Tobias Steinel | Instrument Systems: How does 70 frames per second plus GPU acceleration deliver a 100x metrology speedup for microLEDs?
13:45 - 15:30
Other snippets from this talk
Summary of the clip:
How does 70 frames per second plus GPU acceleration deliver a 100x metrology speedup for microLEDs?
The spectrometer and camera setup is triggered at around 70 frames per second, capturing 70 images per second. That huge amount of data at high speed feeds the iterative measurements described earlier. Within fractions of a second, the system runs through many different test images and can optimize display calibration or CPO calibration. That speeds up the story quite a bit and makes metrology cheaper and faster for the user.
GPU acceleration was used for that, and the overall measurement speedup was on the order of a factor of 100. For standard displays that was apparently not so relevant, because metrology there is actually not that sophisticated. With micro-LEDs you must take care of a lot of spectral features, so this frame rate is really a breakthrough, allowing test scenarios and test cycles that were not possible before.
The payoff is that faster, cheaper metrology opens test scenarios and test cycles previously out of reach. Micro-LED demands spectral care that standard display metrology never required, which is why a hundredfold speedup matters more here than it did for conventional panels. The intent is to contribute these new measurement possibilities to help the micro-LED market move forward.
In this short video, you can learn:
* A spectrometer and camera setup triggered at around 70 frames per second supplies the image volume needed for iterative measurement.
* GPU acceleration delivered an overall measurement speedup on the order of a factor of 100.
* Handling micro-LED spectral features makes metrology cheaper and faster and enables test cycles that were not possible before.
📋 **Clip Abstract** A spectrometer and camera rig running at roughly 70 frames per second captures 70 images per second, feeding iterative measurements that step through many different test images in fractions of a second to optimize display or CPO calibration. Combined with GPU acceleration, this yields an overall measurement speedup on the order of a factor of 100, making metrology cheaper and faster for the user and enabling test scenarios and test cycles that were not possible before.
About the speaker:
* Speaker: Tobias Steinel
* Company: Instrument Systems
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#MicroLEDMetrology, #SpectrometerTriggering, #GPUAcceleration, #DisplayCalibration, #MicroLEDDisplays, #OpticalMetrology
This is a highlight of the presentation:
Optical Metrology for MicroLED Displays and Wafers: Optimized Color-Matching Functions for Cross-Technology Product Portfolios
More Highlights from the same talk.
05:37 - 07:01
How do you measure peak wavelength, power and polarization of every VCSEL emitter in an optical interconnect array?
How do you measure peak wavelength, power and polarization of every VCSEL emitter in an optical interconnect array?
The co-packaged optics setup starts with VCSEL arrays imaged by a camera system alongside a spectrometer — the standard configuration used across all production equipment. The camera captures the near-IR emitter array, locating each emitter and measuring its radiant power. Spectra are measured and peak wavelengths assigned to individual emitters, mapping the array. So position, power and peak wavelength are resolved per emitter.
Polarization measurement works like an RGB sensor, where different pixels receive light through different polarization filters. That lets the system measure polarization state per emitter. Polarization matters for in-coupling into optical connect devices and for slope efficiency. These capabilities are already solved, with machines in production for VCSEL arrays today, and the same systems can be used at the same time for micro-LEDs or any other light emitter in connect applications.
With position, radiant power, spectra, peak wavelength and polarization addressed, the discussion turns to speed. Why emphasize speed? The machines described are already in production for VCSEL arrays, and the platform can handle micro-LEDs or any other emitter for connect applications. Speed becomes the next topic raised in the talk, immediately after the solved measurement stack.
In this short video, you can learn:
* A camera system plus spectrometer images the near-IR emitter array, locating each emitter and measuring its radiant power.
* Spectra are measured and peak wavelengths are assigned to emitters, mapping the array.
* Polarization is measured like an RGB sensor with different polarizations at different pixels, which matters for in-coupling into optical connect devices and slope efficiency.
📋 **Clip Abstract** Camera-and-spectrometer systems image near-IR VCSEL arrays, resolving each emitter's position, radiant power, spectra, peak wavelength and polarization. These machines are already in production for VCSEL, with polarization important for in-coupling and slope efficiency, and can be used for micro-LEDs or any other light emitter for connect applications.
About the speaker:
* Speaker: Tobias Steinel
* Company: Instrument Systems
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#VCSELArrays, #PeakWavelengthMapping, #PolarizationMetrology, #RadiantPowerMapping, #CoPackagedOptics, #MicroLEDDisplays
10:57 - 12:42
Why do displays matched to the same D65 white point look pink to different observers?
Why do displays matched to the same D65 white point look pink to different observers?
The example compares LCD, OLED and Rec. 2020 spectra, including laser lines. The black reference is the D65 standard white. When the white point is calibrated to D65 and measured with different observers, each observer lands elsewhere in color space, even though the calibrated white point is the same. That sets up metamerism risk across technologies.
With OLED, micro-LED and quantum dots, multiple observers are more sensitive. Different spectra calibrated to the same D65 white point produce strong metamerism failure. The observers see mostly a pink shade, because the calibrated white point is not what they perceive. This was reproduced using a tablet and two smartphones, all calibrated closely to the same white point.
When those tablet and smartphone displays are viewed in a different color space, they lie apart from each other. That matches the visual result: the same nominal white point separates into different observed whites. The talk notes that the Apple 2026 color matching function was recently introduced. The same set of spectra comes from LCD, OLED, micro-LED, quantum dots and Rec. 2020 laser displays.
In this short video, you can learn:
* Calibrating LCD, OLED and Rec. 2020 displays to the same D65 white point still produces strong metamerism failure across different observers.
* The observers in the example see mostly a pink shade, and a tablet plus two smartphones calibrated closely to the same white point lie apart in a different color space.
* An Apple 2026 color matching function was recently introduced, according to the talk.
📋 **Clip Abstract** When LCD, OLED and Rec. 2020 spectra are calibrated to the same D65 white point, different observers end up somewhere else in color space, producing strong metamerism failure. The example shows a mostly pink shade, and a tablet plus two smartphones calibrated closely to the same white point lie apart in another color space, while an Apple 2026 color matching function has recently been introduced.
About the speaker:
* Speaker: Tobias Steinel
* Company: Instrument Systems
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#D65WhitePoint, #MetamerismFailure, #Rec2020Spectra, #ColorMatchingFunction, #OpticalMetrology, #DisplayColorScience
06:58 - 09:05
Why is microLED pixel compensation harder than OLED, and what must be measured per emitter?
Why is microLED pixel compensation harder than OLED, and what must be measured per emitter?
MicroLED arrays and optical connect devices contain millions of emitters — thousands in the optical connect case — and each single emitter can carry its own faults, cracks and defects, requiring individual treatment. Metrology therefore runs as a circle: measure, calibrate the image, analyze, pass-fail, then repeat many times until a very inhomogeneous starting display becomes homogeneous. The same holds for CPOs and optical connect arrays.
Why speed? Because these cycles become more numerous or longer for microLED applications. MicroLEDs are more tricky to compensate than OLED, where mura compensation is well known; microLED compensation needs more complex models. Within each cycle, every individual pixel is calibrated for luminance, chromaticity, dominant wavelength, size and position, and this is done in iterative loops. Speed is the enabling requirement.
The end goal is uniform brightness and color for every pixel or emitter. A further display application cited is the multiple observer application: in display, we all have different color matching functions in our eyes. That drives the need for optimized color matching functions across cross-technology product portfolios, while per-emitter luminance, chromaticity, dominant wavelength, size and position still must be measured.
In this short video, you can learn:
* Each single microLED emitter can have its own faults, cracks and defects, so it needs individual treatment rather than array-level correction.
* Metrology runs as an iterative measure-calibrate-analyze-pass/fail loop that repeats many times before an inhomogeneous display becomes homogeneous.
* Per-pixel calibration covers luminance, chromaticity, dominant wavelength, size and position, and microLED needs more complex compensation models than OLED mura.
📋 **Clip Abstract** MicroLED arrays, CPOs and optical connect devices are built from millions of emitters, each with individual faults needing individual correction, so metrology becomes an iterative measure-calibrate-analyze loop. MicroLED compensation is more complex than OLED mura, forcing speed into the loop while every pixel is calibrated for luminance, chromaticity, dominant wavelength, size and position, with color matching functions differing between observers.
About the speaker:
* Speaker: Tobias Steinel
* Company: Instrument Systems
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#MicroLEDMetrology, #PixelCalibration, #MuraCompensation, #ColorMatchingFunctions, #MicroLEDDisplays, #DisplayMetrology