Active Constellation Extension (ACE)

Unlike clipping-based Crest Factor Reduction (CFR) methods, ACE introduces no in-band distortion to the transmitted signal. The technique operates entirely within the tolerable EVM margin defined by the modulation scheme's constellation boundaries. By only extending outer points into regions where the decision threshold is not compromised, ACE achieves PAPR reduction while preserving modulation accuracy and bit error rate performance. This makes it particularly valuable for high-order QAM schemes where EVM budgets are tight.

ACE dynamically modifies the transmitted symbol vector when the time-domain signal envelope exceeds a predetermined threshold. The algorithm projects the required peak-reduction signal onto the allowable extension region for each active subcarrier:

• Outer constellation points are extended outward radially
• Inner constellation points remain unmodified to preserve decision boundaries
• Corner points receive the most extension freedom in square QAM constellations This selective modification ensures that only symbols with sufficient margin contribute to peak reduction.

ACE is typically implemented as an iterative clipping-and-projection loop in the digital baseband:

1. Peak Detection: Identify time-domain samples exceeding the amplitude threshold
2. Clipping: Apply hard or soft amplitude limiting to those samples
3. FFT to Frequency Domain: Transform the clipped signal back to subcarrier symbols
4. Constellation Projection: Map each modified symbol back to its allowable extension region
5. IFFT to Time Domain: Return to the time domain for the next iteration This process repeats until the PAPR target is met or convergence is achieved, typically within 4-8 iterations.

A critical advantage of ACE over filtering-based CFR is the absence of out-of-band spectral regrowth. Because constellation projection constrains modifications to remain within the occupied subcarrier set, no energy is introduced into adjacent frequency channels. This preserves the Adjacent Channel Leakage Ratio (ACLR) and ensures compliance with spectral mask requirements without additional bandpass filtering stages. The technique is inherently spectrally contained, simplifying transmitter chain design.

The PAPR reduction capability of ACE is fundamentally bounded by the available EVM margin of the modulation scheme:

• QPSK: Large extension regions enable 3-5 dB of PAPR reduction
• 16-QAM: Moderate margins yield 2-4 dB reduction
• 64-QAM: Tighter constellations limit gains to 1-3 dB
• 256-QAM and above: Minimal extension freedom restricts practical ACE application Higher-order modulations with dense constellations offer less room for outer-point extension, creating a direct trade-off between data rate and PAPR reduction gain.

Advanced ACE implementations employ gradient-based optimization rather than simple iterative clipping. The Smart Gradient Projection (SGP) method formulates PAPR reduction as a convex optimization problem, minimizing peak power subject to constellation constraints. Key benefits include:

• Faster convergence compared to conventional ACE iterations
• Optimal power allocation across subcarriers for peak reduction
• Joint optimization with other PAPR techniques like Tone Reservation SGP-ACE achieves near-optimal PAPR reduction within the theoretical limits of the constellation extension approach.

The fundamental constraint governing ACE operation. EVM quantifies the deviation of measured constellation points from their ideal reference positions. ACE exploits the EVM budget specified by standards like 3GPP—extending outer constellation points outward only within the tolerable distortion limit. This ensures that while peak reduction is achieved, the modulation accuracy remains compliant. Higher-order QAM schemes (e.g., 256-QAM, 1024-QAM) have tighter EVM requirements, directly limiting the aggressiveness of ACE.

The broader category of signal conditioning to which ACE belongs. While techniques like Clipping and Filtering apply amplitude thresholds indiscriminately, ACE is a distortion-aware method. It intelligently modifies only outer constellation points, avoiding the in-band distortion and peak regrowth typical of hard clipping. ACE is often cascaded with conventional CFR in multi-stage architectures, where ACE provides a first pass of distortion-minimized reduction before more aggressive pulse injection handles residual peaks.

An alternative PAPR reduction method that reserves a subset of unused subcarriers specifically to carry a peak-canceling signal. Unlike ACE, which modifies data-bearing constellation points, TR isolates all distortion to reserved tones that are discarded by the receiver. The trade-off is a direct loss of spectral efficiency (throughput). ACE and TR are complementary—ACE can reduce peaks on data subcarriers while TR handles residual excursions, preserving more data capacity than TR alone.

A probabilistic PAPR reduction scheme that generates multiple candidate OFDM symbols from the same data block by applying different phase rotation sequences. The candidate with the lowest PAPR is transmitted, with the phase sequence index sent as side information. Unlike ACE, SLM introduces zero in-band distortion because it selects an unmodified representation. However, it requires side information overhead and multiple IFFT computations, making ACE more attractive for low-latency, computational-efficient implementations.

The statistical tool used to evaluate ACE performance. A CCDF curve plots the probability that a signal's instantaneous power exceeds a given threshold relative to average power. Engineers assess PAPR reduction gain by comparing CCDF curves before and after ACE processing at a specific probability point (e.g., 10⁻⁴). A leftward shift of the CCDF curve indicates successful peak suppression. CCDF analysis reveals whether ACE is effectively reducing the tail of the power distribution without excessive constellation distortion.

The regulatory metric that constrains ACE's out-of-band behavior. While ACE primarily introduces in-band distortion (measured by EVM), the nonlinear extension of constellation points can generate spectral regrowth into adjacent channels. ACE algorithms must be designed with spectral awareness, often incorporating frequency-domain constraints or post-processing filtering to maintain ACLR compliance. Standards like 3GPP TS 38.104 specify ACLR limits (typically -45 dBc) that ACE implementations must satisfy alongside PAPR reduction targets.