Selected Mapping (SLM)

Unlike Crest Factor Reduction (CFR) techniques such as clipping or companding, SLM does not introduce in-band distortion or out-of-band spectral regrowth. The transmitted signal is an exact representation of the original data, preserving Error Vector Magnitude (EVM) and Adjacent Channel Leakage Ratio (ACLR). This makes SLM ideal for modulation schemes with strict constellation fidelity requirements, such as 256-QAM and 1024-QAM in 5G NR.

SLM operates by multiplying the original frequency-domain OFDM symbol vector by U statistically independent phase rotation sequences. Each sequence applies a different pseudo-random phase shift to each subcarrier:

• Common phase sets: {±1}, {±1, ±j}, or unit-circle phases
• The IFFT is computed for each candidate, generating U distinct time-domain signals
• The candidate with the minimum PAPR is selected for transmission

The receiver must know which phase rotation sequence was selected to correctly demodulate the data. This side information (SI) must be transmitted alongside the payload:

• Typically encoded as log₂(U) bits per OFDM symbol
• SI is critical—if corrupted, the entire symbol is lost
• Robust channel coding or dedicated control channels protect SI integrity
• Advanced variants embed SI via pilot scrambling or blind detection to eliminate explicit overhead

SLM requires U parallel IFFT operations per OFDM symbol, making it computationally intensive compared to clipping-based methods. The complexity scales as O(U · N log N) where N is the FFT size. Practical implementations use:

• Reduced-complexity SLM with simplified phase sets
• Suboptimal selection algorithms that avoid exhaustive PAPR calculation
• Hardware-accelerated IFFT banks on FPGA or ASIC platforms
• Hybrid SLM-CFR architectures that combine a small U with mild clipping

SLM achieves PAPR reduction gain that improves with the number of candidates U, following a diminishing returns curve. At a CCDF probability of 10⁻⁴:

• U=4: approximately 2–3 dB reduction
• U=8: approximately 3–4 dB reduction
• U=16: approximately 4–5 dB reduction Performance depends on the phase set design and the statistical independence of candidate sequences. The technique is most effective for OFDM systems with a large number of subcarriers.

SLM belongs to the family of distortionless PAPR reduction techniques alongside Partial Transmit Sequence (PTS) and Tone Reservation (TR):

• SLM vs. PTS: PTS partitions subcarriers into blocks and optimizes phase per block; SLM applies per-subcarrier phases. PTS offers finer control but requires more complex optimization
• SLM vs. TR: TR reserves dedicated subcarriers for peak cancellation; SLM uses all subcarriers for data, avoiding spectral efficiency loss
• SLM is preferred when spectral efficiency and signal fidelity are paramount

A closely related probabilistic PAPR reduction method that partitions an OFDM symbol into disjoint sub-blocks. Unlike SLM, which generates full candidate sequences, PTS applies independent phase rotations to each sub-block and combines them. The combination with the lowest PAPR is transmitted. PTS typically requires transmitting side information about the chosen phase factors, similar to SLM's index signaling overhead.

A critical overhead in SLM systems. The receiver must know which phase rotation sequence was selected by the transmitter to correctly demodulate the data. This selected mapping index is transmitted as side information. Robust SI encoding is essential because errors cause catastrophic block failure. Techniques include embedding the index in reserved tones or using SI-free detection via maximum likelihood estimation at the receiver.

A distortion-less PAPR reduction technique that reserves a small subset of subcarriers specifically for a peak-canceling signal. Unlike SLM, TR does not modify data-bearing subcarriers and requires no side information. The reserved tones carry an optimized cancellation kernel that subtracts from detected peaks. TR is often combined with SLM in hybrid architectures to achieve aggressive PAPR targets without EVM degradation.

The primary statistical metric for evaluating SLM performance. The CCDF curve plots the probability that a signal's instantaneous PAPR exceeds a given threshold. SLM shifts the CCDF curve leftward, indicating reduced probability of high peaks. Performance is typically quoted as the PAPR reduction gain at a specific probability point, such as 10⁻⁴, comparing the original and SLM-modified CCDF curves.

A PAPR reduction method that dynamically extends outer constellation points outward within acceptable EVM limits. Unlike SLM's phase rotation approach, ACE exploits the tolerable decision margins in the constellation. Outer points are moved further from decision boundaries to reduce peaks without increasing bit error rate. ACE is compatible with SLM and can be applied as a post-processing step to further suppress residual peaks.

A central challenge in practical SLM implementation. Generating U candidate sequences requires U separate IFFT operations, which is computationally expensive. Reduced-complexity techniques include:

• Conversion matrices to generate candidates from a single IFFT
• Sub-optimal phase sequences with structured search
• Pruned candidate sets using prior CCDF knowledge These methods trade marginal PAPR performance for significant complexity savings.