Crest Factor Reduction (CFR)

The most fundamental CFR technique. The signal magnitude is hard-limited to a target threshold, which directly cuts off signal peaks. This brute-force approach creates severe out-of-band spectral regrowth. A subsequent low-pass filter is mandatory to suppress adjacent channel leakage, but filtering causes peak re-growth, requiring iterative clipping stages. Best suited for low-cost implementations where EVM degradation is tolerable.

Instead of hard-clipping, peak windowing multiplies the signal by a smooth window function (e.g., Hanning, Kaiser, or Gaussian) centered around detected peaks above the threshold. This smooths the transition and dramatically reduces spectral regrowth compared to hard clipping. The window length is a critical parameter: longer windows improve spectral containment but degrade EVM over a wider symbol span.

A subtractive method that detects peaks and subtracts a pre-designed, spectrally clean cancellation pulse at each peak location. The cancellation pulse is typically a sinc function or a filtered impulse that occupies the same bandwidth as the original signal. This technique preserves the in-band signal integrity better than clipping because only the peak region is modified, but it requires precise peak alignment and complex pulse scheduling logic.

A technique specific to multi-carrier systems like OFDM where a subset of subcarriers is reserved exclusively for peak reduction. These reserved tones carry no data; instead, they are modulated with optimized cancellation signals that destructively interfere with the time-domain peaks. The key advantage is zero in-band distortion on data subcarriers, as the reserved tones are orthogonal. The trade-off is reduced spectral efficiency due to the reserved bandwidth overhead.

A distortion-less technique that reduces PAPR by intelligently extending outer constellation points outward without crossing decision boundaries. For QAM modulations, corner points can be moved further from the origin to create anti-peak signals. Since no constellation boundaries are violated, the symbol error rate is unaffected. This method is most effective for high-order QAM with many corner points and is often combined with other CFR techniques.

A non-linear transformation inspired by audio noise reduction. The signal is compressed at the transmitter using a μ-law or A-law function to reduce dynamic range, then expanded at the receiver to restore the original signal. While effective at PAPR reduction, companding introduces non-linear distortion that is difficult to fully reverse in the presence of channel noise, making it less common in modern wideband systems compared to clipping-based methods.

The fundamental metric that Crest Factor Reduction seeks to minimize. PAPR is the ratio of the peak instantaneous power to the average power of a transmitted waveform, expressed in dB.

• High PAPR forces power amplifiers to operate with large back-off, drastically reducing efficiency.
• OFDM signals are notoriously high-PAPR due to the summation of many independent subcarriers.
• CFR directly targets PAPR reduction to enable operation closer to the amplifier's compression point.

A critical trade-off metric in CFR design. EVM quantifies the deviation of the actual transmitted symbols from their ideal constellation points.

• Aggressive CFR degrades EVM by distorting the signal peaks.
• The goal is to minimize PAPR while staying within a strict EVM budget defined by the wireless standard (e.g., 3.5% for 256-QAM).
• Modern CFR algorithms use peak windowing to confine distortion energy in time, minimizing EVM spread.

The simplest and most intuitive CFR method. The signal amplitude is hard-limited to a threshold, and the resulting out-of-band spectral regrowth is suppressed by a filter.

• Hard clipping is computationally cheap but causes severe spectral regrowth.
• The subsequent filter must be sharp, which can cause peak re-growth.
• Iterative clipping and filtering (ICF) repeats the process multiple times to converge on the target PAPR while maintaining the spectral mask.

An advanced CFR technique that multiplies the signal by a smooth window function centered on detected peaks, rather than hard-clipping.

• Uses windows like Gaussian, Kaiser, or raised-cosine to minimize spectral splatter.
• The window length controls the trade-off between PAPR reduction and EVM.
• Shorter windows preserve EVM but cause more out-of-band leakage; longer windows contain the spectrum but distort more symbols.

A peak-cancelling CFR method that subtracts a pre-computed cancellation pulse from the signal at each detected peak location.

• The cancellation pulse is designed to have the same spectral shape as the transmit filter, so it adds no out-of-band energy.
• Successive peak cancellation iteratively detects and cancels peaks until the target PAPR is met.
• This method is highly efficient in FPGA implementations using a peak detection engine and a bank of scaled, delayed cancellation pulses.