Search

The Volterra framework plays an important role in digital pre-distortion (DPD) for power amplifier (PA) linearization. In practice, the full-Volterra (FV) model is avoided due to the so-called curse of dimensionality, which has motivated several Volterra pruning techniques in the literature. However, choosing the pruned-Volterra structure with a given complexity that most accurately describes a PA or DPD with unknown characteristics often requires trial and error. The motivation of this article is to propose a reduced-Volterra model that extends traditional memory polynomials, e.g., the generalized memory polynomial (GMP), by including higher-dimensional terms in a flexible and controlled way, avoiding the curse of dimensionality of the Volterra model and resulting in a model that can apply to a wide range of PA and DPD. Furthermore, the article investigates the parsimonious sizing of the proposed model using least absolute shrinkage and selection operator (LASSO) and sparse-group LASSO (SGL) convex optimization, significantly reducing the model’s running cost, while preserving its performance. In our experimental validation, the sparse versions of the proposed model achieved 3 dB better adjacent channel power ratio (ACPR) than the conventional GMP model for the same or lower DPD running cost, at the expense of a more costly estimation, which is performed off-line.

This paper proposes a novel predistorter architecture suitable for hybrid baseband/RF implementation. The distortion function is split into a memory polynomial and an AM/AM only look-up table (LUT). The memory polynomial function is to be applied in the baseband digital domain whereas the LUT predistorter is suitable for low complexity implementation in either the baseband digital or the RF analog domains. The proposed model was compared to benchmark models using 20MHz and 40MHz wide 5G test signals, and found to be superior in terms of adjacent channel leakage ratio performance and complexity. Experimental results showed that the model can achieve ACLR levels of better than −50dBC with a relatively low number of coefficients. The proposed model requires around 20% less coefficients than the reverse twin-nonlinear two-box model, and achieves slightly better ACLR while easing the analog implementation of the static nonlinearity predistortion function. Most importantly, the low-complexity hybrid implementation of the proposed model can drastically reduce the speed requirements in the digital to analog converter of the signal transmit path by up to 80% when compared to traditional DPD systems.

In this paper, an efficient neural-network-based adaptive DPD design which performs well under power varying conditions is presented. The DPD design is derived on the basis of the envelop time-delay neural network (ETDNN). The redefined ETDNN-DPD requires the part of parameter updates, which enables to adapt it to the rapid change of power amplifier (PA) distortion. Additionally, the redefined ETDNN-DPD also maintains the stability of the compensation performances under varying power condition while its structure is pruned by the structured pruning. Furthermore, to verify its practical use, we also propose the weight scaling technique, which reduces multiplications of the redefined ETDNN-DPD, and applied it to the implementation of the redefined ETDNN-DPD on FPGA. Compared FPGA-implemented ETDNN-DPD with FPGA-implemented conventional memory polynomial DPD, we verified that our proposed DPD achieved 3.2 dB better error vector magnitude (EVM) while lower hardware resource utilization at the fixed power level. Moreover, our proposed DPD kept better performance under the varying power condition only by the partial update of its parameters than memory polynomial DPD.

With the rise of 5G and beyond, the ever-increasing data-rates demanded by mobile access are severely challenging the capacity of optical fronthaul networks. Despite its high reliability and ease of deployment, legacy digital radio-over-fiber (RoF) technologies face an upcoming bandwidth bottleneck in the short term. This has motivated a renewed interest in the development of analog RoF alternatives, owing to their high spectral efficiency. However, unlike its digital counterpart, analog RoF transmission requires a highly linear transceiver to guarantee signal fidelity. Typical solutions exploited in recent research works tend to adopt the use of bulky benchtop components, such as directly modulated lasers (DML) and photodiodes. Although this provides a convenient and quick path for proof-of-concept demonstrations, there is still a considerable gap between lab developments and commercial deployment. Most importantly, a key question arises: can analog-RoF transceivers meet the 5G requirements while being competitive in terms of cost and footprint? Following this challenge, in this work we exploit the use of a low-cost commercial off-the-shelf (COTS) small form-factor pluggable (SFP) transceiver, originally designed for digital transmission at 1 Gbps, which is properly adapted towards analog RoF transmission. Bypassing the digital electronics circuitry of the SFP, while keeping the original transmitter optical sub-assembly (TOSA) and receiver optical sub-assembly (ROSA), we demonstrate that high-performance 5G-compatible transmission can be performed by reusing the key built-in components of current low-cost SFP-class transceivers. Particularly, we demonstrate error vector magnitude (EVM) performances compatible with 5G 64QAM transmission both at 100MHz and 400MHz. Furthermore, employing a memory polynomial model for digital pre-distortion of the transmitted signal, we achieve 256QAM-compatible performance at 100MHz bandwidth, after 20 km fronthaul transmission.

This paper proposes an efficient neural-network-based digital predistortion (DPD), named as envelope time-delay neural network (ETDNN) DPD. The method complies with the physical characteristics of radio-frequency (RF) power amplifiers (PAs) and uses a more compact DPD model than the conventional neural-network-based DPD. Additionally, a structured pruning technique is presented and used to reduce the computational complexity. It is shown that the resulting ETDNN obtained after applying pruning becomes so sparse that its complexity is comparable to that of conventional DPDs such as memory polynomial(MP) and generalized memory polynomial (GMP), while the degradation in performance due to the pruning is negligible. In an experiment on a 3.5-GHz GaN Doherty power amplifier (PA), our method with the proposed pruning had only one-thirtieth the computational complexity of the conventional neural-network-based DPD for the same adjacent channel leakage ratio (ACLR). Moreover, compared with memory-polynomial-based digital predistortion, our method with the proposed pruning achieved a 7-dB improvement in ACLR, despite it having lower computational complexity.

In this paper, a software-defined radio (SDR) based transceiver system is designed and implemented on the system-on-chip (SoC) platform, which consists of a high-speed Arm embedded processor and a reconfigurable field-programmable gate array (FPGA). In the proposed SDR transceiver, the real-time baseband signal generation and adaptive digital predistortion (ADPD) units are implemented on the SoC platform. Memory polynomial model based ADPD solution is implemented to linearize the radio frequency (RF) power amplifiers (PAs). The implementation of the ADPD on a reconfigurable FPGA platform makes the system flexible and cost-effective. The PA characterization, in terms of model extraction and coefficient calculation, is done in real-time. These calculated coefficients are updated in the transmission path to precondition the transmitted signal before it is applied to the PA. The proposed ADPD is applied at the baseband level. Therefore, it can be used for different classes of PA operating at different RF carrier frequencies. A long-term evolution (LTE) signal with 20 MHz bandwidth and 11 dB peak to average power ratio (PAPR) is used for simulation and measurement purposes. The LTE signal is amplified using a GaN-based harmonically tuned continuous Class-F PA in measurement. The performance of the implemented ADPD scheme is analyzed in terms of NMSE, ACPR and EVM.

This paper introduces the measurement/modeling of a nonlinear RF power amplifier (PA) model extraction of spectral overlaps and peak-to-average power ratio (PAPR). The modeling consists of a weighted cubic-spline basis approach that preserves a relationship by its generic asymptotic properties under adequate PAPR regime to estimate signaling conditions. To jointly provide a data-driven model a field-programmable gate array (FPGA) testbed is proposed, in which a precomputed approach with deterministic signals is used to perform: (i) parameter estimation; (ii) adequate PAPR levels; (iii) reinforcement of sparsity data with extrapolation fitting; and (iv) FPGA implementation. Moreover, a technique for parameter identification is introduced to provide insights of a digital predistortion (DPD) PA model extraction with cubic-spline to efficiently improve the linearization performance. A theoretical analysis that states on the benefits from coefficients precomputation to preserve multiple-input multiple-output (MIMO) relationship with antenna selection technique estimation, is also proposed. Also, a Cholesky FPGA algorithm implementation to matrix inversion is validated, which aims to show the good numerical and computational complexity for up to $64\times 64$ MIMO arrays. Experimental results prove a good accuracy and close agreement between the modeling and estimation yielding a reliable model with a little overfitting.

We demonstrate experimentally, for the first time, a distributed digital predistortion model for radio-over-fiber (RoF) downlink suitable for broadband 5G signals. The model reduces the analog-to-digital converter sampling frequency up to 1/3 of the required sampling in the example of the conventional memory polynomial DPD model, hence reducing demodulator costs and hardware complexity. This model also provides better intercarrier interference by reducing the adjacent channel leakage ratio (ACLR). This leads to reduced crosstalk, which impacts the multi-core and multimode RoF links. Long-term evolution advanced signals of a 256-quadratic-amplitude modulation of different bandwidths are used to validate the model. We achieved an average reduction of 2.5 dB in ACLR for signal bandwidths in the range 100–300 MHz while maintaining minimum error vector magnitude (EVM). The RoF downlink can deliver a peak bit rate of 2.016 Gb/s with a spectral efficiency of 6.72 b/Hz and EVM less than 4%.

The analog front of the direct conversion transmitter suffers from the effects of various unavoidable nonlinearities, such as local oscillator (LO) leakage, power amplifier (PA) nonlinearities, and in-phase and quadrature (I/Q) branch amplitude and phase imbalance, etc. To overcome these nonlinear effects, an accurate three-input nonlinear model is proposed for joint compensation of frequency-dependent I/Q imbalance and PA nonlinear distortion in this paper. The proposed nonlinear model is an augmented version of the modulator I/Q imbalance compensation model, where the magnitudes of the input signal are included as the model input, and a three-input nonlinear compensation model with linear parameters is developed. The advantage of the proposed model is that the augmented envelope-dependent terms can produce the basis function sets for compensating the PA nonlinear distortions in the transmitter. Furthermore, the computational complexity of the proposed model is analyzed in detail for practical digital implementation. To verify the validity of the proposed model, the imperfect transmitters based on single-device gallium nitride (GaN) PA and GaN Doherty PA, which are driven by different communication signals, are used for experimental verification and analysis. The experimental results show that the proposed model is able to give improved modeling and distortion mitigation capability than the other reported I/Q imbalance compensation methods when the effects of I/Q imbalance and nonlinear characteristics of PA are considered together.