High Accuracy Optical Metrology for MicroLED Displays and Wafers
- khashayar Ghaffarzadeh
- Sep 9
- 4 min read
Updated: Sep 12
Author: Tobias Steinel, Steinel@instrumentsystems.com, Instrument Systems GmbH, Munich, Germany
Background
MicroLED (µLED) displays promise high contrast, fast response, wide color gamut, and long lifetime. However, production faces critical challenges:
Massive parallelization of testing: Millions of tiny µLEDs must be characterized quickly.
Metrology limitations: Narrow emission bandwidth and strong wavelength variability (≈5 nm) demand both speed and spectral accuracy.
Hardware requirements: In order to rapidly test and measure millions of µm sized µLEDs high resolution optics and cameras as well as precise detection algorithms are needed to avoid image artefacts due to low oversampling ratios (Rs).
Traditional LIV measurements with integrating sphere methods are too slow (hours per wafer). To address this, the authors developed a spectrally enhanced imaging light measurement device (ILMD) – the LumiTop system – which combines a high-resolution camera with a traceable spectroradiometer.
Methodology
Color Calibration: Live calibration is performed for every image, using the spectroradiometer to adapt to spectral variations from manufacturing tolerances or drive conditions. [1,2]
Experimental Setup: [3]
Photoluminescence (wafer test): µLED 6”wafer with 17M µLEDs; 165 stitched images captured ≈100,000 µLEDs per frame. Total test time: ~5 min.
Electroluminescence (microdisplay): RGBW microdisplay (1.7M emitters, 5 µm pixel size, 11 µm pitch) tested in single shots (few seconds) using a 150MP camera for imaging microscopy.
Single Pixel/Emitter Evaluation (SPE) Algorithm: [3] Provides per-µLED parameters – dominant wavelength, luminance, chromaticity, tristimulus values (X,Y,Z), emitter size/location, and purity.
Optimized Stepped Kernel Filters: [4] Introducing a new method for adjusting kernel weights so that the filter notch frequency aligns exactly with the sampling ratio 1/Rs. This ensures accurate suppression of pixel-grid periodicity even between Rs = 2 to 5.
Results
Speed: Entire wafers with millions of µLEDs can be tested in minutes rather than hours. Analysis scales linearly with emitter count.
Accuracy: Chromaticity values derived from camera + live spectroradiometer calibration match spectroradiometer only measurements within one color point.
Flexibility: SPE can be tailored (e.g., defect detection only) to further increase speed by up to an order of magnitude.
Relevance for Displays: Results directly support per-pixel calibration and correction (demura) for microdisplays, ensuring uniform brightness and color.
Optimized Stepped Kernel Filters: A new method for adjusting kernel weights so that the filter notch frequency aligns exactly with 1/Rs. This ensures accurate suppression of pixel-grid periodicity even between Rs = 2 to 5.
Impact
The combination of an ILMD and spectrometer system enables fast, accurate, and traceable wafer- and display-level testing of µLEDs, balancing speed (imaging cameras) with spectral precision (spectroradiometers). This approach:
Makes wafer-scale optical testing economically viable.
Provides full spectral, spatial, and colorimetric analysis per emitter.
Supports yield improvement and uniformity correction in µLED display mass production.
Conclusion
By combining high-resolution cameras, spectroradiometers, and live calibration, the proposed method dramatically reduces testing time. In the case of µLED wafers testing times go down from hours to few minutes while maintaining high accuracy. It is a powerful solution for industrial µLED metrology, essential for scaling µLEDs into mass production of displays.
Calibration choice critically affects chromaticity accuracy for LED-based displays. The Live Calibration method, leveraging real-time spectrometer referencing and DUT-specific calibration, offers the most reliable results across diverse display technologies. It significantly reduces error budgets compared to traditional methods, supporting precise, traceable colorimetric measurements in both R&D and mass production.
The optimized stepped-kernel MWA (Moving Window Average) filter provides a practical, mathematically sound solution to aliasing and averaging challenges at low sampling ratios. It ensures precise suppression of display pixel matrix modulations, reducing the need for oversampling and improving both cost efficiency and measurement reliability in display characterization.
References
[1] Schanz, R., Fischer, F. and Steinel, T. (2024), 58-3: Impact of Calibration Sources on Accuracy of Chromaticity Measurements of LED based Displays. SID Symposium Digest of Technical Papers, 55: 801-804. https://doi.org/10.1002/sdtp.17649
[2] Steinel, T. and Wolf, M. (2021), 58-3: Invited Paper: Color Uniformity of μLED Displays: New Color Calibration Concept for Fast and Accurate Optical Testing. SID Symposium Digest of Technical Papers, 52: 822-825. https://doi.org/10.1002/sdtp.14809
[3] Tobias Steinel, Habib Gahbiche, Pooja Baisoya, Roland Schanz, (2023), Invited Paper: Rapid Testing of µLEDs and Microdisplays on Wafer. ICDT China, Session 24.3.
[4] Becker, M.E. and Steinel, T. (2025), 30-3: Matched Moving-Window Averaging Filter. SID Symposium Digest of Technical Papers, 56: 397-400. https://doi.org/10.1002/sdtp.18176
To learn more about MicroLED and AR/VR displays please join the show in Eindhoven on 24 and 25 Sept 2025. Learn more [here]
Comments