MIMO-OTFS vs MIMO-OFDM

Head-to-Head: MIMO-OTFS vs MIMO-OFDM

By this point, the reader knows the MIMO-OTFS architecture, detection, and fundamental limits. The practical question remains: when is it worth the added complexity over MIMO-OFDM — the incumbent standard that 5G inherited from 4G? This section compares the two paradigms on a common operational footing: capacity, diversity, detection complexity, standardization readiness. The answer is not universal but velocity-dependent — for high Doppler, MIMO-OTFS dominates; for low Doppler, MIMO-OFDM remains simpler.

,

Definition:

MIMO-OFDM Primer

MIMO-OFDM applies MIMO processing per subcarrier:

  1. OFDM converts the time-dispersive channel into NscN_{sc} parallel flat subcarriers (cyclic prefix handles ISI).
  2. On each subcarrier, the channel is a flat MIMO matrix Hk∈CNr×Nt\mathbf{H}_k \in \mathbb{C}^{N_r \times N_t} for subcarrier kk.
  3. Per-subcarrier SVD or MMSE precoding/detection.
  4. Recombine across subcarriers.

Advantages: simple per-subcarrier structure; mature standardization; efficient implementation.

Weakness: assumes the channel is constant over the OFDM symbol duration TOFDMT_{\text{OFDM}}. Breaks at high Doppler: νmax⁡TOFDM>0.1\nu_{\max} T_{\text{OFDM}} > 0.1 causes inter-carrier interference (ICI) that destroys orthogonality.

Theorem: MIMO-OFDM Under High Doppler

For MIMO-OFDM with subcarrier spacing Δf\Delta f under a channel with maximum Doppler νmax⁡\nu_{\max}, the per-subcarrier SINR is SINRMIMO-OFDM  ≈  SNR1+(νmax⁡/Δf)2⋅π2/3⋅SNR.\mathrm{SINR}_{\text{MIMO-OFDM}} \;\approx\; \frac{\text{SNR}}{1 + (\nu_{\max}/\Delta f)^2 \cdot \pi^2/3 \cdot \text{SNR}}. Consequence. At νmax⁡/Δf=0.1\nu_{\max}/\Delta f = 0.1 (e.g., 120 km/h at 28 GHz mmWave): SINR saturates at ∼10\sim 10 dB regardless of actual SNR. No amount of MIMO antennas recovers the lost capacity.

In contrast, MIMO-OTFS preserves full SINR: SINRMIMO-OTFS  =  SNR(no ICI penalty).\mathrm{SINR}_{\text{MIMO-OTFS}} \;=\; \text{SNR} \quad \text{(no ICI penalty).}

This is the quantitative failure mode of MIMO-OFDM at high mobility. The ICI term (ν/Δf)2(\nu/\Delta f)^2 grows quadratically with Doppler. Once it exceeds unity, the SINR ceiling eats into the MIMO multiplexing gain. Even Nt=64N_t = 64 antennas cannot push rate above the ICI-limited ceiling. MIMO-OTFS, working in DD, sees no such ceiling.

Proof
ICI power

Per-subcarrier ICI power in MIMO-OFDM: σICI2≈(ν/Δf)2π2/3⋅Ps\sigma_{ICI}^2 \approx (\nu/\Delta f)^2 \pi^2/3 \cdot P_s per stream, where PsP_s is signal power. Derived from Fourier analysis of time-varying channel.

SINR formula

SINR=Ps/(σ2+σICI2)=SNR/(1+(ν/Δf)2π2/3⋅SNR)\mathrm{SINR} = P_s / (\sigma^2 + \sigma_{ICI}^2) = \text{SNR} / (1 + (\nu/\Delta f)^2 \pi^2/3 \cdot \text{SNR}).

Saturation

As SNR→∞\text{SNR} \to \infty: SINR→3/(π2(ν/Δf)2)\mathrm{SINR} \to 3/(\pi^2 (\nu/\Delta f)^2) — finite ceiling. Cannot escape by increasing power. ■\blacksquare

,

Key Takeaway

MIMO-OFDM hits a SINR ceiling at high Doppler. Above νmax⁡/Δf∼0.1\nu_{\max}/ \Delta f \sim 0.1, SINR saturates and MIMO multiplexing stops paying off. MIMO-OTFS has no such ceiling — the DD-domain processing handles arbitrary Doppler. This is the decisive architectural advantage at V2X / LEO operating points.

MIMO-OTFS vs MIMO-OFDM: Feature Comparison

FeatureMIMO-OTFSMIMO-OFDM
Channel viewDD-domain (sparse)Per-subcarrier (dense)
Max DopplerNo ceiling (OTFS extends DD)∼0.1Δf\sim 0.1 \Delta f (ICI)
Diversity orderNtNrPN_t N_r P (multiplies by paths)NtNrN_t N_r (classical MIMO)
Precoder designDD-joint (or per-DD-cell)Per-subcarrier
Detection complexityO(MN⋅P⋅Nr)\mathcal{O}(MN \cdot P \cdot N_r)O(MN⋅Nr⋅Nt2)\mathcal{O}(MN \cdot N_r \cdot N_t^2)
Pilot overhead (high mobility)~1% (DD-sparse)~10-30%
Implementation maturityResearch / prototype (2024)Mature (2010+)
Standardization6G candidate (2028+)4G/5G baseline

Example: MIMO Comparison Across Velocity Regimes

Compare MIMO-OTFS vs MIMO-OFDM capacity for Nt=16N_t = 16, Nr=4N_r = 4, P=8P = 8 paths, at 28 GHz, Δf=120\Delta f = 120 kHz, TOFDM=8.3μsT_{\text{OFDM}} = 8.3 \mu s. Velocities: 3 km/h, 30 km/h, 120 km/h, 300 km/h.

Solution
Doppler spreads

νmax⁡\nu_{\max} at each velocity: 78 Hz, 780 Hz, 3.1 kHz, 7.8 kHz.

MIMO-OFDM SINR

(ν/Δf)2π2/3(\nu/\Delta f)^2 \pi^2/3 at each velocity: 1.4×10−61.4 \times 10^{-6}, 1.4×10−41.4 \times 10^{-4}, 0.0030.003, 0.020.02. SINR at 20 dB SNR: 19.999 dB, 19.9 dB, 16.6 dB, 10.6 dB. Capacity: 88%, 82%, 72%, 42% of MIMO-OTFS.

MIMO-OTFS capacity

Full capacity regardless of velocity. C≈min⁡(Nt,Nr)log⁡(1+SNR)=4log⁡(1+100)=26.6C \approx \min(N_t, N_r) \log(1 + \text{SNR}) = 4 \log(1 + 100) = 26.6 bits/s/Hz.

Capacity gap

3 km/h: ~0 dB gap. 120 km/h: ~4 dB. 300 km/h: ~13 dB. The gap grows monotonically with velocity. Above 100 km/h, MIMO-OTFS is decisively better.

Capacity vs Velocity: MIMO-OTFS vs MIMO-OFDM

Plot ergodic capacity as a function of UE velocity (0-500 km/h) for both schemes. Sliders: antenna counts, subcarrier spacing, PP.

Parameters
16
4
120
20

Theorem: MIMO-OTFS Complexity Advantage at High Mobility

For MIMO-OTFS and MIMO-OFDM at the same target rate under high Doppler, the detection complexity ratio is CostMIMO-OTFSCostMIMO-OFDM  ≈  P⋅NrNt⋅Nr2⋅νmax⁡TOFDM.\frac{\text{Cost}_{\text{MIMO-OTFS}}}{\text{Cost}_{\text{MIMO-OFDM}}} \;\approx\; \frac{P \cdot N_r}{N_t \cdot N_r^2 \cdot \nu_{\max} T_{\text{OFDM}}}. For P=8P = 8, Nt=16N_t = 16, Nr=4N_r = 4, νmax⁡TOFDM=0.1\nu_{\max} T_{\text{OFDM}} = 0.1 (high mobility): cost ratio =8⋅4/(16⋅16⋅0.1)=32/25.6=1.25= 8 \cdot 4 / (16 \cdot 16 \cdot 0.1) = 32 / 25.6 = 1.25.

Interpretation: at high mobility, MIMO-OTFS is roughly equal in compute to MIMO-OFDM (which must also handle ICI). At low mobility, MIMO-OFDM wins the compute race. The crossover is at νmax⁡TOFDM≈0.05\nu_{\max} T_{\text{OFDM}} \approx 0.05 (medium mobility).

Low mobility: MIMO-OFDM is a well-oiled standard; MIMO-OTFS adds the DD transform overhead for no benefit. High mobility: MIMO-OFDM must handle ICI (itself expensive), and the DD structure of MIMO-OTFS becomes competitive on both capacity and compute.

The crossover point νmax⁡TOFDM∼0.05\nu_{\max} T_{\text{OFDM}} \sim 0.05 corresponds to ∼60\sim 60 km/h at 28 GHz with 5G NR numerology — entirely realistic for vehicular deployments. Above this threshold, MIMO-OTFS is the right choice.

Proof
MIMO-OTFS cost

Per frame: MP detection O(MN⋅P⋅Nr)\mathcal{O}(MN \cdot P \cdot N_r), precoding O(Nt3)\mathcal{O}(N_t^3). Dominant: MP.

MIMO-OFDM cost

Per frame with ICI correction: per-subcarrier detection O(MN⋅Nt⋅Nr2)\mathcal{O}(MN \cdot N_t \cdot N_r^2) + ICI tracking O(MN⋅Nt⋅Nr⋅(νT)2)\mathcal{O}(MN \cdot N_t \cdot N_r \cdot (\nu T)^2).

Ratio

At high Doppler, ICI correction dominates: MIMO-OFDM cost ∝Nt⋅Nr2⋅(νT)2\propto N_t \cdot N_r^2 \cdot (\nu T)^2. MIMO-OTFS cost ∝P⋅Nr\propto P \cdot N_r. Ratio ≈P/(Nt⋅Nr⋅(νT)2)\approx P / (N_t \cdot N_r \cdot (\nu T)^2).

Crossover

When the ratio equals 1: νT=P/(NtNr)\nu T = \sqrt{P / (N_t N_r)}. For representative numbers: ∼0.05\sim 0.05. ■\blacksquare

Key Takeaway

Use MIMO-OTFS when v>∼60v > \sim 60 km/h at 28 GHz. Below this velocity, MIMO-OFDM wins on simplicity and standardization. Above, MIMO-OTFS wins on capacity, diversity, and latency. The operating point for 5G/6G vehicular, LEO, and V2X applications is uniformly above this threshold.

MIMO-OTFS vs MIMO-OFDM Under Doppler

Side-by-side animation showing how MIMO-OFDM's per-subcarrier structure breaks under increasing Doppler (ICI spreads signal energy across subcarriers), while MIMO-OTFS's DD-domain representation remains sparse and stable. Doppler slider drives the animation from pedestrian to LEO speeds.
🔧Engineering Note

Migration Path: 5G MIMO-OFDM → 6G MIMO-OTFS

Industry consensus on the 5G → 6G MIMO transition:

5G (current): MIMO-OFDM on sub-6 GHz and mmWave. Low mobility dominant use case. Per-subcarrier MIMO processing. 64-256 QAM constellations.

5G Rel. 18-19 (2024-2026): Limited OTFS overlay for high- mobility scenarios (V2X, high-speed rail). Dual-mode UEs support both.

6G Rel. 21+ (2028+): MIMO-OTFS as primary waveform for mmWave and sub-THz. MIMO-OFDM retained for sub-6 GHz (low-mobility, legacy devices).

Full 6G MIMO-OTFS: 2030+. Network-wide, including cell-free architecture (Chapter 17) and LEO integration (Chapter 18).

Migration cost: Primarily chip-level. OTFS baseband requires ISFFT/SFFT hardware + DD-domain processing + new channel-estimation path. Modern semiconductor foundries can deliver this at marginal cost premium (∼10\sim 10-20%20\%) vs current 5G chips.

Practical Constraints
  • •

    5G: MIMO-OFDM baseline

  • •

    5G Rel. 18+: OTFS overlay for high-mobility

  • •

    6G: MIMO-OTFS primary for mmWave, sub-THz

  • •

    Full migration: 2030+

,

Common Mistake: MIMO-OTFS Is Not Universally Better

Mistake:

Assuming MIMO-OTFS should replace MIMO-OFDM in all scenarios. For low-mobility, low-Doppler scenarios (indoor, stationary cellular), MIMO-OFDM is simpler and standardized. Adding OTFS machinery without Doppler-induced pain is unnecessary.

Correction:

Design rule: deploy MIMO-OTFS when expected νTframe>0.05\nu T_{\text{frame}} > 0.05 at the operating frequency. This threshold captures vehicular, LEO, HST, and V2X scenarios. For indoor WiFi, urban microcell with pedestrian UEs, fixed-wireless access: MIMO-OFDM remains the right choice. The transition is gradual and use-case- driven, not a wholesale replacement.

Why This Matters: Connection: Telecom Ch 19 MIMO-OFDM

Telecom Chapter 19 developed MIMO-OFDM: the per-subcarrier MIMO processing that underlies 4G/5G. This chapter's MIMO-OTFS is the delay-Doppler-domain analog of that framework — same goal (spatial multiplexing + diversity), different channel representation (DD vs. per-subcarrier). The two are not competitors but complements: OFDM for low-mobility, OTFS for high-mobility. Both are first-class waveforms in the 6G vision.

Spatial Multiplexing and DiversityChapter Summary