ieice communications express, vol.7, no.5, 154 performance … · user multiple-input...

Performance evaluation ofdistributed antenna in uplinkMU-MIMO using amplitudeinformation and many-to-oneswitch

Soichi Ito1a), Sho Yoshida1, Kentaro Nishimori1,Tomoki Murakami2, Koichi Ishihara2, and Yasushi Takatori21 Graduate School of Science and Technology, Niigata University,

8050 2-no-cho Ikarashi, Nishi-ku, Niigata 950–2181, Japan2 Nippon Telegraph and Telephone Corporation, NTT Access Network Service

Systems Laboratories, 239–0847, Japan

a) [email protected]

Abstract: This paper proposes a hardware configuration for uplink multi-

user multiple-input multiple-output (MU-MIMO) transmissions with dis-

tributed antenna systems (DASs) in real indoor environments. Beam-forming

(BF) technology is used in the DAS with massive MIMO. In massive MIMO

transmission, signal processing becomes complicated because of the massive

antennas. Therefore, in general, a technique known as digital beamforming

(DBF) is adopted, in which the weight values are calculated through digital

signal processing. Massive MIMO systems applying DBF, which requires

receivers for all massive antennas, have problems related to power con-

sumption and cost because the access point becomes large.

We propose an analog–digital hybrid configuration, which selects anten-

nas using a many-to-one switch connected to the antenna. In the proposed

configuration, it is possible to reduce the number of receivers required, by

grouping antennas using many-to-one switches. In this paper, we evaluate

the effect of the proposed configuration through computer simulations using

the propagation channels obtained from experiments conducted in real

indoor environments. The effectiveness of the proposed configuration is

demonstrated by comparing the basic characteristics, when antennas are

selected according to the conventional full-digital configuration and the

proposed analog–digital hybrid configuration.

Keywords: multiuser MIMO, distributed antenna system, antenna selec-

tion, channel capacity, zero-forcing

Classification: Antennas and Propagation

References

[1] G. J. Foschini and M. J. Gans, “On limits of wireless communications in afading environment when using multiple antennas,” Wirel. Pers. Commun.,

© IEICE 2018DOI: 10.1587/comex.2018XBL0012Received January 23, 2018Accepted February 9, 2018Publicized February 23, 2018Copyedited May 1, 2018

154

IEICE Communications Express, Vol.7, No.5, 154–159

vol. 6, no. 3, pp. 311–335, Mar. 1998. DOI:10.1023/A:1008889222784[2] Q. H. Spencer, C. B. Peel, A. L. Swindlehurst, and M. Haardt, “An introduction

to the multi-user MIMO downlink,” IEEE Commun. Mag., vol. 42, no. 10,pp. 60–67, Oct. 2004. DOI:10.1109/MCOM.2004.1341262

[3] D. Gesbert, M. Kountouris, R. W. Heath, Jr., C. B. Chae, and T. Salzer, “Shiftingthe MIMO paradigm,” IEEE Signal Processing Magazine, vol. 24, no. 5,pp. 36–46, Sep. 2007. DOI:10.1109/MSP.2007.904815

[4] A. Ghosh, R. Ratasuk, B. Mondal, N. Mangalvedhe, and T. Thomas, “LTE-advanced: Next-generation wireless broadband technology,” IEEE WirelessCommun., vol. 17, no. 3, pp. 10–22, June 2010. DOI:10.1109/MWC.2010.5490974

[5] B. Bellalta, “IEEE 802.11ax: High-efficiency WLANS,” IEEE WirelessCommun., vol. 23, no. 1, pp. 38–46, Feb. 2016. DOI:10.1109/MWC.2016.7422404

[6] E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzetta, “Massive MIMOfor next generation wireless systems,” IEEE Commun. Mag., vol. 52, no. 2,pp. 186–195, Feb. 2014. DOI:10.1109/MCOM.2014.6736761

[7] E. G. Larsson, “Very large MIMO systems,” Proc. in IEEE ICASSP 2012Tutorial Text.

[8] F. Rusek, D. Persson, B. K. Lau, E. G. Larsson, T. L. Marzetta, O. Edfors, andF. Tufvesson, “Scaling up MIMO - Opportunities and challenges with very largeMIMO -,” IEEE Signal Processing Magazine, vol. 30, no. 1, pp. 40–60, Jan.2013. DOI:10.1109/MSP.2011.2178495

[9] J. Hoydis, S. ten Brink, and M. Debbah, “Massive MIMO in the UL/DL ofcellular networks: How many antennas do we need?” IEEE J. Sel. AreasCommun., vol. 31, no. 2, pp. 160–171, Feb. 2013. DOI:10.1109/JSAC.2013.130205

1 Introduction

Multiuser multiple-input multiple-output (MU-MIMO) transmissions have been

attracting much attention, as a technique for improving the channel capacity of the

entire system by generating a large virtual channel between the access point (AP)

and multiple user terminals (UTs) [1, 2, 3].

Downlink MU-MIMO transmissions have been introduced as one of the main

technologies in the LTE and IEEE 802.11ac standards. Uplink MU-MIMO trans-

mission is currently being standardized under LTE-advanced [4]. In IEEE 802.11ax

[5], the standardization of uplink MU-MIMO and orthogonal frequency division

multiple access (OFDMA) has been decided, and it is expected that their impor-

tance will increase in the future.

Massive MIMO transmissions have been attracting much attention in 5th

generation mobile communication systems (5G) and IEEE 802.11ay, in order to

improve the performance of MU-MIMO transmissions [6, 7, 8, 9].

In massive MIMO transmission, the antenna configuration used for communi-

cation is usually selected by the digital part is adopted. The signal processing

becomes complicated because of the presence of many antennas. In this config-

uration, when massive MIMO transmission is assumed, the scale of the access point

becomes large, as receivers are required for all antennas. The increase in the access

point size is problematic in terms of power consumption and cost.


155


http://dx.doi.org/10.1023/A:1008889222784

http://dx.doi.org/10.1023/A:1008889222784

http://dx.doi.org/10.1109/MCOM.2004.1341262




http://dx.doi.org/10.1109/MSP.2007.904815




http://dx.doi.org/10.1109/MWC.2010.5490974
















http://dx.doi.org/10.1109/JSAC.2013.130205




In this paper, we propose a hardware configuration for uplink MU-MIMO

transmission, to solve this problem. In the proposed configuration, the distributed

antennas are first divided into multiple groups. Then, the antennas with the highest

received powers are selected using amplitude detectors. Finally, the antennas are

selected using many-to-one switches. This configuration is named the analog–

digital hybrid configuration. The proposed configuration is suitable for implemen-

tation in massive MIMO transmission. In order to clarify the effectiveness of the

proposed configuration, we performed computer simulations using the propagation

channels obtained through experiments conducted in real indoor environments.

The remainder of this paper is organized as follows. Section 2 describes the

conventional and proposed configurations and explains the advantages of the

proposed configuration. Section 3 shows the measurement environment and meas-

urement parameters. In Section 4, the effectiveness of the proposed configuration is

demonstrated by comparing the basic characteristics when antennas are selected

according to the conventional full-digital configuration and the proposed analog–

digital hybrid configuration. The paper is concluded in Section 5.

2 Conventional configuration and proposed configuration

Fig. 1(a) shows the conventional access point (AP) configuration of the DAS in

uplink MU-MIMO. In the conventional configuration, the antennas used for

communication are selected through digital signal processing. Down converters

and analog-to-digital (A/D) converters are necessary for the receivers, and re-

ceivers are connected to all antennas. In the case of massive MIMO, the required

number of receivers increases as the number of antennas increases, and the

hardware is scaled up.

Fig. 1(b) shows the proposed analog–digital hybrid configuration. The pro-

posed configuration uses many-to-one switches and amplitude detectors. The N

antennas are divided into NS groups. NL represents the number of antennas in each

group. All antennas in a group are connected to one many-to-one switch. Therefore,

in the proposed configuration, one antenna is selected by each switch. The received

signal of each group is input to the amplitude detector via the directional coupler.

Fig. 1. Hardware configuration


156


The amplitude information is measured by the amplitude detectors, and the antenna

with the maximum amplitude is selected by each many-to-one switch. In this

configuration, one receiver is connected to one many-to-one switch. Therefore, the

required number of receivers is reduced from N to NS, when compared to the

conventional AP configuration, and the hardware scale is reduced. Compared to

the conventional configuration in which the antenna is selected in the digital part, in

the proposed configuration, the antenna is selected in the analog part by the many-

to-one switch, and thus, the antenna selection process becomes simple.

3 Antenna selection methods and measurement environment

Fig. 2(a) show the real indoor measurement environment. In Fig. 2(a), the blue

circles are the UTs, and the AP antennas are represented by the colors A, B, C, and

D. Each color indicates one group of AP antennas. The number of AP antennas is

32 and the number of UTs is 32. From 32 users, four users are selected for each

trial. A total of 35,960 trials are conducted for different UT combinations. At this

time, the numbers 1∼4 are assigned to the UTs. Then, #UT 1∼4 are paired with

groups A∼D. A single antenna with the highest signal-to-noise ratio (SNR) is

selected from each group (A, B, C, and D at the AP in Fig. 2(a)): four AP antennas

are selected in total in Fig. 2(a). The resulting 4 � 4 uplink MU-MIMO trans-

mission is evaluated in Fig. 2(a).

Fig. 2(b) shows our measurement parameters. The carrier frequency and

bandwidth are 2.425GHz and 20MHz, respectively. The transmission signal is

an orthogonal frequency division multiplexed (OFDM) signal. The transmission

power is 0 dBm. The transmitters and receivers are connected with the UTs and AP

antennas, respectively, and the uplink channel state information (CSI) between the

UTs and AP antennas is measured. The heights of the AP and UT antennas are 2.3

and 0.7m, respectively. Patch and sleeve antennas are used for the AP and UT

antennas, respectively. The thermal noise power is determined so that the average

received power versus the thermal noise power, i.e., SNR, is 20 dB when consid-

ering all the antennas and UTs.

We now evaluate the Shannon capacity and achievable bit rate using zero

forcing (ZF) while considering the actual propagation channel. The Shannon

capacity in the uplink channel Cs can be written as:

Fig. 2. Measurement environment and parameter


157


Cs ¼Xk4

log2 1 þ P

4�2�k

� �; ð1Þ

where �k is the k-th eigenvalue, and P and �2 are the transmit power and noise

power, respectively. The achievable bit rate using ZF CZF can be written as:

CZF ¼Xk4

log2 1 þ SNRZF;k

4

� �; ð2Þ

where SNRZF;k is the k-th SNR obtained from the interference cancellation due to

the ZF for the k-th user.

In the proposed configuration, the antennas are selected in the analog section.

Using the signal sent to the amplitude detector via the directional coupler, the

antenna with the highest SNR is selected. The selected antennas are switched using

the many-to-one switch. This is named DASH.

In the conventional configuration, the antennas are selected in the digital part.

Therefore, the antennas with the highest SNR are selected from each group (A, B,

C, and D) at each subcarrier of OFDM, similar to DASH. This is named

DASH (sub).

4 Comparison of channel capacity

Fig. 3 shows the transmission rates and channel capacities of DASH and

DASH (sub). The solid line represents the transmission rate when ZF is applied

to DASH (sub) and DASH, and the dashed line represents the channel capacity

when applying the Shannon limit to DASH (sub) and DASH.

In the conventional configuration DASH (sub), antennas are selected in the

digital part; therefore, antennas can be selected for each subcarrier of OFDM. It is

thus possible to select a more optimal antenna, when compared to the proposed

configuration that selects the antenna in the analog part. The transmission rate is

also high, accordingly. This difference can also be confirmed from the graph.

However, since the difference is very small, the performance degradation of DASHis also small, when compared to that of DASH (sub).

Fig. 3. Achievable bit rate (DASH and DASH (sub))


158


Antenna selection in the conventional configuration is a very complicated

process, and the scale of the base station increases on the hardware side. On the

other hand, the antenna selection process by the proposed configuration is relatively

simple. Furthermore, in the proposed configuration, the scale of the AP can be

reduced by using many-to-one switches. From this, it can be confirmed that the

proposed configuration is very effective as a realistic hardware configuration in

which the performance hardly deteriorates.

5 Conclusion

In this paper, we proposed an analog–digital hybrid hardware configuration using

amplitude detectors and many-to-one switches for uplink MU-MIMO transmission

with DAS.

In the proposed configuration, the distributed antennas were divided into

multiple groups. Each group of antennas were connected to a many-to-one switch,

and an amplitude detector selected the antenna with the highest received power in

the group. The selected antennas were switched by each many-to-one switch.

Compared to the conventional full-digital hardware configuration, the proposed

configuration could reduce the size of the AP by using many-to-one switches. In the

conventional configuration, the processing becomes complicated as many antennas

are used for communication. On the other hand, in the proposed configuration, the

antennas can be selected through relatively simple processing.

In this study, we experimented in real indoor environments assuming DAS and

acquired the propagation channel. From the obtained propagation channel, the

performance was evaluated by comparing the conventional configuration and the

proposed configuration, in terms of the transmission rate and channel capacity.

From the results, it was confirmed that the performance degradation in the proposed

configuration was small, when compared with the conventional configuration

selecting antennas for each subcarrier. It was shown that the proposed configuration

was effective as it could be realized through simple calculations and the sizes of AP

devices could be reduced.

Acknowledgments

We would like to thank the members of the Nishimori Laboratories and NTT

Access Network Service Systems Laboratories for their assistance in the indoor

measurements. Part of this work was supported by KAKENHI, Grant-in-Aid for

Scientific Research (B) (17H03262).


159


Clarification ofaccommodatable number offunctional split base stationsin TDM-PON fronthaul

Daisuke Hisanoa), Hiroyuki Uzawa, Yu Nakayama,Kenji Miyamoto, Hirotaka Nakamura, Jun Terada,and Akihiro OtakaNTT Access Network Service Systems Laboratories, NTT Corporation,

1–1 Hikarinooka, Yokosuka-shi, Kanagawa 239–0847, Japan


Abstract: Mobile fronthaul (MFH) is the optical link between a central

unit (CU) and a distributed unit (DU) in centralized radio access network

(C-RAN) architecture. To suppress the cost of MFH, we have studied MFH

networking based on a time division multiplexed passive optical network

(TDM-PON). We have proposed low-latency uplink forwarding with a

mobile dynamic bandwidth allocation (M-DBA) scheme. When the M-

DBA scheme is employed, the uplink latency can be suppressed compared

with fixed bandwidth allocation (FBA). In this paper, we clarify the number

of ONUs that can be accommodated by the M-DBA scheme when also

accommodating a novel functional split based C-RAN.

Keywords: TDM-PON, mobile fronthaul, dynamic bandwidth allocation,

5G

Classification: Fiber-Optic Transmission for Communications

References

[1] 3GPP, TR 38.801, v1.0.0, “Study on new radio access technology; Radioaccess architecture and interfaces (Release 14),” Dec. 2016.

[2] T. Tashiro, S. Kuwano, J. Terada, T. Kawamura, N. Tanaka, S. Shigematsu, andN. Yoshimoto, “A novel DBA scheme for TDM-PON based mobile fronthaul,”Optical Fiber Communication Conference, paper Tu3F.3, Mar. 2014. DOI:10.1364/OFC.2014.Tu3F.3

[3] Y. Nakayama, K. Maruta, T. Shimada, T. Yoshida, J. Terada, and A. Otaka,“Utilization comparison of small-cell accommodation with PON-based mobilefronthaul,” J. Opt. Commun. Netw., vol. 8, no. 12, pp. 919–927, Dec. 2016.DOI:10.1364/JOCN.8.000919

[4] H. Nomura, H. Ou, T. Shimada, T. Kobayashi, D. Hisano, H. Uzawa, J. Terada,and A. Otaka, “First demonstration of optical-mobile cooperation interface formobile fronthaul with TDM-PON,” IEICE Commun. Express, vol. 6, no. 6,pp. 375–380, Jun. 2017. DOI:10.1587/comex.2017XBL0030

[5] H. Uzawa, H. Nomura, T. Shimada, D. Hisano, K. Miyamoto, Y. Nakayama, K.Takahashi, J. Terada, and A. Otaka, “Practical mobile-DBA schemeconsidering data arrival period for 5G mobile fronthaul with TDM-PON,”


160


http://dx.doi.org/10.1364/OFC.2014.Tu3F.3





http://dx.doi.org/10.1364/JOCN.8.000919




http://dx.doi.org/10.1587/comex.2017XBL0030



Proc. of IEEE/OSA European Conference on Optical Communication(ECOC), pp. M. 1. B. 2, Sep. 2017.

[6] CPRI, “eCPRI specification V1.0,” http://www.cpri.info/spec.html, AccessedDec. 19, 2017, Aug. 2017.

[7] CPRI, “CPRI specification V7.0,” http://www.cpri.info/spec.html, AccessedFeb. 8, 2018, Oct. 2015.

[8] G. Piro, N. Baldo, and M. Miozzo, “An LTE module for the ns-3 networksimulator,” Proc. of the 4th International ICST Conference on Simulation Toolsand Techniques. ICST (Institute for Computer Sciences, Social-Informatics andTele-communications Engineering), pp. 415–422, Mar. 2011.

[9] ns-3, available: http://www.nsnam.org/.[10] 3GPP, TR 36.814, v9.2.0, “Evolved universal terrestrial radio access (E-

UTRA); Further advancements for E-UTRA physical layer aspects (Release9),” Mar. 2017.

1 Introduction

A centralized RAN (C-RAN) architecture is employed for a mobile base station

(MBS) to reduce costs and make MBSs cooperate each other. With the C-RAN

architecture, a radio frequency (RF) functional block and part of the physical

processing unit are located in a distributed unit (DU) at the antenna site [1]. The

other upper layer processing units are consolidated in a centralized unit (CU). The

transmission link between the CU and the DU is connected by an optical fiber and

called mobile fronthaul (MFH). MFH networking has been widely studied to limit

the MFH link cost, and we have studied MFH networking based on a time division

multiplexed passive optical network (TDM-PON) [2, 3]. Specifically, we have

proposed low-latency uplink forwarding with a mobile dynamic bandwidth allo-

cation (M-DBA) scheme. When the M-DBA scheme is employed, the optical line

terminal (OLT) and the DU cooperate as regards the uplink transmission through a

dedicated interface defined as a cooperative interface (C-IF) [4]. Moreover, we have

demonstrated experimentally that the M-DBA can reduce uplink forwarding

latency compared with the conventional fixed bandwidth allocation (FBA) scheme

[5]. However, the number of DUs that can be accommodated has not been

estimated. In general, the cost reduction becomes greater as we increase the number

of accommodatable optical network units (ONUs) connected to the DUs. In this

paper, we clarify the number of ONUs that can be accommodated based on

statistical information regarding the amount of MFH data, with the M-DBA or

the conventional FBA when accommodating functional split based MBSs [1, 6].

2 Overview of mobile-DBA scheme

The bandwidth allocation scheme is detailed in [2], [4], and [5]. The CU calculates

a wireless schedule for a wireless uplink transmission from user equipment (UE).

The CU then allocates the wireless bandwidth to the UE. After 4 transmission time

intervals (TTIs), the UE transmits a wireless signal to the DU. With a long term

evolution (LTE) system, 1 TTI equals one millisecond. In the M-DBA scheme, the

CU transmits the calculated wireless scheduling information to the OLT through the


161


http://www.cpri.info/spec.html








http://www.nsnam.org/



C-IF. The OLT calculates the bandwidth for each ONU based on the wireless

scheduling information and allocates the bandwidth to each ONU. After 4 TTIs,

when the DU forwards the MFH signal to the ONU, the ONU has already been

allocated a suitable bandwidth and can forward its signal to the OLT without any

transmission waiting time. Thus the transmission waiting latency is minimized.

3 Functional split-based MBS

If the MBS employs a conventional functional split point (e.g. common public

radio interface (CPRI) [7]), the MFH link will have to cope with a huge amount of

data in 5G era. Therefore a novel functional split point has been discussed [1, 6] to

reduce the amount of MFH data. In [1], there are several candidate functional split

points. Fig. 1(a) shows the functional blocks with the novel functional split point.

The option number in Fig. 1(a) refers to [1]. A MAC-PHY split (option 6) and an

Intra-PHY split (option 7–2) are particularly suitable with TDM-PON since the

MFH traffic is packetized in a layer-2 frame. For options 6 and 7–2, the amount of

MFH data is following,

Dmfh ¼NlayStbs ðif option 6Þ (1a)

NlayNiqNqNrbNscNsym ðif option 7{2Þ (1b)

(

where Nlay is the number of multiple-input and multiple-output (MIMO) layers, Stbsis the transport block size, Niq is the number of in- and quadrature-phase parts

(¼ 2), Nq is the number of quantization bits of the in- and quadrature-phase parts,

Nrb is the number of resource blocks in a wireless system bandwidth, Nsc is the

number of subcarriers in a resource block, and Nsym is the number of orthogonal

frequency division multiplexed (OFDM) symbols in a physical uplink shared

channel (PUSCH).

In addition, the MFH link has bursty traffic distribution characteristics. With

option 6 in Fig. 1, the wireless signal is demodulated and decoded with respect to

each TTI. The bursty MFH traffic is periodically generated at every TTI. For option

7–2, two types of bursty traffic distribution are conceivable. First is bursty MFH

traffic with a resource block interval. When a wireless signal is forwarded by every

resource block, the interval of the bursty MFH traffic is 0.5 TTI. The other is the

bursty MFH traffic generated by every OFDM symbol. For example, for an LTE

system, 1 TTI is equivalent to 1 millisecond and 14 OFDM symbols are transmitted

during 1 TTI. As a result, bursty MFH traffic with an interval of approximately

71.4 µs is generated. Fig. 1(b)–(d) summarizes the three types of bursty MFH

traffic distribution.

4 Calculation method and analysis result

We estimate the probability P that all DUs generate MFH signals at the same time

without exceeding the latency requirement Treq. We assume the three types of MFH

traffic distribution shown in Fig. 1(b)–(d). The estimation method is divided into

two steps. First we calculate the probability mass function of the discrete proba-

bility distribution for the amount of MFH data. Then we calculate the number of

accommodatable DUs taking the M-DBA into consideration.


162


4.1 Probability mass function for MFH data amount

We use the LTE module [8] of a network simulator 3 (ns-3) [9]. The LTE module

generates the TBS and modulation coding scheme (MCS) information per TTI and

we simulate multiple times with different UE deployment locations and different

mobile traffic generation conditions. This is because the wireless throughput

changes depending on the propagation environment between the DU and the UE.

And, in [10], the parameter for the number of the UEs is described as 2, 5, 8, 10,

and 14. Since the cell size was smaller than that in [10], in this paper we assumed 4

UEs as half of the median described in [10]. We then converted the TBS and MCS

information about the LTE system into MFH data assuming a 5G system in

accordance with Eq. (1). We produced a histogram of the MFH data. The class

of the histogram indicates the MFH data volume vmfh. The frequency of the

histogram is equivalent to the discrete probability pðvmfhÞ. We assume that all the

DUs generate an MFH signal according to its histogram.

4.2 Number of accommodatable DUs based on M-DBA

We define the DU identifier as i, and the number of ONUs (¼ DUs) as Nonu. The

request size ri of the uplink transmission from the CU to the OLT through C-IF is

equivalent to the MFH data volume vmfh. We need to calculate the transmission

waiting time Di for all the DUs. The allocated bandwidth bi [byte/polling cycle]

is as follows,

bi ¼ riXNonu

k¼1rk

Cpon; ð2Þ

where the bi unit is bytes per polling cycle and Cpon is the PON link capacity. The

polling cycle means the uplink transmission interval of the TDM-PON system. The

number of polling cycles Nipoll to forward the request size ri is,

Nipoll ¼ ceil

ribi

� �: ð3Þ

ceilðxÞ represents a ceiling function. Therefore the transmission waiting time Di is,

Di ¼ NipollTpoll � Li þ Tfix; ð4Þ

where Tpoll is the polling cycle and Li is the original frame length. Tfix is the fixed

latency caused by processing delay and frequency/synchronization errors. Next, we

(a)

(d)

(b)

(c)

Fig. 1. Functional split point [1]. (a) Functional blocks, burstyMFH traffic distribution with (b) TTI, (c) resource block, and(d) OFDM symbol intervals.


163


calculate the probability P.

P ¼XMs1¼0

XMs2¼0

� � �XM

sNonu¼0ðp1ðrs1Þp2ðrs2Þ � � �pNonuðrsNonu

ÞÞ; ð5Þ

ðs1 ≠ s2 ≠ � � � ≠ sNonuÞwhere the identifier si is the class of the histogram and M is the maximum value of

its class. The number of summations depends on the number of ONUs Nonu. When

the element p1ðrs1Þp2ðrs2Þ � � �pNonuðrsNonuÞ is not under the following condition, the

calculation result is excluded from P in Eq. (6).

XNonu

i¼1Di < Treq: ð6Þ

We calculate the probability P for different Nonu values according to Eq. (3)–(6).

4.3 Calculation result and discussion

The calculation parameters are shown in Table I. The exponential distribution

model is defined in [10]. For the exponential distribution model, the average burst

Table I. (a) Calculation parameters for mobile system.

Item Value

Traffic model Exponential distribution

Number of UEs 2 UEs/DU

Wireless data rate from UE 100Mbps

UE deployment Uniform distribution

Transmission power from UE 23.0 dBm

Noise figure from UE 7.0 dB

Nrb 100

System bandwidth 20MHz/CA

Number of carrier aggregations 5 CAs

Nlay 2

Radius of small cell 10.0m

Minimum distance between UE and DU 1.0m

Simulation time 30.0 s

Number of iterations 1000

Fading model Extended typical urban (ETU) model

Path loss model Urban Micro (UMi)

(b) Calculation parameters for TDM-PON.

Item Value

Cpon 10.0Gbps

Tpoll 31.25 µs

Minimum allocated bandwidth 320 byte/Tpoll

Burst overhead 1024 byte/Tpoll

FEC RS(248,216)© IEICE 2018DOI: 10.1587/comex.2018XBL0008Received January 22, 2018Accepted February 5, 2018Publicized February 26, 2018Copyedited May 1, 2018

164


duration was 1.0 s and the average burst interval was 5.0 s. The latency requirement

Treq was set at 150µs. To compare the performance, we undertook the calculation

when employing an FBA scheme. For the FBA, the bandwidth bi is,

bi ¼ 1

NonuCpon: ð7Þ

We show the calculation result and the histogram in Fig. 2(a)–(f ). From Fig. 2(a),

(b), the functional split point is assumed to be option 6. For example, the P value is

improved more than 20% when 8 DUs are accommodated. For option 7–2,

Fig. 2(e), (f ) show the bursty MFH traffic distribution with a resource block

interval. Fig. 2(c), (d) show the bursty MFH traffic distribution with an OFDM

symbol interval. Since we expect statistical multiplexing effects, we assumed that it

was not necessary for the result to be 100%. The probability threshold value Pth for

determining the number of accommodatable ONUs is set at 90%. Then, from

Fig. 2(c), (d), the number of ONUs that can be accommodated is improved from 2

to 4. Moreover, for Fig. 2(e), (f ), the number is improved from 4 to 6.

5 Conclusion

We calculated the number of accommodatable ONUs when our proposed M-DBA

is applied. We evaluated the number of ONUs based on the mobile traffic generated

by the LTE module of the network simulator 3. For option 6, the probability is

improved by more than 20% when 8 DUs are accommodated. For option 7–2, the

accommodation efficiency is improved 1.5 and 2.0 times in the cases of a bursty

(c)

(a)

(d)

(b)

(f)(e)

Fig. 2. Calculation result. (a) Probability P and (b) histogram foroption 6, (c) and (d) for option 7–2 with OFDM symbolinterval, and (e) and (f ) for option 7–2 with resource blockinterval.


165


MFH traffic distribution with OFDM symbol and resource block intervals, respec-

tively when the probability threshold value Pth is set at 90%.


166


Second-order spectrumextraction based onionospheric clutter

Anyun Yina) and Zili Lib)

School of Electronic Engineering, GuangXi Normal University,

Guilin City, the Guangxi Zhuang Autonomous Region, 541004, China


b) [email protected]

Abstract: The reflection and scattering effects of ionosphere can cause

clutter interference to the receiving signal of high frequency surface wave

radar (HFSWR), which makes it difficult to analyze the spectrum of signal

echo. In order to separate the second-order spectrum with wave information

component from spectrum containing ionospheric clutter interference, we

propose a hybrid echo signal processing algorithm. That is, analyzing the

ionospheric signal characteristics for a specific water’s latitude and on the

basis of the existing elimination algorithm, according to the characteristics

of large scale steady state of the ocean, we introduce parameters such as

coherence time and distance value, making an intensive study of the diving

method of second-order spectrum area, and obtaining a new extraction

algorithm with regional characteristics and strong pertinence. Experimental

results show that this method is effective.

Keywords: ionospheric clutter, mixed echo, signal processing, specific sea

area, second-order spectrum extraction

Classification: Fundamental Theories for Communications

References

[1] S. T. Gille, “An introduction to ocean remote sensing,” Eos Trans. Am.Geophys. Union, vol. 86, no. 12, p. 125, 2005. DOI:10.1029/2005EO120009

[2] R. Shah and J. L. Garrison, “Application of the ICF coherence time method forocean remote sensing using digital communication satellite signals,” IEEE J.Sel. Topics Appl. Earth Observ. in Remote Sens., vol. 7, no. 5, pp. 1584–1591,2014. DOI:10.1109/JSTARS.2014.2314531

[3] B. Zhang and Y. J. He, “Research progress of the ocean remote sensinginformation extraction technology under high sea states,” J. Ocean Technol.,2015.

[4] W. Huang and E. W. Gill, “An alternative algorithm for wave informationextraction from X-band nautical radar images,” IET International RadarConference 2013, pp. 1–5, 2013. DOI:10.1049/cp.2013.0298

[5] T. Wen-Long, L. Gao-Peng, and X. Rong-Qing, “Ionospheric clutter mitigationfor high-frequency surface-wave radar using two-dimensional array and beamspace processing,” IET Radar Sonar Navig., vol. 6, no. 3, pp. 202–211, 2012.DOI:10.1049/iet-rsn.2011.0121

[6] Y. Su, Y. Wei, R. Xu, and Y. Liu, “Ionospheric clutter suppression using

© IEICE 2018DOI: 10.1587/comex.2017XBL0183Received November 30, 2017Accepted February 5, 2018Publicized March 12, 2018Copyedited May 1, 2018

167


http://dx.doi.org/10.1029/2005EO120009

http://dx.doi.org/10.1029/2005EO120009

http://dx.doi.org/10.1109/JSTARS.2014.2314531




http://dx.doi.org/10.1049/cp.2013.0298

http://dx.doi.org/10.1049/cp.2013.0298

http://dx.doi.org/10.1049/cp.2013.0298

http://dx.doi.org/10.1049/cp.2013.0298

http://dx.doi.org/10.1049/iet-rsn.2011.0121




wavelet oblique projecting filter,” IEEE Conference on Radar 2017, pp. 1552–1556, 2017. DOI:10.1109/radar.2017.7944454

[7] M. Ravan, R. J. Riddolls, and R. S. Adve, “Ionospheric and auroral cluttermodels for HF surface wave and over-the-horizon radar systems,” Radio Sci.,vol. 47, no. 3, pp. 1–12, 2012. DOI:10.1029/2011rs004944

[8] X. L. Chu, J. Zhang, J. I. Yong-Gang, et al., “The second order spectrumextraction from high frequency ground radar sea echoes based on the outline ofthe RD spectrum,” Adv. Marine Sci., 2016.

[9] B. Wang, G. Zhang, L. I. Zhi, et al., “Wavelet threshold denoising algorithmbased on new threshold function,” J. Comput. Appl., vol. 34, no. 5, pp. 1499–1502, 2014.

[10] W. H. Zhu, W. An, L. H. You, et al., “Wavelet threshold denoising algorithmbased on improved wavelet threshold function,” Appl. Comput. Syst., vol. 25,no. 6, pp. 191–195, 2016.

[11] A. M. Atto, D. Pastor, and G. Mercier, “Wavelet shrinkage: Unification of basicthresholding functions and thresholds,” Signal Image Video Process., vol. 5,no. 1, pp. 11–28, 2011. DOI:10.1007/s11760-009-0139-y

1 Introduction

The use of high frequency surface wave radar (HFSWR) in the technology of

remote sensing [1, 2] for detecting the ocean can be seen in recent years, the sea

state information associated with the waves is mainly included in the two spectral

range of the radar echo spectrum, the existing ocean wave information acquisition

technology mainly aims at the analysis and processing of the second-order

spectrum and the extraction of the characteristic parameters to obtain the related

ocean information, such as the effective wave height and the ocean wave spectrum

[3, 4]. HFSWR is affected by ionospheric clutter with different morphological

characteristics in time, space and frequency bands, it shows obvious non-stationary

signal characteristics in frequency spectrum, and its energy is so strong that it often

completely drowns out useful ocean echo spectrum [5, 6, 7]. So it is difficult

to extract the second-order spectrum of ocean echo on the distance element with

ionospheric signal interference, and then it is a great obstacle to the subsequent sea

state information retrieval. Therefore, how to separate the second-order spectra

from the ionospheric echo spectrum effectively is the key to obtain the sea state

information of the region. To resolve the above problem, this paper proposes a

hybrid echo signal processing algorithm. The method of dividing the second-order

spectral region under ionospheric clutter is studied in-depth, and the validity of the

algorithm is verified by the measured data. The new algorithm solves the problem

of two order spectrum extraction under ionospheric interference, and improves the

extraction accuracy, which lays the foundation for the next research.

2 The present situation of second-order spectrum extraction

In the Doppler spectrum of radar signal echo, the second-order spectrum is usually

distributed on both sides of the first order spectrum. The signal-to-noise ratio (SNR)

is obviously lower than that of the first order spectrum, and has a certain spectrum

range. Because of the noise, clutter and other kinds of interference signals in


168


http://dx.doi.org/10.1109/radar.2017.7944454




http://dx.doi.org/10.1029/2011rs004944

http://dx.doi.org/10.1029/2011rs004944

http://dx.doi.org/10.1007/s11760-009-0139-y

http://dx.doi.org/10.1007/s11760-009-0139-y

Doppler spectrum, the difficulty of extracting second-order spectra is increased. At

present, the extraction of the second-order spectra is mostly aimed at the Doppler

spectrum without ionospheric disturbances and it is extracted by constructing range

Doppler (RD) spectral contour [8]. However, when dealing with the Doppler

spectrum with ionospheric disturbances, the interference signal has obvious influ-

ence on the outline of the RD spectrum, and cannot reflect the general location of

the second-order spectrum.

3 Second-order spectrum extraction under ionospheric clutter

Since the study of second-order spectrum extraction usually takes place without the

ionosphere interference, in the case of ionospheric disturbances, the previous

extraction methods have great limitations, for example, the peaks of the second-

order spectrum and the second-order spectral boundary cannot be determined, and

the second-order spectrum is completely disturbed by the ionosphere. In order to

solve the second-order spectrum extraction in the background of ionospheric

interference, this paper attempts to improve the probability and accuracy of the

second-order spectrum extraction from the spectrum of ionospheric interference

from three directions as follows.

3.1 The prolongation of coherent time

In order to improve the accuracy of the second-order spectrum extraction, the paper

pretreated the echo data and extended the coherence length of the data to achieve

the purpose of broadening the spectrum. According to the initial coherence time of

the echo data, the cumulative increment of the integer double data acquisition time

is performed.

N ¼ t � k ð1ÞIn the upper formula, N is the coherence length of the extended data, and t is the

minimum cumulative time, and k is the coherent multiple.

3.2 The cancellation of ionospheric clutter

In HFSWR echo data, the first-order peak signal information mainly contained

in the low frequency component of the scale coefficient, the ionospheric echo

information is present in the high frequency component of the wavelet coefficients.

The wavelet threshold contraction algorithm [9] will deal with the interference

signal with strong correlation in the wavelet coefficients, and it will be normalized

to the elimination result.

Similar to [10], the threshold function uses the following expression:

ypðxÞ ¼ sgnðxÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðjxjj � �jÞj

q; jxj � �

0; jxj � �

8<: ð2Þ

Adopted from [11], the threshold selection uses the following expression:

� ¼ �ffiffiffiffiffiffiffiffiffiffiffiffiffiffi2 lnðNÞp

lnðj þ 1Þ ð3Þ

j is the decomposition scale.


169


3.3 The choice of threshold for signal-to-noise ratio of second-order

spectrum boundaries

The threshold is used to judge the boundary of the second-order spectrum, and the

selection of the threshold must have some ambiguity after the ionospheric elimi-

nation operation. In order to solve the problem of threshold ambiguity, applying

two order boundary spectrum average value dynamic calibration of adjacent

distance element without ionospheric disturbance, and reference spectra of the first

substrate noise and signal-to-noise ratio, to determine the second-order spectrum

threshold.

SNRave ¼ 1

n

Xni¼1

SNRi ð4Þ

SNRave is the mean value of the second-order spectrum boundary; SNRi is the

signal-to-noise ratio of the ith data for the second-order spectrum boundaries.

4 Second-order spectrum extraction and analysis

The experimental data of this paper are taken from the spectrum of a distance

element at 11 am of September 26, 2013 in the Beibu Gulf (08° 13.30B E, 21°

30.30B N) as shown in Fig. 1(a).

4.1 Coherent accumulation of ionospheric elimination

In the time to eliminate the interference on the ionosphere, considering the iono-

spheric cancellation algorithm that will cause loss of signal-to-noise ratio for the

echo spectrum data, due to less number of points in every game, it is easy to cause

the cancellation after the second-order spectrum is not obvious or cancellation of

the elimination of the ionosphere, which is on the order of two subsequent

extractions caused by the spectrum great influence. In order to deal with this

problem, this paper uses the method of increasing coherent accumulation time, and

uses more coherent data to reduce the influence of the ionospheric cancellation

algorithm on the second-order spectrum.

From the comparison of Fig. 1(b)∼(c), it is observed that the spectrum of the

data at 256 sampling points is significantly less than that of the 1024 sampling

points after coherent integration. In the process of ionospheric elimination, it is

obvious that the possibility of extracting two order spectra before coherent

accumulation is relatively small. However, after the coherent accumulation,

although the ionospheric cancellation algorithm has affected the signal-to-noise

ratio of the two order spectrum, the range of the approximate data points contained

in the two order spectrum can be distinguished from the spectrum, which reflects

the advantages of the more data sampling points.


170


4.2 The extraction of second-order spectrum extraction parameters

After the coherent accumulation, the signal-to-noise ratio of the spectrum changes

with the ionosphere elimination. In this paper, according to the state parameters of

the ocean echo spectrum of the distance elements which are not affected by the

ionospheric disturbances as the reference, the two order spectra are divided after the

ionospheric disturbances are eliminated. The Table I gives the reference mean

values of the related main parameters of adjacent interference free range elements,

and takes the transformation data before and after the first peak cancellation as the

reference basis. The acquisition of the table data parameters is calculated by the

echo data received within one hour, and two distance elements and two adjacent

distance elements are calculated.

4.3 Extraction results

According to the statistics of the above parameters, echo spectrum correlation

parameters are based on adjacent undisturbed range elements, and combined with

the influence of ionospheric cancellation algorithm on SNR of echo spectrum,

ultimately they determine the ionosphere on consumption, increase the coherent

time spectrum of second-order boundary spectral signal-to-noise ratio, and then the

scope of second-order spectrum is determined, finally, according to the peak

parameters of second-order spectrum, the first-order peak area and two spectral

region is separated. The Fig. 2 is the result of the second-order spectral partition of

the coherent accumulation time spectrum after the above processing.

(a)

(b) (c)

Fig. 1. The spectra of two groups before and after coherence


171


5 Conclusion

Ionospheric disturbances are one of the most difficult problems to be solved in the

radar signal analysis, just in a resource limited environment, analysis of echo signal

of the ionospheric disturbance under the conditions of the ionosphere and the

elimination of the echo spectrum in a certain extent. In this paper, the good weather

conditions and the characteristics of the states in a specific area are basically the

same, by increasing the coherent time, and combining with the spectrum parameters

of adjacent distance elements without ionospheric disturbances, we develop the

second-order spectrum feature analysis and extraction using echo spectrum with

ionospheric interference by mathematical statistical analysis and approximate

method. Finally, the calculation results show that the echo spectrum of ionospheric

disturbances can be analyzed effectively and reasonably based on certain ocean

condition and the stability principle.

Fig. 2. The result of two order spectrum extraction

Table I. Reference values of related parameters

DI-LSNR DI-RSNR D-N LSNRave RSNRave LBSNRave RBSNRave

(dB) (dB) (dB) (dB) (dB) (dB) (dB)

6.96 14.22 1.98 33.66 40.09 17.28 12.7

7.42 11.74 4.75 32.57 37.24 10.78 13.73

6.67 10.76 5.34 36.78 17.63 28.79 3.83

8.05 13.97 3.07 31.13 14.73 14.48 2.74

7.14 14.07 3.84 29.60 12.40 12.75 8.64

7.59 11.32 4.91 36.07 17.82 23.63 6.44

5.73 15.48 1.79 35.97 18.52 21.48 11.43

9.03 15.02 2.65 35.64 17.86 28.27 12.13

DI-LSNR: Difference of SNR of left first order peak after clutter suppression.DI-RSNR: Difference of SNR of right first order peak after clutter suppression.D-N: Noise base difference before and after interference suppression.LSNRave: Mean value of SNR of left first order peak without interference.RSNRave: Mean value of SNR of right first order peak without interference.LBSNRave: Mean value of SNR of left boundary without interference.RBSNRave: Mean value of SNR of right boundary without interference.


172


Acknowledgment

The authors would like to thank many teachers for useful suggestions and

discussions. This work was partially supported by the National Natural Science

Foundation of China (61661009) and Innovation Project of Guangxi Graduate

Education (YCSW2017056).


173


Optimization design methodof arbitrarily shaped elementsfor wideband reflectarrays

Shuki Wai, Hiroyuki Deguchia), and Mikio TsujiDept. of Electronics, Doshisha University,

10–3 Tatara-Miyakotani, Kyotanabe, Kyoto 610–0321, Japan


Abstract: In this paper, the authors have proposed an optimization design

method for arbitrarily shaped resonant elements with wideband property and

have constructed an offset-feed reflectarray by the optimized elements in the

X band. Its radiation pattern has been also investigated numerically and

experimentally. Usefulness of the proposed method has been proved from

comparison between the measured radiation pattern and the calculated ones.

Keywords: wideband reflectarray, genetic algorithm, antenna, method of

moments

Classification: Antennas and Propagation

References

[1] C. R. Dietlein, A. S. Hedden, and D. A. Wikner, “Digital reflectarrayconsiderations for terrestrial millimeter-wave imaging,” IEEE Antennas WirelessPropag. Lett., vol. 11, no. 3, pp. 272–275, March 2012. DOI:10.1109/LAWP.2012.2189545

[2] R. E. Hodges, D. J. Hoppe, M. J. Radway, and N. E. Chahat, “Novel deployablereflectarray antennas for CubeSat communications,” Digest 2015 IEEE MTT-SIntern’l Symp., Phoenix AZ, USA, pp. 1–4, May 2015. DOI:10.1109/MWSYM.2015.7167153

[3] T. Asada, S. Matsumoto, H. Deguchi, and M. Tsuji, “Reflectarray with arbitrarilyshaped elements having four-axial symmetry,” Proc. 2014 Asia PacificMicrowave Conf., Sendai, Japan, pp. 1238–1240, FR1G-27, Dec. 2014.

[4] H. Matsumoto, H. Yamada, H. Deguchi, and M. Tsuji, “Reflectarray witharbitrarily shaped elements for linear-to-circular polarization,” Digest 2016Intern’l Symp. Antennas Propagat., Okinawa, Japan, pp. 650–651, D3-5, Nov.2016.

[5] D. Higashi, H. Deguchi, and M. Tsuji, “Omega-shaped resonant elements fordual-polarization and wideband reflectarray,” Digest 2014 IEEE AP-S Intern’lSymp., Memphis TN, USA, pp. 809–810, July 2014. DOI:10.1109/APS.2014.6904733

1 Introduction

Recently, many studies of reflectarrays have been carried out for various applica-

tions such as millimeter imaging systems, and deployable antennas for satellite

© IEICE 2018DOI: 10.1587/comex.2017XBL0191Received December 23, 2017Accepted January 11, 2018Publicized March 12, 2018Copyedited May 1, 2018

174


http://dx.doi.org/10.1109/LAWP.2012.2189545




http://dx.doi.org/10.1109/MWSYM.2015.7167153




http://dx.doi.org/10.1109/APS.2014.6904733




systems [1, 2]. The GA (genetic algorithm) optimization for reflectarrays is one of

useful design methods [3]. However, the conventional GA-optimized reflectarray

with dual-polarization or low cross-polarization properties was limited in the

narrow frequency band because of adopting experiential phase functions. In this

paper, we propose a new GA procedure to expand the frequency bandwidth. Its

usefulness is confirmed by evaluating radiation properties of the designed reflec-

tarray in the X band numerically and experimentally. Our design method proposed

here to get a broadband property could be useful for various reflectarrays, for

example, the circular polarization conversion reflectarray [4], the multiband reflec-

tarray, the reflectarray with resonant elements having four axial symmetry [5] and

so on.

2 Optimization design method

In the conventional GA-optimized method [3], we determine the element geometry

by fitting the frequency property of its reflection phase into a straight line with

some slope over the specified frequency range. This procedure is repeated for the

parallel straight lines with the same slope in the range of 360 degrees at the center

frequency. Whereas, in the proposed optimization, we adopt two-step optimization

and also search the element geometry by fitting its reflection-phase property into a

curved line corresponding to the resonant characteristic of an element. Furthermore,

optimization is not performed under the fitting line one by one, but done under all

the fitting lines at a time. In the first step, we sort the GA-generated reflection-phase

curves into the group of the N units at intervals of 360=N degrees, based on the

phase value at the center frequency, continue this sort until the given number M of

reflection-phase curves gathers every unit, and select the N initial geometries giving

the suitable phase curve in each unit. Next, one of their phase curves is chosen as

the basic fitting curved line and the N parallel fitting phase curves based on it are

set. In the second step, the initial geometries are modified slightly to be matched

with the fitting curves, and this procedure is repeated until fitness of the error

function is satisfied. It should be noted here that the phase curve of the modified

geometry is compared with all the N fitting curves, so that this process makes it

possible to obtain the optimized geometry from the initial geometry of the different

unit. Thus, this proposed optimization can shorten computation time and extend the

specified frequency range. Fig. 1 illustrates an algorithm of the proposed design

procedure.

Fig. 1. Algorithm of the proposed optimization design method.


175


3 Reflection properties of designed elements

The reflection properties are analyzed as infinite array by the method of moments in

spectral domain imposing the periodic boundary condition. The dimension of the

unit cell is 12.0mm, and the thickness of the substrate with relative permittivity

"r ¼ 1:67 is 3.0mm. The strip width of the element is 0.3mm. We design resonant

elements as M ¼ 20 and N ¼ 12 in the frequency range from 6.5GHz to 13.5GHz.

Fig. 2(a) shows each geometries obtained by the optimization. Fig. 2(b) and (c)

shows the calculated reflection phases and amplitude properties of the cross

polarization for the TE incidence, and also the properties for the TM incidence is

very similar to them for the TE one. The phases vary almost in the range of 360

degrees over the specified frequency. The cross polarization level is less than about

−30 dB except a few elements in the specified range. To evaluate degree of parallel

between the reflection-phase curves of the optimized elements, Fig. 2(d) shows

their deviations by normalizing the phase value of each element to 0 degree at the

center frequency 10GHz. You can see from this figure that the deviations is

suppressed within 45 degrees in the higher frequency region and 90 degrees in

the lower one. Fig. 2(e) shows the calculated reflection phases of “Initial Geo-

metries” obtained in the first step for the TE incidence. Although their curves is

similar to those in Fig. 2(a), they are not parallel comparing with Fig. 2(a) and does

not give the phase difference more than 360° in the lower frequency region. So it is

expected that the designed elements are useful for constructing a reflectarray by

their appropriate arrangement.

4 Radiation patterns of reflectarray

To confirm usefulness of the proposed elements, we designed an offset feed

reflectarray antenna with a dimension 180mm � 180mm (15 � 15 cells). Fig. 3(a)

shows the fabricated reflectarray. The offset angle of the primary radiator is 30

Fig. 2. Properties of the designed elements.


176


degrees and the main beam is radiated in the specular direction. The distance from

its phase center to the center of the reflectarray antenna was chosen to be 290mm.

Fig. 3(b) shows comparison of the radiation patterns between the calculated results

and the experimental ones at the center frequency 10GHz. The radiation pattern is

calculated by AFIM (aperture field integration method) and HFSS (Ansys). You

can see from this figure that main-beam pattern agrees well with the calculated

ones. The cross polarization level using AFIM is not shown there, because it is

suppressed less than −50 dB. The measured level is supressed less than −45 dBexcept the specular direction. Fig. 3(c) shows the aperture efficiency from 6GHz

to 16GHz, although the experimental results are given only within the specified

frequency range. The aperture efficiency is kept to the level of more than 50% over

a wide range from 7.5GHz to 15.0GHz. It is clear from this experimental

verification that the proposed optimization design method is useful.

5 Conclusion

We have proposed the GA-optimization method for designing resonant elements of

a broadband reflectarray. By adopting two-step optimization, the desirable resonant

elements with broadband property are obtained efficiently. Then the reflectarray

constructed by using them has been evaluated from radiation property for the dual

polarized wave, numerically and experimentally. As a result, the measured main-

beam pattern agrees well with the calculated one and the cross-polarization is

suppressed in the sufficiently low level. The broadband property of the proposed

reflectarray has been confirmed from the aperture efficiency more than 50% over a

wide range from 7.5GHz to 15.0GHz.

Acknowledgments

This work was supported in part by a Grant-in-aid for Scientific Research (C)

(15K06090) from Japan Society for Promotion of Science.

(a) Photograph of the fabricated reflectarray.

(b) Radiation patterns at 10GHz for the TE incident wave.

(c) Comparison between the calculated and the measured aperture efficiencies for both the incident waves.

Fig. 3. Radiation properties.


177


ieice communications express, vol.7, no.5, 154 performance … · user multiple-input...

Documents