Spectral Regrowth

Spectral regrowth originates in the power amplifier (PA) when operated near its saturation point. The amplifier's AM-AM and AM-PM distortion curves cause the modulated envelope to clip and warp, generating intermodulation products that spread energy beyond the intended channel bandwidth. This non-linear behavior is a deterministic function of the specific transistor physics and biasing of each individual amplifier.

The shape and extent of spectral regrowth are heavily influenced by the modulation scheme's peak-to-average power ratio (PAPR). High-PAPR signals like OFDM drive the PA into non-linear regions more frequently, producing wider and more pronounced spectral shoulders. In contrast, constant-envelope modulations like GMSK exhibit minimal regrowth. This interaction creates a modulation-specific signature that can be used to infer transmission parameters.

ACLR is the primary metric for quantifying spectral regrowth, defined as the ratio of transmitted power within the assigned channel to the power leaking into an adjacent channel. Regulatory bodies like the 3GPP and FCC mandate strict ACLR limits to prevent interference. A device's specific ACLR value, measured across multiple offset frequencies, forms a reproducible numerical fingerprint that varies subtly between units of the same model.

Real-world PAs exhibit memory effects caused by thermal dynamics, bias circuit impedance, and trapping phenomena in the transistor. These effects create an asymmetry in the spectral regrowth profile, where the upper and lower sidebands are not mirror images. This asymmetry is highly sensitive to the specific semiconductor material properties and amplifier circuit layout, providing a rich, device-unique feature for fingerprinting systems.

Digital Pre-Distortion is a linearization technique that intentionally distorts the baseband signal to cancel out the PA's non-linearity. While DPD reduces spectral regrowth to meet emission masks, the residual uncorrected distortion still contains unique hardware signatures. The DPD coefficient set itself, which adapts to each PA's specific non-linear curve, can be treated as a compact, high-dimensional feature vector for device identification.

The spectral regrowth profile is not perfectly static; it drifts with junction temperature and component aging. As the PA heats up, its gain and phase response shift, altering the intermodulation products. Over years of operation, gate oxide degradation in the transistor permanently changes the non-linear transfer function. Robust fingerprinting systems must implement drift compensation algorithms that track these slow variations to avoid false rejections.

The primary physical cause of spectral regrowth. When a power amplifier operates near its saturation point, it exhibits AM-AM (amplitude-to-amplitude) and AM-PM (amplitude-to-phase) distortion. This non-linear behavior compresses the signal's envelope and generates intermodulation products that spill into adjacent channels. The specific non-linear transfer function is unique to each amplifier due to microscopic manufacturing variances in the semiconductor transistors, making the resulting regrowth pattern a powerful, unclonable hardware fingerprint.

The primary metric for quantifying spectral regrowth. ACLR is defined as the ratio of the total power transmitted within a signal's assigned channel to the power that leaks into an adjacent channel. It is typically measured in dBc (decibels relative to the carrier). Regulatory bodies like the 3GPP and FCC mandate strict ACLR limits to prevent interference. For fingerprinting, the precise ACLR value and its asymmetry between upper and lower sidebands serve as a stable, device-specific identifier.

A linearization technique designed to counteract spectral regrowth. A DPD system applies an inverse model of the power amplifier's non-linearity to the baseband signal before transmission. The goal is to cancel out the distortion, resulting in a cleaner output spectrum. However, the residual error after DPD correction—the uncorrected distortion—is itself a unique signature. The specific DPD coefficients required to linearize a given amplifier reveal its underlying hardware impairments.

The mathematical mechanism behind spectral regrowth. When a complex modulated signal with multiple frequency components passes through a non-linear device, the components mix to create new signals at sum and difference frequencies. Third-order intermodulation products (IMD3) are particularly problematic because they fall directly into adjacent channels. The amplitude and phase of these IMD products are deterministic functions of the amplifier's non-linear coefficients, providing a rich feature space for Specific Emitter Identification (SEI).

A critical complication in spectral regrowth analysis. Memory effects occur when the amplifier's current output depends not only on the present input but also on past inputs, due to thermal dynamics, bias circuit impedance, and charge trapping in the semiconductor. This causes the spectral regrowth pattern to be asymmetrical between the upper and lower sidebands and to vary with signal bandwidth. The unique time-constant of these memory effects adds another dimension to a device's RF fingerprint.

A rigorous mathematical framework for modeling non-linear systems with memory, including the power amplifiers that cause spectral regrowth. A Volterra series expands the system's output as a sum of multi-dimensional convolution integrals, capturing both instantaneous non-linearity and time-dependent memory effects. The extracted Volterra kernels form a comprehensive, high-dimensional feature vector that uniquely characterizes a transmitter's analog front-end for deep learning-based fingerprinting systems.