A Fair Comparison Between Spatial Modulation and Antenna Selection in Massive MIMO Systems

Bernhard Gäde, Ali Bereyhi, Saba Asaad, Ralf R. Müller

Abstract

Massive MIMO systems offer significant data rate improvements but are often constrained by the high cost and power consumption of dedicated radio frequency (RF) chains. This paper presents a "fair" head-to-head comparison between two low-complexity alternatives: Antenna Selection (AS) and Spatial Modulation (SM). By incorporating realistic pulse shaping and spectral mask constraints, the study identifies a performance crossover point. While AS excels at low power efficiency due to transmit diversity, SM emerges as the superior architecture for massive MIMO uplinks as the number of receive antennas scales, leveraging its inherent spectral efficiency gains.

Introduction: The RF Chain Bottleneck

In the quest for higher data rates, massive MIMO systems have become a standard. However, the requirement for a dedicated RF chain for every antenna remains a significant economic and technical hurdle. Techniques like Antenna Selection (AS) and Spatial Modulation (SM) address this by utilizing multiple antennas with only a single RF chain.

Comparing these techniques has historically been an "apples-to-oranges" affair because SM involves rapid antenna switching at every symbol interval. This switching causes "spectral regrowth," which can bleed signal into adjacent frequencies and violate spectral masks unless the symbol rate is significantly reduced. This paper aims to provide a level playing field by accounting for these implementational overheads through realistic pulse shaping and band-limitation constraints.

Methodology: System Model and Pulse Shaping

The authors evaluate a system with a single RF chain and $N_t$ transmit antennas communicating with $N_r$ receive antennas across a frequency-flat Rayleigh fading channel.

System model of a single RF-chain MIMO transmitter and receiver architecture.

Two primary strategies are compared:

  1. Antenna Selection (AS): The transmitter uses Channel State Information (CSI) to select the single "best" antenna for an entire coherence interval, maximizing diversity gain.
  2. Spatial Modulation (SM): The transmitter switches antennas for every symbol. The specific index of the active antenna carries additional data bits, increasing spectral efficiency without requiring extra transmit power.

To mitigate the spectral regrowth inherent in SM, the authors propose using Slepian windows for pulse shaping. Unlike standard Root-Raised-Cosine (RRC) filters, Slepian windows are designed to maximize energy within a specific bandwidth. This innovation allows SM to meet spectral masks with a symbol rate reduction factor of only 2.5x, a nearly fourfold improvement over the 9.25x reduction required by RRC filters.

Key Insights: Diversity vs. Spectral Efficiency

The comparison reveals a fundamental trade-off between diversity and efficiency. AS benefits from transmit diversity, which is vital when signal strength is low. In contrast, SM lacks this diversity but gains "free" bits from the antenna index. This means that for a fixed energy budget, the actual transmitted symbol in SM can be "louder" than in AS.

As the number of receive antennas ($N_r$) increases, the system enters the "massive" regime. In this environment, the additional receive antennas provide enough diversity to compensate for SM's lack of transmit diversity. Consequently, the power efficiency threshold ($E_b/N_0$) at which SM begins to outperform AS moves to lower SNRs.

Comparative Results and the Crossover Point

Numerical simulations demonstrate the scaling behavior of both techniques as the receive array grows from 4 to 64 antennas.

Bit Error Rate (BER) performance comparison between SM and AS across different receive antenna configurations.

As shown in the BER curves, for a configuration with 32 receive antennas, SM begins to outperform AS at a power efficiency of approximately $-13$ dB. This advantage becomes even more pronounced as the modulation order increases (e.g., from QPSK to 8QAM), highlighting SM's potential for high-capacity uplink transmissions in massive MIMO arrays.

Conclusion and Future Work

This study concludes that Spatial Modulation is a highly promising architecture for massive MIMO uplinks, particularly as the number of receive antennas grows. While hardware constraints like finite antenna switching time and the overhead of CSI estimation remain challenges, the use of optimized pulse shaping like Slepian windows makes SM a viable and efficient alternative to traditional Antenna Selection. Future research will likely focus on the power consumption of high-speed switching networks and the impact of imperfect CSI on these low-complexity MIMO strategies.

Summary updated

BibTeX
@inproceedings{gade2019fair,
  author = {Bernhard Gäde and Ali Bereyhi and Saba Asaad and Ralf R. Müller},
  title = {{A Fair Comparison Between Spatial Modulation and Antenna Selection in Massive MIMO Systems}},
  booktitle = {WSA},
  year = {2019},
  url = {https://openreview.net/forum?id=W00UZMy16Y},
}
Figure