Electronic warfare (EW) is the struggle for control of the electromagnetic spectrum. Electronic support (ES) intercepts, identifies, and locates adversary emitters; electronic attack (EA) disrupts or deceives those emitters; and electronic protection (EP) defends friendly systems from adversary EW. Radar and EW are deeply intertwined — every radar is a potential EW target, and many EW techniques exploit the same signal processing principles that radar depends on. This article examines the signal processing foundations of modern electronic warfare.

Wideband Digital Receivers for ES

Electronic support begins with the receiver. Unlike radar receivers optimized for a specific waveform and bandwidth, ES receivers must cover enormous frequency ranges — typically 0.5 to 18 GHz, extending to 40 GHz and beyond for millimeter-wave threats — with sufficient instantaneous bandwidth to capture frequency-agile signals. Channelized receivers using polyphase filter banks divide the wideband input into many narrowband channels, each processed independently for detection and parameter measurement.

Key measured parameters include: carrier frequency (with resolution of a few megahertz or better), pulse width (from tens of nanoseconds to continuous wave), pulse repetition interval (PRI) and its modulation pattern, amplitude, and angle of arrival (AOA). Interferometric and monopulse techniques provide AOA accuracy better than one degree, while time-difference-of-arrival (TDOA) and frequency-difference-of-arrival (FDOA) between multiple platforms enable precision geolocation.

Signal Classification and Emitter Identification

Raw pulse descriptor words (PDWs) — tuples of frequency, pulse width, amplitude, AOA, and time of arrival — are the currency of ES processing. Deinterleaving algorithms sort interleaved pulse trains from multiple emitters into separate tracks, using PRI analysis, scan pattern recognition, and clustering techniques. Once deinterleaved, the emitter’s parameters are compared against electronic order of battle (EOB) databases to identify the specific radar type, platform, and operating mode.

Machine learning is transforming emitter identification. Deep neural networks trained on large libraries of measured and simulated emitter signatures can classify emitters with accuracy exceeding 95%, even in dense signal environments where traditional template matching fails. Specific emitter identification (SEI) goes further, using unintentional modulation features — subtle variations in pulse shape, frequency stability, and transient behavior caused by manufacturing tolerances — to identify individual radar units, a capability of immense tactical value.

Electronic Attack Waveform Generation

Electronic attack degrades or deceives adversary sensors. Noise jamming raises the noise floor, reducing detection range. Deceptive jamming produces false targets by generating realistic radar returns with manipulated range, Doppler, and angle parameters. Digital radio frequency memory (DRFM) technology captures, stores, and retransmits adversary radar pulses with precise modifications, creating coherent false targets that are extremely difficult to discriminate from real ones.

Modern DRFMs digitize the received signal at RF or IF, store it in high-speed memory, apply the desired modulation, and retransmit through a DAC and power amplifier — all within a few tens of nanoseconds. Advanced DRFMs can generate multiple false targets with independent kinematics, simulate complex scattering behaviors, and adapt jamming techniques in real time based on the adversary radar’s observed response.

The EW Processing Challenge

EW processing operates under extreme time pressure. The entire detect-identify-locate-decide-jam cycle must complete in milliseconds to be tactically relevant. This demands hardware acceleration for the throughput-intensive front-end processing and tight integration between ES, command-and-control, and EA subsystems. As radars become more agile and adaptive, the EW processor must keep pace — an arms race that shows no sign of slowing.