Murphy Lee | Rayleigh Vision Intelligence Co. Ltd.: Can you really use "bad" epi-wafers to make a "good" microLED display?
09:25 - 11:03
Other snippets from this talk
Summary of the clip:
Can you really use "bad" epi-wafers to make a "good" microLED display?
In the microLED manufacturing process, a critical and costly stage is the transfer of individual chips from a carrier wafer to the final display backplane. This process uses a laser-based transfer technique to selectively pick and place individual red, green, and blue microLEDs. This selective placement is not random; it follows a unique, pre-calculated pattern that is optimized for every single source wafer based on its specific defect and performance map.
A key technical and commercial breakthrough is the system's ability to tolerate a much wider wavelength variation from the incoming epi-wafers. While conventional processes may be restricted to a tight Ā±2 nanometer wavelength window to ensure color purity, this AI-enhanced platform can accept a Ā±5 nanometer variation. This significantly relaxes the specifications for the expensive GaN epi-wafer growth process, a major driver of overall microLED cost.
This wider tolerance is made possible by the AI-driven bin mixing algorithm. The system measures the specific wavelength of each chip and intelligently distributes them across the display area. By mixing slightly different shades of red, green, and blue in a calculated pattern, it can still achieve the target white balance and color uniformity, effectively turning lower-grade, cheaper wafers into high-quality, uniform displays.
In this short video, you can learn:
* How laser transfer enables selective, pattern-based placement of RGB microLEDs.
* The concept of widening the acceptable epi-wafer wavelength specification from Ā±2nm to Ā±5nm.
* The use of AI algorithms to mix different wavelength chips to achieve a uniform white balance.
š **Clip Abstract** This clip reveals a key strategy for reducing microLED production costs: relaxing the stringent specifications on the epi-wafer. By using an AI-powered bin mixing algorithm, their system can accept a wider wavelength variation (Ā±5nm), turning lower-cost wafers into high-performance displays.
š Link in comments š
#MicroLEDTransfer, #AIBinMixing, #WavelengthTolerance, #GaNEpiwafers, #MicroLEDDisplays, #DisplayManufacturing
This is a highlight of the presentation:
MicroLEDs, AR/VR Displays, Micro-Optics 2025: Innovations, Start-Ups, Market Trends
Online | TechBlick platform
Organised By:
TechBlick
MicroLED Connect
More Highlights from the same talk.
06:25 - 09:02
How can AI solve the microLED yield crisis by intelligently mixing good and bad chips?
How can AI solve the microLED yield crisis by intelligently mixing good and bad chips?
The primary bottleneck in microLED manufacturing is the complexity of bin mixing and the low initial yield of "perfect" chips from the epi-wafer. As chips shrink to the micro-scale, traditional sorting methods become impractical for the millions of components on a single display. This challenge makes mass production economically unviable if one relies solely on pristine LEDs, as the vast majority of the wafer would be discarded.
Rayleigh Vision's solution leverages AI to tackle this challenge head-on. The first application is in "pathfinding," which calculates the most efficient physical path for the laser transfer head to pick up selected LEDs from the source wafer. This AI-driven optimization minimizes stage movement and cycle time, directly addressing the throughput issue inherent in picking and placing millions of individual components one by one or in small groups.
The most critical AI application is an advanced bin mixing algorithm. Instead of only picking perfect chips, the system intelligently selects a combination of "good" and "imperfect" chips based on pre-inspected data (AOI, PL, EL). It then calculates the optimal placement pattern on the final display backplane to ensure visual uniformity, effectively compensating for minor variations in wavelength or brightness to achieve the desired overall performance.
In this short video, you can learn:
* The fundamental yield and binning challenges in microLED mass production.
* How AI-driven pathfinding optimizes the physical transfer process for speed.
* The concept of using algorithms to mix LEDs of varying quality to create a uniform display.
š **Clip Abstract** Rayleigh Vision Intelligence explains how their AI-powered platform addresses the core microLED manufacturing bottlenecks of low yield and complex bin mixing. Their solution uses intelligent pathfinding and advanced algorithms to mix LEDs of varying quality, enabling high-throughput, cost-effective mass transfer.
š Link in comments š
#MicroLEDYield, #AIBinMixing, #AIPathfinding, #LaserMicroTransfer, #ARDisplays, #DisplayManufacturing
17:58 - 19:01
How does AI practically speed up microLED transfer? Here's a 4x improvement explained.
How does AI practically speed up microLED transfer? Here's a 4x improvement explained.
This clip provides a concrete example of how AI optimization directly improves manufacturing throughput for microLED displays. For a watch-sized panel, a conventional, non-optimized transfer process might require 200 to 300 distinct "steps" or movements of the transfer stage to pick up all the necessary chips. By implementing an AI pathfinding algorithm, the process can be streamlined to fewer than 50 steps, representing a greater than 4x improvement in efficiency.
This dramatic reduction is achieved by solving a complex optimization problem rooted in a key physical constraint: the limited Field of View (FOV) of the transfer equipment. The machine can only "see" and access a small portion of the source wafer at any given time. The AI calculates the most efficient sequence of moves and pick-up patterns within each FOV to minimize the total number of stage movements required to populate the entire display.
The algorithm's goal is to maximize the number of useful, high-quality chips transferred in each step, based on the pre-inspected wafer map. This "path of least resistance" approach significantly cuts down the cycle time for each display panel. This demonstrates a tangible, quantifiable improvement in manufacturing efficiency, moving the discussion of AI from an abstract concept to a practical, performance-enhancing tool.
In this short video, you can learn:
* A quantifiable metric for AI's impact: reducing transfer steps from ~200 to under 50.
* The critical role of the transfer equipment's limited Field of View (FOV) as a process constraint.
* How AI pathfinding optimizes the pick-and-place sequence to maximize efficiency within the FOV.
š **Clip Abstract** This clip provides a tangible metric for the value of AI in microLED manufacturing, explaining how it can reduce the number of transfer steps by over 75%. The speaker details how the algorithm optimizes the pick-and-place path within the equipment's limited field of view, directly boosting throughput.
š Link in comments š
#MicroLEDTransfer, #AIPathfinding, #PickAndPlaceOptimization, #TransferEquipmentFOV, #ARDisplays, #WearableElectronics