Conventional radio-frequency (RF) power amplifiers operating with wideband signals, such as wideband code-division multiple access (WCDMA) in the Universal Mobile Telecommunications System (UMTS) must be backed off considerably from their peak power level in order to control out-of-band spurious emissions, also known as "spectral regrowth." Adapting these amplifiers to wideband operation therefore entails larger size and higher cost than would otherwise be required for the same power output. An alternative solution, which is gaining widespread popularity, is to employ digital baseband predistortion ahead of the amplifier to compensate for the nonlinearity effects, hence allowing it to run closer to its maximum output power while maintaining low spectral regrowth. Recent improvements to the technique have included memory effects in the predistortion model, which are essential as the bandwidth increases. In this paper, we relate the general Volterra representation to the classical Wiener, Hammerstein, Wiener-Hammerstein, and parallel Wiener structures, and go on to describe some state-of-the-art predistortion models based on memory polynomials. We then propose a new generalized memory polynomial that achieves the best performance to date, as demonstrated herein with experimental results obtained from a testbed using an actual 30-W, 2-GHz power amplifier

A Generalized Memory Polynomial Model for Digital Predistortion of RF Power Amplifiers

Most recent semi-supervised video object segmentation (VOS) methods rely on fine-tuning deep convolutional neural networks online using the given mask of the first frame or predicted masks of subsequent frames. However, the online fine-tuning process is usually time-consuming, limiting the practical use of such methods. We propose a directional deep embedding and appearance learning (DDEAL) method, which is free of the online fine-tuning process, for fast VOS. First, a global directional matching module, which can be efficiently implemented by parallel convolutional operations, is proposed to learn a semantic pixel-wise embedding as an internal guidance. Second, an effective directional appearance model based statistics is proposed to represent the target and background on a spherical embedding space for VOS. Equipped with the global directional matching module and the directional appearance model learning module, DDEAL learns static cues from the labeled first frame and dynamically updates cues of the subsequent frames for object segmentation. Our method exhibits state-of-the-art VOS performance without using online fine-tuning. Specifically, it achieves a & mean score of 74.8% on DAVIS 2017 dataset and an overall score of 71.3% on the large-scale YouTube-VOS dataset, while retaining a speed of 25 fps with a single NVIDIA TITAN Xp GPU. Furthermore, our faster version runs 31 fps with only a little accuracy loss.

Ieee transactions on neural networks and learning systems

In this article, a thorough overview of behavioral modeling and predistortion of dynamic nonlinearities in RF PAs and transmitters was presented. The sensitivity of the DUT behavior to the characteristics of the stimulus was reviewed to ensure appropriate conditions for accurate observation. Nearly all state-of-the-art behavioral models were described and their relative performance and complexity discussed. Similarities and specifics of behavioral modeling and digital predistortion were presented. Thereby, digital predistortion can be seen as a behavioral modeling problem for which performance assessment is much more straightforward. For DUT behavioral modeling, there is no comprehensive metric that allows the model performance evaluation while taking into account the model accuracy in predicting all the three components of the DUT behavior (in-band distortion, static nonlinearity and memory effects). Finally, a software digital predistortion solution that enables closed-loop wideband linearization was briefly presented with excellent linearization capabilities when amplifying a 12-carrier 60-MHZ wide WCDMA signal.

Behavioral modeling and predistortion

This paper presents the design and implementation of the first in-band full duplex WiFi-PHY based MIMO radios that practically achieve the theoretical doubling of throughput. Our design solves two fundamental challenges associated with MIMO full duplex: complexity and performance. Our design achieves full duplex with a cancellation design whose complexity scales almost linearly with the number of antennas, this complexity is close to the optimal possible. Further we also design novel digital estimation and cancellation algorithms that eliminate almost all interference and achieves the same performance as a single antenna full duplex SISO system, which is again the best possible performance. We prototype our design by building our own analog circuit boards and integrating them with a WiFi-PHY compatible standard WARP software radio implementation. We show experimentally that our design works robustly in noisy indoor environments, and provides close to the expected theoretical doubling of throughput in practice.

/pdf/full-duplex-mimo-radios-3lv4rh6pon.pdf

Full duplex MIMO radios

This paper proposes a new behavioral model to treat memory effects in nonlinear power amplifiers (PAs) Phenomena such as asymmetries in lower and upper intermodulation terms, and variation of AM/AM and AM/PM, depending on signal history, are often observed in high-power PAs To treat these phenomena, this paper presents a model based on the previously developed memory polynomial model The contribution made in this paper is to augment the memory polynomial model to include a sparse delay tap structure that reduces the parameter space required for accurate model identification A figure-of-merit, called the memory effect ratio, is defined to quantify the relative level of distortion due to memory effects, as compared to the memoryless portion Another figure-of-merit is defined as the memory effect modeling ratio, which quantifies the degree to which the PA memory effects have been accounted for in the model This new technique is validated using a variety of RF PAs, including an 880-MHz and a 21-GHz high-power laterally diffused metal-oxide semiconductor PA and various signals such as two-tone, eight-tone, and IS-95B signals

Behavioral modeling of nonlinear RF power amplifiers considering memory effects

Power amplifiers (PAs) are inherently nonlinear devices and are used in virtually all communications systems. Digital baseband predistortion is a highly cost-effective way to linearize PAs, but most existing architectures assume that the PA has a memoryless nonlinearity. For wider bandwidth applications such as wideband code-division multiple access (WCDMA) or wideband orthogonal frequency-division multiplexing (W-OFDM), PA memory effects can no longer be ignored, and memoryless predistortion has limited effectiveness. In this paper, instead of focusing on a particular PA model and building a corresponding predistorter, we focus directly on the predistorter structure. In particular, we propose a memory polynomial model for the predistorter and implement it using an indirect learning architecture. Linearization performance is demonstrated on a three-carrier WCDMA signal.

A robust digital baseband predistorter constructed using memory polynomials

We have developed a novel technique for measuring the dielectric constant and loss tangent of a thin film dielectric material up to 5 GHz. The dielectric film needs to be deposited on a metal layer and capped with a metal electrode layer. The bottom metal layer does not have to be very conductive, as long as its sheet resistance is uniform and known. Only one step lithography on the top metal layer is required. No dc electrical contact to the bottom metal layer is necessary. The measurement is taken with a Vector Network Analyzer and a coplanar-wave-guide miniature wafer probe.

RF measurement technique for characterizing thin dielectric films

Impairments in the analog quadrature modulator can adversely impact the performance of a digital predistortion linearization system. In this paper, we focus on modeling and compensation of frequency-dependent gain/phase imbalance and dc offset when the input and output of the direct upconverter are available. We first propose channel models to describe the effects of these impairments on transmitted I/Q data streams. We then develop algorithms to estimate the parameters of these channel models and to construct I/Q compensators. Performance of the algorithms and the I/Q compensators is shown in conjunction with digital predistortion.

Compensation of Frequency-Dependent Gain/Phase Imbalance in Predistortion Linearization Systems

Power amplifiers (PAs) are inherently nonlinear devices and are used in virtually all communications systems. Digital baseband predistortion is a highly cost effective way to linearize PAs, but most existing architectures assume that the PA has a memoryless nonlinearity. For wider bandwidth applications such as WCDMA, PA memory effects can no longer be ignored, and memoryless predistortion has limited effectiveness. In this paper, instead of focusing on a particular PA model and building a corresponding predistorter, we focus directly on the predistorter structure. In particular, we propose a memory polynomial model for the predistorter and implement it using an indirect learning architecture. Linearization performance is demonstrated on a 3-carrier UMTS signal.

Zhengxiang Ma

Papers

A Generalized Memory Polynomial Model for Digital Predistortion of RF Power Amplifiers

A robust digital baseband predistorter constructed using memory polynomials

RF measurement technique for characterizing thin dielectric films

Compensation of Frequency-Dependent Gain/Phase Imbalance in Predistortion Linearization Systems

Memory polynomial predistorter based on the indirect learning architecture