中文 English

Manufacturing Bits: June 1

Frozen finFETs; deepfake detection; photonics ink; printed radar.

popularity

Frozen finFETs
The Defense Advanced Research Projects Agency (DARPA) has launched the Low Temperature Logic Technology (LTLT) program, an effort to develop finFETs that operate at temperatures close to liquid nitrogen (~77K or minus 321 degrees Fahrenheit).

The goal of LTLT is to develop low-temperature finFET transistors at the 14nm node and below by making modifications to the manufacturing process. The resulting device technology should be capable of achieving a 25X improvement in performance/power compared to today’s central processing units (CPUs) at room temperature. LTLT also seeks to develop and demonstrate an SRAM cell that can operate at 77K.

Kelvin is a base unit of temperature, which uses absolute zero as its null point. On the Kelvin scale, pure water freezes at 273.15 K, and it boils at 373.15 K.

Today, the semiconductor industry is grappling with a major problem–the traditional way to advance a chip design is showing some cracks. Traditionally, to advance a design, IC vendors develop a system-on-a-chip (SoC) and then implement a concept called chip scaling. The idea is to shrink the transistors at each node and cram more of them on an SoC, enabling more functions at lower costs. But chip scaling is slowing and the price/performance benefits are shrinking at advanced nodes.

That’s where LTLT fits in. “Today, we’re aggressively reaching the end of Moore’s Law scaling and are faced with the inability to scale power density much further in order to improve computing performance,” said Jason Woo, a program manager in DARPA’s Microsystems Technology Office (MTO). “A viable solution is cold computing. While microelectronics is typically designed to operate at room temperature, we know that device characteristics improve significantly at reduced temperatures. Very low temperature devices – those operating at 77K or below – have the potential to overcome the power scaling limit, but challenges exist when you apply them to very large scale integration.”

LTLT aims to exploit the device/material characteristics and performance of today’s finFETs. The big difference is that the transistor technology is operating at very low temperatures. This in turn will enable chips with better performance/power.

The program is broken out into two research areas. The first will focus on researching and developing a fabrication technology, which can enable low-voltage devices operating at 77K. The technology will be able to integrate low- temperature transistors and SRAM cells with 25X lower switching energy at 77K.

The second area in the program will explore advanced research concepts on low-temperature finFETs. Three specific challenges will be explored, which include ultra-low power, high-speed scaled transistors with new switching mechanisms; low-energy SRAM cells; and new circuit techniques.

Deepfake detection
Researchers at the U.S. Army Combat Capabilities Development Command (DEVCOM) and the University of Southern California have developed a deepfake detection method called DeFakeHop, adding important knowledge in AI, intelligent scene understanding, and face biometrics.

Deepfake is an AI-synthesized, hyper-realistic software that modifies existing images and videos and replaces them with fake ones. This fake likeness is then manipulated to say or do things. Deepfake poses a significant threat to society and national security, since it increases the realism of fake content.

“AI-driven deepfake advances so rapidly that there is a scarcity of reliable techniques to detect and defend against deepfakes,” said Suya You, a researcher at the Army Research Laboratory. “There is an urgent need for an alternative paradigm that can understand the mechanism behind the startling performance of deepfakes and develop effective defense solutions with solid theoretical support.”

DeFakeHop’s key innovation is called Successive Subspace Learning (SSL). SSL is a new mathematical framework for neural network architectures that is based on signal transform theory. SSL is suitable for high-dimensional data, and is a new tool for image processing and understanding face biometrics. Using SSL, DeFakeHop has several advantages over existing deepfake detection software, such as robustness, scalability and portability.

The military plans to use DeFakeHop’s low size–weight–power vision-based devices on the battlefield. Using DeFakeHop, the military hopes to detect fake content and add to its knowledge of intelligent scene understanding and face biometrics. “(DeFakeHop) has quite a few desired characteristics, including a small model size, requiring limited training data, with low training complexity and capable of processing low-resolution input images. This can lead to game-changing solutions with far reaching applications to the future Army,” You said.

Photonics ink
The Air Force Research Laboratory has approved a Cooperative Research and Development Agreement with Iris Light Technologies to develop hybrid silicon lasers for use in silicon photonics.

Iris has developed a photonic ink, which is used to print laser gain materials onto passive silicon chips. The ink is capable of emitting light over a broad spectrum from visible to the near infrared.

Silicon photonics, a promising field for the data center and other apps, makes use of photonic integrated circuits (PICs). PICs use photons, which move at the speed of light. “This greatly increases the bandwidth (the data transfer rate) and speed of the circuit, without big energy losses making PICs significantly more efficient than their IC counterparts,” according to Wevolver, a hi-tech Website.

But there are various challenges with the technology. Integrating silicon-based integrated circuits with photonic elements like lasers on the same device is challenging.

“Light sources (lasers, the engine of photonic components) are very challenging to develop in silicon photonics due to the indirect bandgap of silicon (a horizontal shift between the valence and conduction band of the material),” according to Synopsys. “For light to be generated, a material needs to have a direct bandgap. Therefore, other materials with direct bandgap, such as indium phosphide (InP), are used to create the lasers, and they are integrated in the silicon photonics wafer (chip) to drive the photonic components within the photonic circuit.”

AFRL is investigating silicon photonics with a goal of finding a better way to fabricate hybrid devices that make possible the integration of lasers onto silicon chips. The goal is to find a less expensive and more reliable way to manufacture the on-chip lasers. One method AFRL is investigating involves a type of “photonic ink” developed by Iris Light Technologies.

“The novelty with this technology is that we are taking a plain silicon chip, which is easy to make, with no gain on it, and no light sources. During the processing of the chip, an ink material is printed onto the chip,” said AFRL research scientist Steven Mckeown. “The ink will be what converts the energy into the laser light. The thing that actually kind of shapes the light and carries it and guides it is the silicon. It interacts with the ink in that it converts energy into laser emission.”

Printed radar
The U.S. Naval Research Laboratory has developed and tested 3D-printed antennas and arrays to advance radar technology.

Source: U.S. Naval Research Laboratory. Photo by: Jonathan Steffen

Radar is critical for maritime navigation and defense. Parts for antennas and arrays may break or wear out requiring replacement. Traditionally, parts are ordered or intricately machined out of metal, sometimes taking weeks to produce. The lightweight and rapid production of 3D-printed parts make it an attractive alternative to traditional manufacturing that often requires expensive materials and specialized equipment.

“3D printing is a way to produce rapid prototypes and get through multiple design iterations very quickly, with minimal cost,” said NRL electrical engineer Anna Stumme. “The light weight of the printed parts also allows us to take technology to new applications, where the heavy weight of solid metal parts used to restrict us.”



Leave a Reply


(Note: This name will be displayed publicly)