Selection combining

Selection combining (SC) operates by continuously estimating the instantaneous signal quality—typically the signal-to-noise ratio (SNR) or received power—on each of N diversity branches. The receiver selects the branch with the highest metric and connects it to the demodulator. This is a switched diversity method that uses only one receiver chain, reducing hardware cost and power consumption relative to combining techniques that require multiple coherent receivers. The selection process can be performed at the RF level using a fast switch, or at baseband by comparing digitized samples. In practice, a threshold-based variant (switched diversity) avoids continuous monitoring by switching only when the current branch falls below a preset threshold, reducing switching overhead.¹ The statistical distribution of the output SNR in SC is the maximum of N independent, identically distributed (i.i.d.) Rayleigh fading SNRs, yielding a cumulative distribution function that improves with N.

The average output SNR of selection combining is given by the sum of the per-branch SNR weighted by harmonic numbers, leading to a diversity gain that grows as the number of branches increases but with diminishing returns. For N branches in Rayleigh fading, the average SNR improvement is approximately 1.5 dB for N=2 and 2.5 dB for N=4, compared to a single branch. This is significantly lower than the gain of maximal-ratio combining (MRC), which achieves an SNR equal to the sum of all branch SNRs. However, SC requires no amplitude or phase alignment, making it attractive for non-coherent modulation schemes. Bit error rate (BER) performance of SC with coherent modulation is roughly 3 dB worse than MRC for the same number of branches at high SNR. Equal-gain combining (EGC) falls between SC and MRC in performance but requires phase alignment.

Selection combining is widely used in legacy cellular systems, wireless local area networks (WLANs), and satellite communications where receiver simplicity is prioritized. It is also employed in Bluetooth receivers, which often use two antennas and a simple selection algorithm to improve link reliability.² A notable variation is the switched-and-stay combiner, which keeps the selected branch until its SNR drops below a threshold, then tries another. This reduces switching transients and is common in handset receivers. Another variant, hybrid selection combining, combines best branches using MRC, balancing complexity and performance. In cognitive radio, SC is used for spectrum sensing diversity, where the sensing device selects the sub-band with the highest received energy.³ Space-time block codes (STBC) can be used with SC at the receiver to further improve diversity without multiple RF chains.

While selection combining is often described as optimal for non-coherent receivers, its performance in correlated fading channels is less well known. When branches are correlated—e.g., due to insufficient antenna spacing—the diversity gain degrades significantly, and the output SNR distribution approaches that of a single branch.⁴ A less common implementation uses a codebook of antenna patterns and selects the best pattern, known as selection combining with beamforming. Another overlooked nuance is the impact of estimation error: practical SC receivers must estimate the SNR over a finite window, and incorrect selection due to fast fading can cause a penalty of up to 1 dB. The technique was first proposed by David G. Brennan in 1959 in his seminal paper on diversity combining, where he derived the exact outage probability for SC under Rayleigh fading.⁵ Despite its maturity, SC remains a building block in modern massive MIMO systems, where the number of antennas is large, and a subset may be selected for further processing to reduce complexity.

Diversity gain

The improvement in signal-to-noise ratio (SNR) at the output of a diversity combiner relative to a single branch.

Branch

An individual signal path in a diversity receiver, typically from a separate antenna or polarization.

Fading

Variation in signal amplitude and phase caused by multipath propagation or shadowing.

Maximal-ratio combining (MRC)

A diversity combining technique that weights and phases each branch to maximize the output SNR.