Analysis and Implementation of a Novel Single Channel Direction Finding Algorithm on a Software Defined Radio PlatformJohn Joseph KeavenyCapter1A radio direction finding (DF) system is an antenna array and a receiver arranged in a combination to determine the azimuth angle of a distant emitter. Basically, all DF systems derive the emitter location from an initial determination of the angle-of-arrival (AOA).Radio direction finding techniques have classically been based on multiple-antenna systems employing multiple receivers. Classic techniques such as MUSIC 12 and ESPRIT use simultaneous phase information from each antenna to estimate the angle-of-arrival of the signal of interest. In many scenarios (e.g., hand-held systems), however, multiple receivers are impractical. Thus, single channel techniques are of interest, particularly in mobile scenarios. Although the amount of existing research for single channel DF is considerably less than for multi-channel direction finding, single channel direction finding techniques have been previously investigated.When considering single channel direction finding systems, we find that there are two distinct types of DF systems. The first type of DF system is the amplitude-based DF system. Amplitude-based systems determine the bearing of the signal (or the AOA) by analyzing the amplitudes of the output voltages from each antenna element. Amplitude DF systems include the Watson-Watt technique using an Adcock antenna array .The second type of DF system is the phase-based DF system. Phase-based systems use three or more antenna elements that are configured in a way so that the relative phases of their output voltages are unique for every wavefront angle-of-arrival. Phase-based DF systems include the Pseudo-Doppler technique with a commutative switch based antenna array .Since both of the above techniques are primarily analog techniques and have been analyzed in previous work, we will investigate a new single channel direction finding technique that takes specific advantage of digital capabilities. Specifically, we propose a phase-based method that uses a bank of Phase-Locked Loops (PLLs) in combination with an eight-element circular array. Our method is similar to the Pseudo-Doppler method in that it samples antennas in a circular array using a commutative switch. In the proposed approach the sampled data is fed to a bank of PLLs which tracks the phase on each element. The parallel PLLs are implementedin software and their outputs are fed to a signal processing block that estimates the AOA.This thesis presents the details of the new algorithm and compares its performance to existing single channel DF techniques such as the Watson-Watt and the Pseudo-Doppler techniques. We also describe the implementation of the algorithm on a DRS Signal Solutions Incorporated (DRS-SS), WJ-8629A Software Definable Receiver with Sunrise . Technology and present measured performance results. Simulations on a signal with 10dB SNR have shown that the Watson-Watt algorithm and the Pseudo-Doppler algorithm have an accuracy that is worse than the proposed technique by approximately an order of magnitude.The algorithm was implemented on a single-channel DSP-based software radio with a homemade eight-element circular antenna array. The WJ-8629A software defined radio receiver was provided by DRS-SS in order to implement our algorithm. The implementation was tested using a CW signal at 1.57068 GHz in a low multipath laboratory environment and outdoors. The performance of the prototype is compared to the data provided by the simulations in Matlab.Implementation results focus on CW measurements in a relatively benign laboratory environment for proof-of-concept testing. This document will show that the basic version of the algorithm can result in a significant computational burden, thus we investigate a low-complexity approach and demonstrate its performance. It will be shown that a significant computational reduction can be achieved with minimal performance penalty.1.1 Software IntroductionDuring our research, all of the single-channel direction finding simulations were performed using the MATLAB 6.1 software. After the simulations were completed, the MATLAB code was then ported to hardware for implementation using the C programming language. The initial C programs were written and tested to prove that the algorithms could be implemented on the TI based software radio. After the C programs were tested and compared to their Matlab counterparts, they were then optimized for the Texas Instruments TMS320C67x Digital Signal Processor.1.2 Hardware Introduction1.2.1 DRS Signal Solutions, Incorporated WJ-8629A Software Definable Receiver with Sunrise. TechnologyThe implementation was performed on a Texas Instruments DSP-based WJ-8629A software defined radio provided by DRS-SS. It has a frequency range from 20 to 2700 MHz with 10-Hz resolution, receiver filtering with 22 filter slots (200 Hz to 1.23 MHz), and 5 reserved slots for user-downloadable custom filters The main processing unit is the Texas Instruments TMS320C6701 DSP processor with a maximum computational rate of nearly 1GFlops. The radio allows one to develop algorithms for certain signal processing modules in the C programming language or the TMS320C67x assembly language. Other details of the radio are not listed here due to their proprietary nature. Throughout this thesis we will include only those details necessary for proper understanding of the implementation.1.2.2 MPRG Antenna ArrayThe antenna baseline is the geometric line of interconnection between antenna elements. Antenna aperture is defined as the plane surface area near the antenna through which most of the radiation flows. The spacing between antenna elements usually determines the aperture of an array, and since we are using circular arrays, the diameter of the entire circular array determines the array aperture .In order to model the antenna array, assuming a single plane wave impinging on thearray, the array manifold vector for a uniform circular array can be written as:where R is the radius of the circular antenna array, is the elevation angle, is the angle of arrival (AOA) of the incoming plane wave, m is the angle of the mth antenna element in the azimuthal plane, and is the wavelength of the center frequency of interest. For simplicity, the elevation angle is set to 90o in order to consider azimuth angles only. We do not consider the effects of different elevations in this study.The MPRG antenna array as seen in Figure 1.1 is an eight-element antenna array with a diameter of 19.1 cm. We desire to have a waveform that completes one wavelength over the diameter of the array which will be discussed in detail in later chapters. Therefore, the frequency of the CW is defined as f = c/or 1.57068 GHz.Chapter2Introduction to Single Channel Direction FindingTo date, the two primary methods that have been examined for single channel direction finding are the Watson-Watt Method using an Adcock antenna array, and the Pseudo-Doppler Method using a commutative switch with a circular antenna array . While little is available in the open literature concerning these two techniques, what is available assumes an analog receiver and operates at relatively low frequencies. Specifically, the Adcock/Watson-Watt algorithm is typically used for frequencies up to about 1000 MHz, while the Pseudo-Doppler algorithm typically has an operational bandwidth from 2-2000 MHz. In this chapter, we will discuss the amplitude-based Watson-Watt technique, the phase-based Pseudo-Doppler method, and an amplitude-based Pseudo-Doppler technique developed as part of the current research. We will discuss their strengths and short-comings and motivate the investigation of new techniques.2.1The Watson-Watt MethodWatson-Watt DF is an amplitude-based method that uses the relative amplitude of the output of two antenna arrays arranged according to the Adcock design. The Adcock design consists of four antenna elements in a perpendicular, crossed-baseline configuration as seen in Figure 2.1.This method can be used for frequencies up to about 1000 MHz. One Adcock pair contains two antenna arrays (four antenna elements) in a perpendicular configuration, with element spacing of less than one half the wavelength at the highest operating frequency. The azimuth gain pattern from each antenna array is obtained by a vector difference of signals from each of two antennas.The signals seen on the four antennas in complex baseband notation are:where r(t) is the received signal, R is the radius of the circular antenna array, is the wavelength of the center frequency of interest, m(t) is a linearly modulated message signal and is the AOA6. The East antenna represents our 0o reference.The N and S antenna pair creates the Y-axis voltage, which has a maximum gain along the Y-axis. In other words when , the east and west signals are equal and thus x(t) =re(t)-rw(t) = 0, whereas y(t) = rn(t)-rs(t) = 2m(t). The E and W antenna pair creates the X-axis voltage, which has maximum gain along the X-axis. In other words when , the north and south signals are equal and thus x(t) = re(t)-rw(t) = 2m(t), whereas y(t) = rn(t)-rs(t) = 0.Figure 2.1 Adcock Antenna Array used for Watson-Watt AlgorithmIn order to pass the AOA data to the single receiver, each of the X and Y axis voltages have to be combined into a composite signal. In our example in Chapter 4, the two signals are linearly combined to form an AM signal with dual tone modulation in order to pass the data to the single receiver.After the linearly combined AM signal reaches the receiver and AM demodulation is performed, the estimated AOA ( ) is calculated by taking the arctangent of the N-S difference divided by the E-W difference. where the approximation holds for small values of , since for small values of x. If we use the antenna array described in Figure 2.1, we will encounter an 180o phase ambiguity since a negative ratio could correspond to either quadrant 2 or 4 whereas a positive ratio could correspond to either quadrant 1 or 3. If a centrally located omni-directional antenna is included in Figure 2.1, then it can provide basic directional sensing to help eliminate the 180o phase ambiguity . In Chapter 4, we will examine the accuracy of the Watson-Watt algorithm using an Adcock array in various conditions via simulation.2.2 Pseudo-Doppler AlgorithmThe Pseudo-Doppler technique is a phase comparison method that exploits the Doppler shift on successive samples of circularly disposed antenna elements. Measurements of phase differences between the elements of a multi-element antenna array enable the azimuth angle of the arriving signal to be determined. One system of this type is the Pseudo-Doppler method. In principle, an antenna element could be moved in a circular path so that the instantaneous frequency of the received signal would be modified.Alternatively, a rotating commutative RF switch is used to couple a receiver in rapid sequence to the elements of the array, thereby introducing a frequency shift on the received signal which is extracted by a frequency discriminator. As the antenna moves, it imposes a Doppler shift on the arriving signal. The magnitude of the Doppler shift is at a maximum as the antenna moves directly toward and away from the direction of the incoming wavefront. There is no apparent frequency shift when the antenna moves orthogonal to the wavefront 9. The azimuth angle is given by the angular position of the rotor at which zero instantaneous frequency shift occurs. Ambiguity can be removed by taking account of the angles at which maximum positive and negative frequency shifts occur.As in Figure 2.2, the value of r changes with the sampling position which results in a frequency shift of 0o when 1 is exactly coincident with the incoming signal azimuth angle with an 180o phase ambiguity. Therefore, near zero frequency shift occurs at angles () and (). The ambiguity can be resolved by finding the maximum negative frequency shift at () and the maximum positive frequency shift at ().Figure 2.2 Pseudo Doppler Frequency ShiftConsider a linearly modulated signal impinging on an Na-element circular arrayAssume that the receiver switches from the ith antenna to the (i + 1)th antenna every Ts seconds. Now each antenna imposes a phase shift ofwhere R is the radius of the circular array, is the wavelength of interest, is the angle-ofarrival and i = 0.Na 1. Now if the switch changes to the neighboring antenna every Ts seconds, it imposes a time varying phase shiftwhere u(t) is the unit step function. The received signal is thenIgnoring for the moment the message signal, the output of an FM discriminator isNow, since this is not a true differentiator, but a discrete approximation, there is a delay of Ts/2:Now, after down-conversion, we can ignore the carrier term. Thus, we haveThe samples for every Na values can be entered into a vector:Now, the FFT of this vector isIn the expression above, each sum will be zero for all values of k except k = 1. Further, for k = 1,Thus the estimated AOA is,2.3 Advantages and Disadvantages of Direction Finding Systems2.3.1 Watson-Watt2.3.1.1 AdvantagesWith the appearance of low-cost, wide frequency receivers, many manufacturers realized that .stand alone. DF bearing processors could be interfaced with the new low-cost receivers at minimal cost. A well designed Watson-Watt direction finding array can be interfaced with almost any receiver with good results 8. The Adcock antenna array.s diameter is small in size. Therefore, the array is beneficial in mobile and transportable DF applications. Since the DF antenna tone modulation technique is AM, FM listen-through capability is excellent due to the high AM rejection of most receiver FM limiter/discriminators. Listen- through capability is also good for AM signals as a result of the fact that the DF antenna modulation tone frequencies are well below the low end of the voice spectrum and can thus be easily attenuated in the audio output channel 8.The Adcock/Watson-Watt system is suited mainly for mobile applications especially if budgetary constraints dictate the use of low-cost receivers.2.3.1.2 Disadvantages The Adcock antenna array is inherently a narrow aperture. Since it is a narrow aperture, the DF resolution is affected. If a center antenna is not used, then the algorithm then suffers from an 180o phase ambiguity. A narrow aperture antenna is quite susceptible to multipath and reflection errors.The Adcock array requires balanced sum and difference hybrids, balanced modulators, phase-matched cables, and circuits for phase or gain imbalance. All of these components can escalate the cost of the array 7.The Adcock/Watson-Watt algorithm has a limitation on the maximum frequency. Due to the more complex electronics circuitry required by the Adcock antenna, it is not feasible to manufacture a wideband DF antenna capable of good and consistent performance at frequencies over 1000 MHz 8. The Adcock/Watson-Watt algorithm also does not provide elevation measurements, which have greater influence over the azimuth at higher frequencies.2.3.2 Pseudo-Doppler2.3.1.1 AdvantagesWhen compared to the Adcock/Watson-Watt DF system, the pseudo-Doppler systems have advantages in the areas of site error suppression, DF antenna economy, and extended high frequency performance. Due to the circular arrangement of the antenna elements, the array can be constructed as a wide aperture array. A wide aperture array can increase AOA resolution and reduce site errors, but the size of the array then becomes an issue as the number of antenna elements increases. Because spacing between antenna elements should be /2 or more, as the number of elements increases, the size of the array increases. If the size of the array becomes too large, the feasibility of mobility diminishes 8.The electronic circuitry required to implement a pseudo-Doppler system consists of GaAs FET high frequency RF switches, the necessary driver circuitry, and phase-matched cables 7. The simpler pseudo-Doppler DF antenna array is more easily and economically designed and manufactured when compared to the Adcock/Watson-Watt system. Please note that the economic impact of the cheaper pseudo-Doppler system only applies to narrow-aperture designs. As the aperture becomes larger and more antenna elements are added, the cost of the design will increase, but they are no more as expensive as the Adcock arrays 8.In contrast to the Adcock/Watson-Watt system, the Pseudo-Doppler algorithm should work at frequencies up to 2000 MHz and beyond due to the simplicity of the electronics associated with a pseudo-Doppler manufacturable wideband DF antenna. This allows for a greater range of applications such as cellular applications 8.2.3.2.2 DisadvantagesBecause the wide aperture pseudo-Doppler arrays can suppress site errors, the large circular arrays can limit mobility and covertness. In addition to the large arrays, the quality of the receiver needs to be more complex than the Adcock/Watson-Watt because the pseudo-Doppler receiver requires more sensitivity than an Adcock/Watson-Watt receiver, and it also needs to control the switching circuit which chooses the correct antenna element. Another disadvantage to pseudo-Doppler systems is that the listen-through capability is a problem. Because there is a desire to obtain an accurate AOA, a high commutative switching rate is needed. When there is a high commutative switching rate, the switching rate is placed in the audio range. FM voice audio is badly distorted due to the high switching rate because the commutative process creates FM modulation at the commutative rate. AM also suffers as a consequence of the soft-commutation switches 8.2.4 Motivation for a New Single Channel DF MethodAfter reviewing the above algorithms, we decided to try and develop an algorithm that is loosely based on the pseudo-Doppler system. The goal of our algorithm and system was to provide the AOA resolution of a pseudo-Doppler system or better while maintaining a small aperture mobile circular antenna array. Our algorithm will be implemented on a software-defined radio which will take advantage of a digital implementation. It should provide equal or better performance than the algorithms or methods described above. The DF system should allow for listen-through capability which should be equal or better than the above algorithms. And last, the algorithm should be different from existing algorithms.The Phase-locked loop (PLL) algorithm that is proposed in Chapter 3 is loosely based on the pseudo-Doppler system in that we will use a similar switched antenna array system. The advantage of the PLL algorithm is that we are able to maintain a small aperture array while increasing the AOA resolution. Therefore, the PLL algorithm is an accurate direction finding system with mobile capabilities and listen through capabilities.一種新型的分析與單通道尋找在軟件無線電通信平臺(tái)的實(shí)現(xiàn)算法的方向第1章一個(gè)無線電測(cè)向（DF）的天線陣列系統(tǒng)是在一個(gè)組合安排，以確定一個(gè)遙遠(yuǎn)的發(fā)射方位角一個(gè)接收器?；旧希械臏y(cè)向系統(tǒng)，推導(dǎo)出從該角的落地（AOA）的初步測(cè)定發(fā)射器的位置。無線電測(cè)向技術(shù)已經(jīng)是基于經(jīng)典的多天線系統(tǒng)采用多個(gè)接收器。如音樂1 2和ESPRIT經(jīng)典技術(shù)使用每個(gè)天線同步相位信息估計(jì)角的，有用信號(hào)的到來。在（例如，手持系統(tǒng)）很多情況下，但是，多個(gè)接收器是不切實(shí)際的。因此，單聲道技術(shù)的興趣，特別是在移動(dòng)的情況。雖然現(xiàn)有的研究單通道東風(fēng)金額大大低于多通道測(cè)向少，單通道測(cè)向技術(shù)此前已進(jìn)行調(diào)查。當(dāng)考慮單通道測(cè)向系統(tǒng)，我們發(fā)現(xiàn)有兩個(gè)不同類型的DF系統(tǒng)。對(duì)DF系統(tǒng)的第一類是振幅的東風(fēng)系統(tǒng)。振幅為基礎(chǔ)的系統(tǒng)，通過分析確定每個(gè)天線單元的輸出電壓的幅值的信號(hào)（或AOA）的影響。振幅東風(fēng)系統(tǒng)包括屈臣氏瓦技術(shù)，使用一阿德科克天線陣列。對(duì)DF系統(tǒng)第二類是逐步的東風(fēng)系統(tǒng)。第一階段為基礎(chǔ)的系統(tǒng)使用三個(gè)或更多的天線，是要素配置的方式，使它們的輸出電壓的相對(duì)相位是每個(gè)波前角的，獨(dú)特的到來。相的東風(fēng)系統(tǒng)包括天線陣列與基于交換開關(guān)偽多普勒技術(shù)。由于上述兩種技術(shù)主要是模擬技術(shù)，并已在以往的工作分析，我們將探討新的單通道測(cè)向技術(shù)，它利用數(shù)字功能的特定優(yōu)勢(shì)。具體來說，我們提出了一個(gè)階段為基礎(chǔ)的方法，它使用一八元的鎖相環(huán)（PLL）的合并圓陣。我們的方法是類似的偽多普勒方法，在一個(gè)圓陣天線進(jìn)行采樣用一個(gè)交換開關(guān)。在新方法中的采樣數(shù)據(jù)送入一個(gè)鎖相環(huán)儲(chǔ)備，跟蹤每個(gè)元素的階段。實(shí)施平行鎖相環(huán)在軟件和它們的輸出輸入到信號(hào)處理