A novel joint satellite-transmitter and ground-receiver (JSG) digital pre-distortion (DPD) (JSG-DPD) technique is proposed to improve the linearity and power efficiency of the space-borne active phased arrays (APAs) in low Earth orbit (LEO) satellite communications. Different from the conventional DPD technique that requires a complex RF feedback loop, the DPD coefficients based on a generalized memory polynomial (GMP) model are extracted at the ground-receiver and then transmitted to the digital baseband front-end of the LEO satellite-transmitter via a satellite–ground bi-directional transmission link. The issue of the additive white Gaussian noise (AWGN) of the satellite–ground channel affecting the extraction of DPD coefficients is tackled using a superimposing training sequences (STS) method. The proposed technique has been experimentally verified using a 28 GHz phased array. The performance improvements in terms of error vector amplitude (EVM) and adjacent channel power ratio (ACPR) are 7.5% and 3.6 dB, respectively. Requiring limited space-borne resources, this technique offers a promising solution to achieve APA DPD for LEO satellite communications.

This paper proposes a robust digital signal recovery (DSR) technique to tackle the high signal-to-noise ratio (SNR) variation and transmitter memory effects for broadband power efficient down-link in next-generation low Earth orbit (LEO) satellite constellations. The robustness against low SNR is achieved by concurrently integrating magnitude normalization and noise feature filtering using a filtering block built with one batch normalization (BN) layer and two bidirectional long short-term memory (BiLSTM) layers. Moreover, unlike existing deep neural network-based DSR techniques (DNN-DSR), which failed to effectively take into account the memory effects of radio-frequency power amplifiers (RF-PAs) in the model design, the proposed BiLSTM-DSR technique can extracts the sequential characteristics of the adjacent in-phase (I) and quadrature (Q) samples, and hence can obtain superior memory effects compensation compared with the DNN-DSR technique. Experimental validation results of the proposed BiLSTM-DSR with a 100 MHz bandwidth OFDM signal demonstrate an excellent performance of 11.83 dB and 9.4% improvement for adjacent channel power ratio (ACPR) and error vector magnitude (EVM), respectively. BiLSTM-DSR also outperforms the existing DNN-DSR technique in terms of the ACPR and EVM by 2.4 dB and 0.9%, which provides a promising solution for developing deep learning-assisted receivers for high-throughput LEO satellite networks.

This brief proposes a digital signal recovery (DSR) method to compensate the nonlinear distortion introduced by power amplifiers (PAs) under dynamic nonlinear operating states. Unlike conventional PA linearization methods that extract the nonlinearity based on the baseband I/Q PA input and output signal samples, the proposed method attempts to derive the memory polynomial (MP) model parameters based on PA operating states using a deep neural network (DNN). This method allows the receiver to achieve DSR by tracking the operating states of the PA effectively with a few telemetry data. Validation results from simulations and experiments based on a GaN PA operating at 3.5 GHz reveal that the proposed method can maintain satisfactory DSR performance in terms of adjacent channel power ratio (ACPR) and error vector magnitude (EVM) while the transmitter PA is operating with fluctuating average input/output power, supply voltage, and bias voltage. The training data size and time are further reduced by using a transfer learning (TL) approach.

Geosynchronous satellite (GEO) communications are highly susceptible to interference from environments such as rain, clouds, and hail. This exhibits a technical challenge to accurately detect the status of the weak Q-band (39.4 GHz) satellite beacon signals, which is critical for ensuring reliable satellite pointing using the ground station antenna. This paper exploits recent advances in deep learning to cope with this challenge. The proposed approach based on deep neural networks (DNN) and filtering (Filter-DNN) classifies rare events such as cloud, mist, rain as Non-line of sight (NLOS) and ordinary clear skies as Line of sight (LOS). Beacon data from a GEO satellite (Alphasat) ground station under two different attenuation conditions is used for validation. The experimental results show that our method can detect the rare event with an accuracy score of 92% by using only 2000 data sample points, while conventional approaches such as MUSIC require about 100k sample points to maintain detection. These results indicate that the proposed technique could be a promising tool for achieving satisfactory results in space exploration and gain insights into GEO satellite communication problems.

This letter presents a continual transfer learning-assisted (CTLA) digital predistortion (DPD) (CTLA-DPD) method for linearizing active phased arrays (APAs) with dynamic nonlinearities. Unlike existing transfer learning-assisted (TLA) DPD methods, the proposed method does not rely on a specific reference operating state. Instead, whenever the working conditions of the APA are updated, deep neural networks (DNNs) utilized in the previous state serve as the initial model to initiate a new training process until the retrained DNN effectively adapts to the new operating state. Experiments are conducted on an APA in the 28 GHz band with dynamic configurations including bandwidth, average input power level, and beam steering angle. The experimental results show that the proposed method can reduce the required data amount for the online transfer learning phase by 60% and achieve performance improvements in adjacent channel power ratio (ACPR) and normalized mean squared error (NMSE).

In this letter, a transfer learning-assisted (TLA) digital signal recovery (DSR) method is proposed for the linearization of active phased arrays (APAs) for satellite communications to handle the transition to different communication scenarios. The frozen layers of a deep neural network (DNN) are pretrained to extract the main nonlinear characteristics of the APA, and the fine-tuning layers are adopted to achieve prompt update of the regression parameters according to the variations in average input power levels and steering angles. This allows satisfactory reception performance to be maintained in time when the nonlinearity of the APA changes. Experimental results on a $4 \times 4$ APA antenna at 28 GHz show that the proposed method improves the adjacent channel power ratio (ACPR) and the error vector magnitude (EVM) performance. Compared with the existing fully relearned DNN-DSR method, the proposed method achieves almost as good performance with only 1/10 of the data.