Published in Military Embedded Systems
Two major challenges confront developers of military radar-processing systems. The first challenge is that modern multiband radar sensors produce huge amounts of data that need to be brought into the system's digital-processing stage as accurately and rapidly as possible in order to generate actionable data for the warfighter. The second challenge in this arena is the rapid rate of change that missions must respond to, as adversaries continually morph and evolve their tactics and develop more sophisticated technologies.
As recently as ten years ago, legacy radar sensors might have only supported a single band, such as KU-Band, X-Band, or S-Band. These single-band point solutions produced a particular data rate which had specific downstream data and processing effects. Each radar system would have a fixed set of functions and attack only a particular band of operation within which intelligence had identified that adversaries were operating.
As threats have become more sophisticated, radar sensors have, too. Now, radar signal detection must be performed over multiple bands, meaning that the sensors have had to become much smarter and more sensitive. The result? Radar sensors today are collecting and sending significantly more data to the radar processor. To support these new sensors, radar-system developers want to use underlying digital processing technology that can accept and handle data from a broad variety of different sources. This approach, compared to the earlier point solution, effectively eliminates the cost and time needed to develop a unique digital processor every time a new threat is identified.
The point solution
Sensor requirements for radar, electronic warfare (EW), and signal intelligence (SIGINT) applications usually differ from application to application. The investment required to develop a new system from the ground up for every application can be costly and time-intensive. For instance, every antenna design differs depending on the application space or where in the electromagnetic spectrum the system designer is trying to attack, sense, or operate.
FPGA [field-programmable gate array] devices, a key component in radar-processing systems, are continuing to evolve and add more capability such as floating-point math, increased local memory, faster sensor I/O channels, local processors, and embedded RF analog I/O. Functions such as digital down converters (DDC) are particularly well suited for FPGAs, enabling extraneous data to be removed or dynamically scanned, ensuring that later processing chains are not flooded with data that slows the system down. FPGAs almost uniquely have the ability to support the high processing speeds needed to handle and process the vast sensor data bandwidths typical of radar-system applications. That capability makes these devices exceptional technology for the front end of any high-performance system.
FPGAs are ideal for performing math-intensive algorithms such as Fast Fourier Transforms (FFT) on the incoming raw sensor data stream. After the key data has been extracted by the FPGAs, it can be sent to devices such as DSPs [digital signal processors] or GPGPUs [general-purpose graphics processing units] that can provide even more sophisticated processing, but on smaller data sets better suited to the throughput limits of those device types.
For many years, radar system designers turned to in-house-designed FPGA module solutions that targeted a specific application. In the past, radar systems often used costly custom or semicustom FPGA technology, in a system designed with a specific program or purpose in mind. While dedicated point solutions have their place, such systems lack the flexibility needed to address a wide variety of applications. To support multiband radar, a processing system needs to be flexible and – ideally – reconfigurable. One downside to custom and semi-custom FPGA module developments is that they tend to require large amounts of resources to develop and maintain. Moreover, because they are typically designed for a specific point solution, it’s rarely practical to leverage the investment in these devices across multiple applications.
Read the full article here