[Paper Review] Antenna Count for Massive MIMO: 1.9 GHz versus 60 GHz

Erik G. Larsson, Thomas L. Marzetta+2 more|arXiv (Cornell University)|Feb 20, 2017

Millimeter-Wave Propagation and Modeling4 citations

TL;DR

This paper compares Massive MIMO performance at 1.9 GHz (PCS) and 60 GHz (mmWave) under line-of-sight propagation and perfect channel state information. It shows that while 128,000 mmWave antennas are needed in a noise-limited single-cell scenario to match 128-antenna PCS performance, only 215 mmWave antennas are required in a highly interference-limited multi-cell environment due to improved channel orthogonality and reduced impact of path loss differences.

ABSTRACT

If we assume line-of-sight propagation and perfect channel state information at the base station -- consistent with slow moving terminals -- then a direct performance comparison between single-cell Massive MIMO at PCS and mmWave frequency bands is straightforward and highly illuminating. Line-of-sight propagation is considered favorable for mmWave because of minimal attenuation, and its facilitation of hybrid beamforming to reduce the required number of active transceivers. We quantify the number of mmWave (60 GHz) service antennas that are needed to duplicate the performance of a specified number of PCS (1.9 GHz) service antennas. As a baseline we consider a modest PCS deployment of 128 antennas serving 18 terminals. We find that, to achieve the same per-terminal max-min 95%-likely downlink throughput, 10000 mmWave antennas are needed. To match the total antenna area of the PCS array would require 128000 half-wavelength mmWave antennas, but a much reduced number is adequate because the large number of antennas also confers greater channel orthogonality. The principal alleged benefit of mmWave technology--vast amounts of inexpensive spectrum--is at least partially offset by the complexity of possibly unwieldy amounts of hardware.

Motivation & Objective

• To quantify the number of mmWave (60 GHz) antennas required to match the performance of a given number of PCS (1.9 GHz) antennas in Massive MIMO systems.
• To analyze the trade-off between path loss, array gain, and channel orthogonality in mmWave versus sub-5 GHz bands.
• To evaluate how interference limitations in multi-cell environments reduce the required mmWave antenna count compared to single-cell noise-limited scenarios.
• To assess the impact of channel state information quality and array geometry on spectral efficiency and SINR performance.
• To clarify misconceptions about mmWave Massive MIMO scalability by providing a rigorous performance comparison under idealized but realistic propagation conditions.

Proposed method

• Uses a line-of-sight (LoS) propagation model with perfect channel state information at the base station, assuming slow-moving terminals.
• Applies the Shannon-Hartley capacity formula to model spectral efficiency, with capacity scaling as B log₂(1 + P/(BN₀)), where B is bandwidth and P is received power.
• Evaluates performance using max-min fairness in both uplink and downlink, with system-wide power control to model interference-limited scenarios.
• Compares three array configurations: 128-element linear, circular, and 8×16 rectangular arrays at 1.9 GHz, and equivalent mmWave arrays at 60 GHz.
• Analyzes SINR distributions across different array geometries and propagation conditions using Monte Carlo simulations.
• Performs a comparative analysis between single-cell (noise-limited) and multi-cell (interference-limited) deployments to isolate the effect of interference on required antenna count.

Experimental results

Research questions

• RQ1 How many 60 GHz mmWave antennas are required to achieve the same per-terminal max-min 95%-likely downlink throughput as a 128-antenna 1.9 GHz Massive MIMO system in a single-cell, noise-limited environment?
• RQ2 What is the impact of channel orthogonality on the required number of mmWave antennas when compensating for higher path loss compared to sub-5 GHz bands?
• RQ3 How does intercell interference in a multi-cell scenario affect the required number of mmWave antennas relative to a single-cell setup?
• RQ4 To what extent does the power advantage of 1.9 GHz systems become irrelevant in interference-limited mmWave deployments?
• RQ5 How does the effective aperture scaling with frequency (1/f²) influence the required number of antennas to maintain link budget?

Key findings

• In a noise-limited single-cell scenario, 20,000 mmWave (60 GHz) antennas are required to match the per-terminal max-min 95%-likely downlink throughput of a 128-antenna 1.9 GHz Massive MIMO system.
• The theoretical link budget prediction of ~128,000 mmWave antennas (based on (f₂/f₁)² scaling) is reduced by a factor of ~6.4 due to improved channel orthogonality from large array size.
• In a highly interference-limited multi-cell environment with system-wide max-min fairness power control, only 215 mmWave antennas are needed to match the performance of a 128-antenna 1.9 GHz system.
• In interference-limited regimes, the performance becomes independent of carrier frequency because intercell interference dominates, rendering the path loss and array gain differences negligible.
• The improvement in channel orthogonality due to large-scale arrays is the key factor that reduces the required mmWave antenna count below what link budget analysis alone would predict.
• Array geometry (linear, circular, rectangular) has a minor impact on SINR performance in LoS conditions, with all configurations showing similar trends in both PCS and mmWave bands.