Ferkans — Interactive Telecom Tutor

Coded Modulation for Massive Arrays

The point is that massive arrays change the CHANNEL statistics. Channel hardening (large $n_t$ averages out fading) and favourable propagation (user channels become near-orthogonal) narrow the gap between capacity and achievable rate for BICM. Coded modulation design for massive arrays blends the classical theory (BICM from Ch 5, STBC from Ch 11) with architectural constraints (low-res ADCs from §1, RF-chain count, calibration).

Definition:
Channel Hardening

Channel hardening is the phenomenon that, as the number of transmit/receive antennas grows, the variance of the effective per-user SNR shrinks relative to its mean. Formally, for an $n_t \times 1$ system with matched filtering, the SNR variance scales as $1/n_t$ . As $n_t \to \infty$ , the channel becomes "deterministic" from the user's perspective.

Theorem: Channel Hardening Rate

For an $n_t$ -antenna matched filter receiver with i.i.d. Rayleigh channel, the per-user matched-filter output SNR satisfies $\frac{\mathrm{Var}(\mathrm{SNR})}{\mathbb{E}[\mathrm{SNR}]^2} = \frac{1}{n_t}.$ The coefficient of variation of SNR decreases as $n_t^{-1/2}$ ; the channel becomes essentially deterministic for large arrays.

Proof

Matched-filter SNR

With $n_t$ antennas and matched filtering, effective SNR = $\frac{\|\mathbf{h}\|^2 P}{\sigma^2}$ where $\|\mathbf{h}\|^2 = \sum_{i=1}^{n_t} |h_i|^2$ .

Mean and variance

Each $|h_i|^2 \sim \chi^2_2 / 2$ with mean 1 and variance 1. So $\mathbb{E}[\|\mathbf{h}\|^2] = n_t$ and $\mathrm{Var}[\| \mathbf{h}\|^2] = n_t$ (independent sum).

Coefficient of variation

$\mathrm{Var}/\mathbb{E}^2 = n_t/n_t^2 = 1/n_t$ . As $n_t \to \infty$ , the ratio tends to zero — the channel "hardens". $\blacksquare$

CM-to-Capacity Gap Narrows with Array Size

As $n_t \to \infty$ , channel hardening makes the effective channel approximately AWGN with fixed SNR. BICM with LDPC then operates near the AWGN Shannon limit — the 0.5-1 dB gap that BICM has to Shannon on fading channels (Ch 5) narrows substantially for massive arrays. The gap is set by BICM capacity rather than by fading outage probability, and the 1.53 dB shaping ceiling becomes the next design target.

Coded Modulation Gain Across Array Sizes

Sum-rate vs SNR for $n_t = 8, 64, 256$ . As array size grows, channel hardening narrows the CM gap. For $n_t = 256$ , CM is within 0.3 dB of Shannon at moderate SNR.

Parameters

⚠️Engineering Note

Hybrid Beamforming for Massive Arrays

Massive arrays at mmWave (FR2) use HYBRID beamforming:

Digital baseband precoding (complex multiplicative) on a modest number of streams ( $n_{\rm str} \ll n_t$ ).
Analog RF beamforming (phase shifters) across all $n_t$ antennas. This reduces RF chains from $n_t$ (fully digital) to $n_{\rm str}$ (hybrid). For $n_t = 256$ and $n_{\rm str} = 4-8$ : 32-64× reduction in power-hungry RF hardware. Coded modulation design must account for the PHASE QUANTISATION in the analog beamformer (typically 2-4 bits) — another form of low-resolution constraint.

Example: Hybrid Beamforming Rate with 256-Antenna Array

A mmWave base station at 28 GHz has 256 antennas and 8 digital streams. Per-stream SNR after hybrid precoding = 15 dB. What BICM-LDPC rate can be achieved, and how close is it to the Shannon limit?

Solution

Per-stream capacity

$C = \log_2(1 + 10^{1.5}) = \log_2(32.6) = 5.03$ bits/use.

BICM-LDPC gap

With channel hardening (large $n_t$ ), BICM gap to Shannon is ~0.5 dB, so achievable rate $\approx 4.75$ bits/use per stream.

Sum rate

8 streams × 4.75 = 38 bits/use (single user). Dual-pol would double this to 76 bits/use. On a 200 MHz channel: $\sim 15$ Gbps. mmWave 5G NR achievable.

Why This Matters: See Also MIMO Book

A complete treatment of massive MIMO — including pilot contamination, JSDM, channel estimation, and precoding — is in the Ferkans MIMO book. This chapter focuses on the CM-layer implications of massive arrays.

Massive MIMO Architectures

Architecture	RF chains	Performance	Cost
Fully digital	$n_t$	Best (MRC/ZF)	High
Hybrid (analog + digital)	$n_{\rm str} \ll n_t$	~1 dB loss	Medium
1-bit ADCs (fully digital)	$n_t$	~2 dB loss (lowSNR)	Low
Spatial modulation	1	Much lower rate	Very Low
Fully analog (single beam)	1	Single-stream only	Very Low

Key Takeaway

Massive-MIMO CM design leverages channel hardening to narrow the BICM-to-capacity gap to <0.5 dB. Hybrid beamforming reduces RF chain count by 10-100×, introducing phase-quantisation constraints. The three low-resolution constraints — ADC bits, phase bits, and RF chain count — must be jointly optimised. This is an active engineering design space at mmWave / sub-THz frequencies for 6G.

Coded Modulation Design for Massive Arrays