LINEARIZATION TECHNOLOGY OF HIGH-SPEED MEASUREMENT SYSTEMS

Authors

  • Peng Zhang (Corresponding Author) Ceyear Technologies Co., Ltd, Qingdao 266555, Shandong, China.

Keywords:

Nonlinear compensation, Complex valued recurrent neural network, High speed signal acquisition, System identification, Analog to digital converter

Abstract

Receiver sensitivity, dynamic range, and adjacent channel interference all degrade when nonlinear distortion is present, which remains a persistent concern in high speed signal acquisition. The distortion is generated as the signal passes through the low noise amplifier, the filters, the mixers and, ultimately, the analog to digital converter (ADC); once these stages are cascaded, their joint nonlinear behaviour becomes hard to characterise, and memory effects make the task harder still. In practice, conventional polynomial models often fail to compensate this kind of distortion to the accuracy a system actually requires. The present study therefore examines a complex valued recurrent neural network (CRNN) to model and then compensate nonlinear errors within a single channel acquisition chain. To capture the input output relationship, the network is fed delayed input samples together with output feedback, and the trained model is subsequently inverted to recover the undistorted signal. Both two tone and quadrature phase shift keying (QPSK) signals are used to run the simulations. For the two tone cases the spectral performance metric gains 3.73 dB and 3.72 dB, while the QPSK case shows a gain of 7.357 dB. Compensation, however, is not uniform: the CRNN suppresses some distortion components more effectively than others.

References

[1] Schetzen M. The Volterra and Wiener Theories of Nonlinear Systems. New York, NY, USA: Wiley, 1980.

[2] Peng L, Ma H. Design and implementation of software-defined radio receiver based on blind nonlinear system identification and compensation. IEEE Transactions on Circuits and Systems I: Regular Papers, 2011, 58(11): 2776-2789. DOI: 10.1109/TCSI.2011.2151050

[3] Chu J, Chen W, Chen L, et al. A cascaded memory polynomial-neural network behavior model for digital predistortion. Proc. 2020 IEEE MTT-S Int. Conf. Numerical Electromagnetic and Multiphysics Modeling and Optimization (NEMO), 2020. DOI: 10.1109/NEMO49486.2020.9343656.

[4] Khan M T, Ding Y, Goussetis G. Next-gen digital predistortion from hardware acceleration of neural networks: Trends, challenges, and future. IEEE Trans. Neural Netw. Learn. Syst., 2026, 37(8): 3503-3523. DOI: 10.1109/TNNLS.2026.3656642.

[5] Kalra G S, Devi R V S, Dimlioglu D. Neural network-based digital predistortion for wideband power amplifiers: A comparative study of architectures and performance–complexity tradeoffs. IEEE Microwave Magazine, 2026. DOI: 10.1109/MMM.2026.3725608.

[6] Goh S L, Mandic D P. A complex-valued RTRL algorithm for recurrent neural networks. Neural Computation, 2004, 16(12): 2699-2713. DOI: 10.1162/0899766042321779.

[7] Goh S L, Mandic D P. An augmented CRTRL for complex-valued recurrent neural networks. Neural Networks, 2007, 20(10): 1061-1066. DOI: 10.1016/j.neunet.2007.09.015.

[8] Baruch I S, Arellano-Quintana V, Reynaud E P. Complex-valued neural network topology and learning applied for identification and control of nonlinear systems. Neurocomputing, 2017, 233: 104-115. DOI: 10.1016/j.neucom.2016.09.109.

Downloads

Published

2026-11-12

How to Cite

Peng Zhang. Linearization Technology Of High-Speed Measurement Systems. Eurasia Journal of Science and Technology. 2026, 8(5): 34-37. DOI: https://doi.org/10.61784/ejst3171.