Multi-Stage CFR

Multi-stage CFR employs a series of clipping stages arranged in cascade, where each successive stage applies a progressively tighter clipping threshold. The first stage performs coarse peak reduction with a higher clipping ratio, while subsequent stages refine the signal envelope with lower ratios. This graduated approach distributes the distortion burden across multiple stages rather than concentrating it in a single aggressive operation, preventing excessive EVM degradation.

Between each clipping stage, spectrally shaped filters are inserted to suppress out-of-band emissions generated by the preceding clipping operation. These filters are typically designed to match the transmit spectral mask requirements (e.g., 3GPP TS 38.104 for 5G NR). The filtering prevents spectral regrowth accumulation across stages, but introduces peak regrowth—a phenomenon where filtered peaks partially reappear—which the next clipping stage addresses.

Peak regrowth is an inherent challenge in multi-stage CFR where filtering after clipping causes previously suppressed amplitude peaks to partially reconstruct. Each stage is designed with an understanding of the regrowth ratio—typically 0.3 to 0.6 of the original peak reduction. The cascaded design explicitly accounts for this by setting intermediate clipping targets that overshoot the final PAPR goal, allowing the filter response to settle at the desired level.

A critical design parameter in multi-stage CFR is the allocation of the total allowable EVM budget across stages. Typical allocations follow a weighted distribution:

• Stage 1 (coarse): 40-50% of EVM budget
• Stage 2 (intermediate): 30-35% of EVM budget
• Stage 3 (fine): 15-25% of EVM budget This ensures that early aggressive stages do not consume the entire distortion allowance, leaving headroom for refinement.

Multi-stage CFR is particularly well-suited for FPGA and ASIC implementation due to its pipelined nature. Each stage can be mapped to a dedicated hardware pipeline stage, enabling high-throughput, low-latency processing. The resource-sharing opportunity between stages—such as reusing filter coefficient storage and clipping threshold lookup tables—reduces overall logic utilization compared to implementing a single, complex CFR block.

The number of stages represents a trade-off between PAPR reduction performance and implementation complexity. Empirical studies for 5G NR signals show:

• 2 stages: 4-6 dB PAPR reduction, suitable for moderate efficiency targets
• 3 stages: 6-8 dB PAPR reduction, the sweet spot for most base station applications
• 4+ stages: >8 dB reduction, diminishing returns with increased latency and gate count Optimal stage count depends on the signal bandwidth, modulation order, and target ACLR.

The parent signal conditioning technique that deliberately limits the peak amplitude of a transmit waveform. Multi-stage CFR is an advanced implementation of this concept, applying successive clipping stages with progressively tighter thresholds to achieve aggressive PAPR reduction while distributing distortion across multiple stages rather than concentrating it in a single harsh operation.

The fundamental iterative process at the core of each multi-stage CFR stage. Each stage applies hard clipping to truncate peaks above a threshold, followed by spectral filtering to suppress out-of-band emissions. The key challenge—peak regrowth—occurs when filtering causes previously suppressed peaks to reappear, which is why cascading multiple stages with decreasing clipping ratios is necessary for aggressive targets.

The ratio of peak instantaneous power to average power that multi-stage CFR is designed to reduce. Modern OFDM signals in 5G and LTE systems exhibit PAPR values of 10-12 dB, forcing power amplifiers to operate at significant back-off. Multi-stage CFR targets aggressive reduction to 6-8 dB, directly translating to improved amplifier efficiency and reduced operational expenditure.

The primary in-band distortion metric that constrains how aggressively multi-stage CFR can operate. Each clipping stage introduces constellation distortion that degrades modulation accuracy. The cascaded architecture distributes this distortion budget across stages—early stages handle coarse peak reduction with higher EVM contribution, while later stages apply finer correction within the remaining EVM margin specified by 3GPP standards.

The critical out-of-band emission metric that the filtering component of each CFR stage must control. Hard clipping generates spectral regrowth that spills into adjacent channels, violating regulatory spectral masks. Multi-stage architectures apply progressively narrower filtering at each stage, balancing ACLR suppression against peak regrowth to maintain compliance while achieving PAPR targets.

An alternative CFR technique that multiplies detected peaks by a smooth time-domain window function (e.g., Gaussian, Kaiser, or raised-cosine) rather than hard-clipping. When integrated into a multi-stage architecture, peak windowing can replace hard clipping in later stages to reduce spectral splatter while maintaining controlled peak suppression. The window length trades off between PAPR reduction and pulse overlapping effects.