Tone Reservation (TR)

Unlike clipping-based Crest Factor Reduction (CFR) methods, TR generates a peak-canceling signal that occupies only reserved tones. These tones are orthogonal to data-carrying subcarriers, meaning the cancellation signal is completely removed by the receiver's FFT processing.

• Zero In-Band Distortion: No EVM degradation on data subcarriers
• No Spectral Regrowth: Reserved tones act as a guard band, containing cancellation energy
• Receiver Transparency: Standard OFDM receivers require no modification

The core of TR is designing a time-domain signal c[n] that, when added to the original OFDM symbol x[n], reduces peaks above a target threshold. This is formulated as a convex optimization problem minimizing the peak magnitude subject to the constraint that c[n] has zero energy on data subcarriers.

• Time-Domain Kernel: A pre-computed reference pulse p[n] derived from the reserved tone set
• Iterative Peak Reduction: Algorithmically scales and shifts p[n] to cancel peaks sequentially
• Gradient-Based Methods: Modern approaches use fast gradient projection for real-time convergence

System performance depends critically on which subcarriers are reserved. The number and placement of reserved tones represent a trade-off between PAPR reduction capability and throughput loss.

• Typical Overhead: 5-15% of total subcarriers are reserved
• Randomized Patterns: Avoids spectral periodicity that can regenerate peaks
• Standardized Sets: 3GPP LTE and 5G NR define specific reserved tone indices for TR-based CFR
• Adaptive Allocation: Dynamic reassignment based on channel conditions and modulation order

TR algorithms must operate at sample rates exceeding 100 MHz for wideband 5G signals. Efficient implementation is critical for real-time baseband processing.

• FFT-Based Kernel Generation: Leverages existing OFDM modulator hardware
• Peak Detection: Requires magnitude comparison at every sample
• Iteration Count: Typically 4-8 iterations per OFDM symbol for 4-6 dB PAPR reduction
• Hardware Acceleration: FPGA implementations use pipelined CORDIC processors for magnitude and phase operations

Both TR and Active Constellation Extension (ACE) are distortionless PAPR reduction methods, but they operate on fundamentally different principles.

• TR: Adds signal on reserved, unmodulated subcarriers; no impact on constellation points
• ACE: Extends outer constellation points outward within acceptable EVM margins
• TR Advantage: Works with any modulation order, including QPSK where ACE has limited extension room
• ACE Advantage: Requires no dedicated subcarriers, preserving full throughput
• Hybrid Approaches: Modern systems often combine TR for coarse peak reduction with ACE for fine adjustment

TR and Digital Pre-Distortion (DPD) are complementary techniques in the transmitter chain. TR reduces the PAPR before the power amplifier, while DPD linearizes the amplifier's response.

• Cascade Order: TR is applied before DPD to prevent the peak-canceling signal from being distorted
• Combined Benefit: TR reduces the PA back-off requirement; DPD corrects residual nonlinearity
• Joint Optimization: Emerging research explores co-designing the TR kernel and DPD coefficients for optimal end-to-end linearity
• Efficiency Gains: Together, they can improve PA efficiency by 10-15 percentage points in Doherty architectures

The fundamental metric that Tone Reservation addresses. PAPR quantifies the ratio of a signal's instantaneous peak power to its average power, expressed in dB. OFDM signals exhibit inherently high PAPR due to the coherent summation of independently modulated subcarriers. High PAPR forces power amplifiers to operate with significant back-off to avoid nonlinear distortion, drastically reducing efficiency. Without PAPR reduction, a typical OFDM transmitter may require 8-12 dB of back-off, translating to power amplifier efficiencies below 25%.

A distortionless PAPR reduction method that intelligently extends outer constellation points outward within acceptable EVM margins. ACE exploits the fact that exterior constellation points can be moved away from decision boundaries without increasing symbol error rate. This creates additional degrees of freedom for peak cancellation. Unlike Tone Reservation which uses reserved subcarriers, ACE modifies data-bearing subcarriers but only in directions that improve or maintain detection margin. ACE and TR can be combined for additive PAPR reduction gains.

A probabilistic PAPR reduction technique that generates multiple candidate OFDM symbols representing the same data using different phase rotation sequences. The transmitter selects and transmits the candidate with the lowest PAPR. SLM requires:

• Side information transmission to inform the receiver which phase sequence was used
• Multiple IFFT operations per symbol, increasing computational complexity
• No in-band distortion, but reduced spectral efficiency due to side information overhead

SLM achieves PAPR reduction without reserved tones but at the cost of increased processing and overhead.

A non-uniform signal transformation that compresses high-amplitude signal components and expands low-amplitude ones, reducing PAPR at the cost of introduced distortion. Derived from speech processing (e.g., μ-law companding), this technique:

• Reduces PAPR without reserving subcarriers or transmitting side information
• Introduces companding noise that degrades BER performance
• Requires an inverse expansion at the receiver that amplifies channel noise

Companding trades signal fidelity for PAPR reduction, making it less suitable for high-order QAM constellations where EVM requirements are stringent.

A repeated signal conditioning process that alternately clips signal peaks exceeding a threshold and applies frequency-domain filtering to remove out-of-band distortion. Each iteration:

1. Clips time-domain peaks to the target amplitude
2. Transforms to frequency domain via FFT
3. Filters out-of-band components to zero
4. Transforms back to time domain via IFFT

ICF progressively reduces PAPR while controlling spectral regrowth, but introduces in-band distortion that accumulates with iterations. Unlike TR, ICF cannot achieve distortion-free peak reduction.