Fourier Rotman Lens Enabled Directional Modulation Transmitter

A 10 GHz Fourier Rotman lens enabled dynamic directional modulation (DM) transmitter is experimentally evaluated. Bit error rate (BER) performance is obtained via real-time data transmission. It is shown that Fourier Rotman DM functionality enhances system security performance in terms of narrower decodable low BER region and higher BER values associated with BER sidelobes especially under high signal to noise ratio (SNR) scenarios. This enhancement is achieved by controlled corruption of constellation diagrams in IQ space by orthogonal injection of interference. Furthermore, the paper gives the first report of a functional dual-beam DM transmitter, which has the capability of simultaneously projecting two independent data streams into two different spatial directions while simultaneously scrambling the information signals along all other directions.


Introduction
Directional modulation (DM) technology, as a keyless physical layer security means, has developed in many aspects in recent years . Generally speaking, DM enabled transmitter arrays have the capability of transmitting artificially distorted signals along all spatial directions other than an a priori specified communication direction in free space, in this way lowering the possibility of unwanted interception by eavesdroppers.
The characteristics of DM systems were mathematically described via a vector model presented in [12,13]. DM system performance assessment and cross system comparison metrics for DM system evaluation were established in [14], wherein DM systems were categorized as dynamic or static according to whether the array excitations for each unique transmitted symbol were updated at the transmitted data information rate or not. In [4,5,[14][15][16][17][18][19] several DM synthesis approaches, which placed emphasis on different DM system properties, for example, extra energy consumed [14], bit error rate (BER) spatial distributions [4,5,15,16], far-field radiation patterns [17,18], and interference spatial distributions [19], have been proposed. DM physical realization has been the subject of various studies. In [1][2][3] a DM architecture, which relies on the near-field coupling effects, was proposed. However, the deployment of a large number of parasitic array elements, as well as the unavailability of effective synthesis methods, may hinder its further development. Actively driven static DM arrays [4][5][6][7][8][9]15] and antenna subset selection mechanisms [20][21][22][23] have been proposed. These DM architectures are synthesis-friendly, but all use reconfigurable devices in the radio frequency (RF) stage or at antenna level, for example, RF switches in [5,[20][21][22][23], and RF reconfigurable attenuators and phase shifters in [4,[6][7][8]15]. These devices that have to operate at a speed much higher than the information rate require additional control and bias circuits and introduce loss. Critically no examples of physical dynamic DM transmitters have until this paper been reported.
There are two DM architectures that avoid the use of RF switchable devices and which are compatible, from the practical implementation perspective, with both static and dynamic DM systems. One is the digital DM architecture presented in [13]. The implementation of this digital structure requires careful system calibration, especially frequency and phase synchronization between each of the RF chains. Another promising DM architecture exploits the properties of Fourier 2 International Journal of Antennas and Propagation transform beam-forming networks (FT-BFNs) or Fourier lens [24,25]. The beam orthogonality characteristic possessed by FT-BFNs [25] prevents the signals projected along the main information beam direction from being contaminated by interference injected at other beam ports of the Fourier lens. As was shown in [26], the additional phase adjustment could make classical Roman lens [27] exhibit Fourier transform function, enabling beam orthogonality in beam space. Furthermore, as pointed out in [24], when compared with the Butler matrix [28][29][30], the Fourier transform constrained Rotman lens [26] uses only two-dimensional (2D) printed circuit board (PCB) technology and is readily scalable by both frequency of operation and array port aperture size. Thus, we employ a Fourier Rotman lens to construct an experimentally evaluated dynamic DM transmitter.
Several system level experiments were conducted in an anechoic chamber. One is for general demonstration purpose, the video of which can be found in [31]. A real-time video stream with analogue amplitude modulation excited the 7th beam port of a 13-by-13 Fourier Rotman lens, while the other ports were excited with −25 dB sine-wave interference at 10 GHz, to yield a secure communication direction at boresight. In the video, it is observed that when compared with the conventional non-DM system, where the real-time video signal could be recovered in large spatial sectors, the Fourier Rotman DM system successfully confined the decodable region into a small area around boresight. The second experiment employed the Wireless Open-Access Research Platform (WARP) [32], which allowed digital modulation characteristics to be evaluated. This experiment forms the basis for this paper and the experimental setup used is described in Section 2, leaving experiment results and discussion presented in Section 3. Conclusions are drawn in Section 4.

Experimental Setup
A 13-by-13 Fourier Rotman lens DM transmitter for 10 GHz operation was built according to the design rules in [24] and experiments were conducted for real-time data transmission in a 10 × 6 × 6 m 3 anechoic chamber.
The block diagram of the Fourier Rotman lens DM experimental setup is illustrated in Figure 1, and photographs are shown in Figure 2. Here WARP boards are used for modulation, demodulation, and BER measurement.
The workflow for experimental characterization of the Fourier Rotman lens DM system is now presented.  and then MAX2829 and digitized through analogue to digital converters (ADCs) and finally stored in the receive buffer on the WARP board b. This raw data is read into MATLAB for analysis. In order to achieve the frequency synchronization between the transmitter and the receiver, a common 80 MHz sampling reference clock and RF reference clock for both WARP boards a and b are provided externally by the AWG5012. However, the signal generators Agilent 83650B and Rohde&Schwarz SMR20 used to generate RF continuous waves LO1, LO2, and LO3, all at 7.563 GHz, for upconversion and downconversion, are frequency synchronized by sharing a common 10 MHz reference clock; see Figure 1. (f) Artificial AWGN is added onto the received raw data in MATLAB to facilitate various SNR scenarios.
The system intrinsic AWGN is sufficiently low to be ignored. This was validated through an experiment where zero BERs after 10 +6 data bits transmissions were detected along any spatial direction in a conventional non-DM system positioned in the anechoic chamber. (g) Iterate steps (d) to (f) until the required number of data bits is transmitted. 10 +6 was used in the experiment. The BER along can then be computed.  (h) Scan within 0 ∘ to 180 ∘ and repeat steps (c) to (g). BER spatial distribution is obtained under prespecified system configurations and SNR conditions.
Before presenting sample experimental results in Section 3, the characteristics of the 13-by-13 Fourier Rotman lens and the 1D 13-element microstrip patch antenna array for 10 GHz operation, used in the experiment, are now presented.

13-by-13 Fourier Rotman Lens at 10 GHz. A 13-by-13
Fourier transform constrained Rotman lens [26] for 10 GHz operation was designed and fabricated on a substrate of Taconic RF-60A [36], thickness 0.64 mm, and relative dielectric constant 6.15, Figure 3. The overall board dimension, including the extended 50 Ω microstrip lines, is 250 mm × 250 mm. Different to the Fourier Rotman lens used in [24], the array ports are aligned with half wavelength spacing, which enables direct connection between the lens and the microstrip patch antenna array, eliminating cable induced magnitude and phase errors. Using measured -parameters at 10 GHz, the normalized array factors (AFs) for each beam port excited separately were computed and are presented in Figure 4. The beam orthogonality properties of the lens can be observed; for example, along the directions of the 7th and the 9th beams, 90 ∘ and 72 ∘ , the interference beam leakage from other beam ports is below −20 dB and −14 dB, respectively. The fabrication errors and nonuniform insertion loss between each beam port and array port pair contribute to the nonorthogonality, particularly for beams away from boresight.

1D 13-Element Microstrip Patch Antenna Array.
A 1D 13element microstrip patch antenna array with half wavelength spacing for 10 GHz operation was designed and fabricated on the FR4 substrate [37], thickness 1.57 mm, and relative dielectric constant 4.7, Figure 5. The fabricated antenna array is 240 mm in length and 60 mm in width. The -parameters of the patch antenna array were measured and are depicted for the frequency range from 8 GHz to 12 GHz in Figure 6. It can be observed that at 10 GHz array elements are well matched (|S | < −11 dB, = 1, 2, . . . , 13), and cross coupling is less than −18 dB.
The far-field active element pattern (AEP) [38], both magnitude and phase in the azimuth plane of each antenna element, was measured in an anechoic chamber and is presented in Figure 7. Here we normalize the measured power patterns with respect to the magnitude value of the center element, that is, antenna 7, along boresight (90 ∘ ) to International Journal of Antennas and Propagation The far-field radiation pattern for each Fourier Rotman lens beam port excitation when the array ports are connected to the antenna array is shown in Figure 8. Compared with the AFs in Figure 4, along the direction of the main beam of each pattern in Figure 8, there is additional leaked interference energy caused by the remaining patterns. This is due to the nonidentical AEP of each antenna element; see Figure  7, particularly along the spatial angles further away from boresight. However, it can be seen that more than 10 dB difference between the th ( = 3, 4, 6, 7, . . . , 11) main beam and leaked energy from other patterns can still be obtained.

Experimental Results and Discussion
In this section, the Fourier Rotman lens DM experimental results are provided. Firstly, the experimental setup is validated by comparing measured BER results with simulated BER results that are calculated based on the measured farfield radiation patterns in Figure 8 for a conventional beamsteering array, that is, one where no interference is injected into the system. Secondly, the orthogonality between the information and the interference in the beam space is checked by injecting interference under relatively low SNR scenarios. Next, results that show DM spatial BER advantages over the conventional array are presented under high SNR conditions. Finally, some experimental results on the properties of a dualbeam DM system where two separate information beams are projected simultaneously into two different directions are presented.

Validation of Experimental Setup and BER Measurement.
In Figure 9(a), the measured BER spatial distribution and the calculated one that is based on the measured patterns Spatial direction ( ∘ ) Figure 8: Normalized far-field radiation pattern for each Fourier Rotman lens beam port excitation when the array ports are connected to the microstrip patch antenna array. The far-field patterns for the 7th and 9th beam port excitations are highlighted.
in Figure 8 are depicted. At the transmit side, only RF1 was used for data transmission, while RF2 and RF3 were disabled. This makes the transmitter a conventional beam-steering array. In the experiment 10 +6 random DQPSK symbols with 2.5 Msps symbol rate were injected into the 7th beam port of the Fourier Rotman lens; that is, the main BER beam was projected along boresight direction. The power of the artificial AWGN added onto the received raw data in MATLAB was adjusted to make the BER along 90 ∘ 10 −4 , which corresponds to a SNR of approximate 13.8 dB. When the SNR was increased to 25.8 dB, which was achieved by decreasing the applied AWGN, the BER results obtained are shown in Figure  9(b).
For further validation similar BER measurement and simulation were conducted with the 9th beam port excited with 10 +6 random DBPSK data signals; see Figure 10. Here at the transmit side only RF2 was used. In order to normalize the BER value along 72 ∘ to 10 −4 , the SNR was adjusted to 9.2 dB.     Only the 9th beam port of the Fourier Rotman lens was excited with DBPSK data streams.
From Figures 9 and 10, it can be seen that the measured BER results well resemble their simulated counterparts, validating the experimental setup and BER measurement. In order to facilitate the discussion about the effects of the injected artificial noise on the information signal in the remaining section, the power normalized detected constellation diagrams in IQ space along 90 ∘ and 72 ∘ , respectively, for DQPSK and DBPSK cases are shown in Figure 11. The SNRs are 13.8 dB and 9.2 dB, respectively. We observe that the constellation diagrams along other directions are similar, but each symbol cluster is more spread due to lower signal power projected.

Validation of Orthogonality between Information and
Interference in Beam Space. The orthogonality between information and interference in beam space was checked by measuring BERs with and without interference injected. In order to route artificial interference to the remaining 12 beam ports of the Fourier Rotman lens, a 1-to-16 Wilkinson power splitter for 10 GHz operation was designed and built on FR4 substrate [37]; see Figure 12. Only 12 output ports were used in the experiment.
The interference routed to the remaining 12 beam ports of the Fourier Rotman lens was generated by the RF3 on the WARP node a. This was kept at a constant magnitude, either −5 or 0 dB at each interference port relative to the information signal power, but was updated with random phases selected within the range 0 ∘ to 360 ∘ at the symbol rate. The magnitude of interference was chosen to be constant in order to minimize the nonlinearity that active devices may introduce. In Figure 13 we present the measured BER results for the cases of DQPSK and DBPSK information signals, which were applied at the 7th and the 9th beam ports, respectively and alternately. It can be observed that the interference injected into the system has the capability of narrowing the decodable spatial region. Also it has little effect on the detected BERs along the selected communication directions, 90 ∘ and 72 ∘ for each case, which indicates that the orthogonality between information and interference in beam space is being achieved by the Fourier Rotman lens.

DM System Performance under High SNR Scenarios.
In order to better visualize the secrecy performance advantages of DM systems over conventional non-DM systems, the BER spatial distributions were obtained under high SNR scenarios, 25.8 dB for the DQPSK transmitters with main beam towards boresight and 21.2 dB for the DBPSK transmitters with main beam projected along 72 ∘ . Instead of applying interference at each of the remaining 12 beam ports, only two ports next to the information ports were excited with interference, that is, the 6th and the 8th ports for the DQPSK case and the 8th and the 10th ports for the DBPSK case. These ports were selected because they project main interference beams closely around the main information beam, which narrows the decodable spatial region in an energy efficient manner. Figure 14 illustrates the measured BER spatial distributions for the Fourier Rotman lens DQPSK and DBPSK DM systems for various interference configurations. It can be concluded that, compared with the conventional Fourier Rotman lens beam-steering systems where no interference is injected and whose BER responses are also provided in Figure 14, the DM system here exhibits narrower main BER beamwidth and higher BER values associated with BER sidelobes. Consequently the more the interference power, the more enhanced the secrecy performance that a DM system can achieve.
In order to further investigate how injected orthogonal interference affects signal formats along unselected spatial directions, the detected constellation diagrams for the DQPSK systems along 122 ∘ , where a noticeable BER sidelobe exists in the conventional system (see Figure 14(a)) under SNR of 25.8 dB are depicted in Figure 15. It can be observed that a typical DQPSK constellation pattern was detected when no interference was injected, Figure 15 , whose radius is determined by the interference power level. With increased interference power, these circles eventually intersect each other. This greatly obscures the constellation patterns and eventually hinders successful data recovery. Similar constellation patterns, but with 2 symbol states, were obtained for the DBPSK systems. These are omitted for brevity.

Dual-Beam DM Systems.
We now consider that two independent data streams are applied at two different beam ports of the Fourier Rotman lens simultaneously, for example, a DQPSK signal at the 7th beam port and a DBPSK signal at the 9th beam port. Figure 16 presents the measured BER results for the dual-beam DQPSK-DBPSK DM system. In the experiment the two independent data stream signals used Figure 15: Normalized received constellation diagrams for the DQPSK system with (a) no interference, (b) −5 dB interference at the 6th and the 8th beam ports, (c) 0 dB interference at the 6th and the 8th beam ports, and (d) 5 dB interference at the 6th and the 8th beam ports along 122 ∘ . The DQPSK data stream with 0 dB power was applied at the 7th beam port. The SNR along boresight was set to 25.8 dB. were kept at the same power strength, but we could choose different symbol rates. For comparison, the BER responses in the corresponding conventional DQPSK and DBPSK systems are also provided in Figure 16. In the dual-beam DM systems, the radiated DBPSK signals act as orthogonal interference to the DQPSK signals and vice versa. By setting different symbol rates for the DQPSK and the DBPSK data streams, higher BER values within unselected spatial regions were returned, which may be caused by preamble distortion. It should be noted that both DQPSK and DBPSK data were appended with the same Barker preamble sequence modulated with BPSK, but different symbol rates made them interfere each other. This system is a dynamic DM system, but unlike the previous dynamic ones the interference can only be updated with a limited number of states, for example, two (or four) interference states, corresponding to the number of distinct DBPSK (or DQPSK) symbols, for the DQPSK (or DBPSK) information signals. This can be observed in the received   Figure 17: Normalized received constellations diagrams in IQ space along directions of (a) 85 ∘ and (b) 67 ∘ . The symbol rates for both the DQPSK and DBPSK data streams were set to 2.5 Msps. SNR for each receiver case was set to 25.8 dB or 21.2 dB along their desired communication directions, that is, 90 ∘ or 72 ∘ .
constellations diagrams in IQ space along directions of 85 ∘ and 67 ∘ , respectively, 5 ∘ away from the each specified secure communication direction; see Figure 17.
In order to further distort the constellation patterns along unselected communication directions in a randomized fashion, the remaining 11 beam ports of the Fourier Rotman lens can be excited with random interference. Measured BER results are shown in Figure 18. Again the same interference was applied at the 11 beam ports, and the magnitude of the interference was kept constant during the entire transmission International Journal of Antennas and Propagation  Figure 19: The detected constellation patterns along (a) 85 ∘ and (b) 67 ∘ in the dual-beam DM systems when the (a) DQPSK and (b) DBPSK type receivers were equipped. The DQPSK signals and the DBPSK signals with the same power were applied at the 7th and the 9th beam ports of the Fourier Rotman lens, respectively. The same interference with −10 dB power excited the remaining 11 beam ports. SNR for each receiver case was set to 25.8 dB or 21.2 dB along their desired communication directions, that is, 90 ∘ or 72 ∘ .