping. Precisely mapping each region in the camera image to each subpixel on the panel, establishing a one-to-one correspondence.
Mura Identification and Compensation Calculation#
Once data processing is complete, the core algorithmic steps follow:
- Establish Target Value: Calculate the average luminance/chromaticity of the entire panel or a designated area as the target.
- Deviation Calculation: Compare the measured value of each subpixel with the target value to calculate deviation.
- Threshold Judgment: Subpixels with absolute deviation exceeding a preset threshold are marked as Mura regions needing compensation.
- Compensation Coefficient Generation: Based on the direction and magnitude of deviation, calculate the driving signal adjustment needed to make the subpixel output reach the target value.
- Gamma Mapping: Map the compensation coefficients onto the panel’s actual Gamma curve, transforming them into digital values that can be written to the driver IC.
For frequency-domain methods, Fourier transforms can be used to decompose images into different frequency components. Mura boundaries are identified by selecting specific frequency bands—high-frequency components correspond to pixel-level details, and low-frequency components correspond to large-area uniformity deviations.
Data Compression#
The raw pixel-by-pixel compensation data is massive and typically needs to be compressed to fit the storage capacity of driver ICs. A common method is to obtain compensation values in units of blocks (e.g., 2x2 or 4x4), restoring pixel-by-pixel precision through interpolation during actual driving. The compression ratio needs to balance storage space and compensation accuracy.
Technical Workflow of Writing Compensation Data to Driver ICs#
Once compensation data is generated, it must be written to the panel’s driver IC through specific interfaces and protocols:
- Data Encoding: Encoding compensation coefficients into the data format required by the driver IC.
- Data Transmission: Downloading the compressed compensation data to the panel’s Flash memory via a dedicated interface (such as I2C, SPI, or the panel’s data channel).
- Run-time Decompression: After the panel powers on, the driver IC reads the compressed compensation data from Flash and decompresses it into a full compensation table in internal SRAM.
- Real-time Superposition: During each frame of display, the driver IC superimposes the compensation data onto the original display data in real-time, generating corrected driving signals for the panel pixels.
This workflow demands automated integration on production lines: detection equipment, data processing hosts, and driver IC programmers must work in coordination, with the overall Takt Time controlled within a range acceptable for the line.
Differences between Luminance Compensation and Chromaticity Compensation#
Demura can be classified into two levels based on compensation dimensions:
Luminance Demura#
Targets only luminance differences. During detection, white or gray frames at high, medium, and low gray levels are typically captured to calculate luminance deviation for each pixel and generate luminance compensation data. This is the most basic form of Demura, relatively simple to implement, requiring fewer detection gray levels and less compensation data.
Luminance compensation effectively improves light/dark patch issues in pure-color frames but is powerless against color non-uniformities (e.g., some areas appearing blue or yellow).
Chromaticity Compensation (Color Demura)#
Compensates for both luminance and chromaticity differences simultaneously. It requires capturing images of R, G, and B color channels at multiple gray levels separately to independently calculate compensation coefficients for each color channel. Chromaticity compensation requires about three times the data volume of luminance compensation and involves higher algorithm complexity.
Chromaticity compensation places stricter requirements on detection equipment; imaging colorimeters need not only precise luminance measurement but also reliable chromaticity separation performance—meaning accurate distinction and measurement of the chromaticity coordinates of each subpixel across different color channels.
In practical applications, high-end OLED products (like flagship smartphone screens) typically use chromaticity compensation; mid-range products or applications with lower color consistency requirements may use only luminance compensation. For MicroLED, chromaticity compensation is almost mandatory due to the significant wavelength dispersion of RGB chips.
Impact of Demura on Yield Improvement#
Demura technology has a significant and direct impact on display panel yield. Without Demura, OLED panel uniformity yield is constrained by manufacturing process control limits—when process capability cannot be further improved, yield hits a ceiling. Demura effectively raises this yield ceiling by providing software-level compensation at the back end.
Specific effects depend on several factors: the degree of the panel’s own non-uniformity, the precision of the compensation algorithm, the compensation bit depth of the driver IC, and the allowed compensation range. General rules are:
- For moderate Mura (e.g., uniformity deviation within 10%~20%), Demura can compress the deviation to within 2%~5%.
- For subtle pixel-level differences (such as OLED “graininess”), high-resolution Demura can bring the post-compensation frame to a visually indistinguishable level.
- For severe defects (such as dark pixels or obvious bad dots in MicroLED), Demura’s compensation range is limited; defects beyond compensation capability still need repair or replacement.
The economic value of Demura is particularly prominent in the MicroLED field. MicroLED manufacturing costs are high, and mass transfer yield directly determines commercial viability. An efficient Demura system can transform panels originally judged as “fail” (due to unmet uniformity standards) into “pass” products, significantly reducing unit manufacturing costs.
Technical Challenges and Future Directions of Demura#
Although Demura technology is widely deployed on OLED production lines, it still faces several technical challenges:
Conflict between Compensation Precision and Efficiency. Higher precision requires more gray-level acquisitions, larger volumes of compensation data, and longer processing times, creating a natural conflict with high-Takt Time production line operation.
Non-linearity of Gamma Curves. The electro-optical conversion characteristics of pixels are not perfectly linear, leading to variations in compensation effects across different gray levels. Multi-gray-level acquisition and segmented compensation can alleviate this, but they increase system complexity.
Driver IC Bit Depth Limits. Compensation precision is ultimately limited by the bit depth of the driver IC’s lookup table (LUT). An 8-bit LUT provides only 256 levels of compensation resolution, which may be insufficient for high-precision compensation; 10-bit or 12-bit LUTs offer finer control but place higher demands on IC design and storage.
Aging Compensation. Performance decay caused by OLED material aging is a dynamic process; compensation data written at the factory may no longer be accurate after some use. Some high-end products have introduced online compensation mechanisms that regularly re-measure and update compensation parameters during use.
Algorithm Evaluation Standards. The industry has not yet formed unified standards for evaluating Demura algorithms. Compensation accuracy (the ability to accurately eliminate different types of Mura) and algorithm efficiency (completion speed and data volume rationality) are two core evaluation dimensions, but specific test methods and pass/fail metrics remain defined by individual manufacturers.
In terms of future directions, higher resolution imaging detection, faster data processing, integration with AI algorithms, and stronger on-IC compensation capabilities are the primary paths for the continuous evolution of Demura technology.
FAQ#
Q1: What is the difference between optical and electrical Demura compensation?#
Optical compensation (external Demura) uses high-resolution imaging equipment to record each pixel’s actual luminance output, calculates compensation values based on measurement results, and directly corrects the final optical output without relying on internal OLED/TFT characteristics, offering better universality and wider adoption. Electrical compensation obtains the actual I/V characteristics of the driving TFT or OLED and compensates based on differences between actual and target values, requiring coordination across panel design, IC architecture, driving circuitry, and algorithms, making implementation more complex.
Q2: Why does MicroLED need Demura technology even more than OLED?#
MicroLED faces more severe uniformity challenges than OLED for three reasons: first, micron-scale chips exhibit inherent dispersion in luminance, wavelength, and voltage characteristics, and their extremely small size (under 100 microns) makes traditional LED binning impractical; second, mass transfer of millions of chips requires precision within plus or minus 1.5 microns, where any offset causes pixel defects; third, large MicroLED displays are spliced from multiple modules with independent uniformity challenges. Additionally, MicroLED manufacturing costs are high, and Demura can convert panels that would otherwise fail uniformity standards into passing products, making its economic value particularly significant.
Q3: How is Demura compensation data written to driver ICs and applied in real time?#
After compensation data is generated, it is first encoded into the format required by the driver IC, then downloaded to the panel’s Flash memory via dedicated interfaces such as I2C, SPI, or the panel’s data channel. Since raw pixel-by-pixel data is massive, it is typically compressed in blocks of 2x2 or 4x4 pixels. When the panel powers on, the driver IC reads the compressed data from Flash and decompresses it into a full compensation table in internal SRAM. During each display frame, the IC superimposes compensation data onto the original display data in real time, generating corrected driving signals for the panel pixels, entirely transparent to the user.
This article is part of the Imaging Colorimeter Technology Knowledge Base series.