Diversity, Smart Antennas and MIMO

Introduction

Wireless channels suffer from attenuation, shadowing, multipath fading, interference and noise. A link based on a single propagation path may therefore experience severe instantaneous signal degradation even when its average received power is adequate. Diversity techniques reduce this vulnerability by combining independently faded replicas. Smart antennas extend this idea by controlling spatial radiation and reception, while multiple-input multiple-output (MIMO) systems use multiple antennas to improve reliability, coverage, spectral efficiency or combinations of these objectives.

Modern cellular, wireless local-area network and fixed-wireless systems rely heavily on beamforming, spatial multiplexing, space–time coding and multiuser MIMO. Their performance depends not only on antenna count but also on channel rank, channel-state information (CSI), propagation correlation, signal-processing complexity, calibration and implementation constraints.

Learning Objectives

After studying this chapter, the reader should be able to:

• derive the output signal-to-noise ratio of linear diversity combiners;
• compare selection, equal-gain and maximal-ratio combining;
• explain adaptive arrays and calculate basic beamforming gains;
• evaluate MIMO capacity using singular values or channel eigenvalues;
• distinguish spatial multiplexing from transmit diversity;
• explain orthogonal space–time block coding;
• assess CSI acquisition, feedback and estimation errors;
• compare maximum-ratio, zero-forcing and regularized precoding;
• explain massive-MIMO benefits and limitations; and
• evaluate capacity, reliability, energy and complexity trade-offs.

1. Diversity in Fading Channels

1.1 Diversity principle

Diversity creates multiple replicas of the same information whose fading is not perfectly correlated. The replicas may be separated in:

• time, through coding and interleaving across coherence intervals;
• frequency, through transmission across independently faded subcarriers or bands;
• space, through multiple antennas;
• polarization, using differently polarized fields; or
• path, as in resolvable multipath combining.

The central requirement is sufficiently low dependence among the fading gains. Closely spaced antennas in a narrow angular spread may exhibit strong correlation, reducing the effective diversity order.

For an asymptotically high average SNR \(\bar\gamma\), diversity order is defined as

\[

d=-\lim_{\bar\gamma\rightarrow\infty}

\frac{\log P_{\mathrm{out}}}{\log \bar\gamma},

\]

where \(P_{\mathrm{out}}\) is outage probability. A system with diversity order \(d\) has an outage curve whose high-SNR slope is approximately \(-d\) on logarithmic axes.

1.2 Linear combining model

Consider \(L\) receive branches:

\[

\mathbf y=\mathbf h x+\mathbf n,

\]

where \(x\) has average symbol energy \(E_s\), \(\mathbf h=[h_1,\ldots,h_L]^T\) is the channel vector, and \(\mathbf n\sim\mathcal{CN}(\mathbf 0,N_0\mathbf I)\). The noise variance per complex branch is \(N_0\).

A linear combiner produces

\[

z=\mathbf w^H\mathbf y.

\]

Its instantaneous output SNR is

\[

\gamma_{\mathrm{out}}=

\frac{E_s|\mathbf w^H\mathbf h|^2}{N_0\mathbf w^H\mathbf w}.

\]

The branch SNR is

\[

\gamma_\ell=\frac{E_s|h_\ell|^2}{N_0}.

\]

These expressions assume spatially white noise. For noise covariance \(\mathbf R_n\), the denominator becomes \(\mathbf w^H\mathbf R_n\mathbf w\).

2. Diversity Combining Techniques

2.1 Selection combining

Selection combining (SC) processes the branch with the largest instantaneous SNR:

\[

\gamma_{\mathrm{SC}}=\max_{1\leq \ell\leq L}\gamma_\ell.

\]

It requires an SNR or signal-quality estimate for every candidate branch but only one active detection chain after selection. SC is simpler than coherent combining, although it discards useful energy from unselected branches.

If the branch SNRs are independent and identically distributed with cumulative distribution function \(F_\gamma(x)\), then

\[

P(\gamma_{\mathrm{SC}}<x)=[F_\gamma(x)]^L.

\]

For independent Rayleigh fading with mean branch SNR \(\bar\gamma\),

\[

F_\gamma(x)=1-e^{-x/\bar\gamma},

\]

and therefore

\[

P_{\mathrm{out,SC}}=

\left(1-e^{-\gamma_{\mathrm{th}}/\bar\gamma}\right)^L.

\]

Under these assumptions, SC achieves diversity order \(L\).

2.2 Equal-gain combining

Equal-gain combining (EGC) co-phases all branches but applies equal magnitudes to the combining weights. With \(\mathbf w\) chosen so that \(w_\ell^*h_\ell=|h_\ell|\), its output SNR is

\[

\gamma_{\mathrm{EGC}}=

\frac{E_s}{LN_0}\left(\sum_{\ell=1}^{L}|h_\ell|\right)^2

=rac{1}{L}\left(\sum_{\ell=1}^{L}\sqrt{\gamma_\ell}\right)^2.

\]

EGC requires phase estimates but not reliable amplitude weighting. By the Cauchy–Schwarz inequality,

\[

\gamma_{\mathrm{EGC}}\leq \sum_{\ell=1}^{L}\gamma_\ell.

\]

Thus, EGC cannot exceed maximal-ratio combining under the stated white-noise model. There is, however, no universal ordering between EGC and SC: EGC can be better or worse than SC depending on how branch energy is distributed.

2.3 Maximal-ratio combining

Maximal-ratio combining (MRC) chooses \(\mathbf w=c\mathbf h\), where the nonzero scalar \(c\) does not affect output SNR. Substitution gives

\[

\gamma_{\mathrm{MRC}}

=\frac{E_s\|\mathbf h\|^2}{N_0}

=\sum_{\ell=1}^{L}\gamma_\ell.

\]

MRC maximizes output SNR among all linear combiners when branch noises are spatially white and the channel is known. With colored noise, the optimum weight is proportional to

\[

\mathbf w=\mathbf R_n^{-1}\mathbf h.

\]

For independent Rayleigh-faded branches with equal mean SNR \(\bar\gamma\), the MRC output follows a gamma distribution. For integer \(L\), its outage probability is

\[

P_{\mathrm{out,MRC}}=

1-e^{-\gamma_{\mathrm{th}}/\bar\gamma}

\sum_{k=0}^{L-1}

\frac{(\gamma_{\mathrm{th}}/\bar\gamma)^k}{k!}.

\]

MRC achieves diversity order \(L\) under independent branch fading.

Worked Example 1: Correct comparison of SC, EGC and MRC

For three branches with instantaneous SNRs

\[

[\gamma_1,\gamma_2,\gamma_3]=[4,0,0],

\]

SC gives

\[

\gamma_{\mathrm{SC}}=4.

\]

EGC gives

\[

\gamma_{\mathrm{EGC}}

=\frac{1}{3}(\sqrt4+\sqrt0+\sqrt0)^2

=\frac{4}{3}\approx1.33.

\]

MRC gives

\[

\gamma_{\mathrm{MRC}}=4+0+0=4.

\]

Here, EGC is worse than SC because equal weighting introduces noise from branches that contain no signal energy. By contrast, if the branch SNRs were \([4,4,4]\), then SC would give 4 while both EGC and MRC would give 12. This confirms that SC and EGC have no fixed universal ordering.

3. Smart Antennas and Adaptive Beamforming

A smart antenna combines an antenna array with signal processing that adapts the array response. It can direct gain toward desired users and suppress interference arriving from other spatial directions.

For a uniform linear array of \(M\) elements with spacing \(d\), a narrowband steering vector may be written as

\[

\mathbf a(\theta)=

[1,e^{-j2\pi d\sin\theta/\lambda},\ldots,

e^{-j2\pi(M-1)d\sin\theta/\lambda}]^T,

\]

where \(\lambda\) is wavelength and \(\theta\) is measured from broadside under this convention. The array response is

\[

B(\theta)=\mathbf w^H\mathbf a(\theta).

\]

For conventional beam steering toward \(\theta_0\),

\[

\mathbf w=\frac{\mathbf a(\theta_0)}{\sqrt M}.

\]

Then \(|B(\theta_0)|^2=M\), corresponding to an ideal coherent array power gain of \(M\) relative to one identically powered isotropic element under consistent normalization.

Adaptive beamformers may optimize different criteria:

• maximum-ratio or matched-filter beamforming maximizes desired received power;
• minimum-variance distortionless response (MVDR) minimizes output interference-plus-noise power while preserving a desired direction;
• zero forcing places nulls toward known interferers;
• minimum mean-square error (MMSE) balances interference suppression and noise enhancement.

The MVDR solution is

\[

\mathbf w_{\mathrm{MVDR}}=

\frac{\mathbf R^{-1}\mathbf a_0}

{\mathbf a_0^H\mathbf R^{-1}\mathbf a_0},

\]

where \(\mathbf R\) is the received covariance matrix and \(\mathbf a_0\) is the desired steering vector. The normalization enforces \(\mathbf w^H\mathbf a_0=1\).

Practical limitations include steering-vector mismatch, mutual coupling, finite-resolution phase shifters, calibration error and rapidly changing channels. Wideband arrays also experience beam squint when frequency-independent phase shifts are used over a large fractional bandwidth.

4. MIMO Channel Model and Capacity

4.1 Point-to-point MIMO model

A narrowband MIMO channel with \(N_t\) transmit and \(N_r\) receive antennas is

\[

\mathbf y=\mathbf H\mathbf x+\mathbf n,

\]

where \(\mathbf H\in\mathbb C^{N_r\times N_t}\), \(\mathbf x\) is the transmitted vector and \(\mathbf n\sim\mathcal{CN}(\mathbf0,N_0\mathbf I)\).

If the receiver knows \(\mathbf H\), the transmitter has no instantaneous CSI, and total transmit power \(P\) is equally distributed, the capacity for a fixed channel realization is

\[

C=\log_2\det\left(

\mathbf I_{N_r}+\frac{\rho}{N_t}\mathbf H\mathbf H^H

\right)\quad\text{bit/s/Hz},

\]

where \(\rho=P/N_0\) under the adopted normalized channel model. Capacity is the logarithm of the determinant, not the determinant itself.

If \(r=\operatorname{rank}(\mathbf H)\) and \(\lambda_i\) are the nonzero eigenvalues of \(\mathbf H\mathbf H^H\), then

\[

C=\sum_{i=1}^{r}

\log_2\left(1+\frac{\rho}{N_t}\lambda_i\right).

\]

Thus, capacity depends on both received strength and the number and balance of spatial eigenmodes. Correlation or a dominant line-of-sight component can reduce rank even when many antennas are present.

Worked Example 2: MIMO capacity

Let

\[

\mathbf H=\begin{bmatrix}2&0\\0&0.5\end{bmatrix},

\quad N_t=N_r=2,

\quad \rho=10.

\]

The eigenvalues of \(\mathbf H\mathbf H^H\) are 4 and 0.25. With equal power,

\[

C=\log_2[(1+5\times4)(1+5\times0.25)]

=\log_2(47.25)

\approx5.56\text{ bit/s/Hz}.

\]

The stronger eigenmode contributes much more than the weaker one, illustrating why antenna count alone does not determine capacity.

4.2 Transmitter CSI and water filling

With transmit covariance \(\mathbf Q=E[\mathbf x\mathbf x^H]\), the capacity optimization is

\[

C=\max_{\mathbf Q\succeq0,\,\operatorname{tr}(\mathbf Q)\leq P}

\log_2\det\left(

\mathbf I+\frac{1}{N_0}\mathbf H\mathbf Q\mathbf H^H

\right).

\]

Using the singular value decomposition

\[

\mathbf H=\mathbf U\mathbf\Sigma\mathbf V^H,

\]

transmission can be performed along the columns of \(\mathbf V\). If \(\sigma_i\) is the corresponding singular value, optimal power allocation is

\[

q_i=\left(\mu-\frac{N_0}{\sigma_i^2}\right)^+,

\]

where \((x)^+=\max(x,0)\), and \(\mu\) is chosen so that \(\sum_iq_i=P\). This water-filling result assumes accurate transmitter CSI and Gaussian signalling.

5. Spatial Multiplexing and Detection

Spatial multiplexing sends independent streams simultaneously through different spatial modes. If the channel has rank \(r\), at most \(r\leq\min(N_t,N_r)\) independent streams can be separated in an ideal narrowband model.

Common receiver structures include:

• maximum-likelihood detection, which offers strong performance but has complexity that grows rapidly with antenna count and constellation size;
• zero-forcing detection, which inverts the channel but can greatly amplify noise near rank deficiency;
• MMSE detection, which regularizes inversion and usually performs better at low or moderate SNR;
• successive interference cancellation, which detects and subtracts streams sequentially but is sensitive to error propagation;
• sphere decoding, which reduces average search complexity but may still be expensive in difficult channels.

Spatial multiplexing maximizes data rate only when channel rank, SNR, CSI quality and receiver capability are sufficient. Sending too many streams over weak eigenmodes can reduce reliability and energy efficiency.

6. Space–Time Coding

Space–time codes introduce structured redundancy across transmit antennas and time. Their principal objective is transmit diversity rather than multiple independent streams.

6.1 Alamouti code

For two transmit antennas, the Alamouti code maps symbols \(s_1\) and \(s_2\) to

\[

\mathbf S=

\begin{bmatrix}

s_1&s_2\\

-s_2^&s_1^

\end{bmatrix},

\]

where rows represent time slots and columns represent transmit antennas. It satisfies

\[

\mathbf S^H\mathbf S=(|s_1|^2+|s_2|^2)\mathbf I_2.

\]

This orthogonality permits simple linear maximum-likelihood decoding. With one receive antenna, two symbols are transmitted over two time slots, giving code rate one and diversity order two, provided the channel remains approximately constant over both slots. With \(N_r\) independent receive antennas, the diversity order is \(2N_r\).

Orthogonal space–time block codes for more than two transmit antennas generally cannot retain rate one for arbitrary complex constellations. More general codes can trade decoding complexity, rate and diversity.

6.2 Diversity–multiplexing trade-off

At high SNR, increasing multiplexing gain generally reduces achievable diversity. For an independent, identically distributed quasi-static Rayleigh MIMO channel, the optimal Zheng–Tse diversity–multiplexing trade-off is the piecewise-linear curve joining

\[

d(r)=(N_t-r)(N_r-r)

\]

at integer values of multiplexing gain \(r\). This asymptotic result depends on the channel and coding assumptions and should not be treated as an exact finite-SNR performance formula.

7. Channel-State Information

CSI may include channel coefficients, spatial covariance, delay profile, Doppler characteristics or quantized beam indices.

• Receiver CSI supports coherent detection and combining.
• Transmitter CSI supports beamforming, link adaptation, user scheduling and power allocation.
• Statistical CSI changes slowly and can support long-term beam design.
• Instantaneous CSI must be refreshed at a rate determined by channel coherence.

In time-division duplex systems, channel reciprocity allows uplink estimates to support downlink precoding, but radio-frequency transmit and receive chains require calibration. In frequency-division duplex systems, downlink pilots and user feedback are normally needed; feedback overhead can become severe for large arrays.

CSI errors arise from thermal noise, pilot contamination, quantization, delay and model mismatch. A useful model is

\[

\mathbf H=\widehat{\mathbf H}+\mathbf E,

\]

where \(\widehat{\mathbf H}\) is the estimate and \(\mathbf E\) is the estimation error. Precoding based on \(\widehat{\mathbf H}\) may cause residual interference. Robust designs use error statistics, worst-case uncertainty sets or conservative rate adaptation.

The number of orthogonal pilot sequences is limited by the channel coherence block. Pilot reuse across cells can contaminate estimates, although coordinated allocation, covariance-aware estimation and advanced multi-cell processing can mitigate the effect.

8. Precoding and Multiuser MIMO

In downlink multiuser MIMO, a base station with \(M\) antennas serves \(K\) users through

\[

\mathbf x=\mathbf F\mathbf s,

\]

where \(\mathbf F\) is the precoding matrix and \(\mathbf s\) contains user symbols.

8.1 Maximum-ratio transmission

For a single user with channel vector \(\mathbf h\), maximum-ratio transmission uses

\[

\mathbf f=\frac{\mathbf h}{\|\mathbf h\|}.

\]

It maximizes desired signal power under a unit-norm constraint. In multiuser operation it is computationally simple but does not explicitly cancel inter-user interference.

8.2 Zero-forcing precoding

For a full-row-rank user channel matrix \(\mathbf H\in\mathbb C^{K\times M}\), with \(M\geq K\), an unnormalized zero-forcing precoder is

\[

\mathbf F_{\mathrm{ZF}}=

\mathbf H^H(\mathbf H\mathbf H^H)^{-1}.

\]

It ideally removes inter-user interference for the estimated channel. Power normalization is required before transmission. ZF can amplify noise or consume excessive transmit power when user channels are nearly parallel.

8.3 Regularized zero forcing

Regularized zero forcing uses

\[

\mathbf F_{\mathrm{RZF}}=

\mathbf H^H(\mathbf H\mathbf H^H+\alpha\mathbf I)^{-1},

\]

where \(\alpha>0\) depends on SNR, power normalization and the adopted system model. It interpolates between matched filtering and zero forcing and is closely related to MMSE precoding.

Nonlinear methods such as dirty-paper coding provide theoretical gains but are difficult to implement. Tomlinson–Harashima precoding and vector perturbation offer intermediate complexity.

9. Massive MIMO

Massive MIMO employs a base station with many more antennas than simultaneously served users. Under favorable propagation, independently distributed user channel vectors become nearly orthogonal as antenna count grows:

\[

\frac{1}{M}\mathbf h_i^H\mathbf h_j\rightarrow0,

\quad i\neq j.

\]

Likewise, normalized channel energy may concentrate around its mean, a phenomenon called channel hardening. These are asymptotic properties and may be weakened by correlation, common scattering clusters or line-of-sight geometry.

Major benefits include:

• coherent array and energy-efficiency gains;
• simultaneous spatial service to many users;
• narrow beams and reduced interference leakage;
• simplified linear processing under favorable propagation; and
• improved macro-diversity when distributed or cell-free architectures are used.

Major challenges include:

• pilot overhead and contamination;
• calibration and synchronization;
• computational load and fronthaul capacity;
• hardware nonlinearities and finite-resolution converters;
• near-field propagation for very large apertures;
• blockage and beam management at millimetre-wave frequencies; and
• energy consumed by radio-frequency chains and baseband processing.

Hybrid beamforming reduces the number of radio-frequency chains by combining analog and digital processing. It lowers cost and power consumption but constrains the set of realizable precoders.

10. Capacity–Complexity Trade-offs

Higher theoretical capacity generally requires better CSI, more radio-frequency chains, more pilots and more sophisticated processing. The appropriate design therefore depends on traffic, latency, mobility, energy and cost.

| Technique | Principal benefit | Main cost or limitation |

|---|---|---|

| SC | Low combining complexity | Discards unselected signal energy |

| EGC | Coherent gain without amplitude weighting | Requires phase estimation; may collect noise from weak branches |

| MRC | Maximum linear-combining SNR in white noise | Requires complex channel estimates and all branches |

| MRT | High array gain | Limited interference suppression |

| ZF | Strong spatial separation | Matrix inversion and poor conditioning |

| RZF/MMSE | Balanced signal and interference control | Regularization and SNR estimation required |

| Maximum-likelihood MIMO detection | Near-optimal detection | Rapid complexity growth |

| Hybrid beamforming | Fewer RF chains | Restricted beamforming flexibility |

Complexity must be considered jointly with coherence time. A matrix inversion that is manageable in a slowly varying fixed-wireless link may be too costly when repeated frequently for high-mobility users. Energy efficiency should include computation, data converters, cooling and RF-chain consumption rather than radiated power alone.

11. Illustrative Application Case: Urban Massive-MIMO Sector

Background

This fictional illustrative case concerns a mobile operator upgrading a congested urban sector. The available bandwidth is 10 MHz, and the proposed time-division duplex base station has 64 antennas serving eight single-antenna users simultaneously.

Problem

The operator seeks higher sector throughput without proportionally increasing radiated power. The measured coherence block contains approximately 200 complex symbols. User channels occasionally become correlated because users occupy the same street canyon.

Analysis

Eight orthogonal uplink pilot symbols require a pilot fraction of

\[

\frac{8}{200}=0.04,

\]

or 4%. Under ideal independent Rayleigh fading and accurate CSI, ZF has an approximate excess spatial degree of freedom of

\[

M-K=64-8=56.

\]

This quantity indicates substantial array gain after interference nulling, although it is not an exact guaranteed SINR in a correlated practical channel.

Suppose link adaptation targets a post-processing SINR of 15 in linear units. The ideal spectral efficiency per user is

\[

\log_2(1+15)=4\text{ bit/s/Hz}.

\]

The gross eight-user sum spectral efficiency is 32 bit/s/Hz. Accounting only for the 4% pilot overhead gives

\[

0.96\times32=30.72\text{ bit/s/Hz}.

\]

Over 10 MHz, the corresponding idealized payload rate is

\[

30.72\times10^7=307.2\text{ Mbit/s}.

\]

This value excludes control signalling, coding gaps, retransmissions, hardware impairments and scheduling losses.

Approach and Solution

The operator selects regularized ZF rather than pure ZF because measured user channels can be poorly conditioned. Users with strongly aligned channels are scheduled in different time-frequency resources. Reciprocity calibration is performed periodically, and pilot assignments are coordinated with adjacent sectors. Rate adaptation includes a margin for CSI aging.

The deployment also monitors RF-chain power and switches off selected chains during low traffic, provided coverage and calibration remain acceptable.

Outcome

Field performance is expected to remain below the idealized 307.2 Mbit/s calculation, but the design provides significantly greater capacity than single-user beamforming. Scheduling and regularization reduce the performance loss caused by channel correlation.

Learning Points

• Antenna count does not directly equal usable spatial-stream count.
• Pilot and control overhead must be included in throughput estimates.
• ZF gains depend on channel conditioning and CSI accuracy.
• Regularization, user scheduling and calibration are integral system components.
• Operational energy consumption may alter the preferred antenna configuration.

12. Governance, Security and Contemporary Developments

Spatial processing creates governance and security concerns. CSI and beam-management measurements can reveal location, motion and environmental information. Access should therefore be controlled, retained only as necessary and protected against spoofing or manipulation. Adversarial pilot transmissions can corrupt channel estimates or redirect beams.

Operators must comply with spectrum licensing, out-of-band emission limits, electromagnetic exposure requirements and equipment certification. Narrow beams can create localized exposure patterns, so compliance assessment must consider realistic beamforming operation rather than only static broadside transmission.

Machine learning is increasingly investigated for channel prediction, beam selection, user scheduling and hybrid-precoder design. It can reduce online search or exploit environmental structure, but models may fail when propagation conditions differ from training data. Reliable deployment requires uncertainty monitoring, explainable performance metrics, fallback algorithms and continuing validation.

Emerging research includes cell-free massive MIMO, reconfigurable intelligent surfaces, extremely large aperture arrays, near-field beam focusing, integrated sensing and communication, and distributed AI-assisted radio control. Their claimed gains must be evaluated with realistic channel estimation, control overhead, hardware loss and lifecycle energy consumption.

Chapter Summary

Diversity combats fading by providing multiple independently faded signal replicas. SC chooses the strongest branch, EGC coherently combines branches with equal weight magnitudes, and MRC maximizes output SNR under white-noise assumptions. MRC produces the sum of branch SNRs, while EGC does not have a fixed performance ordering relative to SC.

Smart antennas shape spatial responses to enhance desired signals or suppress interference. MIMO channels can provide diversity, beamforming gain and spatial multiplexing. Their capacity is determined by the logarithm of a channel-dependent determinant, or equivalently by the sum of capacities of spatial eigenmodes. Space–time coding emphasizes reliability, whereas spatial multiplexing emphasizes rate.

Accurate CSI enables coherent combining, beamforming, scheduling, precoding and water filling, but its acquisition consumes time, bandwidth and energy. MRT, ZF and regularized precoding provide different balances among array gain, interference suppression and computational robustness. Massive MIMO can produce channel hardening and favorable propagation, but practical gains are constrained by correlation, pilot overhead, hardware impairments and energy consumption. Effective engineering therefore requires joint optimization of capacity, reliability, complexity, governance and operational cost.

Key Terms

Diversity order: High-SNR rate at which outage probability decreases with average SNR.

Selection combining: Technique that detects the branch having the largest instantaneous SNR.

Equal-gain combining: Coherent combining using equal weight magnitudes and phase alignment.

Maximal-ratio combining: SNR-maximizing linear combining for a known channel and white noise.

Smart antenna: Antenna array integrated with adaptive spatial signal processing.

Beamforming: Constructive and destructive spatial combination used to control radiation or reception.

Array gain: Improvement obtained by coherent combination across antenna elements.

Spatial multiplexing: Simultaneous transmission of independent data streams over spatial modes.

Space–time coding: Coding across antennas and time to obtain diversity or coding gain.

Channel-state information: Knowledge or estimates describing the wireless propagation channel.

Precoding: Transmit-side spatial processing applied before signals enter the channel.

Water filling: Capacity-maximizing power allocation across known parallel channel modes.

Zero forcing: Processing that suppresses interference by channel inversion or null placement.

Regularized zero forcing: Channel inversion stabilized by a positive regularization term.

Channel hardening: Reduction in relative channel-gain fluctuations as antenna count increases.

Favorable propagation: Near-orthogonality of different users’ channel vectors.

Pilot contamination: Channel-estimation interference caused by reuse or non-orthogonality of pilots.

Massive MIMO: Multiuser MIMO architecture using a large excess of base-station antennas.

Hybrid beamforming: Joint analog and digital beamforming using fewer RF chains than antennas.

Review Questions

Short Answer Questions

1. Define diversity order and explain its physical significance.

2. State the output SNR expressions for SC, EGC and MRC.

3. Why can EGC perform worse than SC for highly unequal branch SNRs?

4. What is channel hardening?

5. Distinguish transmitter CSI from receiver CSI.

Long Answer Questions

1. Derive the MRC weight vector and output SNR for white noise, and extend the result to colored noise.

2. Explain how channel rank, eigenvalue distribution and antenna correlation influence MIMO capacity.

3. Compare spatial multiplexing and space–time coding in terms of rate, reliability and receiver complexity.

4. Critically compare MRT, ZF and regularized ZF precoding for a multiuser downlink.

5. Discuss the benefits and implementation challenges of massive MIMO in contemporary cellular systems.

Analytical / Application Questions

1. Four receive branches have SNRs \([1,4,9,16]\). Calculate the SC, EGC and MRC output SNRs and interpret the results.

2. For a specified \(2\times2\) MIMO channel, use its singular values to compare equal-power capacity with water-filled capacity at a given total power.

3. Design a pilot, scheduling and precoding strategy for a 128-antenna base station serving users with mixed mobility and correlated channels.

Multiple Choice Questions with Answers

Question 1

For branch SNRs [4, 0, 0], what is the equal-gain-combining output SNR?

A. 4/3

B. 4

C. 8/3

D. 12

Correct Answer: A. 4/3

Question 2

Under spatially white noise, which combining method maximizes output SNR among linear combiners?

A. Selection combining

B. Equal-gain combining

C. Maximal-ratio combining

D. Random combining

Correct Answer: C. Maximal-ratio combining

Question 3

For independent Rayleigh-faded branches, selection combining with L branches has which diversity order?

A. 1

B. L

C. L squared

D. log2 L

Correct Answer: B. L

Question 4

Which statement correctly compares EGC and SC?

A. EGC always exceeds SC

B. SC always exceeds EGC

C. They are always equal

D. Their ordering depends on the branch SNR distribution

Correct Answer: D. Their ordering depends on the branch SNR distribution

Question 5

What is the MRC output SNR when branch SNRs are [2, 3, 5]?

A. 5

B. 10

C. 30

D. 100

Correct Answer: B. 10

Question 6

With equal transmit power and receiver CSI, which expression gives narrowband MIMO capacity for total normalized SNR rho?

A. det(I + (rho/Nt)HH^H)

B. log2 det(I + (rho/Nt)HH^H)

C. trace(I + (rho/Nt)HH^H)

D. log2 trace(HH^H)

Correct Answer: B. log2 det(I + (rho/Nt)HH^H)

Question 7

What is the principal objective of the Alamouti space–time block code?

A. Transmit diversity with simple linear decoding

B. Doubling bandwidth

C. Eliminating receiver CSI

D. Creating more than two independent streams

Correct Answer: A. Transmit diversity with simple linear decoding

Question 8

Why can zero-forcing processing perform poorly for an ill-conditioned channel?

A. It prevents coherent reception

B. It always reduces channel rank to zero

C. Channel inversion can amplify noise or require excessive power

D. It cannot operate when the transmitter has multiple antennas

Correct Answer: C. Channel inversion can amplify noise or require excessive power

Question 9

A coherence block contains 200 symbols, of which 8 are pilots. What is the pilot overhead?

A. 2%

B. 4%

C. 8%

D. 25%

Correct Answer: B. 4%

Question 10

Which massive-MIMO property describes different users' channel vectors becoming approximately orthogonal as antenna count grows under suitable propagation?

A. Pilot contamination

B. Beam squint

C. Favorable propagation

D. Water filling

Correct Answer: C. Favorable propagation