Delay-Doppler Domain

In the Delay-Doppler domain, a wireless channel is represented by a small number of discrete scatterers, each characterized by a specific delay (distance) and Doppler shift (velocity). Unlike the time-frequency domain, where a fast-moving reflector causes a complex, smeared response across many symbols, the Delay-Doppler representation collapses this energy into a single, sharp peak. This inherent sparsity is the key enabler for efficient channel estimation and equalization, as the entire channel geometry can be captured with far fewer parameters than the number of time-frequency resources.

The defining advantage of the Delay-Doppler domain is its temporal stability. While the time-frequency response of a high-mobility channel fluctuates rapidly, the underlying physical geometry—the positions and velocities of scatterers—changes slowly relative to the signaling frame. A scatterer's delay and Doppler shift remain approximately constant over a much longer duration. This quasi-static nature means that a single channel estimate in the Delay-Doppler domain remains valid for the entire transmission frame, drastically reducing the pilot overhead required for tracking fast-varying channels.

The mathematical bridge between the time-frequency domain and the Delay-Doppler domain is the Symplectic Finite Fourier Transform (SFFT). This transform maps a 2D grid of symbols from one domain to the other. At the transmitter, the Inverse SFFT (ISFFT) spreads each QAM data symbol across the entire time-frequency frame, creating resilience against narrowband interference. At the receiver, the SFFT converts the received time-frequency signal back to the Delay-Doppler domain, where the channel interaction simplifies to a 2D convolution with a sparse kernel.

Because the ISFFT spreads every information symbol uniformly over the full bandwidth and time duration of the frame, each symbol experiences the full diversity of the channel. This is fundamentally different from OFDM, where a deep fade on a single subcarrier can destroy the symbols assigned to it. In the Delay-Doppler domain, all symbols suffer the same average signal-to-noise ratio. This property provides robustness against frequency-selective fading and makes the modulation scheme naturally resilient to the high Doppler spreads encountered in mmWave and high-speed rail scenarios.

In the Delay-Doppler domain, the effect of the wireless channel is modeled as a 2D circular convolution with a sparse kernel. This kernel is formed directly by the discrete delay and Doppler taps of the physical scatterers. The sparsity of this operator enables the use of low-complexity message-passing algorithms for detection. Because the interaction is localized to a few non-zero taps, the receiver can perform iterative interference cancellation efficiently, achieving near-optimal performance without the prohibitive complexity of inverting a dense time-frequency channel matrix.

Operating in the Delay-Doppler domain provides a direct interface to the physical world for integrated sensing and communication (ISAC). The estimated channel response in this domain is a radar-like image, where each detected peak directly corresponds to a physical reflector's range (delay) and radial velocity (Doppler). This allows a single waveform to simultaneously perform data transmission and environmental sensing. Neural networks processing Delay-Doppler images can natively perform object detection and classification, fusing communication and perception into a single waveform.

Orthogonal Time Frequency Space (OTFS) is a 2D modulation scheme that multiplexes information symbols directly in the Delay-Doppler domain. Unlike OFDM, which places symbols in the Time-Frequency domain, OTFS spreads each symbol across the entire time-frequency frame. This transforms a time-varying channel into a time-invariant, sparse 2D convolution channel, enabling robust performance in high-Doppler environments like vehicle-to-everything (V2X) and high-speed rail. The key advantage is that all symbols experience nearly identical, averaged channel conditions, eliminating the severe inter-carrier interference that cripples OFDM at high speeds.

The Delay-Doppler spreading function, denoted as h(τ, ν), is the complete characterization of a linear time-variant wireless channel in the Delay-Doppler domain. It represents the complex gain of the channel for each specific delay (τ) and Doppler shift (ν) pair. This representation is inherently sparse because a typical wireless channel consists of only a small number of physical reflectors, each with a distinct delay and Doppler shift. This sparsity is the core reason why signal processing in the Delay-Doppler domain is so efficient for channel estimation and equalization.

The Symplectic Finite Fourier Transform (SFFT) is a 2D transform that maps symbols between the Delay-Doppler domain and the Time-Frequency domain. It is the practical, discrete implementation of the Zak transform for OTFS modulation. The transmitter first applies an Inverse SFFT (ISFFT) to convert the Delay-Doppler QAM symbols to a Time-Frequency grid, followed by a conventional OFDM modulator. At the receiver, the SFFT converts the received Time-Frequency samples back to the Delay-Doppler domain for channel estimation and equalization.

The Time-Frequency domain is the standard signal representation used by OFDM and 4G/5G systems, where symbols are placed on a grid of subcarriers (frequency) and time slots. In contrast to the Delay-Doppler domain, a high-mobility channel in the Time-Frequency domain is rapidly time-varying, causing each OFDM symbol to experience a different channel. This destroys orthogonality between subcarriers and leads to inter-carrier interference (ICI). The Delay-Doppler domain is preferred for high-mobility because it transforms this challenging, time-varying interaction into a stable, time-invariant one.

Channel sparsity is the fundamental property that makes the Delay-Doppler domain so powerful. A wireless channel is composed of a few dominant multipath reflectors. Each reflector is characterized by a specific delay (distance) and a specific Doppler shift (relative velocity). In the Delay-Doppler domain, this channel is represented by only a few non-zero taps, making it highly sparse. This sparsity enables the use of efficient compressed sensing algorithms for channel estimation with minimal pilot overhead, a critical advantage over dense estimation in the Time-Frequency domain.