Rareș Buta - Technical University of Cluj-Napoca, Radiocommunications Research Group, Communications Department
Cristian Codău - Technical University of Cluj-Napoca, Radiocommunications Research Group, Communications Department
Andra Păstrăv - Technical University of Cluj-Napoca, Radiocommunications Research Group, Communications Department
Tudor Palade - Technical University of Cluj-Napoca, Radiocommunications Research Group, Communications Department
Paul Dolea - Technical University of Cluj-Napoca, Radiocommunications Research Group, Communications Department
Emanuel Pușchiță - Technical University of Cluj-Napoca, Radiocommunications Research Group, Communications Department
Importance and Applicability
The deployment of smart antennas is a leading technology that has emerged in wireless communications. Digital signal processing capability, along with the antennas, make systems smart. Wireless communication performance of the antenna array improves, and thus the capacity of the network, by using spatial processing techniques such as beamforming. The principle of beamforming is essentially to weight the signals to be transmitted in such a way that the transmitter exhibits a constructive superposition in a predetermined direction. The system forms a radiation pattern in an adaptive manner, depending on certain variables of the control algorithm. However, we can immediately infer the importance in determining the angle of arrival (AoA) of a target, as it is an essential parameter for the later beamforming algorithm. We need to know the direction of the impinging signal for steering the radiation pattern of an antenna array in the direction given by the target user. AoA estimation in array signal processing is an emerging research area. As noted in the journal article Performance Analysis of MUSIC, Root-MUSIC, and ESPRIT DOA Estimation Algorithm, its effectiveness greatly determines the performance of smart antennas. The scope of this work was to deploy a feasible solution that could indicate the direction of an incoming target signal expressed by its AoA. Namely, we implemented and integrated two AoA algorithms (MUSIC and ESPRIT) in using the LabVIEW Communications System Design Suite and tested their performance on a platform constructed of NI devices.
MUSIC and ESPRIT Algorithms
The work reported in this case study proposed two AoA algorithms, MUSIC and ESPRIT, for integration and evaluation. MUSIC is a high-resolution technique based on exploiting the eigenstructure of the input covariance matrix that is built by grouping the samples captured by each antenna in a 2D matrix. It is a subspace-based algorithm, which exploits the orthogonality between the signal subspace and the noise subspace. ESPRIT is a subspace-based AoA estimation algorithm. It does not involve an exhaustive search through all possible steering vectors to estimate AoA, as the MUSIC algorithm does. It dramatically reduces the computational and storage requirements compared to MUSIC. As explained in an article titled Application of the Total Least Square ESPRIT Method to Estimation of Angular Coordinates of Moving Objects, the goal of the ESPRIT technique is to exploit the rotational invariance in the signal subspace, which is created by two arrays with a translational invariance structure .
The laboratory setup comprises the following subsystems: (1) the transmitting subsystem (one USRP-2954), (2) the uniform linear antenna array subsystem (4–8 array elements), (3) the synchronization subsystem (one CDA-2990 and one USRP-N2900), (4) the receiving subsystem (four USRP-2954R), and (5) the data processing unit subsystem (one PXIe-8880 host controller and one CPS-8910 switch). Figure 1 shows the test platform and subsystem interconnection.
The target signal is emitted through a VERT2450 antenna connected through a sub-miniature version A (SMA) cable to another USRP-2954R. The target USRP (Universal Software Radio Peripheral) also connects to the CPS-8910 switch through which it communicates with the host computer.
Uniform Linear Antenna Array Subsystem
The antenna array comprises eight VERT2450 omnidirectional antennas. They are dual-band antennas working between 2.4 and 2.48 GHz, and 4.9 and 5.9 GHz, respectively at 3 dBi gain. The antennas are linearly disposed, which forms a ULA. The distance between the ULA elements is λ/2 (6.20 cm), where λ is the wavelength of the signal (at 2.415 GHz) that is being tracked.
The eight antennas connect to four USRP-2954R software defined radios (SDRs) through coaxial cables with an SMA connector. We synchronize the USRPs with the help of a CDA-2990 OctoClock as described below. We configure the USRPs by software to capture a wave of 2.415 GHz and to extract the modulating signal by digitizing it and sending it to the host computer. The TX/RX port of each channel belonging to each USRP is employed for the reception of both the target and reference signals. The ports connect to the antennas by SMA cables.
A phased antenna array system needs synchronization in time, frequency, and phase. This block comprises one CDA-2990 OctoClock and one USRP-2900. The CDA-2990 OctoClock distributed by NI makes it possible to achieve the synchronization in time and frequency. All USRPs from the receiving site connect to the external clock (OctoClock) to receive the reference signal and the pulse per second (PPS) signal. The OctoClock provides a 10 MHz clock output and a logic-level pulse of 1 s period and typically 20 percent duty cycle as PPS signal for timestamp synchronization. Moreover, the four SDRs are further synchronized in terms of removing the random phase offset induced by the VCO/PLL chains that appear at the tuning (initialization) of the RF front ends by delivering a reference signal with a USRP-2900. The reference signal is also delivered over the air, filtered, and processed to compensate for the initial phase ambiguity that appears at the initial tuning of the USRP boards. The idea of initial phase synchronization among the USRP boards and its implementation is derived from the case study on direction finding and beamforming written by the Communications and Signal Processing Group from Imperial College of London .
Data Processing Unit Subsystem
For a host computer, we used the PXIe-1082 chassis with a PXIe-8880 eight-core embedded controller, which includes an Intel Xeon CPU at 2.3 GHz and 24 GB of installed RAM memory and Windows 7 Professional 64-bit OS. The system is designed for a large range of test and measurement applications, while being compact and lightweight. It currently includes two mounted PXI Express modules and a FlexRIO module. We chose this controller because of its high capabilities and the three slots where devices can be connected. Figure 2 shows the experimental setup of the target signal direction system deployed at the Technical University of Cluj-Napoca.
A detailed configuration procedure sets up the direction-finding system:
- One antenna connects to the USRP-2900 that is going to serve as the reference signal transmitter.
- The reference signal transmitter connects to the host computer through a USB cable.
- One antenna connects to each TX/RX port of each USRP-2954R from the reception block using an SMA cable.
- Another USRP-2954R with an antenna attached provides the target signal.
- The eight receiving antennas of the ULA are placed at approximately half a wavelength from each other, according to the carrier frequency, typically 2.415 GHz.
- All USRPs from the receiving site connect to the external clock (OctoClock) to receive the reference signal and the PPS signal.
- The four receiving USRPs and the USRP that provides the target signal connect to a switch through MXI cables.
- Further, the switch connects to the host controller through an MXI cable.
- All six USRPs, host controller, switch, and OctoClock finally power up to complete the hardware configuration procedure.
Antennas capture the impinging wave and deliver it to the receiving subsystem, which downconverts, samples, and digitizes the wave signal. Then, through the switch, the bit streams are multiplexed and transmitted to the processing unit where the algorithms are run. Based on their results, we can estimate the direction of the target source. The CDA-2990 OctoClock performs synchronization in frequency and time. It synchronizes the LO frequency of the USRP boards with an REF signal. It also synchronizes the sampling clock of the boards by means of the PPS signal delivered by the same OctoClock device. Another issue to overcome is the initial phase offset between the channels, which appears at each reconfiguration of the USRPs. Our solution is to use an external RF synchronization signal, generated by an external signal source (in our case NI USRP 2900). This signal is delivered to each RX2 port of each USRP device and to estimate the phase shift and later for phase correction; thus, obtaining a proper calibration of the entire system. We separate the signal used for phase synchronization from the target signal by using a digital filter.
Testing and Validation
The evaluation scenarios of the implemented AoA algorithms consider 35 positions of the target transmitter at different angles (for example, ∡0°, ∡±15°, ∡±30°, and ∡±45°) and distances (0.5 m, 1 m, 1.5 m, 2 m, and 2.5 m) with respect to the reference antenna array element. We constructed six measurement scenarios to vary the number of antennas from 4 to 6 and 8, and to test both MUSIC and ESPRIT algorithms. We determined the impact of the source positioning and the size of the antenna array along with insight on the performance of the AoA algorithms at the end of the experiments. The purpose of this work was not to certify the accuracy of one algorithm over the other, but to test their performance individually. Quantitatively, they proved to determine in less than 5° in error the direction of the target. Figure 3 shows a schematic picture of the measurement setup.
Conclusions and NI Product Benefits
We found the NI platform that we worked on to be the most suitable for the hardware implementation of our work. Also, the LabVIEW Communications System Design Suite is a useful tool for developing software to control the USRP transceivers and to compile algorithms directly on the Xilinx Kintex-7 FPGA of the NI USRP. We used five USRP boards as wireless transceivers, one PXI as signal processing unit, a CDA-2990 OctoClock, and a CPS-8910 switch as synchronization and interconnection devices, to design a system capable of indicating the direction of an incoming signal with an error of maximum 5°. This accuracy, which we agree is high enough, helps determine in which direction we should further steer the beam of a linear antenna array to ensure an enhanced quality communication to a previously identified target user. In the future, we intend to test the setup for a circular antenna array configuration.
- N. Waweru, D. Konditi, P. Langat, Performance Analysis of MUSIC, Root-MUSIC and ESPRIT DOA Estimation Algorithm, World Academy of Science, Engineering and Technology, International Science Index 85, International Journal of Electrical, Computer, Energetic, Electronic and Communication Engineering, Volume 8(1), pp. 209 – 216, 2014.
- W. Rosloniec, Application of the Total Least Square ESPRIT Method to Estimation of Angular Coordinates of Moving Objects, International Journal of Antennas and Propagation, Volume 2010, Article ID 548953, 9 pages, http://dx.doi.org/10.1155/2010/548953.
- M. Willerton, Angle of Arrival Detection with NI USRP and LabVIEW Communications, Imperial College London, Communications and Signal Processing Group, [Online]. Available: http://sine.ni.com/cs/app/doc/p/id/cs-15016# . [Accessed: February 24, 2018].
Technical University of Cluj-Napoca, Radiocommunications Research Group, Communications Department
No. 26-28 George Barițiu St.
Tel: +40 744 760 356