Methods for boosting efficiency, testing thousands of pixels, and identifying known good emitters still in development.
MicroLED display makers are marching toward commercialization, with products such as Samsung’s The Wall TV and Apple’s smart watch expected to be in volume production next year or in 2024.
These tiny illuminators are the hot new technology in the display world, enabling higher pixel density, better contrast, lower power consumption, and higher luminance in direct sunlight — while consuming less power than OLEDs or LCDs. Commercially, while consumers eagerly await the introduction of AR/VR glasses, smart watches, flexible screens, adaptive headlights, and TVs, engineers tackle the very real challenge of yield improvement.
Early test results indicate yield issues at chip transfer, array-to-driver bonding, and other relatively new processes. High cost for this immature technology is keeping microLED displays from making the prototype-to-production leap. And because probers are not well suited to testing thousands of microLED pixels in densely packed arrays, DFT with self-testing is employed, which enables lifecycle testing — at ATE, post-assembly test, and in the field.
For instance, Dialog Semiconductor, a Renesas Company, developed a testing scheme for a white adaptive headlight module containing a 20,000-microLED array with 40µm pitch. “It’s a very good example of how a DFT circuit is not just overhead and cost to buy quality,” said Hans Martin von Staudt, director of Design-for-Test at Renesas. “Instead, it serves a valuable function over the lifetime of the chip. So we needed a DFT scheme with high-diagnostic coverage of the assembly process for pinpointing process weaknesses while enabling in-field monitoring.”
Inspection and testing methods are improving in their ability to identify and segregate out-of-spec product. Mass transfer methods that remove microLED die from wafers or film carriers and position them on IC drivers (for small AR/VR, watch and headlights) or TFT PCBs (for TVs), must easily separate known good die (KGD) from failures and underperforming die.
Yield targets for most microLED display apps are high (see figure 1) because the human eye can quickly spot missing pixels. To put yield targets in perspective, an 8K TV contains 99 million microLED chips. So if the defectivity rate is 0.5%, 520,000 devices must be removed and replaced. Top Engineering estimates this process would take 144 hours, making it cost-prohibitive until repair cost (removal and replacement of individual microLEDs) can be accelerated.
Fig. 1: Display yield is a product of process step yield, with transfer and bonding accounting for a good portion of failures due to process immaturity. Source: KLA
Inspection, metrology, photoluminescence
Wafer-level optical inspection (AOI) and photoluminescence (PL) measurement are being used at multiple process points, from epitaxial wafer formation to final display signoff.
“Any kind of visual defects such as particles, scratches etc., need to be detected, as they can decrease illumination intensity emitted by LEDs,” said Woo Young Han, director of application engineering at Onto Innovation. “In addition to traditional brightfield and darkfield inspection, exposing LEDs at around 400nm wavelength will excite the microLEDs, triggering them to emit light (PL), which must be verified in-line.”
Fig. 2: Photoluminescence testing provides luminance data including functionality (upper right), brightness distribution (lower left) and wavelength variation (lower right). Source: Toray
MicroLED and display makers currently utilize multiple inspection steps. “The microLED process is really challenging,” said Charlie Zhu, vice president of research and development at CyberOptics. “In order to improve the yield, the customer needs more inspection steps. Some companies are doing five inspections with our SQ3000 or SQ3000+ systems, including bare ball inspection, solder paste inspection, pre-flow offset measurements, and post-flow inspection for co-planarity and tombstone issues. We also take 3D geometry measurements on well areas around the die, which will then be filled in passivation. Our key advantage for production is the higher throughput, which can be 100X faster than competing tools, with accuracy and repeatability of 2µm.”
Efficiency is achieved by doing more at each step. “At the bare ball inspection step, we look for contamination or foreign objects, but also the shape of the pads, and measuring each pad’s location to make sure it’s correct,” said Zhu, noting that for solder connections, ball height and volume are the most critical parameters, while for pre-flow solder ball offset is key. Final patterned wafer inspection is performed both at the wafer fab and on incoming wafers on film frames delivered to display manufacturers.
Photoluminescence systems, offered by Corbeau Innovation, Hamamatsu, InZiv, Toray, and others can be standalone systems or independent modules that work together with process tools.
According to Toray, PL can be mounted to wafer inspection systems to detect crystal defects at epi processes or to carry out integrated prototype and mass production inspection through patterned wafer inspection.
PL measures wavelength, brightness, and defectivity (see figure 2). Toray aims to cover a majority of microLED steps, including AOI/PL, laser trimming and transfer, laser mass transfer (via laser lift-off), and stamp-transfer based bonders. Both stamp-based and laser-based transfer methods for positioning and repair (replacement) are popular today.
KLA’s metrology, inspection, process, and data analytics solutions seek to address the yield challenges at wafer through display levels (see figure 3). KLA’s microLED portfolio provides plasma etching, silicon substrate thinning, plasma dicing (singulation), wafer front and backside metal deposition, wafer inspection and metrology, as well as panel inspection, and metrology and test for the backplane and mass transfer steps.
“A key factor behind the industry’s focus on microLEDs is that LCD and OLED displays are produced on large substrates, where all layers are deposited one after the other,” said Carolyn Short, marketing communications manager at KLA. “But large single displays can be difficult to handle and ship. MicroLEDs displays, however, can be produced using seamless tiling of small modules into larger displays.” [3]
Fig. 3: Comprehensive inspection and yield management from epilayer testing to backplane and display assembly. Source: KLA
At least two inspection methods require improvement for high-volume manufacturing. Solder flux and MicroLED passivation layers are not opaque, so chromatic confocal measurements are taken. “In SMT, people do not inspect flux residues, but they want to for microLED,” said Zhu. “In addition, many of our customers are using chromatic confocal technology to inspect the passivation layer on microLEDs. But chromatic confocal is very slow, so it’s more of a sample check. Everyone is looking at whether there’s any technology available for the high-speed confocal inspection or an alternative for these applications.”
Design boosts quantum efficiency
From a design standpoint, microLEDs are similar to conventional LEDs, microOLEDs and other nanophotonic devices, where the electronic and optical functions of the device are modeled together.
“With microLED design, there are two main issues: one is the extraction efficiency and the other is the illumination pattern,” said Chenglin Xu, product manager of RSoft Photonic Device Tools at Synopsys. “Extraction efficiency can be 15 to 20 percent at most because of the total internal reflection at the device surface. So tools like our LED Utility allow device simulation with a texture pattern on the microLED surface to help disperse the light and improve the external quantum efficiency.”
Other microLED designs tend to be non-flat by nature. For instance, in AR/VR glasses, waveguides on top of the light source channels the beam to where it’s needed. Alternatively, one company, Aledia, designed submicron-diameter light-emitting GaN nanowires that project up from the silicon surface to improve luminous area efficiency.
Planar devices are easier to simulate using analytical models, but 3D simulation of brightness and radiation pattern of microLEDs are becoming common, requiring a more time-consuming algorithm called finite difference time domain (FDTD). FDTD is extremely accurate, but it doesn’t apply to all microLED applications. “FDTD analysis is fine for small microLEDs, such as for AR/VR or iWatch sizes, but for larger microLED displays, our ray tracing software tools can be used to save computation time,” Xu said.
Fig. 4: MicroLEDs of different wavelengths are incoherent, so brightness (greatest at LED bottom) within an array requires multiplanar, far field calculation using FDTD. Source: Synopsys
In simulating microLED arrays, light scatters and interacts in an incoherent manner, so simulation needs to incorporate several surrounding microLEDs to yield an accurate radiation pattern. The light within both the GaN device and air contribute to the radiation pattern, and should be included in the far-field calculation. One improvement Synopsys made about a year ago to its FullWAVE FDTD software was updating its far-field calculations to account for the radiation from all near-fields in both media and air. “This is the first commercial software, to our knowledge, that extends multiplane, far-field illumination calculations to inhomogeneous media,” said Xu.
LED failure modes and DFT strategy
Companies are developing new DFT and testing methods for microLEDs.
Densely packed microLEDs, such as the 20,000-pixel headlamp that Renesas and Lumileds are working on, pose a serious test challenge. “Even though you may be able to make contact to 20,000 I/Os using today’s vertical probes, now you have 20,000 signals and there’s no ATE with 20,000 instruments. So it’s far better to test internally what you can’t properly do outside, which is the whole point of DFT,” said von Staudt.
Adaptive automotive headlamps work with camera-based ADAS to improve safety by sharpening road lines, dimming oncoming lights, or projecting a warning image in case of hazard. Teams of engineers at Lumileds, together with Renesas — as well as Nichia with Infineon — are working with automakers to bring these headlamps to market.
In the Renesas/Lumileds device, the array of 20,000 microLEDs is flip-chip bonded on top of the silicon driver IC, making 20,000 I/O connections, one per pixel. In solid-state lighting, pulse-width modulation is the most energy efficient way of controlling brightness, allowing current sources of the array to work at the same operating level, minimizing headroom voltage. Renesas’s von Staudt emphasized that the project team developed microLED driver test algorithms for ATE self-test based on sensing the forward voltage of the microLED pad.
“Using this self-test approach at ATE, post-assembly test and self-monitoring in the field, the group proved the self-test effectiveness on a product wafer, including the ability to pinpoint opens, shorts and bridge fault locations,” von Staudt said.
Fig. 5: Pixel driver with DFT. (Blue = digital signals, Red = analog signals) Source: Renesas
The current source in the pixel driver with DFT (see figure 5) is switched on and off by a digital control signal that carries the PWM signal. A switch connects the anode pad to the analog test bus with two comparators to distinguish many fault types. These include LED opens (very common) caused by poor connection of LED to driver pad, LED shorts, driver opens or stuck-off driver, stuck-on driver, driver shorts (causing a super bright pixel), or an out-of-spec current. Another fault type may occur between neighboring connectors after bonding — bridge faults. The DFT approach enables self-monitoring during operation, as well as self-test before and after the critical bonding operation.
“The sheer number of 20k pixels raises challenges regarding the analog test bus topology,” said von Staudt. “As area is at a premium for the smaller pixel pitch, you will always end up with a structure where one metal line serves two adjacent columns or rows.”
To reduce the total bus length, it may be cut into sections using analog switches. Implementing parallel self-test with multiple comparator sets reduces capacitive load and shortens diagnostic test time.
Fig. 6: In this white automotive light headlamp, the microLED chip with 20,000 microLEDs is bonded on top of the silicon driver IC, which forms the backplane of the array. This assembly is then COB (chip-on-board) mounted to a PCB. Source: Lumileds
“The biggest challenge is now to bring these 20,000 pixels to the market. You cannot buy it yet. There’s still some engineering work to be done primarily by the people who manufacture the LEDs and do the assembly of the LEDs with the driver IC,” said von Staudt. “And in the next generation, I would expect more functional safety requirements coming top down from the OEMs as well.”
Conclusion
The design, DFT, inspection, and testing methods for microLEDs are addressing the varied needs for display, AR/VR, automotive headlights, and other applications. But the solid infrastructure in semiconductor design and display manufacturing means technologies are being rapidly adapted to microLEDs.
That will affect a broad swath of technologies in the market today, and enable new applications in the future.
References
Related Stories
MicroLEDs Move Toward Commercialization
But as tools and processes advance, there’s still room for breakthrough technologies.
MicroLEDs Moving From Lab To Fab
Bringing the cost down and yield up on microLED is proving to be formidable, but display companies and LED suppliers are working together toward production-worthy solutions.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Leave a Reply