Open: OTFS at Terahertz

Beyond mmWave: The Terahertz Frontier

Every wireless generation has pushed the carrier frequency up — 1G at hundreds of MHz, 5G at tens of GHz, 6G's mmWave at $\sim 30$ GHz. The next frontier is sub-terahertz (100-300 GHz) and terahertz (300 GHz-3 THz): enormous bandwidth (10-100+ GHz) enabling peak rates of $\sim 1$ Tbps. But at terahertz, everything gets harder: phase noise, atomic absorption, hardware nonlinearities, molecular resonances. OTFS at terahertz is largely unexplored territory — a mix of exciting potential and unsolved challenges.

Definition:
The Terahertz Channel

The terahertz channel ( $f > 100$ GHz) differs from mmWave in several ways:

Atmospheric absorption: water vapor and oxygen cause $\sim 10$ - $100$ dB/km attenuation at specific frequencies. Windows exist at $\sim 140$ , $\sim 220$ , $\sim 300$ GHz.
Coherence time: shorter than mmWave. At 300 GHz, $\lambda = 1$ mm. Coherence time: $\sim 1$ - $10$ $\mu$ s for mobile.
Doppler: $\nu = v f/c$ . At 300 GHz, 30 km/h: $\sim 8$ kHz. At 300 km/h: $\sim 80$ kHz. Severe for OFDM.
Path loss: $10\log(4\pi d^2/\lambda^2)$ = $\sim 100$ dB at 10 m. Compensated by narrow beams.
Multipath: sparser than mmWave. $P = 1$ - $2$ dominant paths typical.

Overall: terahertz is a sparse, high-Doppler, short-range channel — a natural fit for OTFS.

Theorem: OTFS at Terahertz: Theoretical Advantage

For a terahertz channel at 300 GHz with 10 GHz bandwidth, mobile UE at 30-300 km/h:

OFDM: subcarrier spacing needs to be $\geq 100$ kHz to avoid ICI. For 10 GHz bandwidth: $10^5$ subcarriers. Cyclic prefix overhead: $\sim 10\%$ for 1 $\mu$ s delay spread. Feasible but complex.
OTFS: DD-sparse channel with $P = 1$ - $2$ paths. Frame size $MN =$ bandwidth × frame duration. At 10 GHz × 100 $\mu$ s: $10^6$ DD cells. OTFS-MP decoder complexity: $\mathcal{O}(10^6 \cdot P) = 10^6$ ops/frame. Manageable.

Conjecture: OTFS at THz is easier than at mmWave because the channel is more sparse. $P = 1$ - $2$ vs $P = 6$ - $10$ at mmWave. Research direction: exploit this for simplified detection.

Challenge: THz hardware nonlinearities are severe. Phase noise at 300 GHz: 3-10 dB worse than 30 GHz. NN detectors handle this but with training-data demands.

Terahertz is OTFS's natural home: extreme Doppler, sparse multipath, and huge bandwidth all play to OTFS's strengths. The challenge is hardware — generating, amplifying, and filtering THz signals is still experimental. Commercial THz deployment: 2030+ with 6G advanced.

Proof

Channel sparsity

THz waves propagate nearly in line-of-sight. Diffraction minimal. Multipath: 1-2 paths typical (LOS + 1 reflection).

Doppler handling

OFDM ICI = $(\nu/\Delta f)^2$ . At 300 GHz + 300 km/h + 100 kHz subcarrier: 60%! Unusable. OTFS: DD-sparse $\Rightarrow$ per-path detection $\Rightarrow$ no ICI issue. Usable.

Complexity

OTFS-MP at $MN = 10^6$ , $P = 2$ : $2 \times 10^6$ ops/frame. At 100 Hz frame rate: $2 \times 10^8$ ops/sec. Real-time feasible on 6G SoC.

Comparison

OFDM-THz requires special numerology (100 kHz subcarrier, 2 ms frame) to avoid ICI. OTFS-THz uses any numerology. $\blacksquare$

Key Takeaway

Terahertz is OTFS's natural operating regime. Sparse multipath

extreme Doppler + large bandwidth all favor OTFS. The barriers are hardware (THz transceivers still experimental) and standards (no 3GPP body yet targeting THz). Commercial THz-OTFS: 2032+.

Definition:
THz OTFS Architecture

A terahertz OTFS system has distinctive architecture:

RF front-end: 300 GHz oscillator, multiplier chains, wideband amplifier. $\sim 100$ - $500$ mW Tx. Currently expensive.
Massive antenna array: $N_t = N_r = 256$ - $1024$ . Array gain compensates path loss.
Hybrid beamforming: far fewer RF chains than antennas. 6-bit phase shifters + digital baseband.
Wideband ADC: 10-40 GSps, 3-4 bits (Chapter 22 §2's low-resolution discussion applies).
OTFS baseband: $MN = 10^6$ + cells. Specialized ASICs.

Range: Typical THz link: 10-100 m (atmospheric + path loss). Not cellular distances.

Applications:

Short-range hotspots (airports, stadiums).
Vehicle-to-infrastructure (RSU at intersection).
Data-center backplane wireless.
Industrial automation.
Close-range human sensing (gesture, biometric).

Theorem: THz OTFS Link Budget

A 300 GHz OTFS link with 10 GHz bandwidth, 10 m range, 256-antenna arrays at both ends, 20 dBm Tx power, atmospheric absorption 10 dB/km:

Free-space path loss: 100 dB.
Atmospheric loss: 0.1 dB (10 m × 10 dB/km).
Total path loss: 100 dB.
Array gain (Tx+Rx): $2 \cdot 10 \log 256 = 48$ dB.
Received power: $20 + 48 - 100 = -32$ dBm.
Noise floor at 10 GHz: $-174 + 100 + 3 = -71$ dBm.
SNR: 39 dB. Ample.

Rate: 10 GHz × 8 bps/Hz (64-QAM) × 0.8 (overhead) = 64 Gbps.

Terahertz OTFS achieves 10-100 Gbps short-range links — matching or beating fiber optic wireless.

The link budget works out because the huge bandwidth + array gain compensate the path loss. Terahertz is a "stadium" technology: short range, very high capacity. OTFS enables this by handling the extreme Doppler that emerges even at pedestrian speeds (8 kHz Doppler at 30 km/h is significant).

Proof

Path loss calculation

$L = (4\pi d/\lambda)^2 = (4\pi \cdot 10 / 0.001)^2 = 1.6 \cdot 10^{10}$ = 102 dB. OK.

Array gain

$N_t N_r = 65536$ elements effective at beamforming peak. Gain: 48 dB per side, compensates majority of path loss.

SNR

$-32 - (-71) = 39$ dB. Comfortable.

Rate

Shannon + real overhead: ~64 Gbps sustained. Matches fiber-wireless. $\blacksquare$

Example: Stadium THz-OTFS Deployment

Design a THz-OTFS system for a large stadium (30,000 attendees) streaming 8K video. Each user at 50 Mbps. Total aggregate: 1.5 Tbps.

Solution

Spectral requirements

At 64 Gbps per user × attendees: impossible in single beam. Spatial multiplexing: 30,000 independent narrow beams.

Architecture

30,000 hybrid-BF THz APs at 300 GHz, 10 GHz band. Massive MIMO: 256-element arrays. Users: laptop/phone with mini-array receiver.

Range

Each AP serves 10-50 users within 10 m range.

Aggregate capacity

30k users × 50 Mbps = 1.5 Tbps. Stadium bandwidth utilization: feasible with spatial reuse.

OTFS role

Users moving (walking in stadium): 3 km/h. Doppler at 300 GHz: $\sim 1$ kHz. Minor ICI for OFDM; OTFS handles cleanly. More importantly: OTFS supports ISAC (crowd-sensing for emergency applications).

THz OTFS Capacity vs Range

Plot OTFS capacity at 300 GHz vs distance. Compare with mmWave (28 GHz) for same setup. Sliders: transmit power, antenna array size.

Parameters

Tx power (dBm)20

N_t = N_r

256

f_0

(GHz)300

🔧Engineering Note

THz-OTFS Timeline

THz-OTFS deployment timeline:

2026: Research prototypes. 140 GHz mostly; 300 GHz rare. Hardware: cascaded multiplier + doubler chains. Expensive.

2028-2030: First 6G lab deployments at 140 GHz. 300 GHz research active. OTFS in experimental testbeds.

2030-2032: Commercial 140 GHz short-range hotspots (airports, industrial IoT). Early OTFS-THz prototypes.

2032-2035: 6G standardization of sub-THz bands (140-220 GHz) in Rel. 22-23. OTFS-capable chips.

2035+: True THz (300+ GHz) commercial for specialized applications.

Use case priorities:

Short-range high-capacity: airports, stadiums, data centers.
Industrial: factory automation (known environment).
Healthcare: close-range human sensing.
V2X: RSU-vehicle short-range high-rate.

Mass cellular THz: possibly 2040+ if commercially viable. Likely niche before that.

Practical Constraints

•
2026-2030: research prototypes
•
2030+: sub-THz 140 GHz commercial
•
2035+: true THz (300+ GHz) specialized
•
Mass cellular: 2040+ if ever

Common Mistake: Don't Extrapolate from mmWave

Mistake:

Assuming THz-OTFS is just "mmWave OTFS at higher frequency". The hardware and propagation differ dramatically: atmospheric absorption, phase noise, narrow beams all change architectural choices.

Correction:

Design THz-OTFS from the ground up, not as a scaling of mmWave. Specific considerations:

Atmospheric absorption limits range to $\sim 10-100$ m.
Arrays must be ultra-high-gain (256-1024 elements).
Narrow beams mean constant beam-tracking (Chapter 14).
Phase noise worse; require more pilots or ML denoising.
Hardware costs dominate (THz is not cheap in 2020s). Sensing applications (ISAC) may be primary use case, not comms.

Open: OTFS with Low-Resolution ADCs The OTFS vs Enhanced-OFDM Debate