[Paper Review] Energy Efficiency Scaling Law for MIMO Broadcasting Channels

Jie Xu, Ling Qiu|arXiv (Cornell University)|Jan 18, 2012

Advanced MIMO Systems Optimization6 references3 citations

TL;DR

This paper derives the energy efficiency (EE) scaling law for MIMO broadcast channels with many users, accounting for non-ideal transmit-independent power. It shows that EE scales as $\frac{M\log_2\ln(NK)}{\alpha}$ when $\alpha > 0$ and as $\log_2(NK)$ when $\alpha = 0$, where $M$ is the number of transmit antennas, $K$ is the number of users, $N$ is the number of receive antennas per user, and $\alpha$ is the normalized independent power. The results reveal that EE improves with more antennas when $\alpha > 0$, but the multiplexing gain vanishes when $\alpha = 0$. The analysis leverages the Lambert $\omega$ function and bounds via uplink-downlink duality.

ABSTRACT

This letter investigates the energy efficiency (EE) scaling law for the broadcasting channels (BC) with many users, in which the non-ideal transmit independent power consumption is taken into account. We first consider the single antenna case with $K$ users, and derive that the EE scales as $\frac{\log_2 \ln K}α$ when $α> 0$ and $\log_2 K$ when $α= 0$, where $α$ is the normalized transmit independent power. After that, we extend it to the general MIMO BC case with a $M$-antenna transmitter and $K$ users each with $N$ antennas. The scaling law becomes $\frac{M \log_2 \ln NK}α$ when $α> 0$ and $ \log_2 NK$ when $α= 0$.

Motivation & Objective

• To understand the fundamental scaling behavior of energy efficiency (EE) in MIMO broadcast channels with a large number of users.
• To analyze how non-ideal hardware, particularly transmit-independent power, affects EE scaling.
• To derive closed-form EE scaling laws for both single-antenna and multi-antenna MIMO BC systems.
• To provide design insights for green wireless networks by quantifying the impact of user count, transmit antennas, and hardware impairments on EE.

Proposed method

• The authors model the MIMO BC with a base station (M antennas) and K users (each with N antennas), assuming Rayleigh fading and homogeneous pathloss.
• They use uplink-downlink duality to express the sum capacity and derive the total power consumption model including PA efficiency, dynamic, and static power components.
• The energy efficiency is defined as sum capacity divided by total power, and the scaling law is derived using the Lambert $\omega$ function to solve the resulting optimization.
• Upper and lower bounds on EE are established using extreme value theory on channel gains, leveraging the fact that the maximum eigenvalue of i.i.d. $\chi^2$ matrices scales as $\ln(NK)$.
• The analysis extends from the SISO case ($M=1$) to the general MIMO case by generalizing the bounds and asymptotic behavior of the Lambert $\omega$ function.
• The key insight comes from asymptotic analysis as $K \to \infty$, showing that the dominant term in the EE scaling is $\log_2\ln(NK)$ when $\alpha > 0$.

Experimental results

Research questions

• RQ1 How does energy efficiency scale with the number of users $K$ in a MIMO broadcast channel with non-ideal hardware?
• RQ2 What is the impact of transmit-independent power $\alpha$ on the scaling law of energy efficiency in MIMO BC systems?
• RQ3 Does increasing the number of transmit antennas $M$ always improve energy efficiency, and under what conditions?
• RQ4 How does the scaling law differ when $\alpha = 0$ (no independent power) versus $\alpha > 0$?
• RQ5 Can the EE scaling law be derived in closed form using the Lambert $\omega$ function under realistic hardware constraints?

Key findings

• When $\alpha > 0$, the energy efficiency scales as $\frac{M\log_2\ln(NK)}{\alpha}$, showing a logarithmic gain with user and antenna count.
• When $\alpha = 0$, the energy efficiency scales as $\log_2(NK)$, indicating no benefit from increasing $M$ due to the absence of independent power.
• The multiplexing gain is lost when $\alpha = 0$, meaning that increasing the number of transmit antennas does not improve EE in this regime.
• The use of the Lambert $\omega$ function enables a tractable closed-form approximation of the optimal EE under power constraints.
• As $K \to \infty$, the EE scaling is dominated by the logarithmic growth of the maximum channel gain, which scales as $\ln(NK)$.
• The analysis confirms that more antennas always improve EE when $\alpha > 0$, even though there is a trade-off between capacity gain and increased power consumption at finite $K$.