Kedar Sathaye | Konica Minolta: Can fractional-pixel measurement and correction coefficients improve yield on already-manufactured microLED displays?
07:46 - 09:45
Other snippets from this talk
Summary of the clip:
Can fractional-pixel measurement and correction coefficients improve yield on already-manufactured microLED displays?
The fractional-pixel method addresses a basic sensor-coverage limitation: when pixels are not fully covered by the sensor, the traditional approach struggles for accuracy and the fill factor remains zero. Fractional-pixel measurement does not depend on that coverage, so it can still compute accurately how the pixels will be tested. This is presented as a software-level method for microLED displays viewed through a microscope.
Uniformity measurement is not limited to a single metric. The discussed techniques cover uniformity and grid placement, and they adapt to pincushion/barrel and other distortion types. That matters for microLED displays measured with a microscope, where software must handle those distortions before judging pixel behavior. The talk frames these capabilities as software-level, building on the fractional-pixel approach rather than requiring a change to the already-manufactured display.
Correction coefficients, already applied on big billboards, can be applied similarly to microLEDs. The significant claim is that an already-manufactured display can be corrected at wafer level, described as a step in the right direction to improve yield. An example shows luminance improving, and the same applies to bright color. Pixel-level and wafer-level correction can limit waste and safeguard efficiency.
In this short video, you can learn:
* Fractional-pixel measurement works even when sensor coverage is incomplete, unlike traditional methods where fill factor remains zero.
* Grid-placement and uniformity techniques adapt to pincushion/barrel and other distortions for microscope-based microLED measurement.
* Correction coefficients can be applied at wafer level to already-manufactured microLED displays, improving luminance and yield while limiting waste.
📋 **Clip Abstract** Fractional-pixel measurement enables accurate testing of microLED pixels even when sensor coverage is incomplete, while software handles uniformity, grid placement and distortion. Correction coefficients, analogous to billboard correction, can then be applied at wafer level to already-manufactured displays, with luminance improvements and yield gains described.
About the speaker:
* Speaker: Kedar Sathaye
* Company: Konica Minolta
* Event: MicroLED Connect 2024
* Location: High Tech Campus, Eindhoven, Netherlands
Join the most event MicroLED, AR/VR display and Optical I/O Connect event of the year
Learn more https://www.microledconnect.com/
#FractionalPixelMethod, #CorrectionCoefficients, #MicroLEDUniformity, #WaferLevelYield, #MicroLEDDisplays, #DisplayMetrology
This is a highlight of the presentation:
Measuring and Correcting MicroLED Display Uniformity
More Highlights from the same talk.
02:15 - 03:47
How much luminance and color variation appears across a 2 cm microLED display?
How much luminance and color variation appears across a 2 cm microLED display?
The non-uniformity challenge begins at pixel level, where luminance variance appears across the display. MicroLED is described as mature but still improving, and uniformity on the luminance level remains a challenge. Displays measured show wide luminance variance because they are extremely bright — easily from 50,000 to 60,000 candela per square meter on a single display less than two centimeters in size.
Such individual pixel variation is of course challenging to manufacturers, because it appears within one small panel rather than between separate displays. Then there is color dependency: variation of 0.01 was seen on the CIE 1931 curve, so significant color differences appear across the display. And this is a display just two centimeters long, which is again a challenge for metrology as well as for the manufacturers.
What does this mean for measurement? Both the luminance spread — 50,000 to 60,000 candela per square meter — and the 0.01 color shift on CIE 1931 must be captured at pixel level before any correction can be applied, and the two-centimeter format makes that a tight metrology problem for manufacturers and measurement engineers alike. Three minutes of correction logic rest on getting these two numbers right first.
In this short video, you can learn:
* One microLED display under two centimeters can carry luminance from 50,000 to 60,000 candela per square meter.
* Color dependency adds variation of 0.01 on the CIE 1931 curve across that same small display.
* The non-uniformity is a challenge for metrology and for manufacturers, and correction follows measurement.
📋 **Clip Abstract** Pixel-level luminance spread of 50,000 to 60,000 candela per square meter and color variation of 0.01 on CIE 1931 are both measured within a single microLED display under two centimeters. These variances are framed as a challenge for metrology and for manufacturers, with measurement presented as the step before correction.
About the speaker:
* Speaker: Kedar Sathaye
* Company: Konica Minolta
* Event: MicroLED Connect 2024
* Location: High Tech Campus, Eindhoven, Netherlands
Join the most event MicroLED, AR/VR display and Optical I/O Connect event of the year
Learn more https://www.microledconnect.com/
#MicroLEDUniformity, #LuminanceVariation, #CIE1931ColorShift, #PixelLevelMetrology, #DisplayMetrology, #EmissiveDisplays
03:01 - 04:41
If microLED pixels range from 2 to 10 million candelas per m², what does that do to the final image?
If microLED pixels range from 2 to 10 million candelas per m², what does that do to the final image?
Because every pixel is itself a light source, each microLED must be considered individually. The talk points to spatial variability in luminance and spatial variability in color across those pixels. Together these variations produce a display that is not completely uniform. That final non-uniformity is called mura, or unwanted brightness dependency, spread across the display. The non-uniformity can be seen on the individual pixel level.
Color variation is illustrated with an exaggerated range between 510 and 590, used only for simplicity to show that pixels in various locations can differ in color. Luminance variation is equally explicit: microLED displays are known to reach millions of candelas per meter squared. If one pixel outputs 2 million candelas and another outputs 10 million candelas, the display already contains a large luminance variation. This is the core spatial variability challenge described for microLED sources.
The challenge grows from small pixel size and high pixel density, because each source must be individually accounted for. Spatial luminance variability and spatial color variability combine into a non-uniform display. The result is a mura condition with unwanted brightness dependency across the display. At the individual pixel level, that non-uniformity is visible. The example of 2 million versus 10 million candelas per meter squared shows how quickly a large luminance range appears, making uniformity a defining problem for microLED displays.
In this short video, you can learn:
* Every pixel is itself a light source, so variation across pixels must be taken into account individually.
* Spatial luminance variability and spatial color variability together create a display that is not completely uniform, called mura or unwanted brightness dependency.
* MicroLED displays can reach millions of candelas per meter squared, and a 2 million versus 10 million candela example already shows a large luminance variation.
📋 **Clip Abstract** MicroLED displays face non-uniformity because each pixel is an individual light source. Spatial luminance and color variations produce mura, or unwanted brightness dependency, with a 2 million versus 10 million candela per meter squared example showing large luminance variation.
About the speaker:
* Speaker: Kedar Sathaye
* Company: Konica Minolta
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#MicroLEDMura, #PixelLuminanceVariation, #SpatialColorVariability, #CandelaPerMeterSquared, #NearEyeDisplays, #ARVRDisplays
04:01 - 06:08
How much wavelength spread appears across a monochromatic microLED display, and what camera resolution is needed?
How much wavelength spread appears across a monochromatic microLED display, and what camera resolution is needed?
Spectral variation appears at emitter level and therefore at display level: a single monochromatic display can show three different wavelengths, exaggerated here from 510 to 580 nm. Manufacturers would not ship that spread, but measured variation of 3 to 5 nm, sometimes 10 nm, is seen across a display. This shifts the uniformity problem beyond mura alone to wavelength-dependent correction.
The proposed approach is called ura, a Japanese term for defect, meaning removal of defects. At hardware level this requires high-resolution cameras and microscopic lenses so each individual pixel can be seen and measured properly. The measurement must resolve the pixel itself, not an average across neighboring emitters, because spectral and spatial non-uniformities need to be captured before correction.
Camera pixel budgeting is explicit: normally three to five camera pixels are required for one display pixel. For a one-micron microLED, five camera pixels means 0.2 micron per camera pixel. That resolution is described as achievable with a high-resolution camera, linking measurable wavelength spread and single-pixel defect removal to the optics and sensor specification rather than to display-level averages.
In this short video, you can learn:
* Measured wavelength variation across a monochromatic microLED display is typically 3 to 5 nm and can reach 10 nm.
* The ura approach treats defects for removal and depends on high-resolution cameras with microscopic lenses to measure individual pixels.
* At one-micron microLED size, five camera pixels per display pixel require 0.2-micron camera pixel resolution.
📋 **Clip Abstract** Spectral variation on a monochromatic microLED display can span 3 to 5 nm, sometimes 10 nm, even though a 510-to-580-nm example is exaggerated. Correcting such defects with the ura approach requires microscopic lenses and high-resolution cameras, with three to five camera pixels per display pixel and 0.2-micron resolution for a one-micron emitter.
About the speaker:
* Speaker: Kedar Sathaye
* Company: Konica Minolta
* Event: MicroLED Connect 2024
* Location: High Tech Campus, Eindhoven, Netherlands
Join the most event MicroLED, AR/VR display and Optical I/O Connect event of the year
Learn more https://www.microledconnect.com/
#MicroLEDUniformity, #SpectralVariation, #CameraPixelResolution, #DefectRemoval, #EmissiveDisplays, #DisplayMetrology
About this talk
→ Measuring and Correcting MicroLED Display Uniformity · Kedar Sathaye · MicroLED Connect 2024
Other talks by Kedar Sathaye
- Accurate display measurement within AR/VR headsets by using a complete optical solution · Printed Electronics Innovation Day | Winter 2024
More highlights from Kedar Sathaye
- ▶ If microLED pixels range from 2 to 10 million candelas per m², what does that do to the final image? · MicroLED Connect 2026
- ▶ Why does demura need a feedback loop of sequential dot-matrix patterns instead of one camera measurement? · MicroLED Connect 2026
- ▶ Which optical losses must be mapped backwards from the eye, not just corrected on the panel? · MicroLED Connect 2026
- ▶ Can demura of a million-pixel microdisplay run fast enough for a production line? · MicroLED Connect 2026
- ▶ How do imaging colorimeters achieve the spectral accuracy required to replicate human color perception? · Printed Electronics Innovation Day 2024
- ▶ Why do standard camera lenses cause a "knot hole effect" when measuring near-to-eye displays? · Printed Electronics Innovation Day 2024
- ▶ How can an optical metrology system focus on virtual images spanning from meters to infinity? · Printed Electronics Innovation Day 2024
