Self-interference suppression improvement by employing circular polarized antennas

Measurement 110 (2017) 53–59

Contents lists available at ScienceDirect

Measurement journal homepage: www.elsevier.com/locate/measurement

Self-interference suppression improvement by employing circular polarized antennas M. Portela Táboas, M. Vera-Isasa, M. García Sánchez ⇑ Department of Teoría do Sinal e Comunicacións, Universidade de Vigo, Vigo, Spain

a r t i c l e

i n f o

Article history: Received 25 March 2017 Received in revised form 12 June 2017 Accepted 15 June 2017 Available online 16 June 2017 Keywords: Full-duplex Self-interference Passive suppression Analogue cancellation Directional antennas Circular polarization

a b s t r a c t This paper presents a novel method for suppressing the self-interference due to multipath in WiFi nodes of full-duplex systems based on employing two circular polarized antennas. Three different techniques have been used in order to suppress the interference generated by the direct wave: antenna separation, antenna pattern and analogue cancellation circuit comprised of a balun, a phase shifter and a variable attenuator. As a result, an initial isolation of 90 dB was achieved inside an anechoic chamber. In order to beat the multipath interference circular polarization was employed. With the aim of analyzing the robustness of the circular polarization in comparison with the linear one, the self-interference between transmitter and receiver was measured in nine different environments. The measured level of selfinterference was also compared with the results achieved inside the anechoic chamber. In this way, the degradation of the self-interference suppression due to multipath was evaluated. Results show better isolation when employing circularly polarized antennas in seven of nine environments. In addition, with the circular polarization self-interference suppression higher than 70 dB was achieved independently of the environment. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction The self-interference (SI) between two antennas (transmitter and receiver) located in the same node of a WiFi system, working simultaneously and at the same frequency is one of the challenges in implementing a full duplex wireless device. SI is mainly defined by two components: the self-interference due to the direct path coupling between the antennas of the node and the interference owing to multipath produced by the signals reflected, diffracted and scattered in the environment where the system is deployed. In the literature, different solutions can be found to address the SI cancellation problem, which can be divided into three different categories: propagation, analogue and digital self-interference suppression [1]. The first one is based on passive techniques such as antenna directivity [2], antenna separation [3], antenna cancellation [4], passive RF circuits [5], or a combination of some of these methods [6], including the use of cross-polarized antennas. The second one implies to generate an attenuated and phase-inverted copy of the transmitted signal which is added to the received one as in [7,8]. Both techniques have been widely used to mitigate the interference produced by the direct wave. Each technique

⇑ Corresponding author. E-mail address: [email protected] (M. García Sánchez). http://dx.doi.org/10.1016/j.measurement.2017.06.022 0263-2241/Ó 2017 Elsevier Ltd. All rights reserved.

presents advantages and disadvantages but, in general, they reach to solve the intended problem. Once the direct wave interference problem has been mitigated the multipath becomes the main interference issue. Many authors agree in applying digital selfinterference cancellation in order to beat the multipath interference [7,8]. Most of them are focused on estimating the radio channel by employing a pilot carrier that transmits a known signal. In such a way, it is possible to distinguish the part of the received signal which is due to the self-interference since we know what the channel response is. This technique implies to waste a part of communication resources on doing the estimation and subsequent correction of the signal. In this context, the goal of this work is based on showing how useful circular polarization is as a method for decreasing the selfinterference due to multipath. One of the most important characteristics of the circular polarization is that, when the transmitted signal is reflected over a plain surface (wall), the sense of rotation is reversed. Thus, this counter-clock-polarized signal will be filtered for a receiver antenna working with the same polarization as the transmitter one. In the paragraph below, we will explain the steps followed to assess the proposed solution. In Section 2, a description of the measurement system is done. Firstly we describe the passive and analogue techniques which were used to cancel the SI generated by the direct wave. Then,

54

M. Portela Táboas et al. / Measurement 110 (2017) 53–59

we explain the characteristics of the antenna employed. In Section 3, the environments where the measurements were carried out are presented and the process followed to do them is explained. The results obtained from the measurement campaign are analyzed in Section 4. Finally, in Section 5 we expose the main conclusions that can be extracted from the analysis done. 2. Measurement system In order to compare the robustness to multipath of the full duplex systems when operating with linear and circular polarization two requirements should be addressed. First of all, the selfinterference generated by the line-of-sight path has to be suppressed. In second place, an antenna capable of working with both polarizations is necessary. In this work, three different techniques have been used in order to eliminate the interference generated by the direct wave. The first one lies in applying 15 cm of separation between antennas. This technique provides an initial isolation of 20 dB. The second technique involves using directional antennas instead of the traditional omnidirectional. In addition, due to its simplicity and good results achieved, we decided to employ the analogue cancellation circuit presented in [7]. It is comprised of a balun, a variable phase shifter and a variable attenuator. Fig. 1 shows the complete block diagram of the WiFi node. We have employed a PNA Network Analyzer N5222A of Agilent Technologies in order to measure the self-interference (S21 parameter) generated in the node. A circular patch antenna was designed for the measurements by using the software CST Studio Suite. Fig. 2 shows a front view of the antenna design. It consists of a circular patch over a ground plane that is at the back face of the printed circuit board. The circular path is fed in two different points by using a 90° hybrid coupler. The antenna will work with circular polarization when just one of the two ports is excited. However, when both ports of the hybrid coupler are fed with the same excitation simultaneously, the antenna will transmit a linearly polarized signal. Measured copolar and crosspolar radiation patterns in both working modes, circular and linear, are shown in Figs. 3 and 4 respectively. In addition, the antenna fulfils the following requirements: it works at WiFi ISM band, i.e. it is matched from 2.4 to 2.48 GHz (Fig. 5). It has a directional beam pattern with high beam width (100° approximately) and a measured gain, at 2.4 GHz, of 5 dBi for linear polarization, and 4 dBi for circular polarization. This kind of antennas provides additional self-interference suppression [2] whereas guarantee good quality signal between WiFi nodes. The final setup is shown in Fig. 6. As a result of applying the mentioned passive and analogue cancellation techniques, an initial isolation of 90 dB was achieved inside an anechoic chamber.

Fig. 2. Front view of the antenna used for the experiments.

Fig. 3. Measured circular polarized co and crosspolar radiation patterns (left-hand, LHCP, and right-hand, RHCP) at 2.4 GHz.

3. Measurement campaign The measurement campaign was performed in nine different environments as shown in the sketch of Fig. 7 and the pictures in Fig. 8. The walls in the first environment (named 1 in plots and table of results) are composed of absorber material, an insulating carpet covers the floor, and the ceiling (as in all cases) is made of plasterboards. Due to this fact, the main characteristic will be the absence of reflections.

Fig. 1. Block diagram of the full duplex system node.

M. Portela Táboas et al. / Measurement 110 (2017) 53–59

Fig. 4. Measured linearly polarized co and crosspolar radiation patterns at 2.4 GHz.

-5

S11(dB)

phase of the variables components in the analogue circuit until obtaining the maximum isolation between antennas [7]. In this manner, we made sure that the interference due to the direct path was successfully suppressed. Once the system was fitted, we moved the portable structure from the anechoic chamber to the nine selected environments. Thereupon, the interference signal level was measured again. We repeated the same process, initially employing circular polarization and then using the linear one. The isolation between antennas achieved in the anechoic chamber is degraded by multipath when the full duplex system is working in a real environment. The goal in this measurement campaign is to evaluate what kind of polarization is more robust when the characteristics of the environment change. In the area where the measurements were done, there were active WiFi systems working in the band of interest which disturb the results. So, in order to avoid the interference generated by these nodes, we decided to evaluate a different range of frequencies. After studying the spectrum with a spectrum analyzer, the nearest jamming-free frequency was at 2.4 GHz, thus, we decided to measure the band from 2.35 GHz to 2.45 GHz and to analyses in greater depth the results at the central frequency of the band.

4. Data analysis and results

-10

-15

-20

-25

-30 2

55

RHCP LHCP

2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3

Frequency(GHz) Fig. 5. Measured S11 parameter when one of the two ports is excited. The two cases are showed: left-hand and right-hand.

In the scenario 2, the situation is very different since we have a concrete wall in front of the antennas and two metallic doors in both sides. In the environments 3 and 4, we should emphasize that there are not big metallic planes even though there is a wide range of objects (tables, chairs, computers. . .) where the signal can be reflected. Talking about reflective areas, scenarios 5, 6 and 7 are very similar. They have one or two metallic elements that consist of doors, radiators or rails. Concrete walls cover the most part of the remaining area and there are tiles on the floor. The main differences between these environments are the distance and the direction where the metallic objects are located. Finally, we describe the scenarios 8 and 9. In the first one, we have two metallic racks in front of the antennas, work benches full of instrumental equipment in both sides and a metallic closet behind the work bench in the left side. Whereas, the last one is a much reduced area (210 cm  220 cm) where there is a metaldoor in each wall. Because of this, in these two last environments, a high number of reflections is expected. The measurement campaign was carried out as follows: the first step is based on placing the measurement system in a portable structure (vector network analyzer included). Then, the equipment was introduced inside the anechoic chamber and the coupling between both antennas was measured. After that, the cancellation circuit was activated and tuned by varying the amplitude and

In order to set a reference level of SI between antennas in a fullduplex WiFi node, the first step was to measure the isolation between two omnidirectional antennas when they are located inside the anechoic chamber and separated 15 cm. The result was the same whatever polarization we use; 15 cm of distance between antennas involves 28 dB of self-interference suppression. Then, we changed the omnidirectional antennas for the directional antennas we presented in Section 2. After that, the interference was measured inside the anechoic chamber again. As a result of changing the antennas, the isolation was increased for both polarizations. Thus, the reference signal when using linear polarization is reduced until 51.84 dB and it decreases until 62.49 dB when the system transmits with the circular polarization. Finally, we have activated and tuned the analogue cancellation circuit and, after that, the values of interference achieved were 91.34 dB and 105 dB for the linear and circular polarization respectively. A difference of more than 10 dB exists between the reference signals of both polarizations. This makes it more difficult to compare them since the power of the multipath signals will have to be higher to degrade the level of self-interference suppression when working with linear polarization. Consequently, we are not able to compare the effects of the polarization in the suppression system. That is why we decided to readjust the phase and attenuation of the analogue cancellation circuit until setting the isolation for the circular polarization to the same level ( 92.45 dB) as the linear one, instead of using the best value as possible. Once the isolation was measured in the anechoic chamber, we move the system to the nine environments described in Section 3 for evaluating how robust the polarizations are. Fig. 9 summarizes the results achieved in environments 1 to 5, and Fig. 10 shows the results for environments 6 to 9. Linear polarization measurements are presented in Fig. 9(a) and in Fig. 10(a), whereas, graphics (b) contain the results for the circular polarization. Table 1 collects the isolation level obtained in each scenario at 2.4 GHz and the level of degradation of the self-interference suppression when we compare the isolation in the reflective environments with the results achieved inside the anechoic chamber. As we can see, there are only two environments where the degradation of the isolation is smaller when the system works with linear polarization (environments 1 and 8). In the scenery 1, as we expected, the circular polarization does not provide any improve-

56

M. Portela Táboas et al. / Measurement 110 (2017) 53–59

Fig. 6. Front view (a) and rear view (b) of the measurement setup.

Fig. 7. Sketch of the scenarios. ► sets antenna pointing direction.

ment, i.e., due to the characteristics of the environment, there is not multipath interference. Thus, even though a slightly better result is achieved by employing linear polarization, the differences are very small. The situation in the environment 8 is different. This is a very reflective environment, where there are a large number of multipath contributions. In this case, we expected to achieve better results by using the circular polarization since it should be more robust to multipath. But, the environment is so complex that probably, double and triple reflections are taking place, so the received signal is depolarized but not always reversed. As a result, to define

the reasons why the linear polarization provides the best results is not an easy task. Environment 5 is a good example of how complex is to devise the behavior of the system in reflective environments. This is the most reflective environment we have measured. It is based on a small room with three metallic planes (one per wall). Here, the result is totally different. Now, the best option is the circular polarization which reduces the degradation approximately 20 dB more if we compare it with the linear one. If we analyses the rest of environments, the results are very similar. The best alternative is to use the circular polarization. In case of the environments 2, 5, 6 and 7, they present one or two

M. Portela Táboas et al. / Measurement 110 (2017) 53–59

57

Fig. 8. Pictures of the nine scenarios.

metallic elements of different sizes. So, the number of multipath signals will be smaller than it was in the environments evaluated before. This is the reason why the characteristics of circular

polarization can be exploded. Thus, in case of environments 2, 6 and 7, the degradation experimented is very similar as it just varies 2.30 dB when using circular polarization. We should emphasize the

58

M. Portela Táboas et al. / Measurement 110 (2017) 53–59

S21(dB)

-50 -60 -70

Linear

-80 -90

(a) -100 2.35

2.36

2.37

2.38

Without canc.(anec. chamber) With canc. (anec. chamber) Environment 1 Environment 2 Environment 3 Environment 4 Environment 5 2.39 2.4 2.41 2.42 2.43 2.44 2.45 Frequency (GHz)

Fig. 9. Results for environments 1 to 5: (a) using linear polarization and (b) using circular polarization. The line type legend is for both figures.

-50

S21(dB)

-60 -70

Linear Without canc. (anec. chamber) With canc. (anec. chamber) Environment 6 Environment 7 Environment 8 Environment 9

-80 -90

(a)

-100 2.35

2.36

2.37

2.38

2.39 2.4 2.41 Frequency (GHz)

2.42

2.43

2.44

2.45

-50

(b)

S21(dB)

-60 -70 -80 -90 -100 2.35

Circular 2.36

2.37

2.38

2.39 2.4 2.41 Frequency (GHz)

2.42

2.43

2.44

2.45

Fig. 10. Results for environments 6 to 9: (a) using linear polarization and (b) using circular polarization. The line type legend is for both figures.

M. Portela Táboas et al. / Measurement 110 (2017) 53–59 Table 1 Isolation and degradation of the self-interference suppression measured. Best solution values are in bold. Environment

Anechoic chamber 1 2 3 4 5 6 7 8 9

Isolation (dB)

Degradation of SIS (dB)

CP

LP

CP

LP

92.45 79.33 74.94 82.07 71.35 93.87 76.16 77.21 77.35 83.23

91.34 78.46 69.25 74.97 68.94 74.80 70.59 69.40 79.97 63.00

– 13.10 17.56 10.36 21.08 1.44 16.31 15.25 15.12 9.23

– 12.85 22.11 16.34 22.42 16.51 20.72 21.97 11.34 28.21

59

method based on using circular polarized antennas shows better performance compared to the use of linearly polarized antennas. In this comparison we have assumed that antennas with vertical polarization are used, as in [4], for transmission and reception. The comparison would not be valid for a system using orthogonal linearly polarized antennas as in [6]. Another advantage of using circular polarized antenna is found for mobile terminals, as the terminal movement would not affect the polarization matching [6]. Using linearly polarized antennas would require the terminal orientation tracking plus some mechanism to keep the polarization match of the antennas. By using the circular polarized antennas, along with three different techniques of self-interference suppression, (distance, directional antennas and analogue cancellation circuit) isolation values higher than 70 dB were achieved. Acknowledgments

result in the fifth case, where we have obtained a negative level of degradation ( 1.44 dB). This result can be achieved when using circular polarization since we did not fit the system to the best level of self-interference suppression possible. Finally, talking about environments 3 and 4, they have no metallic planes but there is a wide range of objects where the transmitted signal can be reflected. In this kind of environments, the circular polarization seems to be the best option too. Now, if we focus on the level of self-interference suppression achieved, we can realize that, by applying the four different techniques evaluated in this work (distance + directional beam pattern + cancellation circuit + circular polarization), the worst level of self-interference is under 70 dB no matter what the characteristics of the environment are. This is a very positive result considering that no digital cancellation was introduced in the system and the analogue cancellation circuit just was fitted inside the anechoic chamber, at the beginning of the measurement campaign. 5. Conclusion The robustness of the self-interference suppression system was assessed in nine different environments by using linear and circular polarized antennas. In 7 of 9 scenarios, the system working with circular polarization has provided better results. In addition, if the scenario does not present multipath, the circular polarization does not introduce any advantage or disadvantage compared to the linear one. Only when the system is located in a very reflective environment it is not possible to conclude that the new polarization is a better option. We have empirically demonstrated that this novel

This work was funded Spanish Government, Ministerio de Economía y Competitividad, Secretaría de Estado de Investigación, Desarrollo e Innovación, AtlantTIC Research Center and the European Regional Development Fund (ERDF) under projects TEC2014-55735-C3-3-R and TACTICA. References [1] Z. Zhang, K. Long, A.V. Vasilakos, L. Hanzo, Full-Duplex wireless communications: challenges, solutions and future research directions, Proc. of IEEE 7 (104) (2016) 1369–1409. [2] E. Ahmed, A.M. Eltawil, Z. Li, B.A. Cetiner, Full-duplex systems using multireconfigurable antennas, IEEE Trans. Wireless Commun. 11 (14) (2015) 5971– 5983. [3] M. Duarte, C. Dick, A. Sabharwal, Experiment-driven characterization of fullduplex wireless systems, IEEE Trans. Wireless Commun. 12 (11) (2012) 4296– 4307. [4] T. Snow, C. Fulton, W.J. Chappell, Transmit-receive duplexing using digital beamforming system to cancel self-interference, IEEE Trans. Microw. Theory Techn. 12 (59) (2011) 3494–3503. [5] J. Li, S. Yang, Z. Nie, Design of a passive self-interference cancellation network with high cancellation ratio, Chengdu, China, Jul, Proc. Cross Strait QuadRegional Radio Science and Wireless Tech. Conf., 2013, pp. 39–42. [6] E. Everett, A. Sahai, A. Sabharwal, Passive self-interference suppression for fullduplex infrastructures nodes, IEEE Trans. Wireless Comm. 2 (13) (2014) 680– 694. [7] Jain, M., Choi, J.I., Kim, T., Bharadia, D., Srinivasan, K., Seth, S., Levis, P., Katti, S., Sinha, P., ’Practical, real-time, full duplex wireless’ Proc. ACM International Conference on Mobile Computing and Networking, Las Vegas, Nevada, USA, Sep. 2011 pp. 301–312. [8] S. Hong, J. Mehlman, S. Katti, ’Picasso: flexible RF and spectrum slicing’ Proc, Hensilki, Finland, August, ACM Special Interest Group on Data Communication, 2012, pp. 37–48.