Chapter Summary

Chapter Summary

Key Points

• 1.

  Low-resolution ADCs/DACs change CM design. At mmWave with 256 antennas, 10-bit ADCs per antenna consume 25 W — prohibitive. 3-5 bit ADCs cut power by 10-100× at modest rate cost. 1-bit ADCs are the asymptotic limit: 2/π (~1.96 dB) capacity loss at low SNR, saturation at 2 bits per complex channel use at high SNR.

• 2.

  One-bit quantised MIMO recovers most of the capacity. Jacobsson- Durisi-Coldrey-Studer (2017) show that 1-bit massive MIMO with joint BS processing achieves ~85% of full-resolution rate at moderate SNR. Learned constellations and QPSK-matched schemes close the remaining gap at high SNR.

• 3.

  Spatial Modulation uses antenna index. Mesleh et al. (2008): rate $\log_2 n_t + \log_2 M$ bits/use with only ONE RF chain. Diversity is $n_r$ (not $n_t n_r$). Best for power-constrained IoT where RF-chain count is the dominant cost.

• 4.

  GSM generalises SM. Basar-Wen-Mesleh-Di Renzo (2017): activate $n_a$ of $n_t$ antennas. Total rate includes index bits (from activated subset) plus modulation bits (per active antenna). Optimum $n_a^* \approx n_t/2$ balances index bits and modulation bits.

• 5.

  Index modulation framework. Any discrete resource (antennas, subcarriers, time slots) can carry "index bits". Research-stage; no production standards adopt IM yet.

• 6.

  Channel hardening + massive arrays. As $n_t \to \infty$, the effective channel becomes deterministic (coefficient of variation $\propto n_t^{-1/2}$). BICM-to-capacity gap narrows to <0.5 dB, making CM design essentially an AWGN design problem.

• 7.

  Hybrid beamforming is the pragmatic architecture. Digital baseband precoding on $n_{\rm str}$ streams + analog RF beamforming across $n_t$ antennas. 32-64× RF hardware reduction with ~1 dB rate cost.