Compound Photonics (CP) and GLOBALFOUNDRIES (GF) have announced a strategic partnership to manufacture IntelliPix, CP’s microdisplay technology platform. IntelliPix was announced last year as an integrated CMOS backplane and display driver IC for monolithic MicroLED or Liquid Crystal on Silicon (LCoS) displays. The platform provides a real-time video pipeline and drives pixels as small as 2.5?m, for resolutions up to 2048 × 2048.
IntelliPix also uses in-pixel memory to only activate the pixels that need refreshing, while conserving power in inactive regions. This should be particularly useful for Augmented Reality (AR) or Virtual Reality (VR) applications that are battery powered.
According to the announcement, IntelliPix will be manufactured using GF’s 22FDX specialty semiconductor platform, which is a 22nm process node on Fully Depleted Silicon On Insulator (FD-SOI) wafers. FD-SOI was developed to reduce parasitic capacitance and current leakage between the source and drain in each transistor. It was originally targeted at Internet of Things (IoT) components that required low power consumption. Aside from the low-power benefits, density and interconnected features of the 22FDX platform, CP was able to customize additional features for the display and to simplify their manufacturing flow.
Edmund Passon, Co-CEO and CTO of Compound Photonics, and Ed Kaste, VP Industrial and Multi-Market Business Unit of GLOBALFOUNDRIES, have accepted an interview with DSCC to answer a few questions.
DSCC: Some of our readers may not be familiar with semiconductor nodes. Can you tell us what type of products are currently manufactured in 22FDX fabs?
Ed Kaste (GF): Electronic products built using GF’s 22FDX semiconductor platform are all around us – from Internet of Things devices including smart speakers, security cameras, WiFi6/BLE connectivity, to 5G wireless communications devices, to computer systems in your car and satellites in orbit.
Looking ahead, there are a range of exciting applications built on 22FDX currently in development by our customers: Next generation AI chips and AI accelerators, automotive radars, and a host of new wearables and hearables from a needle-free continuous glucose monitor to next-gen earphones.
And of course, augmented reality applications made possible by CP’s incredible IntelliPix platform!
With its best-in-class performance, power consumption, and broad feature integration capability, GF’s differentiated 22FDX platform is clearly the solution of choice for designers and innovators working in IoT, wearables, Edge AI, and other exciting applications. By empowering these customers to take advantage of the 22FDX’s inherent advantages, – increasing the performance, efficiency, and battery life of their devices – GF is enabling the industry to advance the frontier of connectivity toward a fully realized, global Internet of Things.
DSCC: Are there any specific challenges to meet the requirements for display backplane manufacturing?
Ed Kaste (GF): Semiconductor manufacturing is already so complicated, that the new challenges from display backplane feels normal to us. The interaction between optical modules with CMOS wafers further increased the complexity. In addition, we need to monitor optical performance, as well as our standard electric performance.
DSCC: What were the specific features of 22nm FD-SOI that were particularly attractive for a backplane design?
Edmund Passon (CP): There is a balance with respect to the increased logic and transistors under the pixel 22nm FD-SOI while still being able to handle and produce the larger voltage and current needs of driving pixels versus more standard digital logic applications. Enabled by GF’s leading 22FDX semiconductor platform, that balance was optimally maintained in GF’s 22nm FD-SOI while also providing lower power consumption given the feature set and performance. CP’s IntelliPix as instantiated by GF’s 22FDX platform sets a new benchmark for performance and power as applied to extremely demanding real-time AR based display systems.
DSCC: Can you tell us more about how you integrate the driver functionality to the backplane? Does it provide additional capabilities that are not possible with a discrete driver IC?
Edmund Passon (CP): The integration of the driver into the backplane helps on many fronts associated with power consumption and performance.
Firstly, it is enabling high speed signals between the drive and the backplane, for example a simple 1 bit/dumb pixel model can result in a huge bandwidth (~38Gbps for 1080p) and associated power consumption. Our integrated backplane plus logic under the pixel reduces this bandwidth and associated power while increasing both the feature set and performance.
Second, integrating the driver into the pixel allows for increased logic and storing of pixel information. This reduces the bandwidth and associated power.
Third, the increased logic enables key features like OnDemand Pixels™ where we can address only the active pixels. This can benefit the system in numerous ways: 1) Power Saving: Only active pixels are touched so power is only burned for those active pixels. 2) Performance Increase: Given the sparse nature of content for augmented reality (AR), we can focus all the bandwidth on active pixels objects. This allows for even higher frame rates for objects that do not take up the full resolution of the display.
Moreover, reducing cost and physical volume of the display module are also experienced, given a reduced chipset count and the lack of physical interconnect between the driver and backplane.
DSCC: Display backplanes are much larger than typical semiconductor IC. How can you minimize the impact on yield?
Ed Kaste (GF): Such display backplanes are actually not the largest IC we produce. GF offers best-in-class production yield, and we’re continuing along our roadmap to improve this yield while at the same time ramping up production. Display backplane will be a successful product line in our portfolio.
DSCC: Driving LCoS and MicroLED is usually very different. Can you tell us how you can support both types of displays with the same technology? Did you have to make compromise in some areas?
Edmund Passon (CP): We have previous experience applying our drive technology to both voltage-driven LCoS and current-driven microLED (IntelliPix-vDrive™ & IntelliPix-iDrive™) in our current NOVA microdisplay platform. Everything up to the final pixel drive circuitry is the same from a video pipeline standpoint. The final driver transistors are then configured for either vDrive or iDrive. This allows for a platform approach enabling an equivalent feature set and software development with only minimal parameter changes associated to the specific pixel drive. IntelliPix on 22nm FD-SOI allows for as much as 100x digital modulation speed to fully take advantage of the ultra-fast switching speeds of microLEDs. This huge increase in bandwidth product can be applied to features such increased bit depth, frame rates, gamma, depth planes and or other compensations required in the system thus reducing the traditional need to compromise on performance and or power consumption.
DSCC: For MicroLED, are there additional steps to facilitate bonding to the LED wafer? Is it important to manufacture the backplane on 12″ (300mm) wafers?
Edmund Passon (CP): For economies of scale, ideally we can work at 12” (300mm). There can be some intermediate stages to getting there but the full cost potential will be reached when both the light modular and backplane arrays exist at 12” (300mm).
DSCC: What is GlobalFoundries current production capacity with 22FDX? Do you have plans to increase the capacity?
Ed Kaste (GF): GF has tremendous growth potential ahead. While we don’t release production capacity for specific platforms, we can say that GF, to date, has realized more than $4.5 billion in design wins for 22FDX and shipped more than 350 million 22FDX chips to customers around the world.
All of our global Fabs, including Fab 1 in Germany where 22FDX is manufactured, are shipping record numbers of wafers and are at full capacity or even over-capacity. To meet increasing demand from our customers, we are increasing our production levels, accelerating our run rates, and installing new tools as quickly as possible.
In 2021, GF is investing to accelerate our capacity expansion, doubling our yearly average in capital expenditures to address the demand-supply mismatch and better serve our customers. Our CEO has said we will spend $1.4 billion this year to increase capacity at all of our fabs, including Fab 1 where 22FDX is manufactured.
DSCC: There have been chip shortages in some industries recently. Many displays for augmented and virtual reality will be manufactured on silicon backplanes, which will put more pressure on semiconductor foundries. How can the semiconductor industry prepare for increasing demand from the display manufacturers?
Ed Kaste (GF): GF is investing in our capacity growth to meet the world’s increasing demand for chips. We are doubling our yearly average in capital expenditures to further expand our capacity to better serve our customers and participate in the growth across our target markets.
Looking ahead, GF believes tighter partnerships across the semiconductor supply chain will benefit all parties. Tighter integration will create more supply chain certainty and security. Over time, co-investments and stronger partnerships between manufacturers/OEMs and semiconductor manufacturers will change the industry dynamic and help to mitigate the supply-demand mismatch we’re seeing today.
DSCC: Can you tell us more about the first product that will be based on a 22nm backplane? What type of application will it be?
Edmund Passon (CP): CP’s IntelliPix technology was developed to fill the demanding high performance and low power requirements of AR glasses. The first products based on CP’s IntelliPix as instantiated in GF’s 22nm FD-SOI are expected to be available in Q1 2023.
By modeling the human vision system in nature, CP has created the IntelliPix™ microdisplay platform that enables AR by addressing the challenge of combining the virtual and real worlds together and such that virtual objects are projected seamlessly in context of the real world. The platform design focuses on providing optimal performance in a compact form factor, while significantly reducing power consumption. With intelligence in the pixels, data can be encoded to the lowest bandwidth level while modulating what pixels are required and then giving only those the pixels the attention, they deserve.
