D6.1 Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial spectrum access

Transcription

1 Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial spectrum access Grant Agreement nº: Project Acronym: SANSA Project Title: Shared Access Terrestrial-Satellite Backhaul Network enabled by Smart Antennas Contractual delivery date: 31/01/2018 Actual delivery date: 1/02/2018 Contributing WP WP6 Dissemination level: Public Editors: FRA Contributors: FRA, VIA, OTE Abstract: demonstrates performance of the hybrid (analog-digital) antenna array that was manufactured within SANSA project for the small scale demonstration of the spectrum sharing. Firstly, the deliverable discusses the importance of the antenna array calibration and challenges related to it. It is shown that the conventional calibration approaches are too complex and very challenging. An alternative feedback based calibration using iterative particle swarm optimization developed within WP3 was then verified by measurements. Exploiting this calibration, the antenna array could properly steer one beam for the terrestrial link communication and one null to protect the satellite link operating at the same frequency. The spectrum sharing was successfully proven by using a SW tool suitable for throughput measurements and by using a video server that streamed video over a satellite link only when it was protected by the null steering of the hybrid antenna array.

2 Document History Version Date Editor Modification V0.0 01/07/2017 FRA Initial ToC V0.1 06/8/2017 FRA Description of the envisaged demonstration setup V0.2 16/9/2017 VIA Antenna array description V0.3 1/11/2017 FRA Description of the antenna calibration using conventional calibration V0.4 1/12/2017 FRA Description of PSO calibration V0.5 12/01/2017 FRA Pre-final version with description of the beam an null steering in shared access V1.0 29/01/2017 FRA Final version afterqa Contributors Name Company Contributions included Rudolf Zetik FRA The whole document Przemyslaw Gorski VIA Chapter 2, Quality assurance Markus Landmann FRA Chapter 3, 4 Pragadeeshwaran Pandianrajan FRA Chapter 3, 4 Xavier Artiga CTTC Quality assurance : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 2 of 32

3 Table of Contents List of Figures... 4 List of Acronyms... 6 Executive Summary Introduction Hybrid (analog-digital) antenna array prototype Antenna array performance and its calibration Performance of the uncalibrated antenna array Antenna array calibration Conventional calibration of the analog antenna array Power far field calibration of the analog antenna array using iterative optimization Performance of the calibrated hybrid antenna array Over-the-air demonstration of spectrum sharing Conclusions References : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 3 of 32

4 List of Figures Figure 1 Prototype of the mmwave antenna array with 64 elements arranged in 2 side-by-side subarrays... 8 Figure 2 Measured beam patterns of the antenna array prototype steering one beam to 0 by means of analog only beamforming (blue) and forming one beam at 0 and one null at 8 by means of hybrid beamforming (red)... 8 Figure 3 Structure of the analog part of the antenna array Figure 4 Combination of the analog antenna array with the baseband precoder Figure 5 Measured beam pattern of the uncalibrated antenna array steering one beam to 0 by means of analog only beamforming Figure 6 Measured beam pattern of the uncalibrated antenna array steering one beam to 45 by means of analog only beamforming Figure 7 Measured beam pattern of the uncalibrated antenna array steering one beam to 0 and one null to 8 by means of analog only beamforming Figure 8 Setup used for calibration of the antenna array in the anechoic chamber Figure 9 Measured pattern of a single antenna element Figure 10 Beam pattern of the uncalibrated antenna array steering one beam to 0 measured in the nearfield Figure 11 Coherent summation of individual 2D pattern and the 2D pattern of one subarray.. 15 Figure 12 Antenna array leakage (black) and its comparison to the first antenna element aperture with (solid red line) and without leakage (dashed red line) Figure 13 Antenna array leakage (black) and its comparison to the second antenna element aperture with (solid red line) and without leakage (dashed red line) Figure 14 Antenna coupling estimated from measured date with the leakage Figure 15 Antenna coupling estimated from measured date without the leakage Figure 16 Broadband measurement of the antenna array leakage (blue) at the boresight and its comparison to the gain of the first antenna element (red) and to the gain of the whole antenna subarray (yellow) Figure 17 Setup used for the far field calibration of the antenna array based on a received power only Figure 18 Measured beam patterns of the analog antenna array steering one beam to 0, blue before calibration, red after calibration Figure 19 Measured beam patterns of the analog antenna array steering one beam to 45, blue before calibration, red after calibration Figure 20 Measured beam patterns of the hybrid antenna array which steers one beam to 0 (blue line, no null steering) and one null to 8 after the baseband calibration (red) and a comparison to the calibrated null steering at -5 (green) Figure 21 Upper view of the demonstration setup Figure 22 Side view of the demonstration setup Figure 23 Block diagram of the terrestrial link : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 4 of 32

5 Figure 24 Block diagram of the satellite link Figure 25 Vision of the demonstration with hybrid antenna array properly steering the beam towards the terrestrial antenna and forming a null towards the VSAT antenna Figure 26 Antenna constellation and the setup for null steering calibration Figure 27 Measured pattern of the emulated directional antenna pointed towards the terrestrial antenna at 0 with the satellite receiver at Figure 28 Measured pattern of the hybrid antenna array forming a beam towards the terrestrial antenna at 0 and a null towards the satellite receiver at Figure 29 Spectrum of the terrestrial interference (using antenna array without null steering) received at the sat. antenna Figure 30 Spectrum of the terrestrial interference (using antenna array with null steering) received at the sat. antenna Figure 29 Spectrum of the signal received at the sat. antenna which is a composition of the transmitted satellite signal and the interference from the terrestrial link Figure 30 Spectrum of the signal received at the sat. antenna which is a composition of the transmitted satellite signal and the interference from the terrestrial link mitigated by the hybrid antenna array Figure 31 GUI of the satellite modem at the VSAT side for unmitigated interference Figure 32 GUI of the satellite modem at the VSAT side for mitigated interference Figure 33 GUI of the JPERF : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 5 of 32

6 List of Acronyms ABF ACU BF BFN BPF CDF COTS CSI DBF ENOB HPA IF LNA LO LS LSB MMIC MSB PCB PS PSO RF RMS Rx SINR SUCA SUPA T/R TTDU Tx UCA ULA UPA VSAT XPD Analog Beamforming Antenna Control Unit Beam Forming Beam Forming Network Band Pass Filter Cumulative Distribution Function Commercial Off The Shelf Channel State Information Digital Beamforming Effective Number Of Bits High Power Amplifier Intermediate Frequency Low Noise Amplifier Local Oscillator Least Square Least Significant Bit Monolithic Microwave Integrated Circuit Most Significant Bit Printed Circuit Board Phase Shifter Particle swarm optimization Radio Frequency Root Mean Square Receive Signal over Interference plus Noise Ratio Stacked Uniform Circular Array Stacked Uniform Planar Array Transmit/Receive True-Time Delay Unit Transmit Uniform Circular Array Uniform Linear Array Uniform Planar Array Very small aperture terminal Cross Polarization Discrimination : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 6 of 32

7 Executive Summary The main goal of this deliverable is to demonstrate interference mitigation using a multiantenna hybrid (analog-digital) beamforming antenna array, to evaluate its performance and performance of selected beamforming algorithms. Such an antenna array is a prerequisite for the microwave backhaul radio links in a scenario where the terrestrial and the satellite communication sector share the same frequency band. The role of such an antenna array is to steer beams among communicating terrestrial backhaul nodes and to steer nulls towards other terrestrial nodes and satellite receivers in order to protect them from the interference. The deliverable describes: Performance evaluation of the key enabling component for interference mitigation the prototype mmwave beamforming antenna operating at 19.5GHz (output from WP5) Evaluation of a selected beamforming algorithm (output from WP3) o hybrid beam and null steering to predefined directions which was described in D3.1 Chapter Evaluation of antenna calibration algorithms o Based on conventional calibration o Based on Particle swarm optimization (PSO) which was described in D3.1 Chapter 5.1 Over-the-air demonstration of a shared access terrestrial-satellite communication in realistic but well controlled scenario. The antenna array which was designed and manufactured within SANSA as a proof of concept operates in the Ka band (~18-20 GHz). It consists of 64 transmitting antenna elements that are arranged in two side-by-side linear subarrays. Each single antenna element is connected to a phase shifter and an attenuator that are controlled by 8 bits for analog beamforming. Two RF chains that drive the two subarrays were connected to a channel emulator that executed the digital base band precoding. In this constellation, the analog antenna array and the digital channel emulator were capable to perform hybrid (analog-digital) beamforming that is required especially for null steering. Figure 1 shows the antenna array mounted over a motion emulator in an anechoic chamber. The antenna array faced a nearby antenna tower from the anechoic chamber through an EM transparent window. A horn antenna was mounted on the tower for receiving the Ka band signal that was transmitted by the antenna array. Together with the motion emulator (antenna positioner), which enables antenna rotations along 3-axis, beamforming capabilities and the spectrum sharing were demonstrated and evaluated. Figure 2 illustrates beamforming capabilities of the hybrid antenna array. The blue line shows the measured beam pattern of the analog only antenna array. The antenna array steered one beam to the antenna array boresight. Null steering using only analog part was impossible due to the quantization of the beamforming weights with 3 bits in their amplitude and 5 bits in their phase. Such a quantization of the beam forming weights allowed beam steering but prohibited the null steering. The red line in Figure 2 shows measured beam pattern of the : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 7 of 32

8 array obtained by the hybrid beamforming. Even with only two RF chains, i.e. only two complex beamforming weights, the hybrid beamformer could properly form the beam at 0 and a deep null at 8. The decrease of the beam power is due to the normalization that set the maximum power of the digital weights to be equal to one which is related to a maximum power that can input one RF chain. In the case of analog only beamforming, the magnitude of digital weights was equal to one. Thus, the array was working at its maximum power. In the case of the hybrid beamforming, the magnitude of the digital weights usually differs. Thus, if the BF weight with the maximum magnitude is normalized to one. The second weight has a magnitude that is smaller than one which results in reduction of the power of the related subarray and the whole antenna array as well. Even the total power of the array was smaller in the case of the hybrid beamforming; the comparison is realistic and reflects the fact that beamforming weights cannot be set above a predefined maximum level. Figure 1 Prototype of the mmwave antenna array with 64 elements arranged in 2 side-by-side subarrays Azimuth [ ] Figure 2 Measured beam patterns of the antenna array prototype steering one beam to 0 by means of analog only beamforming (blue) and forming one beam at 0 and one null at 8 by means of hybrid beamforming (red) The hybrid beamforming was the basis for the spectrum sharing demonstration with one terrestrial and one satellite link. The forward terrestrial link and the downlink of the satellite link were emulated over the air between the laboratory building and the nearby antenna tower (@ ~100m distance) using the same frequency band at 19.5GHz carrier. The antenna tower carried one receiving horn antenna which was dedicated to the terrestrial link. A satellite dish antenna was also mounted at the tower. It was emulating the downlink VSAT antenna that is interfered by signals transmitted from the prototype antenna array in the laboratory building. The angular separation of the terrestrial forward and the satellite downlink was only about 1.5. Despite such a small link separation the antenna array could steer one beam towards the horn antenna and one null towards the satellite dish antenna by means of the hybrid beamforming. The main beam was used to establish the terrestrial link. The null successfully protected the satellite link which could not be established if the antenna array emulated an antenna without null steering capabilities. Received power [dbm] : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 8 of 32

9 1 Introduction High quality terrestrial communication backhaul links at Ka band frequencies require large antenna aperture antennas at both sides of the link. Since SANSA project aims at selfreconfigurable hybrid terrestrial-satellite backhaul networks, antenna arrays with many antenna elements are required. Large antenna arrays provide high array gains to combat the path loss. Furthermore, in order to increase the spectral efficiency, each base station needs to simultaneously communicate data streams to multiple base stations. Multiplexing different data streams requires appropriate precoding techniques to control interference among these base stations. In [1] and [2] we have analyzed how many antenna elements and which antenna structures are required to fulfil requirements of self-organized hybrid terrestrial-satellite backhaul networks at Ka-band frequencies that are capable of reconfiguring the terrestrial topology and jointly exploiting the terrestrial and satellite links. In such networks, large antenna arrays and appropriate beamforming (BF) techniques are mandatory to overcome the high path loss and to mitigate interferences concurrently, so that the terrestrial as well as satellite receivers are protected from mutual interferences and can operate jointly in one system. In [1] and [2], it is indicated that a realistic antenna array shall have a 3D structure (e.g. a stacked uniform circular array) and the number of elements is at least 1380 for an array gain of 31 dbi and for 42 dbi. This minimum numbers of antenna elements seems to be extremely high, but it is needed inevitably to overcome path loss in realistic network topologies [3]. Due to the large dimensionality of antenna arrays a full digital implementation of such BF array at mmwaves becomes very challenging. A full analog beamforming could be a cost effective solution, but it has high losses in large beamforming networks, that need to be overcome by placing amplifiers, which affect the total power consumption. Its performance is constrained by quantization of beamforming weights, which is given by digital control of the phase shifters and attenuators. Moreover, the constant amplitude assumption is usual in full analog BF which further degrades its BF performance. A possible solution is mixture of the digital and analog BF. Solutions for hybrid digital-analog BF were proposed e.g. in [4]-[8]. One part of the BF operations is performed in the analog domain and the other part in the digital baseband domain. In this way, the number of required RF chains is significantly reduced, which decreases costs and power consumption of such antenna array, whereas the BF performance still remains sufficiently good for the intended applications. The aim of this deliverable is to demonstrate interference mitigation using a multi-antenna hybrid (analog-digital) beamforming antenna array developed in WP5 and its capabilities to steer beams among communicating terrestrial backhaul nodes and to steer nulls towards other terrestrial or satellite receivers in order to protect them from the interference by means of selected BF algorithms developed in Wp3. : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 9 of 32

10 The structure of the deliverable is as follows. The second chapter shortly introduces the prototype of the hybrid (analog-digital) antenna array which was designed and manufactured in WP5. The third chapter firstly indicates performance of an uncalibrated antenna array. Then, it describes calibration of such an antenna array. It shows challenges and possible solutions using conventional calibration procedures and also an alternative approach which is based on particle swarm optimization. At the end of the chapter the performance of the calibrated antenna array is illustrated by measured beam patterns. The fourth chapter describes setup which was used to demonstrate spectrum sharing with one terrestrial and one satellite link. The fifth chapter concludes obtained results. 2 Hybrid (analog-digital) antenna array prototype The antenna array which was designed and manufactured within SANSA as a proof of concept operates in the Ka band (~18-20 GHz) [9]. It consists of 64 transmitting antenna elements that are arranged in two side-by-side linear subarrays. Each single antenna element is connected to a phase shifter and an attenuator that are controlled by 8 bits for analog beamforming. Figure 3 shows the internal structure of the analog part of the antenna array. ATT1 ATTn ATTn+1 ATTna PS1 PSn PSn+1 PSna LO Up1 Up2 Figure 3 Structure of the analog part of the antenna array Its two RF chains that drive the two subarrays were connected to a channel emulator that executed the digital base band precoding. In this constellation, the analog antenna array and the digital channel emulator were capable to perform hybrid (analog-digital) beamforming that is required especially for null steering. Figure 4 illustrates this philosophy of the hybrid : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 10 of 32

11 beamforming for general case of Nrf number of RF chains. In SANSA, we realized a hybrid beamformer using the two RF chains connected to the subarrays in a side-by-side grouping. Digital Analog RF 1 1 Baseband Precoder ww [NN AA x 1] HHHH RF Precoder WW [NN AA x NN RRRR ] HHHH 2 RF N RF N A Figure 4 Combination of the analog antenna array with the baseband precoder 3 Antenna array performance and its calibration 3.1 Performance of the uncalibrated antenna array Firstly, we assessed the performance of the uncalibrated antenna array. We have performed beam pattern measurements using theoretical values for beamforming (BF) weights. The theoretical BF weights were related to a uniform antenna array which consists of 64 omnidirectional antenna elements, operated at 19.5GHz and the antenna spacing matched the manufactured antenna array. The beam pattern measurements were performed in the far field. Fig. 1 shows the antenna array mounted at the motion emulator in an anechoic chamber. The antenna array faced a nearby antenna tower from the anechoic chamber through an EM transparent window. A horn antenna was mounted on that tower for receiving the Ka band signal that was transmitted by the antenna array. Together with the motion emulator (antenna positioner), which enabled antenna rotations along 3-axis, beamforming capabilities and the spectrum sharing could be demonstrated and evaluated. Figure 5 and Figure 6 illustrate BF capabilities of the uncalibrated antenna array which used theoretical values of BF weights for steering of one beam to a predefined direction. Figure 5 shows the measured beam pattern for steering the beam towards 0 and Figure 6 towards 45. It is obvious that the forming of the beam is possible also with the uncalibrated antenna array. However the null steering capabilities are strongly limited. Null steering using only analog : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 11 of 32

12 beamforming with theoretical values is difficult due to the quantization of the beamforming weights with 3 bits in their amplitude and 5 bits in their phase. This quantization of the beam forming weights allowed beam steering but prohibited the null steering. An example is given in Figure 7. The beam was formed at 0, however; the null that was steered using quantized theoretical BF weights is very shallow. A possible solution for the analog null steering using quantized BF weights is to use an optimization algorithm. However due to the large dimensionality of antenna arrays at mmwaves this alternative solution can be computationally complex and time consuming. A solution to avoid such optimization is to properly calibrate the antenna array, which is discussed in the next chapter Received powe [dbm] Received powe [dbm] Azimuth [ ] Figure 5 Measured beam pattern of the uncalibrated antenna array steering one beam to 0 by means of analog only beamforming Azimuth [ ] Figure 6 Measured beam pattern of the uncalibrated antenna array steering one beam to 45 by means of analog only beamforming Received powe [dbm] X: 7.99 Y: Azimuth [ ] Figure 7 Measured beam pattern of the uncalibrated antenna array steering one beam to 0 and one null to 8 by means of analog only beamforming : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 12 of 32

13 3.2 Antenna array calibration Antenna patterns of single antenna elements and the geometrical structure of an antenna array influence the design of BF weights and the performance of the BF antenna array. Theoretical analyses of BF techniques usually assume arrays that contain isotropic antenna elements and ideal BF and feeding networks. However, imperfections of realistic antenna array components provoke deterministic and random errors that degrade performance of an antenna array. These errors reflect geometrical and electrical uncertainties of the antenna array. Geometrical uncertainties arise due to the imprecise locations of the array elements. Electrical uncertainties are caused by imperfections in electronics of the array. In [1], we have already theoretically analyzed such effects on the beam and null steering. In what follows, we present our experiences gained with the antenna array prototype. We describe two calibration approaches. The first one is the conventional approach in a near field of a small anechoic chamber by means of a network analyzer. The second approach is a novel method proposed in [1] which is based on PSO. This calibration is performed in a far field and exploits only measurements done by a power detector (no phases) Conventional calibration of the analog antenna array The setup for the conventional calibration is illustrated in Figure 8. The antenna array that is under test is mounted on a positioner. This is used to rotate the array in its azimuth and elevation. The reference antenna is mounted on another positioner which rotates it in only one axis changing so its polarization. This allows full 3D polarimetric antenna measurements. Figure 8 Setup used for calibration of the antenna array in the anechoic chamber Since the transmitting antenna array contains up convertors that are driven by its own independent local oscillator a more complex network analyzer using more than 2 ports is : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 13 of 32

14 required. Since this was impossible to measure with our network analyzer one requirement that was imposed on the antenna array prototype was to make available measurement points behind the up conversion. This is illustrated also in Figure 8. The network analyzer is connected to the antenna array behind the up convertor. The aim of the conventional calibration was to measure frequency response of each beamforming chain separately. Since the analog antenna array does not offer access to each of these chains as it is in the case of full digital antenna arrays, another requirement of the antenna array design was to include mutes in each beamforming chain. Thus, each BF chain can be switched on and off separately. This allowed separate measurements of all attenuators and phase shifters in their all states. It means we have measured 64 times (64 BF chains) 2 5 states of the phase shifters (controlled by 5 bits) and 2 3 states of the attenuators (controlled by 3 bits). On top of these measurements, we have performed 64 2D cuts of antenna patterns of each single antenna element. Figure 9 shows an example of the antenna pattern of the first antenna element, which was measured at 19.5GHz. In order to obtain more information about the antenna array we have performed broadband antenna measurements for frequencies from 16.5GHz to 20GHz. The goal of these calibration measurements was to obtain enough information about the realistic steering vector of the antenna array that includes all the geometrical and electrical imperfections. Only if the realistic steering vector of the antenna array is known the beamforming algorithms can be implemented properly and the beam pattern of the uncalibrated antenna array illustrated in Figure 10 can be improved. Figure 9 Measured pattern of a single antenna element Figure 10 Beam pattern of the uncalibrated antenna array steering one beam to 0 measured in the nearfield In order to verify calibration measurements the following test was performed. 32 single BF chain measurements performed at a particular angle for phase shifters adjusted to 0 and attenuators set to 0dB were coherently summed together. This emulates measurement of a beam pattern in which the subarray (BF chains 1-32) is used to steer one beam into the boresight direction using theoretical BF weights (zero phases, maximum amplitudes). Theoretically, such emulation shall correspond to a measurement of such beam pattern. : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 14 of 32

15 However, Figure 11 shows that the emulation (blue line) is completely different to the beam pattern measurement (red line). There is no beam formed by the coherent summation of separate single chain measurements. Figure 11 Coherent summation of individual 2D pattern and the 2D pattern of one subarray We have analyzed measured data and came to the conclusion that the reason for this behavior is due to: the radiation leakage and antenna coupling. The ration leakage represents EM waves that are leaking from the antenna array when the BF chains are turned off. Theoretically, no signal should be radiated from the antenna array. However, in what follows we illustrate that the undesired leaking EM waves are sometimes stronger than desired EM waves transmitted by a single antenna element. This is probably the main reason why the coherent summation of separate single element measurements did not correspond to the beam forming measurement of the whole array. In the emulation the leakage is 32 (or 64) times added together. However, in the beamforming measurement it appears only one time. The second reason is the antenna coupling which is different when a particular BF chain is turned on and off. It means that radiation characteristics are different when the beam pattern measurement of the whole (sub-)array were performed with all BF chains turned on in comparison to measurements of separate single antenna elements when only the corresponding BF chain was turned on and the remaining components were turned off. Figure 12 demonstrates the leakage problem in the aperture domain, which was obtained by Fourier Transform of measured patterns along their azimuth axis. The black line represents the : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 15 of 32

16 leakage of the array when the first subarray was turned on and the second subarray was turned off. In comparison, the single element radiation of the first antenna element (the solid red line) is in the order of the leakage. Using the whole set of the calibration measurements we have estimated the leakage and coherently subtracted it from single element measurements. The result of this correction is shown in Figure 12 by the dashed red line and shows the radiation that is concentrated at the position of the first element. -80 Element 1 Panel 1 ON, Other Panel OFF Magnitude [db] Element aperture illumination -150 Element aperture illumination w/o Leackage Leackage Position [cm] Figure 12 Antenna array leakage (black) and its comparison to the first antenna element aperture with (solid red line) and without leakage (dashed red line) This procedure was used to verify proper functionality of other antenna elements. In this way we could identify faulty antenna elements in the array. Figure 13 is an example of a faulty antenna element. The second antenna element radiated less significantly power than the first element. The radiation was somehow distributed over three neighboring antenna elements. Figure 14 and Figure 15 show the antenna coupling matrix before and after the leakage correction. A coupling matrix shows how the neighboring elements are influenced (coupled) with the corresponding element. In the ideal case, when the antenna elements are not coupled, the coupling matrix is a diagonal matrix. If neighboring elements are coupled the diagonal is broadened as illustrated in Figure 14 and Figure 15. Both figures also indicate that the coupling effect is evoked already at the PCBs of the subarrays and not only by the antenna elements themselves. This effect is represented in the figures by division of the coupling matrix into four sectors (rectangles). Each sector is evoked when the corresponding subarray is stimulated by a signal from the corresponding RF chain. : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 16 of 32

17 Note that all the figures in this subchapter are related to a specific frequency at which the antenna array was stimulated. However, the antenna array performance is frequency dependent. This is demonstrated in Figure 20. Here the antenna array leakage (blue) at the boresight and its comparison to the gain of the first antenna element (red) and to the gain of the whole antenna subarray (yellow) is shown for frequencies from 16.5 GHz to 20 GHz. -80 Element 2 Panel 1 ON, Other Panel OFF Magnitude [db] Element aperture illumination Element aperture illumination w/o Leackage Leackage Position [cm] Figure 13 Antenna array leakage (black) and its comparison to the second antenna element aperture with (solid red line) and without leakage (dashed red line) Figure 14 Antenna coupling estimated from measured date with the leakage Figure 15 Antenna coupling estimated from measured date without the leakage : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 17 of 32

18 Figure 16 Broadband measurement of the antenna array leakage (blue) at the boresight and its comparison to the gain of the first antenna element (red) and to the gain of the whole antenna subarray (yellow) Even we could find reasons why our antenna array measurements cannot be straightforwardly used for calibration, we failed in proper calibration of the antenna array by means of the conventional calibration method. Therefore we have followed an alternative approach that is described in the following chapter Power far field calibration of the analog antenna array using iterative optimization Our alternative approach of the analog antenna array calibration is based on far field power measurements and iterative optimization of the antenna array response. The measurement setup suitable for the calibration of the whole hybrid (analog and also its baseband part) is given in Figure 17. The antenna array is mounted at a motion emulator in an anechoic chamber. The antenna array faces a nearby antenna tower from the anechoic chamber through an EM transparent window. A horn antenna is receiving Ka band signal that is transmitted from the antenna array and is driving a power detector. Two RF chains of the antenna array are connected to a channel emulator that executes the digital base band processing. In this constellation the analog antenna array and the digital channel emulator are capable to perform hybrid (analog-digital) beamforming and the hybrid antenna array can be calibrated in both domains. In this subchapter, calibration of only the analog part is explained and demonstrated. Performance of the calibrated hybrid beamformer is shown in the next chapter. : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 18 of 32

19 In order to calibrate the analog part, the channel emulator was used just as a power divider which equally distributed power from the signal generator to both RF chains of the antenna array. Figure 17 Setup used for the far field calibration of the antenna array based on a received power only The far-field power based calibration maximizes signal power at the receiving power detector. No phase information is required. The idea behind this approach is as follows. If the antenna array is properly calibrated then signals transmitted by all antenna elements are coherently superimposed at the receiving antenna. In this case, the power detector measures the maximum power. However, if one of the elements does not possess proper BF weights its transmitted signal does not coherently superimpose at the receiving antenna. In this case, the power at the power detector is smaller. Thus, in order to identify calibrated BF weights for a specific direction of the beam steering, it is necessary to optimize the BF weights so that they maximize the received power at the power detector. Note that for different steering angles different optimizations must be performed. The maximization of the beam power was done in our approach by means of the particle swarm optimization algorithm which was described in [1]. Performance of such optimization is illustrated by Figure 18 and Figure 19 for two different steering angles 0 and 45, respectively. The figures illustrate that the increase of the beam power is about 1-2dB. Another benefit of the calibration is a proper alignment of the antenna array coordinate system to the scenario. In a realistic case when an antenna is mounted at its position it is difficult to pile it properly to the other side of the link. If the feedback from the other e.g. receiving side is available it may be used to electronically correct for the proper steering direction of the antenna array. E.g. : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 19 of 32

20 Figure 5, Figure 6 show beam patterns of the arrays which steered beams to 0 and 45. However, due to the misalignment of the antenna array coordinate system and the coordinate system of the positioner the beams are slightly tilted (~1.2 ). Figure 18 Measured beam patterns of the analog antenna array steering one beam to 0, blue before calibration, red after calibration Figure 19 Measured beam patterns of the analog antenna array steering one beam to 45, blue before calibration, red after calibration 3.3 Performance of the calibrated hybrid antenna array Similar calibration procedure was used to calibrate also the digital part of the hybrid antenna array. However in this second step we have minimized the power radiated towards a predefined direction which was different to the beam direction and represented the null steering direction. This optimization was also based on PSO algorithm. Since our antenna array has only two RF chains, only 2 complex valued BF weights were optimized. This took substantially less time, since the search space was reduced from 128 dimensional (64 complex valued analog BF weights) to only 4 dimensional space (2 complex digital BF weights). : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 20 of 32

21 Figure 20 Measured beam patterns of the hybrid antenna array which steers one beam to 0 (blue line, no null steering) and one null to 8 after the baseband calibration (red) and a comparison to the calibrated null steering at -5 (green) Figure 20 shows the result of the hybrid calibration for beam and null steering. The blue line is represents the beam obtained by the calibration of the analog part only that maximized the power to 0. This was input for the second calibration that optimized the baseband BF coefficients and minimized power at 8 (red line). The green line shows another example of the beam and null steering to 0 and -5 respectively. At the first glance, it seems that the PSO based calibration is just a feedback based optimization of the beam power to a specific direction and it is different to the conventional calibration which extracts a set calibration coefficients that can be used to precompute BF for an arbitrary shape of the antenna array pattern. However, this is true only under assumption that an antenna array consists of antenna elements that are isotropic. However a realistic antenna array consists of antenna elements having different antenna patters. Therefore, the conventional calibration requires calibration measurements of all antenna elements at a sufficient set of angles so that it is possible to interpolate these antenna patterns for arbitrary angle. The same can be achieved by the proposed alternative calibration based on the optimization. The result of one optimization is actually the realistic steering vector for a specific direction. If sufficient number of optimization is performed along different angles it is possible to interpolate the results for an arbitrary angle. The realistic steering vectors are the basis for computation of BF coefficients for arbitrary pattern of the antenna array. 4 Over-the-air demonstration of spectrum sharing The hybrid beamforming shown in the previous chapters was the basis for the over-the-air demonstration of the spectrum sharing with one terrestrial and one satellite link. The forward terrestrial link and the downlink of the satellite link were emulated over the air between the : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 21 of 32

22 laboratory building and the nearby antenna tower ~100m distance) using the same frequency band at 19.5GHz carrier. The antenna tower carried one receiving horn antenna which was dedicated to the terrestrial link. A satellite dish antenna was also mounted at the tower. It was emulating the downlink VSAT antenna that is interfered by signals transmitted from the prototype antenna array in the laboratory building. The angular separation of the terrestrial forward and the satellite downlink was only about 1.5. The upper view and the side view of the demonstration setup is illustrated in Figure 21 and Figure 22. Figure 21 shows also a detailed picture of the tower with the antennas. The antenna in the middle of the cross is the receiving terrestrial Horn antenna. On the right side, there is the satellite dish emulating the VSAT. Figure 22 shows also a picture of the antenna prototype mounted at the motion emulator that was placed in the anechoic chamber and facing the antenna tower. The block diagram of the terrestrial link is depicted in Figure 23. One terrestrial modem was connected to the channel emulator and the analog antenna array that performed hybrid beamforming and up-converted IF signal from the terrestrial modem (~2GHz) to 19.5GHz (~60MHz bandwidth). Terrestrial antenna Satellite antenna Figure 21 Upper view of the demonstration setup : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 22 of 32

23 Figure 22 Side view of the demonstration setup The receiving antenna at the tower was connected to a down converter. It down-converted the 19.5GHz received signal to 140MHz IF signal, which drove the second terrestrial modem. The backward link was just a wired connection between the two terrestrial modems for the sake of simplicity since this link was irrelevant for the spectrum sharing demonstration. Figure 23 Block diagram of the terrestrial link The block diagram of the satellite link is depicted in Figure 24. The downlink of the satellite link was created by one satellite modem (UHP 1000 from Romantis) which was connected to up convertor converting 1GHz IF signal to 19.5GHz downlink signal. The received signal at the tower was down converted by LNB to 1.25GHz IF signal. This signal was connected to the Rx of the second satellite modem. The uplink of the satellite link was operating at 29GHz and was not interfered by the terrestrial link. The vision of the spectrum sharing demonstration that is : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 23 of 32

24 based on the interference mitigation available by a hybrid beamforming antenna array is illustrated in Figure 25. Figure 24 Block diagram of the satellite link The terrestrial backhaul link and the satellite downlink are sharing the same frequency band. The role of the antenna array is to steer beam towards the other terrestrial backhaul node and to steer a null towards the satellite receiver in order to protect it from the interference. For the live demonstration for the final review we decided to generate traffic within a video server that can adjust the traffic according to the throughput of the communication link. The video server will be connected to one side of the satellite link. A PC will be used at the other side of the satellite link to show the quality of the streamed video and to demonstrate so the link quality of the satellite link. The link quality will depend on the beamforming capabilities of the terrestrial antenna array. In the first step, we want to show that if the antenna array emulates just a conventional high directional antenna which cannot steer a null, the video and the satellite link quality will be poor. In the second step the antenna array will be switched to emulation of a hybrid antenna array which can steer a beam and a null. Thus it shall protect the satellite link which can perform at high throughput and offer high quality video streaming. Eventually a video streaming can be used also at the terrestrial side to demonstrate the terrestrial link which is according to the interference analysis performed in [3] unaffected by the interference of the downlink satellite signal. This signals weak. It is coming from a satellite at orbit i.e. at large propagation distance and thus strong attenuation. : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 24 of 32

25 Figure 25 Vision of the demonstration with hybrid antenna array properly steering the beam towards the terrestrial antenna and forming a null towards the VSAT antenna Figure 26 illustrates the antenna constellation that we used for PSO based calibration of the hybrid antenna array for proper null steering. The antenna array was pointed towards the terrestrial antenna. The BF weights of the antenna array were adjusted to steer maximum power to terrestrial antenna which was at 0. The baseband beamformer was optimized to steer minimum power (the null) towards the satellite antenna at Despite such a small link separation the antenna array could steer one beam towards the terrestrial antenna and one null towards the satellite antenna by means of the hybrid beamforming. : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 25 of 32

26 Figure 26 Antenna constellation and the setup for null steering calibration This is illustrated by Figure 27. This figure shows comparison of two measured beam patterns for two setups of the antenna array with and without null steering. The receiving antenna in these measurements was the satellite VSAT antenna. However the coordinate system of the antenna array was defined so that the 0 in azimuth is the direction towards the terrestrial antenna as shown in Figure 26. Therefore, the null is visible at 0 azimuth and the beam is at the -1.5 in Figure 27. Even the position of the satellite VSAT antenna is by chance almost the same as the null of the beam pattern in which the array was not steering the null (blue line in Figure 27), it is not deep enough to protect the satellite link as it will be shown later in this document. The null steering antenna array is required (red line in Figure 27) in order to protect the satellite link with its deep null at the position of the receiving VSAT antenna Received signal power [dbm] Received signal power [dbm] Azimuth [ ] Figure 27 Narrowband beam pattern measurements of the antenna array antenna pointed towards the satellite VSAT antenna at 0 with the terrestrial receiver at -1.5 (blue - antenna array without and red with null steering) Azimuth [ ] Figure 28 Broadband beam pattern measurements of the antenna array antenna pointed towards the satellite VSAT antenna at 0 with the terrestrial receiver at -1.5 (blue - antenna array without and red with null steering The beam pattern measurements presented here were not performed in a conventional way at a certain frequency. Here, the beam patterns were measured by a broadband stimulation signal. The signal was generated by the terrestrial modem with about 60MHz of the frequency : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 26 of 32

27 span. The measurement which had the maximum received signal power within a predefined frequency span was taken to form the measured beam pattern. Figure 27 was measured for the frequency span of 10 khz and it represents a narrowband null steering. The null is in this case very deep. In comparison, Figure 28 depicts the broadband null steering in which the whole operational frequency span of the terrestrial modems was taken into account. In this case the null is very shallow since the null is not properly formed for all the frequencies and the beam pattern contains only the worst case (from the null steering point of view) frequency with the maximum signal power. The broadband null steering requires a frequency dependent beamforming. This can be realized e.g. by means of the digital part of the hybrid beamformer. However this approach was not analysed within this demonstration since is requires further research that was not planned within WP6. These two constellations of the hybrid antenna array (with and without null steering) were used to demonstrate the spectrum sharing enabled the hybrid antenna array. A terrestrial modem was connected to the hybrid antenna array. The antenna array was configured as an antenna with and without the null steering capability. Figure 29 and Figure 30 illustrate the impact of the null steering on the spectrum of the interfering terrestrial signal that was received at the satellite VSAT antenna at the tower. The null reduced the interference up to about 37dB. However it is visible that the null is frequency dependent as it was already indicate in Figure 27 and Figure 28. Only the frequencies around the carrier frequency can drastically reduce the interference. Figure 29 Spectrum of the terrestrial interference (using antenna array without null steering) received at the sat. antenna Figure 30 Spectrum of the terrestrial interference (using antenna array with null steering) received at the sat. antenna Frequencies at the edge of the terrestrial frequency band were reduced only less than 10dB. This shows even more necessity of the hybrid antenna array which digital part can seamlessly implement also frequency dependent null steering. : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 27 of 32

28 The satellite link was firstly tested without interference. The satellite modem was connected to the satellite transmitting antenna. Figure 31 shows spectrum of the signal received at the satellite VSAT antenna at the tower. Since we already anticipated that the frequency selective null can protect satellite link only around the carrier frequencies we have used smaller symbol rate (8 MSps from possibly 32 MSps) which resulted in a narrower spectrum in comparison to the terrestrial link (see Figure 29 and Figure 31). After we successfully established the satellite link without the terrestrial interference the performance of the antenna array and the effect of the null steering was analyzed. Figure 32 illustrates the case when the terrestrial antenna array steers the null towards the satellite VSAT antenna. Spectrum of the desired satellite signal is clearly visible and also its protection from the terrestrial link. Without the null steering capability the satellite signal was completely hidden by the terrestrial interference. Figure 31 Spectrum of the satellite signal received at the VSAT antenna without any interference Figure 32 Spectrum of the signal received at the sat. antenna which is a composition of the transmitted satellite signal and the interference from the terrestrial link mitigated by the hybrid antenna array Figure 33 and Figure 34 illustrate the effect of the antenna array beamforming performance on the satellite modem of the VSAT terminal (Figure 25). In the case of the antenna array without the null steering capability, the VSAT modem could not lock on the modem of the emulated satellite. In the case of the hybrid antenna array with the null steering capability the VSAT modem was locked on the satellite modem. Figure 35 demonstrates the obtained throughput of the satellite link. It shows GIU of the JPERF application which is a frontend for a popular command-line utility for bandwidth measurements named IPERF. Iperf is a reliable utility designed to measure bandwidth performance and to tweak various TCP related parameters, as well as UDP characteristics. Three different curves demonstrating the throughput are visible in Figure 35. The green curve shows the throughput that was achieved without the interference from the terrestrial link. The throughput was about 20Mbps. It was achieved for the maximum symbol rate of the satellite modems with 32 MSps. When the satellite link was interfered by the terrestrial link, the satellite link collapsed even when we tried to protect it by the null : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 28 of 32

29 steering. The reason was the narrowband null steering that was realized and the broad spectrum of the satellite signal. Therefore, we have decreased the symbol rate to 8 MSps. This resulted in narrower spectrumof the satellite signal that could be protected by the null steering as illustrated before by Figure 33 and Figure 34. The blue and white curve in Figure 35 show the throughput that was achieved with the interference from the terrestrial link and with the null steering. The throughput was about 7Mbps. In the case when the null was not steered the link could not be established. Figure 33 GUI of the satellite modem at the VSAT side for unmitigated interference Figure 34 GUI of the satellite modem at the VSAT side for mitigated interference Figure 35 GUI of the JPERF : Review of the Over-the-air test results of antenna prototype for shared satellite-terrestrial Page 29 of 32