Inspecting Distortion in the Power Amplifiers with the aid of Neural Networks

Creative Commons License

Güneş E. O., Aygün S., Kouhalvandi L., Özoğuz İ. S.

IEEE/ICACT2020, 22nd International Conference on Advanced Communications Technology (ICACT2020), Pyeongtaek, South Korea, 16 - 19 February 2020, pp.1-6

  • Publication Type: Conference Paper / Full Text
  • City: Pyeongtaek
  • Country: South Korea
  • Page Numbers: pp.1-6
  • Istanbul Technical University Affiliated: Yes


This paper aims to classify the distortion behavior
of a power amplifier (PA) with the aid of a neural network.
Power amplifiers have quite extensive usage in communication
systems especially with the current developments on 5G and
more. However, distortion in the power amplifiers needs attention
to be pre-distorted with the help of a feedback mechanism using
direct or indirect methods in the digital domain. In the literature,
there are several efforts to understand and reduce distortion in
amplifier devices. Therefore, in this paper, the distortion behavior
in the power amplifier is inspected using the neural networks. In
this work, we have obtained a software-defined network using the
strength of the neural network to inspect the distorted and nondistorted
data as a binary classification on the actual design of the
power amplifier in [1]. For this purpose, a neural network system
is trained. In the tests, more than 96% accuracy can easily be
obtained in an early epoch with the cleverly chosen learning rate
(a) which is optimally outperforming thereabouts after a=0.05
till 0.1. Thus, the linearity and non-linearity response of the PA
is considered with the help of the trained network.