Crest Factor Reduction (CFR)

The most fundamental CFR technique. The signal amplitude is hard-limited to a predefined threshold, directly cutting off peaks. This brute-force approach is simple but generates severe in-band distortion (increasing Error Vector Magnitude) and out-of-band spectral regrowth. A subsequent filtering stage is mandatory to suppress the regrowth, though this filtering can cause peak re-growth, requiring iterative clipping-filtering stages to converge on the target PAPR.

Instead of hard-clipping, peak windowing multiplies the signal by a smooth window function (e.g., Gaussian, Kaiser, Hamming) centered around each detected peak above the threshold. This shapes the clipping noise, concentrating its spectrum more effectively than hard clipping and reducing out-of-band emissions. The window length trades off between spectral containment and the smearing of the distortion across adjacent symbols.

This subtractive method detects signal peaks and subtracts a scaled, pre-computed cancellation pulse from the signal at each peak location. The cancellation pulse is designed to have a spectrum matching the transmit channel's allocated bandwidth, ensuring that the injected distortion remains strictly in-band and does not violate Adjacent Channel Leakage Ratio (ACLR) masks. This is widely used in modern base stations due to its predictable spectral footprint.

A distortion-free CFR method that reserves a small subset of OFDM subcarriers specifically for generating a peak-canceling signal. These reserved tones do not carry user data. An optimization algorithm computes a signal on these reserved tones that, when added to the data-bearing signal, reduces the PAPR without corrupting the data subcarriers. This eliminates in-band distortion entirely at the cost of reduced spectral efficiency.

A technique that intelligently extends outer constellation points outward within their decision regions to reduce signal peaks. By moving constellation points away from the origin, the algorithm creates headroom for peak reduction without crossing decision boundaries, thus maintaining the same bit error rate. ACE is particularly effective for QAM-modulated signals and introduces no out-of-band radiation.

A non-linear companding transform expands low-amplitude signals and compresses high-amplitude signals before transmission, reducing the PAPR. At the receiver, an inverse transform decompands the signal. The classic μ-law and A-law companding algorithms, borrowed from voice telephony, are computationally simple but introduce non-linear distortion that degrades Error Vector Magnitude (EVM) if not carefully designed.

The fundamental metric that CFR aims to reduce. PAPR quantifies the ratio of a signal's instantaneous peak power to its mean power, typically expressed in dB.

• High PAPR signals (e.g., OFDM) force power amplifiers to operate with significant back-off, drastically reducing efficiency
• A 10 dB PAPR means the peak power is 10x the average, requiring the PA to handle peaks far above its efficient operating point
• CFR directly manipulates the waveform to lower this ratio before amplification

The most widely implemented CFR algorithm. The signal amplitude is hard-limited to a threshold, then filtered to remove out-of-band spectral regrowth caused by the abrupt clipping.

• Hard clipping is computationally simple but introduces in-band distortion (EVM degradation) and spectral regrowth
• Iterative clipping and filtering repeats the process multiple times, applying progressively tighter filtering to meet spectral mask requirements
• The filter design is critical: too narrow causes peak regrowth; too wide fails to suppress adjacent channel leakage

An advanced technique that subtracts a carefully shaped cancellation pulse from detected peaks rather than simply clipping them. This preserves spectral integrity more effectively.

• A peak detection block identifies samples exceeding the target threshold
• A pre-computed cancellation pulse with the same bandwidth as the original signal is scaled and subtracted at each peak location
• Unlike clipping, pulse injection minimizes out-of-band emissions by design, reducing the filtering burden and EVM degradation

The primary trade-off metric when applying CFR. EVM measures the deviation of transmitted constellation points from their ideal positions, capturing the in-band distortion introduced by peak reduction.

• Every dB of PAPR reduction comes at the cost of increased EVM
• Regulatory standards (e.g., 3GPP) specify maximum allowable EVM per modulation scheme: 17.5% for QPSK, 8% for 256-QAM
• The CFR algorithm must balance PAPR reduction against staying within the EVM budget for the target modulation order

A distortion-free CFR method that reserves a subset of OFDM subcarriers specifically for peak-canceling signals. These reserved tones carry no data and are orthogonal to the data-bearing subcarriers.

• No in-band distortion is introduced to data subcarriers, preserving EVM perfectly
• The trade-off is reduced spectral efficiency: reserved tones consume bandwidth that could carry data
• Optimal peak-canceling signals are computed via convex optimization, making this method computationally intensive but ideal for high-order QAM where EVM margins are tight