Soft Clipping

Soft clipping applies a continuous, differentiable nonlinear function—such as a hyperbolic tangent (tanh) or arctangent—to the signal envelope. Unlike hard clipping's abrupt truncation, the transition into saturation is gradual. This smoothness eliminates the sharp corners in the time-domain waveform that generate high-frequency spectral splatter. Common functions include the Rapp model and polynomial soft limiters, which can be tuned to balance compression aggressiveness against linearity preservation.

The primary advantage of soft clipping is dramatically reduced out-of-band emission. Hard clipping creates spectral regrowth that decays slowly—approximately 6 dB per octave—requiring aggressive filtering. Soft clipping's smooth transition produces regrowth that decays much faster, often 12 dB per octave or more. This significantly eases the burden on subsequent filtering stages and improves Adjacent Channel Leakage Ratio (ACLR) compliance with 3GPP spectral masks without requiring iterative clipping-and-filtering loops.

Soft clipping distributes distortion more gracefully across the signal. Hard clipping introduces severe in-band distortion concentrated at the clipping threshold, causing constellation point collapse and elevated Error Vector Magnitude (EVM). Soft clipping's gradual compression preserves relative phase relationships for moderately clipped peaks, resulting in lower EVM for equivalent PAPR reduction. This is critical for high-order QAM modulation (64-QAM, 256-QAM) where EVM margins are tight.

The soft clipper's AM/AM transfer function defines the mapping from input envelope magnitude to output envelope magnitude. Below the knee point, the response is linear (unity gain). Above the knee, the gain compresses smoothly toward a saturation level. Key design parameters include:

• Knee sharpness: Controls the abruptness of transition into compression
• Saturation level: The asymptotic maximum output amplitude
• Linear gain region: The input range preserved without distortion These parameters are tuned to meet PAPR targets while minimizing EVM.

Practical soft clipper implementations may introduce unwanted phase shift as a function of instantaneous amplitude—known as AM/PM conversion. This occurs because nonlinear devices often exhibit both amplitude and phase distortion. Advanced soft clipping architectures compensate by applying a phase predistortion curve that counter-rotates the signal phase as the envelope enters compression. Without this compensation, AM/PM conversion adds to EVM and can degrade demodulation performance.

Soft clipping achieves less aggressive PAPR reduction than hard clipping for the same clipping threshold. The smooth transition means peaks are not fully truncated—they are compressed but still extend above the nominal threshold. Typical PAPR reduction gains range from 2–4 dB at 10⁻⁴ CCDF probability, compared to 4–7 dB for hard clipping. However, the EVM penalty is significantly lower, making soft clipping preferable when modulation accuracy is paramount and moderate PAPR improvement suffices.

A memoryless CFR method that applies an instantaneous amplitude threshold to the baseband signal, saturating any sample exceeding the limit. This creates sharp discontinuities in the time-domain waveform, resulting in severe spectral regrowth and high out-of-band emissions. While computationally trivial, the resulting ACLR degradation often requires aggressive post-filtering, which can cause peak regrowth. Soft clipping is the direct evolution of this technique, designed to smooth the transition at the threshold.

A CFR technique that multiplies detected signal peaks by a smooth time-domain window function (e.g., Gaussian, Kaiser, or raised-cosine) rather than applying a hard limit. The window duration and shape are critical parameters: longer windows provide better spectral containment but may degrade EVM by affecting adjacent samples. This approach directly trades off PAPR reduction aggressiveness for reduced ACLR distortion, operating on a similar principle to soft clipping's smooth saturation curve.

A CFR approach that subtracts a pre-designed cancellation pulse from the original signal at each detected peak location. The cancellation pulse is spectrally shaped to match the transmit mask, ensuring out-of-band emissions remain controlled. Key implementation considerations include:

• Pulse design: Must balance cancellation effectiveness with EVM
• Peak detection: Requires accurate identification of peak magnitude and phase
• Overlapping peaks: Needs arbitration logic for closely spaced peaks This method is often preferred over soft clipping in standards-compliant designs.

The ratio of the maximum permitted signal amplitude after clipping to the RMS level of the unclipped signal, expressed as CR = A_max / σ. This dimensionless parameter directly determines the aggressiveness of PAPR reduction:

• Lower CR (e.g., 4 dB): Aggressive clipping, more distortion, higher efficiency
• Higher CR (e.g., 7 dB): Gentle clipping, less distortion, lower efficiency For soft clipping, the CR defines the knee point where the smooth saturation function begins to transition from linear to nonlinear operation.

A nonlinear signal transformation that compresses high-amplitude peaks while expanding low-amplitude valleys to reduce PAPR. At the receiver, an inverse expanding function restores the original signal dynamic range. Common companding laws include μ-law and A-law, originally developed for voice telephony. Unlike soft clipping, companding is a symmetrical process affecting the entire signal envelope, not just peaks, and requires coordinated transmitter-receiver processing.

A cascaded architecture applying successive stages of clipping and filtering with progressively tighter thresholds to achieve aggressive PAPR targets while controlling distortion. A typical configuration:

• Stage 1: Hard clipping at a high CR to remove extreme peaks
• Stage 2: Soft clipping or peak windowing at a moderate CR
• Stage 3: Peak cancellation with spectrally shaped pulses Each stage incrementally reduces PAPR while distributing EVM and ACLR degradation across the processing chain, enabling compliance with stringent spectral masks.