Digital Signal Bandwidth Estimation 

Get Complete Project Material File(s) Now! »

Chapter 2: System Description and Current Literature

System Overview

As introduced previously, the Universal Classifier Synchronizer (UCS) is a Cognitive Radio system/sensor that can detect, classify, and extract the relevant parameters from a received signal to establish physical layer communications using the received signal’s profile. The overall system is composed of different parts and can be identified by the block diagram presented in the figure below.The current implementation is able to identify signals including AM, FM, MPSK, QAM, MFSK, and OFDM. The system is constructed to run on Universal Software Radio Peripheral (USRP) with the GNU Radio software toolkit and also runs on an Anritsu™ signal analyzer. In this chapter, an overview of the system block diagram is introduced. Also, current literature of similar systems and technology is also discussed to see where this system fits into the grand scheme of cognitive radio classifiers and synchronizers. Also, note that the Bandwidth Estimation and the Symbol Timing & Coarse Classification phases are discussed in full detail in later chapters. They are also the focus of this report and the sections of the larger UCS system developed, implemented, and tested for embedded SDR hardware, Lyrtech SFF SDR platform.
The UCS system should ideally be implemented using wide band radio systems that can communicate on different frequency bands. The system can also be implemented on narrow band systems suited for finding signals in a particular band of interest. Therefore, the UCS system is configured to a frequency span to find signals on frequencies of interest, or in a dynamic spectrum scenario, span the region of opportunistic spectrum allocation. The system first performs Spectrum Sensing on all received signals to find signal energy using a Power Spectral Density (PSD) technique. The presence of signal energy provides the location in the frequency domain of the received signal. Spectrum sensing also provides the system with spectrum occupation of signals in the band of interest. In a dynamic spectrum environment, the system can therefore choose unoccupied frequency space to communicate and avoid interference.
The detection of signal energy by the spectrum sensing process starts the cognitive decision process of the UCS system. Wideband suite categorization is first performed. If the signal is wideband for example OFDM, then it is block-based which means that demodulation is performed block by block. If the signal is narrowband, then narrowband analog-digital categorization is performed next. If the signal is determined to be analog, then the suite is sample-based which means it must be demodulated sample by sample, such as FM and AM. If the signal is digital like MPSK or QAM, then it is symbol-based which means that symbol timing and synchronization has to be performed before demodulation. If a signal is identified to belong to one of the three categories described (block-, sample-, or symbol-based), the corresponding decision path to demodulation is taken to estimate the necessary parameters for correct demodulation. The entire structure of UCS prototype can be understood as four branches and three phases. The four branches include multi-carrier digital signal, narrowband digital signal, analog signal, and standard FSK signal based on the different feature extraction scheme for different types of signals. The three phases are briefly concluded as Phase 1: classification, Phase 2: synchronization, and Phase 3: demodulation [Chen, 2008].
The focus of this thesis and system implementation is part of an ongoing process to implement parts of the above mentioned UCS system on the standalone SDR hardware. The aim is to eventually develop the complete UCS CR system on the Lyrtech SFF SDR platform that can act as a standalone portable CR system. The modules created and implanted/implemented on the SDR hardware are the Bandwidth Estimation and Symbol Timing & Coarse Classification modules. This is the system decision paths towards classification, synchronization, and demodulation of digital phase modulated signals (QAM and MPSK signal types) and also analog signals. Let’s assume that an incoming signal is captured by the system and that Spectrum Sensing is first performed to identify the presence of signal energy. Let us also assume that this unknown signal type is either an analog or digital phase modulated signal. Wideband and Narrowband Categorization is then performed which identifies the signal to be narrowband. The system then moves to the next block where Narrowband Categorization is performed to identify the signal to either an analog or digital signal. Regardless of the pre-classification of a signal into analog or digital, Bandwidth Estimation is performed using a technique that involves taking the histogram of the PSD and is discussed in further detail in the next chapter. If the signal is pre-classified as an analog signal, after estimating the bandwidth, the correct analog demodulator can be loaded by the system to complete the demodulation process. The path towards analog signal classification and demodulation is identified in orange in the above system block diagram. If the signal is determined to be a digital signal, the system then also moves to the Bandwidth Estimation block before taking a different route, from that of analog signals, to Symbol Timing & Coarse Classification. The Symbol Timing & Coarse Classification phase is the most important part of the UCS system. Using a fair variance algorithm, the symbol rate of the signal is estimated through a resampling and sample-based variance elimination process based on variance calculations of digital complex IQ samples. A coarse classification of the digital signal is also performed to distinguish it between MPSK or QAM signal schemes. Coarse classification is achieved by analyzing the envelope order of the digital signal. MPSK signals have a single constant envelope, whereas QAM signals’ envelopes are centralized around a few different envelope values. The system can therefore classify the digital signal to one of 3 sets: MPSK, 16QAM, and 64QAM.
The Carrier Synchronization and Fine Classification phases are next. Carrier Synchronization is performed by implementing a Phase Lock Loop (PLL) to achieve frequency and phase synchronizations after signal parameters like modulation type are known. Fine Classification is achieved by removing phase and frequency information from the transmitted signal through the use of the PLL implemented in the Carrier Synchronization stage. With MPSK signals, the amplitude of the transmitted signal is a constant and thus produces a circular constellation plot. Therefore, the phase difference between information bearing elements of the signal achieves fine classification, as the modulation order can now be determined based on the phase information removed from the transmitted signal. With QAM signals, the amplitude of signal varies with the phase, which means that the constellation is uniformly distributed between squares. Fine classification is obtained based on this distribution as the order of the QAM signal is determined by the constellation distribution information that is removed from the transmitted signal. The path of a digital phase modulated signal through the system can be identified by the green blocks in the above system block diagram. After achieving fine classification and the exact type of QAM or PSK signal type is determined, the UCS CR system can load the correct digital demodulator profile to receive the signal to perform demodulation. For a complete explanation of the decision through all four system paths and how the associated blocks affect the decision process, see the original UCS publication [Chen 2008].

Review of Current literature

Signal classification became an attractive research topic in the 1980s as a part of Signals Intelligence or SIGINT, which saw electronic signal interception as early as the Boer War. The Boers captured British radios in order to intercept and interpret the British transmitted signals to provide an edge in the war. In World War II, the United States Marine Corps in their communication units used Native American Navajo speakers, referred to as ‘code talkers’, to speak their coded language. This was a means to prevent the interpretation of possible intercepted radio communications during the war. In the United States and many other countries worldwide, the topic of Signal Interpretation has become of great interest as a part of tactical and military operations. In order to intercept and interpret another’s signal, the signal has to be first detected and classified correctly. In today’s world, the advancement in communications systems has provided many advanced communications systems with a vast array of signal schemes and profiles. Therefore, the task of detecting and classifying over the air signals with little or no prior knowledge of the communicating scheme has proven to be a very difficult and complex task. As a result of the increasing interest in software defined and cognitive radio, signal classification is gaining more attention and is also becoming more practical to solve. Cognitive Radios along with the use of wideband radio hardware have made the task more feasible. With the use of wideband radio hardware, a cognitive radio can be configured to detect signals along many frequency bands. Having detected and captured the signals, further signals analysis and processing can be done in software to compare the signals to the many different profiles available today to match it as closely as possible to the right profile. Therefore a cognitive radio can perform a case by case analysis as it analyses a signal in order to classify it correctly to the right profile.
The methodologies and technologies in the area of signal classification can be roughly divided into three categories: (a) maximum likelihood (ML)-based, (b) feature-extraction based and (c) cyclostationary feature-based [Le, 2007]. Method (a) is classified by comparing the likelihood of candidate signal and modulation types. Polydoros and Kim (1990) is a classic article that discusses the optimal classification rules. Beidas and Weber (1998) is about asynchronous classification for MFSK. Method (b) directly extracts phase or amplitude features from the target signal to differentiate modulations. Zero crossing and wavelet technology are quite frequently involved in this area [Hsue and Soliman, 1990; Jahankhani et al., 2006; Proch´azka et al., 2008]. Some publications combine (a) and (b) to get better performance. For example, in Yucek and Arslan (2007), both ML and extracted features are used for OFDM signal detecting and classification in cognitive radio. Method (c) is attractive for DSA applications because of its ability to detect and classify signals at low SNRs [Kim et al., 2007]. The methods mentioned above have excellent performance in certain scenarios. The scenario conditions include channel types, signal types, and equipment. The objective behind UCS is to design a universal signal classification and synchronization system that can analyze a signal’s physical layer features with minimal prior information and application limits and can demodulate the signal using the acquired information [Chen, 2008].
Apart from the proof of concept prototypes that are implemented in GNU Radio with USRP and the implementation on the Anritsu™ signal analyzer, some work has been done towards the implementation of the UCS system on embedded SDR hardware. The Spectrum Sensing and Wideband/Narrowband categorization blocks were implemented on the embedded SDR platform by another graduate student in our research Lab [Nair, 2009]. These implementations of parts of the UCS system work in conjunction with the modules of this thesis to enable the complete classification of analog and digital phase modulated signals. As discussed earlier, the Spectrum Sensing module first identifies signal energy and starts the decision process. The Wideband and Narrowband blocks pre-classify the signal along with identifying them as analog and digital signals. If the signal is deemed to be analog, the AM-FM detector, which is part of the Narrowband Categorization block, identifies the correct demodulator profile. The Bandwidth Estimation block implementation of this thesis can then be called prior to demodulation. This is identified by the blocks in orange in the UCS system block diagram above. If the signal is deemed to be digital, then the Bandwidth Estimation and Symbol Timing & Coarse Classification blocks of this thesis are called to find the symbol rate of the signal along with the modulation type. Although not implemented in this thesis, the Carrier Synchronization and Fine Classification blocks would then have to be called to synchronize and demodulate the digital phase modulated signal. Therefore the work done by Nair [Nair, 2009], and the implementations of this thesis complement each other. Both of these implementations seem to be the only instance of signal classifier implemented on embedded SDR hardware up to date. Most of the published work seen to date with signal classifiers and cognitive software defined radio is performed with generic SDR hardware such as the USRP running on host computers.

Grant Information 
Chapter 1: Introduction 
1.1 Motivation and Objectives
1.2 Thesis Outline
Chapter 2: System Description and Current Literature
2.1 System Overview
2.2 Review of Current Literature
Chapter 3: Digital Signal Bandwidth Estimation 
3.1 Introduction
3.2 Histogram of PSD Technique
3.3 Module Implementation
3.4 Conclusion
Chapter 4: Symbol Timing 
4.1 Introduction and Overview of Symbol Timing
4.2 Vector Re-sampling
4.3 Symbol Rate Estimation and Coarse Classification
4.4 Reshaping and Variance Among Samples
4.5 Symbol Timing Conclusion
Chapter 5: FPGA Digital Receiver and SDR Platform
5.1 SDR Platform Overview
5.2 Digital Receiver Overview and FPGA Implementation
5.3 Conclusion
Chapter 6: Conclusion and Future Work 
6.1 Summary and Conclusion
6.2 Future Work
Symbol Timing and Coarse Classification of Phase Modulated Signals on a Standalone SDR Platform

Related Posts