Crest Factor

Crest factor (CF) is defined as the ratio of the peak amplitude of a waveform to its root mean square (RMS) value. For a voltage signal v(t), it is expressed as:

• CF = |V_peak| / V_rms
• Equivalent to the square root of the Peak-to-Average Power Ratio (PAPR) : CF = √(PAPR)
• A pure sine wave has a crest factor of √2 (approximately 1.414 or 3.01 dB)
• DC signals have a crest factor of 1 (0 dB), representing the theoretical minimum

High crest factor signals force power amplifiers to operate with significant output back-off (OBO) to maintain linearity, severely degrading efficiency.

• The amplifier must be sized for the peak power, but operates most of the time at the much lower average power
• A 10 dB PAPR signal forces a Class-A PA to operate at ~10% of its peak efficiency
• This creates a fundamental trade-off between linearity and power efficiency
• Crest factor reduction (CFR) techniques are employed before the PA to clip or shape peaks, reducing the required back-off

Spectrally efficient modulation schemes inherently produce high crest factors, making this a critical challenge for 5G NR and Wi-Fi systems.

• OFDM (Orthogonal Frequency Division Multiplexing) signals exhibit high PAPR due to the coherent addition of many independent subcarriers
• A 4G LTE downlink signal typically has a PAPR of 8-12 dB
• 5G NR signals with higher bandwidth and more subcarriers can exceed 12-13 dB
• Single-carrier waveforms like SC-FDMA (used in LTE uplink) have lower crest factors (~4-6 dB), which is why they were chosen for power-limited handsets

The CCDF curve is the standard tool for statistically characterizing the crest factor of complex communication signals that have a non-deterministic envelope.

• It plots the probability that the instantaneous power exceeds a given level above the average power
• A point on the curve reads: 'The signal power exceeds the average power by X dB for Y% of the time'
• The 0.01% probability point is often used as a design target for defining the effective peak power
• CCDF curves are essential for setting clipping thresholds in CFR algorithms and for specifying PA linearity requirements

Crest factor directly influences the complexity and performance requirements of a Digital Predistortion (DPD) system.

• High crest factor signals drive the PA deeper into its nonlinear compression region at the peaks, requiring the DPD to model and invert a wider range of nonlinear behavior
• The DPD model must be accurate across the entire dynamic range from average power to peak power
• Aggressive CFR before the PA simplifies the DPD's job by reducing the peak-to-average ratio, but introduces in-band distortion (EVM) and out-of-band regrowth that must be managed
• The optimal system design co-optimizes CFR and DPD to meet both ACLR and EVM specifications while maximizing PA efficiency

Accurate crest factor measurement requires careful instrumentation setup to avoid underestimating the true peak values.

• Measurement bandwidth must be sufficient to capture the fast transient peaks of wideband signals
• Vector Signal Analyzers (VSAs) with high dynamic range are required
• The measurement duration must be long enough to capture rare peak events with statistical confidence
• Using a sample rate significantly higher than the Nyquist rate prevents peak underestimation due to interpolation errors
• The crest factor of a captured waveform is often reported in dB as CF_dB = 20 * log10(CF_linear)