Spectral Regrowth

The fundamental physical cause of spectral regrowth. When a Power Amplifier (PA) operates near its saturation point for efficiency, its gain compresses. This AM-AM distortion (amplitude-dependent gain) and AM-PM distortion (amplitude-dependent phase shift) create a non-linear transfer function. This non-linearity generates intermodulation products when a modulated signal passes through, causing energy to spill from the intended carrier frequency into adjacent channels. The severity is directly proportional to the PA's deviation from an ideal linear response.

Modern spectrally efficient modulation schemes like OFDM have a high Peak-to-Average Power Ratio (PAPR). This means the signal has infrequent but extreme power peaks. To avoid clipping these peaks and generating severe distortion, the PA must operate at a large output back-off (OBO) from its saturation point. A high PAPR forces the PA into a highly inefficient linear region, but any reduction in back-off to improve efficiency immediately pushes the peaks into the non-linear region, directly causing spectral regrowth.

The PA's output is not solely dependent on the current input; it also depends on past inputs. These memory effects are caused by thermal dynamics, bias circuit impedance, and semiconductor trapping effects. They create an asymmetric distortion pattern in the frequency domain, making the spectral regrowth uneven across the upper and lower adjacent channels. Memory effects invalidate simple memoryless linearization models like basic Look-Up Tables (LUTs), requiring more complex Volterra Series or Generalized Memory Polynomial (GMP) models for correction.

As the bandwidth of the transmitted signal increases (e.g., for 5G NR signals using 100 MHz carriers), the frequency dependency of the PA's non-linearity becomes significant. The gain and phase response of the amplifier can vary across the signal's own bandwidth. This wideband non-linearity causes the spectral regrowth to be a distorted, filtered version of the classic intermodulation pattern. The bandwidth of the regrown spectrum is typically 3-5 times the original signal bandwidth, making wideband linearization a critical challenge for modern systems.

The physical characteristics of a PA's transistors change with junction temperature and over the component's operational lifetime. A DPD model identified at a cold start will become mismatched as the amplifier heats up to its steady-state operating temperature. This drift alters the exact shape of the non-linearity, causing the spectral regrowth to slowly increase over time. This necessitates coefficient adaptation through online training to maintain regulatory compliance for Adjacent Channel Leakage Ratio (ACLR) over long transmissions.

The impedance seen by the PA's output, known as the load impedance, is not constant. It changes with the antenna's environment (the "antenna mismatch"). A deviation from the ideal 50-ohm load alters the PA's gain and phase characteristics, directly changing its non-linear behavior. This is a critical issue for mobile handsets, where the antenna impedance fluctuates based on how the device is held. Over-the-Air DPD techniques are required to capture and compensate for this variable load-dependent spectral regrowth.

The primary regulatory metric for quantifying spectral regrowth. ACLR measures the ratio of transmitted power within the assigned channel to the power that has leaked into an adjacent channel due to intermodulation distortion. A poor ACLR indicates severe non-linearity and can cause interference with neighboring transmissions, leading to compliance failures. Modern DPD systems are explicitly designed to improve ACLR by 15-25 dB.

The root cause of spectral regrowth. When a power amplifier operates near its saturation point to maximize efficiency, its gain compresses, and the output is no longer a linear function of the input. This non-linear transfer function generates intermodulation products (IMDs) that spread the signal's bandwidth. Key manifestations include:

• AM-AM Distortion: Amplitude-dependent gain compression.
• AM-PM Distortion: Amplitude-dependent phase shift.
• Memory Effects: Output dependence on prior signal states due to thermal and electrical dynamics.

A signal characteristic that directly exacerbates spectral regrowth. Modern modulation schemes like OFDM exhibit high PAPR, meaning the signal has extreme amplitude peaks relative to its average power. To avoid clipping these peaks and generating severe out-of-band emissions, the power amplifier must operate with a large back-off from its saturation point, drastically reducing efficiency. Crest Factor Reduction (CFR) is often paired with DPD to lower PAPR before the signal reaches the amplifier.

The physical mechanism behind spectral regrowth. When a multi-tone or modulated signal passes through a non-linear device, the mixing of frequency components creates new spectral products at sums and differences of the original frequencies. Third-order intermodulation products (IMD3) are particularly problematic because they fall close to the original carrier and are difficult to filter. The spectral regrowth pattern is essentially the convolution of the signal's spectrum with itself due to these IMD products.

The primary countermeasure against spectral regrowth. DPD applies an inverse model of the power amplifier's non-linearity to the digital baseband signal before transmission. When the pre-distorted signal passes through the amplifier, the two non-linearities cancel, resulting in a linear output with minimal spectral regrowth. Key architectures include:

• Indirect Learning Architecture (ILA)
• Direct Learning Architecture (DLA)
• Neural Network DPD for complex memory effects

A comprehensive in-band signal quality metric that correlates with spectral regrowth. While ACLR measures out-of-band emissions, EVM quantifies the in-band distortion caused by the same non-linearity. High EVM indicates that constellation points are displaced from their ideal locations, degrading the bit error rate. A well-designed DPD system simultaneously improves both ACLR (spectral containment) and EVM (modulation accuracy).