Baseband Clipping

Baseband clipping operates instantaneously on each complex I/Q sample without regard to past or future signal states. The output depends solely on the current input amplitude relative to the clipping threshold. This lack of memory makes it computationally trivial but introduces sharp discontinuities in the time-domain waveform. These discontinuities are the root cause of severe spectral splatter, as they represent high-frequency components that violate the spectral mask. Unlike peak windowing, there is no smoothing function to taper the transition at the threshold boundary.

The clipping operation corrupts the modulated data symbols within the assigned channel. By truncating amplitude peaks, the process distorts the ideal constellation points, directly increasing the Error Vector Magnitude (EVM) . This in-band distortion is irreversible and sets a hard limit on the modulation order a link can support. For high-order QAM schemes like 1024-QAM, even minor clipping can raise the EVM floor above the receiver's sensitivity threshold, causing bit errors. The clipping ratio must be carefully chosen to balance PAPR reduction against acceptable modulation accuracy loss.

The hard saturation of the signal envelope generates intermodulation products that spill into adjacent frequency channels. This out-of-band emission is quantified by the Adjacent Channel Leakage Ratio (ACLR). The sharper the clipping transition, the wider the spectral splatter. To mitigate this, baseband clipping is almost always followed by a filtering stage to suppress out-of-band energy. However, this filtering process itself causes peak regrowth, where previously clipped peaks re-emerge above the threshold, necessitating iterative clipping and filtering stages in practical implementations.

The aggressiveness of the clipper is defined by the Clipping Ratio (CR) , expressed as the maximum permitted amplitude divided by the RMS level of the unclipped signal. A lower CR yields greater PAPR reduction gain but inflicts more distortion. The trade-off is visualized on a Complementary Cumulative Distribution Function (CCDF) curve, where the probability of high peaks is suppressed. Engineers select a CR that achieves the required power amplifier back-off reduction while keeping the resulting EVM and ACLR within the limits specified by the 3GPP or IEEE standard for the target waveform.

A single stage of clipping and filtering is rarely sufficient. Iterative Clipping and Filtering (ICF) cascades multiple stages, each applying a moderate clip followed by a frequency-domain filter to remove out-of-band regrowth. Each iteration further suppresses persistent peaks while allowing the filter to control the spectral mask. The number of iterations is a critical design parameter: too few leave residual peak regrowth, while too many increase computational latency and can over-distort the in-band signal. This technique is a foundational block in multi-stage CFR architectures.

Baseband clipping is the least computationally intensive CFR method, requiring only a magnitude calculation and a complex multiplication per sample. This makes it suitable for high-sample-rate, resource-constrained FPGA implementations. However, the associated filtering stage, typically an FFT-based frequency-domain filter, dominates the processing latency. The total group delay through the clipping and filtering chain must be strictly bounded to meet the stringent time-division duplex (TDD) timing requirements of 5G NR systems, where the turnaround time between downlink and uplink is specified in microseconds.