PaperlayerView exact revision on arXiv

Source-linked AI summary

An Overview of Signal Processing Techniques for Terahertz Communications

Hadi Sarieddeen, Mohamed-Slim Alouini, Tareq Y. Al-Naffouri

arXiv:2005.13176v3eess.SP

TL;DR

THz communications require signal-processing methods that address severe propagation losses, power constraints, and mismatches between wideband channels and digital baseband hardware. This tutorial synthesizes THz channel, waveform, beamforming, estimation, coding, detection, sensing, and system-design techniques, emphasizing UM-MIMO and reconfigurable surfaces. It reports channel-dependent performance frameworks and optimized modulation approaches while identifying substantial complexity and energy challenges for practical Tbps systems.

  • Problem

    THz systems face short communication distances, severe power limitations, wideband channel effects, and energy-intensive baseband and array processing that constrain practical Tbps operation.

  • Method

    The paper presents a tutorial synthesizing THz channel models, performance frameworks, testbeds, and signal-processing techniques across modulation, beamforming, estimation, coding, detection, sensing, imaging, localization, networking, and security.

  • Results

    Water-filling achieves more than 75 Gbps with 10 dBm over 0.06−1 THz, while distance-aware multi-carrier schemes achieve Tbps rates over 10 m.

  • Takeaways & Limitations

    Efficient, jointly optimized signal processing and hardware-aware architectures are needed to support THz systems beyond 100 Gbps and toward Tbps data rates.

Abstract

from arXiv · show

Terahertz (THz)-band communications are a key enabler for future-generation wireless communications systems that promise to integrate a wide range of data-demanding applications. Recent advances in photonic, electronic, and plasmonic technologies are closing the gap in THz transceiver design. Consequently, prospect THz signal generation, modulation, and radiation methods are converging, and corresponding channel model, noise, and hardware-impairment notions are emerging. Such progress establishes a foundation for well-grounded research into THz-specific signal processing techniques for wireless communications. This tutorial overviews these techniques, emphasizing ultra-massive multiple-input multiple-output (UM-MIMO) systems and reconfigurable intelligent surfaces, vital for overcoming the distance problem at very high frequencies. We focus on the classical problems of waveform design and modulation, beamforming and precoding, index modulation, channel estimation, channel coding, and data detection. We also motivate signal processing techniques for THz sensing and localization.

I. INTRODUCTION

THz communications are being developed for future high-bandwidth systems as advances in electronic, photonic, hybrid, and plasmonic devices close transceiver gaps. Their distinctive propagation, hardware, and scaling constraints require specialized signal processing, efficient architectures, and continued standardization.

  • THz Motivation: THz communications can provide terabits-per-second rates by exploiting broad available spectrum, while shorter wavelengths improve directionality, reduce diffraction and inter-antenna interference, and support smaller footprints.These properties motivate applications including backhaul, data-center connectivity, and kiosk-to-mobile links.
  • THz Devices: The THz gap reflects limited compact signal sources and detectors, although electronic and photonic advances now support more efficient signal generation, modulation, and radiation.Electronic approaches emphasize compact CMOS and SiGe BiCMOS integration, whereas photonic approaches support higher carrier frequencies but have lower integration and output power.
  • THz Devices: Hybrid electronic-photonic and plasmonic technologies extend THz capabilities through reconfigurability, compact arrays, and direct frequency operation, but hybrid systems require more precise transmitter-receiver synchronization.Graphene-based plasmonic devices can provide direct THz sources, modulators, and on-chip antenna arrays while avoiding electronic and photonic frequency-conversion losses.
  • THz Devices: 80% power reduction was reported for a fully digital 140 GHz receiver using 90 nm SiGe BiCMOS, with 2 GHz sampling and power consumption below 2 W.True-THz device power consumption remains insufficiently established, making energy-efficient architectures an important deployment issue.
  • Standardization: The paper addresses THz signal-processing solutions for standardized links using a quasi-deterministic link-level model with known antenna directions, while also considering interference, mobility, and multiple-channel access.IEEE 802.15.3d covers sub-THz channels from 253–322 GHz with bandwidths up to 69 GHz, and proposed waveforms exploit 99.3% of total in-band energy.
  • Signal-Processing Challenges: UM-MIMO is used to address short THz communication distances caused by severe power limitations and propagation losses, while signal processing must also bridge THz-channel and digital-baseband bandwidth mismatch.Because channel coding is computationally demanding, efficient and parallelizable processing across the complete baseband chain is required; sparsity and low-resolution conversion can reduce complexity.

E. Paper Contributions and Outline

The paper surveys THz signal-processing advances and links devices, channel and noise models, and algorithms to communication, sensing, and localization challenges. It emphasizes practical, low-cost, low-power, reliable, and low-latency processing for systems constrained by limited digital baseband capabilities.

  • The overview addresses the gap between advancing THz devices and limited baseband-processing capabilities.
  • It links novel THz devices, channel and noise models, and signal-processing techniques.
  • The paper surveys system-level and algorithmic challenges, emphasizing hardware-aware processing for promised Tbps data rates.
  • Its signal-processing scope includes modulation and waveform design, spatial tuning, reconfigurable arrays, beamforming, precoding, channel estimation, coding, and detection.
  • System model: The adopted 3D UM-MIMO model uses M_t×N_t and M_r×N_r subarrays, each with Q×Q antenna elements, forming a large doubly-massive MIMO system.
  • System model: The generic model assumes highly directional THz propagation dominated by line-of-sight transmission with few reflected multipath components.

III. THZ-BAND CHANNEL MODELING

THz channel modeling must represent spreading and molecular absorption losses alongside LoS, NLoS, reflected, scattered, and diffracted propagation. The literature therefore combines deterministic ray tracing with statistical models and simulators spanning diverse frequencies, environments, and communication distances.

  • Accurate THz channel models must capture spreading loss, molecular absorption loss, and LoS, NLoS, reflected, scattered, and diffracted signals.
  • Deterministic approaches use computationally extensive ray tracing to represent site geometry and propagation paths.
  • Statistical alternatives model measured propagation behavior across indoor, outdoor, LoS, and NLoS scenarios without relying solely on fixed-geometry ray tracing.
  • TeraMIMO provides a 3D end-to-end wideband UM-MIMO simulator modeling coherence time, coherence bandwidth, Doppler spread, and RMS delay spread.

C. Effect of Scattering

THz propagation is shaped by frequency-dependent molecular absorption, atmospheric conditions, surface scattering, and blockage. These effects motivate exact or approximate absorption models and distance-aware spectral analysis.

  • Higher frequencies increase surface roughness effects, diffuse scattering, and backscattering, while smooth surfaces such as drywall remain reasonably modeled as reflective.
  • THz beams can experience stronger losses from snow than rain, while measured rain attenuation above 70 GHz shows no further impact under the reported conditions.
  • Molecular absorption creates resonant path-loss spikes whose strength and width increase with absorbing-molecule density, dividing the spectrum into distance-dependent sub-bands.
  • The HITRAN-based absorption model captures temperature, pressure, gas composition, and absorption cross-sections but is complex and difficult to track analytically.
  • Simplified absorption models fit measured line shapes, and the updated model covers 100-450 GHz with sufficient accuracy for links up to 1 kilometer under standard atmospheric conditions.
  • Three spectral windows appear below 1 THz, while higher frequencies and medium ranges produce more fragmented spectra whose window widths depend on center frequency and distance.

F. Multipath THz Channels

THz channels are generally sparse but can retain multipath indoors and in wideband scenarios. Multipath and molecular absorption introduce frequency selectivity, while ultra-wideband systems additionally face beam split and non-stationarity.

  • Only five multipath components survive at 0.3 THz in a 256×256 UM-MIMO system, 32.5% fewer than at 60 GHz.
  • Multipath paths are modeled through clusters and rays with distributed arrival times, decay factors, and random departure and arrival angles.
  • Wideband THz channels exhibit frequency selectivity from multipath and molecular absorption, with deterministic absorption-induced fading not captured by Rayleigh or Ricean models.
  • Beam split occurs when different subcarriers steer path components toward different spatial directions, causing array-gain loss in ultra-broadband THz signals.
  • Large UM-MIMO arrays and narrow beamwidths worsen beam split, making paths visible to only parts of an array and producing channel non-stationarity.
  • Digital-domain mitigation, true-time delays, sparse delay-matrix factorizations, and beam tracking are proposed to reduce beam-split effects.

H. Spherical Wave Model

THz system modeling must account for near-field propagation, highly directional antennas, misalignment, and transmitter or receiver hardware impairments. These effects alter channel responses and can severely degrade communication performance.

  • At 0.6 THz, the Rayleigh distance grows to approximately 40 m for the reported array dimensions, making near- and far-field selection critical.
  • Spherical waves can be modeled at antenna-element or subarray levels, while retaining plane waves at the element level provides a simpler equivalent array response.
  • Directional antennas are essential to counter high THz propagation losses, but narrow sectors and limited beamwidth constrain coverage.
  • Higher antenna directivity increases gain but also makes pointing errors and connection loss more likely because THz beams are narrow.
  • Misalignment is represented through an effective channel coefficient combining the nominal channel, misalignment fading, and a stochastic path-gain term.
  • Transmitter and receiver hardware imperfections are incorporated as additional Gaussian distortion noises in the impaired system model.

K. THz Noise Modeling

THz noise and performance analyses span molecular absorption noise, thermal noise, phase noise, propagation conditions, interference, and application-specific reliability or throughput. Reported studies therefore use channel-aware and scenario-specific models.

  • K. THz Noise Modeling: THz receivers primarily encounter thermal noise from multiplier and mixer chains and channel-induced absorption noise from water-vapor molecules.
  • K. THz Noise Modeling: Channel-induced noise dominates pulse-based systems in low-noise graphene devices and is colored over frequency.
  • K. THz Noise Modeling: Frequency planning can minimize channel-induced noise over absorption-free spectra, but molecular absorption can dominate propagation loss at longer distances and higher frequencies.
  • K. THz Noise Modeling: Phase noise arises from timing jitter, and precise THz-specific measurements remain lacking despite proposed Wiener-plus-Gaussian models.
  • Water-filling achieves more than 75 Gbps with 10 dBm transmit power over 0.06-1 THz, while delay spread and coherence bandwidth vary with distance and frequency.
  • In dense 0.1-10 THz networks, high antenna gains reduce interference probability but increase interference severity when interference occurs.
  • Use-case studies report reliability gains from proper densification, performance dependence on silicon-layer design, and 95% traffic offloading in THz information showers.
  • At higher altitudes, reported usable bandwidths over 2 km reach 8.218 THz, 9.142 THz, and 9.25 THz under the stated atmospheric conditions.

C. Experimental Demonstrations

Experimental demonstrations show that THz links can deliver multi-Gbps to Tbps rates across electronic and photonic systems, while waveform design must address wideband frequency selectivity, synchronization, and complexity.

  • 11 Gbit/sec was demonstrated over a 0.6 m, 300 GHz line-of-sight link with 1.5 GHz bandwidth using digital beamforming.
  • More than 1 Km transmission was demonstrated at 100 GHz, challenging assumptions that THz links are limited to short distances.
  • C. Experimental Demonstrations: Single-carrier schemes are favored by sparse, frequency-flat channels and lower implementation complexity, potentially requiring deviation from OFDM.
  • C. Experimental Demonstrations: OFDM remains difficult because THz systems require multi-giga- or tera-samples-per-second synchronization and face high PAPR and Doppler-related constraints.
  • C. Experimental Demonstrations: THz frequency selectivity persists because of molecular absorption, receiver characteristics, and multipath, with coherence bandwidth at 0.3 THz ranging from 1 GHz to 60 GHz.
  • C. Experimental Demonstrations: Distance-aware modulation can achieve Tbps rates at 10 m, an order of magnitude above fixed-bandwidth schemes, but requires slightly greater complexity.

C. Pulse-Based Modulation

Pulse-based modulation offers compact, low-complexity THz transmission despite source and power constraints, while spatial tuning adapts antenna separations to improve line-of-sight channel conditioning.

  • C. Pulse-Based Modulation: Pulse-based THz communications can achieve Tbps rates, but wideband pulses are power-limited and typically support low-power, compact sub-band transmissions.
  • C. Pulse-Based Modulation: Iterative joint modulation and power-allocation optimization demonstrates Tbps indoor rates under realistic filters and practical error-rate constraints.
  • A. THz Spatial Tuning: When communication distance is below the Rayleigh distance, adapting SA separation can improve channel conditions and achieve near-orthogonality.
  • A. THz Spatial Tuning: The Rayleigh distance depends on antenna dimensions, operating frequency, and separations; higher frequencies and larger arrays increase it for fixed separation.
  • A. THz Spatial Tuning: Spatial tuning numerically optimizes antenna separations, particularly for plasmonic arrays with uniformly spaced active graphene elements and real-time configurability.
  • A. THz Spatial Tuning: Antenna separations that are too small reduce spatial resolution and multiplexing gains, while separations above λ/2 produce grating lobes.

B. Index Modulation and Blind Parameter Estimation

THz index modulation encodes information through antenna, frequency, and modulation choices, while adaptive hybrid beamforming and blind estimation address hardware, interference, and channel-structure constraints.

  • B. Index Modulation and Blind Parameter Estimation: Spatial modulation maps bits to SA or AE locations, with antenna counts and constellation size tuned for the desired bit rate.
  • B. Index Modulation and Blind Parameter Estimation: Joint spatial-frequency index modulation can exploit many THz AEs and fragmented absorption-free windows, but frequency-hopping speed limits adaptability.
  • B. Index Modulation and Blind Parameter Estimation: Blind parameter estimation uses power-comparison tests, frequency-index detectors, modulation estimators, and modulation-type information bits.
  • A. THz Hybrid Beamforming: Fully digital THz arrays are hindered by prohibitive complexity and power consumption, motivating hybrid architectures with analog subarray beamforming.
  • A. THz Hybrid Beamforming: Dynamic AoSA precoding allocates subarrays through switch selection, but fully dynamic connections require exhaustive searches and thousands of switches.
  • A. THz Hybrid Beamforming: Distance-dependent hybrid beamforming combines user grouping, digital beamforming, power allocation, and subarray selection to share frequencies while avoiding analog-domain interference.

B. One-Bit Precoding

One-bit and related THz precoding approaches reduce converter demands and hardware burden, while NOMA and superposition-based designs address multiplexing under specific channel and user configurations.

  • B. One-Bit Precoding: ADC and DAC power, complexity, and interconnect demands rise sharply in ultra-massive THz systems, motivating reduced-resolution conversion.
  • B. One-Bit Precoding: One-bit quantization requires only simple comparators and eliminates automatic gain-control circuits, lowering converter requirements.
  • B. One-Bit Precoding: One-bit distance-aware multi-carrier systems show achievable rates insensitive to transmit-power changes and robust single-user transmission under phase uncertainties.
  • THz NOMA can superpose streams across overlapping effective channel matrices by assigning different power levels to multiplexed symbol vectors.
  • Low-complexity detection and decoding, particularly efficient successive interference cancellation, are needed for generalized antenna-, frequency-, and power-selection designs.
  • MIMO-NOMA can lose multiplexing gain through full stream decoding with SIC compared with MU-LP or rate splitting, although it benefits over OMA.

VIII. THZ BASEBAND SIGNAL PROCESSING

THz baseband processing must address device impairments and Tbps-scale operation through efficient, parallelizable algorithms and architectures across the full signal-processing chain. The section highlights channel estimation, coding, and detection as central challenges shaped by sparse THz channels.

  • Efficient baseband signal processing is critical for mitigating THz device impairments and enabling operations beyond 100 Gbps.The bottleneck includes energy-efficient transceivers approaching Tbps rates, while current ADCs and DACs operate at tens of gigasamples per second.
  • The complete baseband chain requires joint algorithm and architecture co-optimization for synchronization, channel estimation, decoding, and detection.This co-optimization targets latency, area efficiency, power consumption, and overall throughput.
  • Channel estimation: Accurate channel estimation is difficult because mobile beamforming needs precise CSI, while micrometer wavelengths make even fixed-link environmental changes consequential.The absence of a line-of-sight path further increases the importance of accurate CSI.
  • Channel estimation: Compressive-sensing channel estimation exploits sparsity in dictionary, delay, or beamspace domains through methods including matching pursuit, OMP, and LASSO.In dense multi-user wideband scenarios, however, many paths can require substantial measurements and real-time computational overhead.
  • Channel coding: Channel coding remains the most computationally demanding baseband process, motivating high-throughput Polar and Turbo decoder architectures.Reported approaches combine soft cancellation, majority logic, successive cancellation, and adaptive LLR quantization for low-latency decoding.
  • Channel coding: Coding advances often overlook inherent THz channel characteristics, motivating channel-aware MIMO detection and joint modulation, coding, and detection.

C. Data Detection

THz data detection must overcome inter-channel interference and prohibitive large-dimensional nonlinear-search complexity while meeting stringent latency and energy constraints. The section surveys subspace, non-coherent, iterative, and joint coding approaches suited to sparse or doubly-massive MIMO settings.

  • Inter-channel interference prevents simple linear detectors from decoupling spatial streams and can cause noise amplification.Consequently, THz systems require detectors that preserve near-optimal performance at reasonable complexity.
  • Optimal nonlinear detection is computationally prohibitive at large dimensions, including fixed-complexity sphere decoding for UM-MIMO.Proposed alternatives include local search, tabu search, message passing, Monte Carlo sampling, and lattice reduction.
  • Subspace detectors trade performance and complexity in large highly correlated MIMO channels by using channel puncturing to reduce computation and enhance parallelism.Their computational cost decreases with the number of nonzero channel-matrix elements.
  • Envelope- and energy-based detectors enable direct baseband, non-coherent detection that is inherently robust to phase noise.Compressed detection with orthogonal matching pursuit is also considered for sparse pulse-based multipath THz communications.
  • Coarsely quantized equalization matrices can reduce complexity, power consumption, and circuit area for precoding and data detection.
  • Joint coding, modulation, and detection: Iterative detection and decoding exchanges soft-output LLRs between MIMO detectors and channel decoders, but adds iteration and soft-output computation costs.
  • Joint coding, modulation, and detection: Parallelizable detectors are favored, while higher-order modulation increases complexity and phase-noise sensitivity in THz systems.Deep pipelining can aggregate decoder input but may reduce throughput.
  • THz channel-aware coding and detection can support capacity gains, while programmable surfaces provide active-array and reflective-array options for communication environments.Passive IRS elements introduce phase shifts to focus reflected power and steer beams without additional RF operations.

B. THz IRS Material Properties

THz IRS material choices balance compactness, power, speed, controllability, and reliability. Their signal-processing role is constrained by sparse, correlated high-frequency channels, incomplete performance limits, and the need for efficient channel estimation and beamforming.

  • THz IRS implementations use CMOS, MEMS, graphene, vanadium dioxide, liquid crystals, and other tunable technologies with different footprint, speed, power, and control constraints.
  • Material properties: CMOS offers low power and integration flexibility but faces clock-speed, parasitic-capacitance, and footprint limitations.
  • Material properties: MEMS provide reconfigurability but are limited by switch movement speed, control signaling, power consumption, faults, errors, and relatively large footprints.
  • Material properties: Graphene supports low-power, compact metasurfaces with simple biasing, while its phase control is limited by the voltage-implementing controller.Electrostatic biasing varies complex conductivity to control reflecting-element phase.
  • Material properties: THz metasurfaces support compact, lightweight, wide-angle beam steering and additional functions such as polarization conversion and orbital-angular-momentum generation.
  • Signal processing: High-frequency IRS performance limits remain insufficiently characterized because most analytical studies use lower-frequency models and assumptions.THz multi-user channels are very sparse, so existing analytical frameworks require revision.
  • Signal processing: At high frequencies, IRS channels are highly correlated and low-rank, so IRSs should increase channel rank and suppress interference in addition to increasing signal strength.Increasing channel rank can lead to substantial capacity gains.
  • Signal processing: IRS-assisted THz channel estimation combines cooperative beam training with high dimensionality and poor scattering, while compressive sensing can reduce training overhead and hardware complexity.

X. EXTENSIONS

THz sensing and localization extend communication systems with high-resolution position awareness and material-specific measurements. The paper surveys acquisition, signal processing, machine learning, and UM-MIMO-based methods for joint sensing and communications.

  • Localization: Narrow beams and mobile users make high-resolution localization critical for THz communications.THz directionality, array compactness, bandwidth, and line-of-sight propagation can support localization.
  • Localization: Location information can support channel estimation, spatial multiplexing, beamforming, resource allocation, tracking, and link re-establishment.Dense networks and tiny device footprints make localization a prerequisite for communications.
  • Localization: Massive-array processing enhances 3D angular and triangular accuracy and supports accurate elevation- and azimuth-direction beamforming.Conventional received-signal-strength, time-of-arrival, angle-of-arrival, and time-difference-of-arrival methods can be adapted to THz characteristics.
  • Sensing and imaging: THz sensing exploits spectral fingerprints and material interactions for applications including quality control, food safety, security, water analysis, and gas-composition analysis.THz signals can penetrate several materials, reflect strongly from metals, and interact with water and gases in characteristic ways.
  • Sensing and imaging: Carrier-based gas sensing correlates estimated channel responses with the HITRAN database to identify gaseous constituents in an UM-MIMO array-of-subarrays system.Symmetric frequency tuning makes the channel diagonal, reducing the MIMO problem to multiple single-input single-output problems.
  • Sensing and imaging: Solving for the channel identifies gas types, isotopes, and concentrations, while larger array-of-subarray size provides more observations per channel use and faster decisions.Absorption coefficients K_g(f_n) encode the gas-dependent response used in this inference.

C. Networking and Security

THz networking and security require designs tailored to highly varying environments and state-of-the-art devices. The paper surveys joint MAC–PHY optimization, synchronization, and multi-layer security approaches, while noting vulnerabilities despite THz directionality.

  • Networking: THz-specific MAC protocols must be optimized jointly with PHY signal-processing schemes under state-of-the-art device constraints.The need arises from highly varying THz mobile environments and the dependence of networking and signal processing on device architectures.
  • Networking: Examples include energy-harvesting nanonetwork control, joint power allocation and scheduling, on-demand multi-beam allocation, and receiver-initiated synchronization.These schemes address coordination across THz network and link layers.
  • Security: Higher propagation losses and increased directionality improve THz security relative to lower-frequency paradigms, but do not eliminate eavesdropping risks.Security protocols are therefore studied across multiple levels, including hardware and the physical layer.
  • Security: Scattering objects can redirect radiation toward nearby eavesdroppers, and narrow beams may still cover areas around receivers that expose vulnerabilities.Characterizing channel backscatter and mitigating beam-sector exposure are identified as security considerations.
  • Security: Covert THz communications seek to hide transmission occurrence through network-level techniques and modulation schemes such as distance-adaptive absorption peak hopping.The surveyed examples include dense IoT systems using reflections and diffuse scattering from rough surfaces.
  • Tutorial scope: The tutorial connects THz channel modeling, signal processing, sensing, networking, and security while emphasizing that these techniques will evolve with transceiver and system-model advances.Its coverage includes modulation, beamforming, channel estimation, coding, detection, reflecting surfaces, sensing, imaging, and localization.
Loading 2005.13176v3…