Exercises

Fractional offset means channel falls between DD grid points. Pilot is grid-aligned; mismatch creates estimation error.

Classical methods: ~4 dB penalty. Super-resolution: ~1 dB. Best ML: ~0.8 dB. Open: close the 0.8 dB gap.

~1 dB penalty from CRB with current methods. Research target: 0 dB.

1-bit: 2 dB. 2-bit: 0.9 dB. 3-bit: 0.3 dB. 4-bit: 0.1 dB.

Sparse DD channel + NN detector: recovers most of loss. At 3-bit: 0.3 dB vs optimal. At 1-bit: ~0.5 dB advantage over OFDM.

Massive MIMO: 64 antennas × 1-bit = 100 mW total, vs 12-bit = 64 W. 640× savings.

THz waves are near-line-of-sight. Diffraction minimal. $P = 1-2$ paths. OTFS DD structure: optimal for sparse.

Same velocity gives more Doppler at THz. OFDM suffers ICI. OTFS handles it natively.

10-100+ GHz available. OTFS scales to handle it (MN = 10^6+ DD cells feasible).

THz channels align with OTFS's strengths. Commercial deployment 2032+ for short-range applications.

Sign of each subcarrier corrupted by sum of all others. Per- subcarrier noise: $1/\sqrt{MN}$.

Sign of each DD cell corrupted by sum of all others. But DD is sparse: only $P$ active cells.

OFDM: all $MN$ carriers contribute to per-carrier noise. OTFS: only $P$ active + $(MN - P)$ quiet cells. Quiet cells contribute zero to signal, but their noise spreads. Net: slightly less aggregate noise in OTFS.

~0.5 dB advantage in simulation. Not dramatic but consistent. Compounds with cell-free / multi-sat diversity.

$\lambda = 1$ mm. $L = (4\pi \cdot 10 / 0.001)^2 = 1.6 \cdot 10^{10} = 102$ dB.

$2 \log 256 = 2 \cdot 24 = 48$ dB both sides.

Tx 20 dBm. Rx: $20 + 48 - 102 = -34$ dBm. Noise 10 GHz bandwidth: $-174 + 100 + 3 = -71$ dBm. SNR: 37 dB. Ample.

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

Feasible for stadium/conference-hall scenarios. Research in 2026-2028; commercial 2032+.

Backwards compatible, ecosystem mature, IPR clean, simple. Extension of 5G NR numerology. Sufficient for most 6G use cases.

Fundamentally correct signal space for high mobility. Essential for LEO, V2X, ISAC. Architectural advantage long-term.

Enhanced-OFDM: "incremental is enough". OTFS: "structural matters."

Dual-waveform. Both standardized in Rel. 21. Per-session selection. 80% OFDM, 20% OTFS expected.

OTFS essential for niche but critical use cases (20%). Enhanced OFDM sufficient for mainstream (80%). Both have permanent roles.

Gap of 1 dB from CRB at $\epsilon = 0.3$. ML super-resolution can close. Impact: LEO and V2X deployments. Feasibility: high (ML tools mature).

Tight bounds needed. Impact: massive MIMO cost reduction. Feasibility: medium (theoretical work).

Study item in 2026-2027. Impact: 6G standardization outcome. Feasibility: requires organized advocacy.

Productize Ma-Wang-Caire framework. Impact: 2-3 dB MSE gain. Feasibility: high (existing work).

Each priority delivers concrete gains and is achievable in 2 years. Avoids speculative research.

~80% probability of Rel. 21 standardization given CommIT contributions + operator interest.

Ericsson/Nokia push for OFDM-only. 3GPP consensus difficult. Probability: 20%.

Cohere FRAND disputes. Licensing terms unclear. Probability: 10%.

US-China tensions. Decoupled standards. Probability: 15%.

~30% chance of 1-2 year slippage to Rel. 22. Still 95% chance of eventual standardization.

$I = \sigma_w^{-2} \nabla_\mathbf{h} \mathbb{E}[\mathbf{p}^T \mathbf{U}_\epsilon \mathbf{h}]^T \nabla_\mathbf{h} \mathbb{E}[\cdot]$

$\mathbf{U}_\epsilon$: fractional shift. Determinant: $\prod_k (1 - \sin^2(\pi \epsilon) \cdot f_k^2)$ where $f_k$ are pilot frequencies.

$\mathrm{MSE} \geq 1/I$. Inverse of Fisher. Grows at $\epsilon \to 0.5$.

$\mathrm{MSE} \geq \sigma_w^2 / (MN \cdot (1 - \sin^2 \pi \epsilon))$ for simple channel model.

$\epsilon \to 0$: no penalty. $\epsilon \to 0.5$: infinite penalty.

Each MIMO pair: 1-bit. Per-subcarrier rate bounded by 1-bit capacity ~2 bits per stream. Total: $N_r \cdot N_{subcarrier} \cdot 2 = 128 \cdot 1024 \cdot 2 = 262$ kbps.

Sparse DD channel: $P$ paths. NN detector uses DD-sparsity prior. Effective rate: $P \cdot N_r \cdot 8$ bits per DD cell. $P = 3$, $MN = 10^4$: $\sim 2$ Mbps per stream.

OTFS 1-bit: ~8× OFDM 1-bit rate for sparse channels. Closes for dense channels.

NN-based detection required. Classical 1-bit fails.

• ML-based super-resolution pilot (grad student).
• Atomic-norm unrolling + NN (postdoc).
• Collaborative simulation with industry testbed data. Budget: $200k.

• Joint estimation-detection algorithm development.
• Multi-profile training.
• Real-world channel testing via operator partnership. Budget: $250k.

Publication, conference presentations, patent filings.

1-2 major papers in IEEE TWC/TIT. Benchmark demonstrating <0.5 dB gap to CRB at $\epsilon = 0.4$. Open-source ML-learned pilot library. Patent filing for methodology.

Enables LEO and V2X deployments. 1 dB SNR gain at scale. Feeds into 3GPP Rel. 21.

THz chip fabrication: still specialized. Need: CMOS-compatible THz transistors at <$10 per chip. 2028-2030.

256-1024 element arrays at 300 GHz: $\sim \$100$-$\$1000$ per device. Need: < $\$100$. 2030-2035.

ITU spectrum allocation at 300+ GHz: ongoing. Need: 10 GHz+ bands allocated by 2030.

Commercial use cases identified: stadiums, data centers, healthcare, V2X short-range. Market size: $$1-10B by 2035.

Chip vendor + antenna vendor + system integrator + operator. Current: sparse. Need: coordinated investment, $\sim \$10$B industry total through 2032.

LEO services: limited. High-mobility (HST, V2X): degraded. Coverage gaps: 20% of 6G use cases poorly served.

OTFS deployed outside 3GPP spec. Defense, private networks, specialized IoT. 5-10% of the intended OTFS market.

Chip vendors with OTFS hardware: stranded investment. ~$1B globally.

Rel. 22 (2030-2032): OTFS added as optional. Commercial adoption begins.

2035+: OTFS has niche role. Not transformative. Academic research continues.

Strong CommIT advocacy in Rel. 20. Demonstrable LEO deployments by operator commitments. Reduces rejection risk.

Quantum signal processing for channel estimation and detection. Potential for exponential speedup. Research: 2030+. Speculative.

Post-Shannon: transmit meaning not bits. OTFS as substrate for semantic transport. Active research area.

Unified 3-tier NTN + terrestrial. OTFS as common waveform. Mass deployment 2040+.

OTFS integrated with AI inference at edge. UE-AI collaborative signal processing.

IoT + sustainability: ultra-low-power OTFS. Carbon-neutral 6G+. Research 2030+.

These are 2035+ research topics. Near-term (2026-2030): stay focused on 6G OTFS deployment. 2030+: explore post-6G.

Comprehensive pedagogical treatment. Clear derivations. Open problems list (this chapter). Supports courses and research.

Engineering perspective: compute, hardware, deployment. Clear performance gains. Informs investment decisions.

Balanced OTFS-vs-OFDM debate. Dual-waveform recommendation. Technical basis for 3GPP discussions.

Hands-on: simulations, algorithms, code links. From theory to implementation.

Highlights Mohammadi-Caire, Buzzi-Caire-Colavolpe, Raviteja- Viterbo-Caire contributions. Connects research to impact.

The book is designed to accelerate OTFS adoption. By providing complete technical treatment + quantitative case + realistic standardization view, it helps the community make OTFS a 6G reality.

Fractional-Doppler ML super-resolution. Close gap to CRB. Publishable in IEEE TWC. Impact: LEO, V2X deployments.

Cell-free OTFS AI/ML pipeline. Extend Chapter 17 with learned components. Industry partnership valuable.

Low-resolution ADC OTFS detectors. NN-based. Fun ML + EE. Applied focus.

THz-OTFS hardware. Engineering-heavy. Industry-partnership essential. 5-year horizon.

3GPP OTFS advocacy. Internship at Ericsson/Qualcomm. Learn standardization process. Career in telecoms engineering.

Pick a problem that matters. Collaborate with industry. Read widely outside pure OTFS. Publish in IEEE. Attend conferences where CommIT group presents. Stay curious, stay rigorous, stay honest about what's solved and what's not.