Concurrent Multi-Band DPD

The core innovation of concurrent DPD is its ability to synthesize and inject correction signals that specifically cancel cross-band intermodulation distortion (IMD). Unlike parallel single-band DPDs, this architecture models the interaction between carriers. It generates a predistorted signal that pre-compensates for distortion products falling both in-band and in adjacent transmit bands, preventing spectral regrowth that would otherwise violate Adjacent Channel Leakage Ratio (ACLR) masks.

Concurrent DPD relies on multi-dimensional models like the 2D Memory Polynomial (2D-MMP) or Multi-Band Generalized Memory Polynomial (MB-GMP). These models index predistortion coefficients not just on the instantaneous power of a single carrier, but on a vector of envelope magnitudes from all concurrent bands.

• 2D-DPD: Uses a two-dimensional address based on |x1(n)| and |x2(n)|.
• Cross-Terms: Include terms like x1(n)|x2(n-m)|^k to capture cross-band memory effects, where the past envelope of one band influences the current distortion in another.

To function correctly, the coefficients for all bands and their cross-terms must be estimated simultaneously. Joint Coefficient Estimation solves a single, large optimization problem—typically using least-squares (LS) or recursive least-squares (RLS)—on a composite matrix of basis functions. This ensures that the correction for Band 1 does not inadvertently degrade the linearity of Band 2. The Multi-Band Indirect Learning Architecture (MB-ILA) is a common closed-loop topology for this extraction, where a post-distorter is trained on the PA output and then copied to the forward path.

To meet the stringent latency and power requirements of radio units, complex polynomial models are often mapped to a 2D Look-Up Table (2D-LUT). Instead of computing high-order polynomials in real-time, the predistorter uses the instantaneous magnitudes of the two baseband signals to form a 2D address. This address indexes a pre-computed complex gain value. Adaptive algorithms update the LUT contents periodically, trading off memory depth for computational complexity, making it ideal for FPGA-Based DPD Implementation.

Concurrent DPD is often tightly integrated with Multi-Band Crest Factor Reduction (MB-CFR). A composite multi-band signal has a significantly higher Peak-to-Average Power Ratio (PAPR) than individual carriers. MB-CFR jointly processes the combined waveform before the predistorter to clip and filter peaks intelligently, preventing the power amplifier from being driven into deep saturation. This co-design ensures the DPD model operates within a linearizable region of the PA, preventing correction failure.

For extremely wideband multi-carrier scenarios, Subband DPD or Frequency-Selective DPD decomposes the total bandwidth into smaller slices. Independent DPD blocks operate on each subband at a lower sample rate, significantly reducing the total computational load. This architecture is critical for linearizing mmWave and massive carrier aggregation signals where the total linearization bandwidth exceeds the processing capability of a single high-speed digital path, directly addressing the Wideband Signal Linearization challenge.

A predistortion model that synthesizes the correction signal using a two-dimensional indexing structure. The address is typically derived from the instantaneous magnitudes of the two concurrent baseband signals. This allows the model to directly account for the joint envelope behavior driving nonlinearity, making it more accurate than independent single-band DPD for concurrent transmissions.

Nonlinear interference products generated by the interaction of multiple carrier signals within a power amplifier. These unwanted components fall on top of or near the desired transmit bands. Key types include:

• Intermodulation Distortion (IMD): Sum and difference frequency products.
• Cross-Modulation: Transfer of the modulation envelope from one carrier onto another. Mitigation requires cross-band predistorters that actively cancel these specific products.

A behavioral model extending the memory polynomial to two dimensions. It includes cross-terms dependent on the envelope magnitudes of both concurrent bands to capture cross-band memory effects. The model balances complexity and accuracy by including:

• In-band memory terms for each carrier.
• Cross-band envelope coupling terms.
• Sample-crossing products between bands. This structure is foundational for many practical multi-band DPD implementations.

A closed-loop DPD adaptation method where a post-distorter model is identified from the attenuated PA output and then copied to the predistorter in the forward path. This architecture avoids the need for a direct inverse model of the PA. The process involves:

1. Capturing multi-band input and feedback signals.
2. Training a post-distorter to minimize the error between the desired and observed output.
3. Copying the trained coefficients to the forward predistorter.

A signal conditioning technique that jointly reduces the peak-to-average power ratio (PAPR) of a composite multi-band signal. High PAPR drives the PA into deep compression, exacerbating nonlinear distortion. MB-CFR algorithms apply coordinated clipping and filtering across all bands to prevent peak regrowth while managing error vector magnitude (EVM) degradation, ensuring the PA operates in a more linear region.

A hardware-efficient predistorter implementation where complex gain correction values are indexed by a two-dimensional address. The address is derived from the instantaneous magnitudes of two concurrent input signals. This approach avoids real-time polynomial computation, making it ideal for FPGA or ASIC implementation in high-bandwidth systems. Adaptive update mechanisms can modify table entries based on feedback.