Self-organizing Maps for Digital Pre-distortion
(2020) EITM02 20182Department of Electrical and Information Technology
- Abstract
- Power amplifiers are very important components in the area of wireless communications. However, they are non-linear devices and they seem to achieve high efficiency when they are operating in the non-linear regions, at the cost of being more power-consuming. In order to linearize their behavior, designers have been considering many methods. Among those, the digital pre-distortion tends to be the most popular one.
There are different ways of applying the digital pre-distortion. Two of the most common are the pseudo-inverse and gradient descent methods. Both those two methods require too much power, as a result of the high amount of computation. For this reason, new power-efficient methods are under research.
The main focus of this... (More) - Power amplifiers are very important components in the area of wireless communications. However, they are non-linear devices and they seem to achieve high efficiency when they are operating in the non-linear regions, at the cost of being more power-consuming. In order to linearize their behavior, designers have been considering many methods. Among those, the digital pre-distortion tends to be the most popular one.
There are different ways of applying the digital pre-distortion. Two of the most common are the pseudo-inverse and gradient descent methods. Both those two methods require too much power, as a result of the high amount of computation. For this reason, new power-efficient methods are under research.
The main focus of this master thesis is to explore the possibility of applying the digital pre-distortion with the use of self-organizing maps, a machine learning algorithm, to develop a more efficient digital pre-distortion (DPD) unit. The algorithm implementation and the simulations were performed with the use of MATLAB and the neural network toolbox.
During this research, the performance, the accuracy and the computational complexity of the implemented design were measured before and after the fine tuning of some of the system’s parameters. The target is to show if the use of this algorithm is an efficient method of implementing digital pre-
distortion. (Less) - Popular Abstract
- The power amplifier is an electronic device that is used in order to increase the magnitude of the power of a given input signal. It is used for devices like speakers, headphones and RF transmitters. It is considered as an essential element within the wireless communication area, due to the fact that it is used for the transmission and the broadcasting of the signals to the users. In addition to those, with the use of power amplifiers and the increase of the power levels, higher data transfer rates became available. What basically a power amplifier does, from a computation perspective, is that it receives an input signal and multiplies it with the desired gain. But that is an ideal model of a power amplifier. However, the power amplifiers... (More)
- The power amplifier is an electronic device that is used in order to increase the magnitude of the power of a given input signal. It is used for devices like speakers, headphones and RF transmitters. It is considered as an essential element within the wireless communication area, due to the fact that it is used for the transmission and the broadcasting of the signals to the users. In addition to those, with the use of power amplifiers and the increase of the power levels, higher data transfer rates became available. What basically a power amplifier does, from a computation perspective, is that it receives an input signal and multiplies it with the desired gain. But that is an ideal model of a power amplifier. However, the power amplifiers are non-linear sources for a communication system, so the real model differs a lot from the ideal one. They tend to be non-linear as their output power increases and reaches close to its maximum value, which can create an in-band distortion within the system. For this reason, the linearization of power amplifiers is a very important topic under research in the digital communications field.
The most common method for linearizing a power amplifier’s behavior is the digital pre-distortion. This method is very power efficient as well as cost-
saving. Ideally, with the pre-distortion, the characteristics of a power amplifier are inverted in order to compensate for the non-linearities. The role of a pre-distorter unit inside a digital communication system is to correct any possible gain and phase nonlinearities and in combination with the system’s amplifier to produce a “clear” out of distortion signal. With the use of the pre-distortion and the gain stability that it can provide in the output of the amplifiers, the construction of bigger, more expensive and less efficient amplifiers is no longer necessary. Despite the fact that the pre-distortion is widely and successfully used, it has been observed that the current ways of applying the pre-distortion are very power-consuming due to their complexity and the amount of computational power they require in order to perform the pre-distortion. For that reason, new approaches are under the scope.
Nowadays, there is a trend in electronics where many concepts are being implemented with the use of machine learning in order to replace the number of computations with simpler logic, such as the classification of the data and the prediction of the desired values. It is easy to understand, that a high amount of computations inside a system means more power which leads to a
higher operating cost. As a result, a more power-efficient system could lead to a big saving in terms of power and money.
This project evaluates the impact of the self-organizing maps algorithm as an adaptive algorithm for the digital pre-distortion. The main goal is to reduce computations and provide a more power-efficient system. It could be a
guideline for future researches with new approaches that might lead to more energy-saving results. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9030131
- author
- Isaakidis, Michail LU
- supervisor
-
- Liang Liu LU
- Ove Edfors LU
- organization
- course
- EITM02 20182
- year
- 2020
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Self-organizing maps, Digital pre-distortion
- report number
- LU/LTH-EIT 2020-789
- language
- English
- id
- 9030131
- date added to LUP
- 2020-10-13 14:48:21
- date last changed
- 2020-10-13 14:48:21
@misc{9030131,
abstract = {{Power amplifiers are very important components in the area of wireless communications. However, they are non-linear devices and they seem to achieve high efficiency when they are operating in the non-linear regions, at the cost of being more power-consuming. In order to linearize their behavior, designers have been considering many methods. Among those, the digital pre-distortion tends to be the most popular one.
There are different ways of applying the digital pre-distortion. Two of the most common are the pseudo-inverse and gradient descent methods. Both those two methods require too much power, as a result of the high amount of computation. For this reason, new power-efficient methods are under research.
The main focus of this master thesis is to explore the possibility of applying the digital pre-distortion with the use of self-organizing maps, a machine learning algorithm, to develop a more efficient digital pre-distortion (DPD) unit. The algorithm implementation and the simulations were performed with the use of MATLAB and the neural network toolbox.
During this research, the performance, the accuracy and the computational complexity of the implemented design were measured before and after the fine tuning of some of the system’s parameters. The target is to show if the use of this algorithm is an efficient method of implementing digital pre-
distortion.}},
author = {{Isaakidis, Michail}},
language = {{eng}},
note = {{Student Paper}},
title = {{Self-organizing Maps for Digital Pre-distortion}},
year = {{2020}},
}