Saturday, February 1, 2003
Chips That Think: RF Design
Radio Frequency design is on the verge of a revolution at the component level, which could produce self-adapting, "intelligent" microsystems.
In the 1980s radio frequency (RF) engineers developed microwave and millimeter-wave integrated circuits that revolutionized the reliability and controllability of RF systems, while significantly reducing their size and cost. But now a second revolution is in view: active, chip-sized "microsystems" that automatically monitor the RF environment and adjust themselves to operate over a wide frequency band.
Radically new "intelligent" components could migrate to operational systems in the next decade. On the horizon: multifunctional RF systems for the largest aircraft or surface ship to the smallest microdrone.
Defense firms, universities and government centers have pursued RF component adaptability for years. But last year the Defense Advanced Research Projects Agency (DARPA) focused efforts in a new Intelligent RF Front-End (IRFFE) program. IRFFE aims to develop "intelligent" building blocks, such as power amplifiers, that can adapt themselves in real time to changing mission requirements.
"We’re trying to achieve levels of polymorphism in RF components, changing the form and function of RF systems," says Edgar Martinez, IRFFE program manager. The new devices typically will integrate micro electromechanical systems (MEMS), RF transistors and digital circuitry as self-contained units.
Increased integration and miniaturization will produce smaller, more flexible RF systems. Electronic warfare (EW) front-end electronics, for example, could be made five times smaller, claims John Windyka, BAE Systems’ director of advanced microwave devices. Adaptable components will pave the way to enhanced radar, EW, missile seeker, satcom and_ground communication systems, as well as multifunctional arrays. Self-calibration at the component level also could help compensate for aging effects, transient faults or design and manufacturing errors, increasing reliability and decreasing costs.
DARPA wants to develop "chips that can think," Martinez says. Solid state microwave chips today–known as a monolithic microwave integrated circuits (MMICs)– are "dumb devices," tuned to perform at narrow bands.
"What we’re going after is chips with the capability of self-assessment and self-optimization, in real time, for wide adaptability to different operational demands," Martinez adds. "We’re trying to design tunable, adaptable components with very narrowband performance over a very broad bandwidth."
Tunable networks mean more power out, less power lost, and greater ability to extract small signals from large interfering signals. IRFFE approaches range from the integration of sensor, actuator and control functions into a two-dimensional system to the use of optical-like processing in a three-dimensional, cubic array.
IRFFE players include: BAE Systems (EW), Rockwell Scientific (communications, seekers), Lockheed Martin (shipboard multifunctional RF), Raytheon (multifunctional RF, radar), Massachusetts Institute of Technology/_Lincoln Lab (EW, multifunctional radar/com), Mayo Foundation (dynamic testing techniques) and TRW (millimeter-wave transmitters and receivers), as well as Colorado University (adaptable amplifiers), the University of Virginia (tunable transmission line) and Wright State University (intelligent control and self-test). Lockheed Martin and Raytheon also are working on the U.S. Navy’s Advanced Multifunction Radio Frequency-Concept (AMRF-C) program, which focuses on nearer-term multifunctional shipboard RF technology.
The first, 12- to 18-month phase of the DARPA program develops concepts and demonstrates subcomponents, such as impedance matching networks and control and optimization algorithms. The second phase (funding for which is contingent on Phase 1 success) integrates the pieces into proof-of-concept prototypes. And the third phase demonstrates the components in military applications.
Impedance matching networks, critical to maximize amplifier gain (or minimize signal loss), are typically designed around a specific center frequency. If an amplifier operates at different frequencies, the network could be adaptable. Or there could be multiple impedance matching networks, corresponding to various frequencies, and one or several MEMS switches to select the right network.
Five of the contractors take two-dimensional RF integrated circuits as the baseline. Rockwell Scientific, however, is using three-dimensional, waveguide structures and optical-like processing techniques for applications in the 25- to 45-GHz range. The company is "projecting an RF signal into an array of amplifiers," using active waveguides "to concentrate energy through free space to generate a higher-power signal," Martinez explains. Raytheon is working in the 2- to 18-GHz band, TRW is in the 20- to 50-GHz band, and BAE Systems and MIT/Lincoln Labs are in lower-frequency EW bands.
BAE Systems is designing transmit and receive amplifiers whose performance can be tuned anywhere in the 6- to 18-GHz band. "For example, we can tune the receiver, so it is operating at X-band, but tune out interfering signals at other frequencies, and then reconfigure at a different frequency–all autonomously," Windyka says. On the receiver side, the company wants an amplifier that provides around 200-MHz instantaneous bandwidth that can be tuned over the frequency range."
BAE Systems wants to develop high power amplifiers for EW transmitter applications, tunable for higher power and efficiency than a wideband amplifier can provide today. Impedance matching can be optimized for the exact frequency of operation, maximizing power and efficiency, compared with a typical wideband design. BAE aims for an approximately 25 percent increase in output power and efficiency.
Some functions–such as the detection of small signals in the presence of large signals–now are performed digitally in the signal processing section. In new intelligent microsystems, these functions could be performed at the front-end, right behind the antenna. "We’re trying to do prefiltering and presorting by applying some intelligence to the signals coming through, so that the signal processor just has to focus on the signals of interest," Windyka says.
BAE is looking at both "monolithic" and "heterogeneous" integration approaches, where elements are combined using the same or different semiconductor processes, respectively. The company’s monolithic approach would combine transistors, sensor probes, tuning networks and amplifiers on an approximately 5-by-5-mm (0.2-inch-by-0.2-inch) surface, using external digital logic. The heterogeneously integrated module would be 1.5 to two times larger.
MEMS devices basically can perform three functions in RF microsystems:
Switching, a simple on/off process that can connect elements into the circuit or put them off line;
Tuning circuits in frequency, bandwidth, capacitance and other electrical characteristics; and
Reshaping the path of the electrical signal in a circuit.
"The challenge is integrating [the functions] into active circuits with long-term reliability," Martinez says. "The fabrication processes have not been demonstrated to be compatible." Electrical and electromechanical devices, made with different materials and behaving differently, could interfere with each other.
Another challenge is local digital control of analog circuits. Based on RF sensor input, the digital controller decides whether and how to reconfigure the microsystem. MEMS-based or other actuators then execute the controller’s decisions.
Analog signal parameters have to be converted to digital, assessed and then reconverted to analog. Factors such as signal amplitude and frequency, amplifier temperature, voltage, current level, humidity and power in/out have to be expressed digitally. And digital control decisions have to be made much faster than the microsecond [millionth of a second] time scale for MEMS switching.
Designers will have to make tradeoffs between the control circuitry speed, the control precision and the complexity of the algorithms. The controller will be closely integrated with the analog functions, as a microprocessor, field programmable gate array (FPGA) or embedded digital circuitry.
Rockwell Scientific is taking a different tack. It plans a 10-watt power amplifier tunable from 25 to 45 GHz, a five-fold improvement over what is available commercially at the high end of that range. Applications include satcom, seeker and ground communications. While the focus is on the transmit side, the architecture could lend itself to the receiver side, as well, says Mark Rosker, manager of Rockwell Scientific’s RF Circuits and Applications Department.
Rockwell Scientific’s approach is based on the idea of waveguide, Rosker explains. Waveguide typically is a passive material, shaped like a hollow pipe, which moves energy from one point to another through free space. Rockwell, however, uses waveguide sliced into rectangular cubes and studded with active electronics (see illustration, page 29). The cubes are stacked in a three-dimensional assembly a few centimeters long. Each modular, four-element, active waveguide is about 12 millimeters on a side. A central control unit is implemented externally.
Although the DARPA program will revolutionize component design, it envisions a "more or less conventional view of the interconnect between the component and the rest of the world," says Rosker. Instead of wire-like microstrip to move signals from input to output, Rockwell Scientific uses waveguide to couple signals through free space and optical-like processing techniques to provide at least 10 dB gain.
At the center of the stack is an amplifier array, the core "quasi-optic" grid. This array of 100 or more amplifiers, each operating on a small part of the signal, is more reliable than traditional microwave circuits, which may fail if they lose a single gate, Rosker claims. It will continue to operate even though up to 10 percent of the devices fail or are out of spec.
The approach is called quasi-optics because the signal is transmitted in a wave propagating in space and acted on by elements separated by distances that are small, compared to a wavelength. This situation has much in common with optics, as compared to traditional, transmission line electronics, Rosker explains.
Sandwiching the core are the active waveguides, which, like a conventional impedance matching network, couple energy into and out of the amplifier. Additionally, the waveguides control the signal’s amplitude and phase to create a constant wave front impinging on the amplifier array. Ideally, each amplifier sees exactly the same signal and performs the same operation in parallel. Rockwell uses 2-by-2-element waveguide arrays to demonstrate power scalability. The waveguide (and surrounding system) can scale modularly to 100 watts output.
There is a cost benefit, as well. Rockwell’s active waveguide enables designers to separate impedance matching from amplifier circuitry, allowing the fabrication of lower-cost homogeneous amplifier and impedance matching arrays.
On either side of the active waveguide elements is an array of tunable elements, such as MEMS devices, forming part of the amplifier’s impedance matching network. Because hundreds, or even thousands of MEMS devices can be placed on a small surface, an individual device will see less than 1 percent of the signal’s total power. If the devices in this section are MEMS structures, this would provide "a superb means of overcoming the power and speed limitations inherent in MEMS devices," Rosker says. At the far front and back of the stack are modular waveguides that channel the signal in and out of the system. These contain probes to measure power in and out, a key part of the dynamic feedback loop required for high performance.
Lockheed Martin is working on self-adaptive amplifiers in the receiver chain of phased array antennas, operating in the 3- to 25-GHz band. A key application is multifunctional shipboard RF, but the company continually assesses the technology’s readiness to insert into other applications, such as ballistic missile defense and anti-air warfare weapons. It aims for an architecture with inherent self-adaptivity to allow peak performance even in the complex battle group environment, where sensors are exposed to a barrage of interference from other transmitters, according to Tom McNellis, Lockheed Martin’s manager of radar technologies programs.
The company plans to achieve this architecture, while maintaining critical performance parameters, such as a spurious-free dynamic range of 50 to100 dB and noise suppression of 30-50 dB. This will allow multifunctional radar/communications systems to detect a weak signal in the presence of heavy clutter or jamming. The first statistic refers to the system’s ability to accurately receive very small signals in the presence of signals up to 10 billion times stronger. The noise suppression measure refers to the system’s ability to reduce internal noise as well as spurious signals from the outside by up to 100,000 times in power.
Lockheed Martin relies on an all-electronic implementation rather than MEMS to tune amplifier performance over the frequency band, "nulling" out signals outside of the band of interest. This provides higher performance and inherently higher reliability, the company claims.
Lockheed Martin’s approach to the DARPA challenge of designing chips that "think" is to use an all-electronic GaAs (gallium arsenide), adaptable MMIC that is heterogeneously integrated in a module with a low-power, external CMOS (complementary metal oxide semiconductor) mixed-mode controller element.
The company aims at very broadband receivers that handle multiple frequencies at the same time, with very broadband input and very wide instantaneous bandwidth. This is a challenging application, McNellis stresses. "Numerous interference sources combine with other limitations, such as system phase noise, intermodulation products, power supply stability, device fabrication tolerances, thermal variations and even design model inaccuracies."
Related DARPA Programs
Technology for Efficient Agile Mixed-Signal Microsystems (TEAM) — developing low-cost modules that integrate radio frequency (RF) and digital circuitry onto a single semiconductor device.
Technology for Frequency Agile, Digitally Synthesized Transmitters (T-FAST) — exploring the scaling limits of indium phosphide bipolar devices for very high-speed and mixed-signal applications, including five-fold reduction in power consumption.
Vertically Interconnected Sensor Array (VISA) — developing advanced interconnect concepts to bridge the gap between analog and digital domains in three-dimensional, mixed-signal microsystems.
NeoCad — developing new design methodologies and tools for mixed-signal systems.
The Advanced Multifunction Radio Frequency-Concept (AMRF-C) program, sponsored by the Office of Naval Research (ONR), targets the development of multifunctional shipboard RF systems. Funding component research and development at the rate of $9 million to $10 million a year, the program is increasing expertise in areas such as low noise amplifiers, high power amplifiers, circulators (signal isolators) and filters.
AMRF-C is looking at some of the same applications and technologies (micro electromechanical systems, for example) as the longer-term Intelligent RF Front-End (IRFFE) program, sponsored by the Defense Advanced Research Projects Agency (DARPA). Ideas generated from a recent ONR study contract on "Version 2" of the AMRF-C test bed will be examined by the IRFFE program.
AMRF-C is attempting to realize radar, communications and electronic warfare (EW) functions in a single, small set of multifunctional transmit and receive arrays, shrinking topside "antenna farms" and aggregate radar cross-section and reducing maintenance costs. Candidate ships include next-generation aircraft carriers, cruisers and destroyers, as well as the new-concept littoral combat ship. A smaller ship’s "island" could permit future carriers to significantly increase their sortie rate.
Each of the functional areas could benefit from the AMRF-C approach. Communications systems today, for example, "are relatively narrowband–hundreds of megahertz," says Joseph Lawrence, director of ONR’s Surveillance, Communications & Electronic Combat Division. "We’re going to broaden that to an operating bandwidth of multiple gigahertz." Shipboard electronic warfare systems would have better accuracy and the ability for simultaneous transmissions. And the radar function, which could use very long pulses or coded, continuous wave (CW) transmissions, could more easily pick out targets against clutter.
AMRF-C plans to launch a test bed in 2004. The initial test bed system will employ a 1,000-element transmit array–producing more than 1 megawatt of power–and a 1,000-element receive array.
In two related projects, Raytheon and Northrop Grumman are developing tunable filters that operate over the 6- to 18-GHz and the 4- to 20-GHz bands, respectively, in both the transmit and receive chains, says Ingham Mack, program manager in ONR’s Electronics Division. These filters can be programmed to let a desired frequency band through (bandpass filter) or block undesired frequencies (band rejection filter). "We have to get a tunable filter because we want to dynamically allocate the functions on the apertures," Lawrence says.
The tunable bandpass filter contains multiple sections, each of which is tuned to resonate, or pass a signal through, at a certain frequency. Each "resonator" is composed of multiple MEMS devices, which help to tune the "resonant frequency" by switching between high and low capacitance values.
ONR wants maximum flexibility and performance over the band to:
Keep the center frequency constant and change the bandwidth,
Keep the bandwidth constant and move the center frequency, and
Simultaneously move the center frequency and vary the bandwidth.
MEMS-based tunable filters will be smaller and less power-hungry than today’s designs. Equally important, they will be "linear," so that a signal coming in at a certain frequency will come out only at that frequency. While ONR aims for proof-of-concept prototypes with a small number of switches within each resonator element, many hundreds or more switches may be needed in the overall filter for wideband performance.