Peak-to-Average Power Ratio (PAPR)

PAPR is defined as the ratio of the peak instantaneous power to the average power of a complex baseband signal over a given observation interval. For a discrete-time signal x[n] of length N, it is expressed as:

• PAPR(dB) = 10 log₁₀( max(|x[n]|²) / E[|x[n]|²] )
• For voltage waveforms, the square root yields Crest Factor (CF)
• Typically measured over one OFDM symbol period or multiple frames
• A constant-envelope signal like GMSK has PAPR = 0 dB; an OFDM signal can exceed 12 dB

PAPR is not a single deterministic value but a statistical phenomenon. The Complementary Cumulative Distribution Function (CCDF) is the standard tool for characterizing it:

• CCDF plots the probability that instantaneous power exceeds a given threshold above average power
• The 10⁻⁴ probability point (0.01%) is the industry-standard reference for PAPR specification
• Steep CCDF curves indicate well-confined envelope statistics
• Used to determine required power amplifier back-off and CFR aggressiveness

High PAPR directly forces the power amplifier to operate at a large output back-off (OBO) from its compression point to maintain linearity:

• PA efficiency is maximum near saturation (P1dB) but nonlinearity causes spectral regrowth
• Required back-off ≈ PAPR − (acceptable distortion margin)
• A 10 dB PAPR signal forces a Class-AB PA from ~45% peak efficiency down to ~15% average efficiency
• This efficiency loss translates to higher operating expenditure (OPEX) for base stations and reduced battery life in handsets
• Envelope tracking and Doherty architectures partially recover this loss

Orthogonal Frequency Division Multiplexing (OFDM) is the dominant waveform in 4G LTE, 5G NR, and Wi-Fi, and it inherently produces high PAPR:

• OFDM sums many independent, modulated subcarriers via IFFT
• When subcarriers align constructively in phase, coherent addition produces extreme amplitude peaks
• PAPR scales approximately with the number of subcarriers: N subcarriers can theoretically produce PAPR = 10 log₁₀(N) dB
• 5G NR with 256-QAM on 3300 subcarriers can exhibit PAPR exceeding 12 dB
• This is the primary motivation for Crest Factor Reduction (CFR) in modern transmitters

PAPR is the problem statement; Crest Factor Reduction (CFR) is the solution space:

• CFR algorithms deliberately reduce PAPR before the power amplifier to improve efficiency
• Clipping applies a hard amplitude threshold but generates in-band distortion (EVM) and out-of-band spectral regrowth (ACLR)
• Peak windowing smooths clipped peaks with a time-domain window to control spectral splatter
• Peak cancellation subtracts spectrally shaped pulses at peak locations
• Tone reservation and active constellation extension avoid data-bearing subcarrier distortion
• The trade-off is always: PAPR reduction vs. EVM degradation vs. ACLR compliance

Accurate PAPR measurement requires careful instrumentation setup to avoid underestimation:

• Vector signal analyzers (VSAs) must capture sufficient samples to observe rare peak events
• Measurement bandwidth must exceed the signal bandwidth to capture overshoot from filtering
• Sample rate must satisfy Nyquist for the post-CFR signal, which may have expanded bandwidth due to clipping nonlinearity
• CCDF measurements typically require millions of samples for statistical confidence at the 10⁻⁴ probability point
• Cubic Metric (CM) is an alternative figure of merit that better correlates with PA power de-rating than raw PAPR for 3GPP signals

The CCDF is the primary statistical tool for characterizing PAPR. It plots the probability that a signal's instantaneous power exceeds a given threshold relative to the average power. Engineers use the CCDF curve to determine the power amplifier back-off required to achieve a target probability of clipping. A typical specification might read: 0.01% probability at 8 dB PAPR.

Crest Factor Reduction is a baseband signal processing technique that deliberately limits peak amplitudes before the power amplifier. Key methods include:

• Hard Clipping: Saturates the envelope at a fixed threshold, causing sharp discontinuities.
• Peak Windowing: Multiplies peaks by a smooth window (e.g., Gaussian, Kaiser) to control spectral regrowth.
• Peak Cancellation: Subtracts a spectrally shaped cancellation pulse at each detected peak. The goal is to reduce PAPR while balancing EVM degradation and ACLR compliance.

EVM quantifies the in-band distortion introduced by CFR and power amplifier nonlinearity. It measures the Euclidean distance between measured constellation points and their ideal reference positions. Aggressive PAPR reduction increases EVM, degrading modulation accuracy. Standards like 3GPP specify maximum EVM limits (e.g., 3.5% for 256-QAM) that constrain how much CFR can be applied.

ACLR measures the ratio of transmitted power within the assigned channel to power leaking into adjacent channels. CFR nonlinearity generates spectral regrowth that degrades ACLR. Regulatory bodies like the FCC and ETSI mandate strict spectral masks. Effective CFR algorithms must suppress peaks while containing out-of-band emissions to avoid interference with neighboring carriers.

Power amplifier back-off is the intentional reduction of input drive level to operate a PA in its linear region, away from the compression point. The required back-off is directly proportional to the signal's PAPR. A signal with 10 dB PAPR requires roughly 10 dB of output back-off, reducing efficiency from a theoretical 78% (Class B peak) to below 30%. Reducing PAPR by 3 dB can nearly double efficiency.

The Cubic Metric is a figure of merit that estimates the power de-rating required for a PA to handle a given signal. Unlike PAPR, which only considers peak-to-average ratio, the CM accounts for third-order nonlinearity in the PA transfer function. It provides a more accurate prediction of achievable power output for modulated signals and is standardized in 3GPP specifications for handset power class determination.