IEICE Communications Express, Vol.6, No.11, 615 Delay ...

Delay-aware sleep control ofoptical network unit in EPONsusing disturbance observer

Takahiro Kikuchia) and Ryogo KuboDepartment of Electronics and Electrical Engineering, Keio University,

3–14–1 Hiyoshi, Kohoku-ku, Yokohama-shi, Kanagawa 223–8522, Japan

a) [email protected]

Abstract: Recently, quality-of-service (QoS)-aware cyclic sleep controllers

for optical network units (ONUs) have been proposed to guarantee the QoS

and energy efficiency in passive optical networks (PONs). Proportional-

integral-derivative (PID)-based cyclic sleep controllers can maintain the

average downstream queueing delay in an optical line terminal (OLT) at a

constant level. However, conventional PID-based controllers generate errors

between a target queueing delay and an actual delay because of the nonlinear

characteristics of the cyclic sleep mechanism. This letter proposes a feedback

controller with a disturbance observer (DOB) that includes a linearized cyclic

sleep model to improve the QoS in terms of the average downstream

queueing delay. Simulations confirm that the proposed DOB-based controller

can reduce errors effectively compared with a PID-based controller.

Keywords: passive optical network, optical access network, cyclic sleep,

energy efficiency, quality of services

Classification: Fiber-Optic Transmission for Communications

References

[1] IEEE Std 1904.1-2013, “Standard for service interoperability in Ethernet passiveoptical networks (SIEPON),” June 2013.

[2] R. Kubo, J. Kani, Y. Fujimoto, N. Yoshimoto, and K. Kumozaki, “Adaptivepower saving mechanism for 10 Gigabit class PON systems,” IEICE Trans.Commun., vol. E93-B, no. 2, pp. 280–288, Feb. 2010. DOI:10.1587/transcom.E93.B.280

[3] A. R. Dhaini, P. H. Ho, and G. Shen, “Toward green next-generation passiveoptical networks,” IEEE Commun. Mag., vol. 49, no. 11, pp. 94–101, Nov.2011. DOI:10.1109/MCOM.2011.6069715

[4] Y. Maneyama and R. Kubo, “QoS-aware cyclic sleep control with proportional-derivative controllers for energy-efficient PON systems,” J. Opt. Commun.Netw., vol. 6, no. 11, pp. 1048–1058, Nov. 2014. DOI:10.1364/JOCN.6.001048

[5] T. Kikuchi and R. Kubo, “Proportional-integral (PI) and proportional-integral-derivative (PID)-based cyclic sleep controllers with anti-windup technique forenergy-efficient and delay-aware passive optical networks,” Jpn. J. Appl. Phys.,vol. 55, no. 8S3, 08RB02, Aug. 2016. DOI:10.7567/JJAP.55.08RB02

© IEICE 2017DOI: 10.1587/comex.2017XBL0112Received July 10, 2017Accepted July 28, 2017Publicized August 18, 2017Copyedited November 1, 2017

615

IEICE Communications Express, Vol.6, No.11, 615–620

http://dx.doi.org/10.1587/transcom.E93.B.280





http://dx.doi.org/10.1109/MCOM.2011.6069715




http://dx.doi.org/10.1364/JOCN.6.001048




http://dx.doi.org/10.7567/JJAP.55.08RB02




1 Introduction

Passive optical networks (PONs) have been deployed as fiber-to-the-home (FTTH)

communication infrastructures. PONs have the advantage that broadband Internet

services can be provided at a low cost. The Ethernet PON (EPON), which is a time-

division-multiplexing (TDM)-PON, consists of an optical line terminal (OLT), a

power splitter, and multiple optical network units (ONUs). The OLT is located at a

central office. The ONUs are installed on customer premises.

The increase of power consumption in optical access networks is an important

issue. The power-saving mechanisms for ONUs in EPON systems, e.g. TRx sleep

mode or cyclic sleep mechanism, were specified by IEEE 1904.1 [1]. The cyclic

sleep mechanism disables both the transmitter and receiver of the ONU during the

sleep state. It is also important to balance the quality of service (QoS) and energy

efficiency from the viewpoint of service diversification [2, 3]. Some researchers

proposed delay-aware ONU cyclic sleep controllers in EPON systems. Maneyama

et al. [4] proposed a delay-aware cyclic sleep controller to reduce the power

consumption in ONUs on the basis of feedback control techniques. In addition,

we proposed a proportional-integral-derivative (PID)-based cyclic sleep controller

with an anti-windup technique to maintain the average downstream queueing delay

in the OLT at a constant level and to reduce ONU power consumption [5].

However, conventional PID-based controllers generate errors between a target

queueing delay and an actual delay because of the nonlinear characteristics of

the cyclic sleep mechanism.

This letter proposes a feedback controller with a disturbance observer (DOB)

that includes a linearized cyclic sleep model to improve the QoS in terms of

the average downstream queueing delay. The DOB estimates the variations in

queue length (QL) in the OLT downstream buffer caused by the ONU cyclic

sleep mechanism on the basis of the linearized model. Specifically, the DOB

calculates and compensates the variation as an equivalent disturbance in the

dimension of the sleep period. Simulations confirm that the proposed DOB-based

controller can reduce the error effectively compared with a conventional PID-based

controller.

2 ONU power-saving mechanism

The ONU power-saving mechanism is illustrated in Fig. 1. A system configuration

of the EPON system is shown in Fig. 1(a), and a block diagram of the delay-aware

ONU cyclic sleep control system is shown in Fig. 1(b). The OLT has a downstream

traffic monitor and cyclic sleep controller. The OLT notifies the ONU of sleep

period Ts and makes the ONU enter the sleep state. The ONU enters the sleep state

during sleep period Ts. Tqd , the information of the target queueing delay requested

by executed applications, is sent from the ONU to the OLT in advance.

The downstream traffic monitor in the OLT measures the downstream QL qout

and the average downstream frame arrival rate λ. The cyclic sleep controller

calculates the target QL qcmd as

qcmd ¼ Tqd�: ð1Þ© IEICE 2017DOI: 10.1587/comex.2017XBL0112Received July 10, 2017Accepted July 28, 2017Publicized August 18, 2017Copyedited November 1, 2017

616


The cyclic sleep controller calculates the sleep period Ts from qcmd and qout. First,

the average downstream QL qave is calculated from qout by using an exponential

moving average. Then, the error signal between qcmd and qave is given by

e ¼ qcmd � qave: ð2ÞThe temporal sleep period Ts;tmp, calculated from the error signal e by the cyclic

sleep controller described in the following sections, is input into the sleep period

limit. The relationship between Ts;tmp and Ts is expressed as

Ts ¼0 if Ts;tmp < Ts;min

Ts;tmp if Ts;min � Ts;tmp < Ts;max

Ts;max if Ts;max � Ts;tmp

8><>: ; ð3Þ

where Ts;min and Ts;max denote the minimum sleep period and the maximum sleep

period, respectively.

3 Conventional PID-based controller

The PID-based controller with an anti-windup technique [5] determines the

temporal sleep period Ts;tmp as

Ts;tmp ¼Kp þ Ki

sþ Kds

� �e þ Ts if x ¼ 0

ðKp þ KdsÞe þ Ts if x ≠ 0

8><>: ; ð4Þ

x ¼ Ts;tmp � Ts; ð5Þwhere s, Kp, Ki, and Kd denote the Laplace operator, proportional gain, integral

gain, and derivative gain, respectively. The PID-based controller can maintain the

average queueing delay in the OLT at a constant level regardless of the amount of

downstream traffic. However, it generates an error between the target queueing

delay and actual delay.

(a) Configuration of EPON system

(b) Block diagram of delay-aware cyclic sleep control system

Fig. 1. Delay-aware ONU cyclic sleep control scheme.


617


4 Proposed DOB-based controller

In this letter, a feedback controller with a DOB that includes a linearized cyclic

sleep model is proposed. The cyclic sleep mechanism can be modelled in the time

domain as

dqoutðtÞdt

¼ �ðtÞ � �0ðtÞ; ð6Þ

�0ðtÞ ¼ TaTsðtÞ þ Ta

�; ð7Þ

where μ, �0, and Ta denote the downstream link rate, downstream traffic rate, and

active period of the ONU, respectively. When λ is a constant value, the linearized

model of the cyclic sleep mechanism in the frequency domain can be represented as

Ts � Ts0 ¼ Mnðqout � qcmdÞs; ð8Þ

Mn ¼ ðTs0 þ TaÞ2Ta�

; ð9Þ

where Ts0 and Mn denote the equilibrium point and nominal model of the cyclic

sleep mechanism, respectively.

The proposed DOB-based controller is illustrated in Fig. 2. Fig. 2(a) shows a

block diagram of the entire control system, and Fig. 2(b) shows the internal

structure of the DOB. The DOB-based controller adopts a proportional (P) con-

troller with proportional gain Kp as a feedback controller. The DOB estimates the

variation of QL in the OLT downstream buffer caused by the cyclic sleep mechan-

ism on the basis of the linearized model, i.e. (8) and (9). The variation can be

estimated as the disturbance d in the dimension of the sleep period, which is given

by

d ¼ gdiss þ gdis

ððM �MnÞqqves þ dlimitÞ; ð10Þ

where gdis, M, and dlimit denote the cut-off frequency of the DOB, actual dynamics

of the cyclic sleep mechanism, and disturbance caused by the sleep period limit,

respectively. The DOB estimates the QL variation through the low-pass filter to

reduce the effect of noise. In the proposed DOB-based controller, the temporary

sleep period Ts;tmp is calculated as

Ts;tmp ¼ MnKpe þ d: ð11Þ

5 Performance evaluation

Simulations were performed to compare the conventional PID-based and proposed

DOB-based controllers. We assumed that the controllers were implemented in a

10-Gbps EPON (10G-EPON) OLT. In the simulations, the cyclic sleep controller

calculated the average QL at time k, qave½k�, asqave½k� ¼ �qout½k� þ ð1 � �Þqave½k � 1�; ð12Þ

where α is a smoothing factor.

The distance between the OLT and ONU was set to 20 km. The transition time

from the sleep state to the active state was set to 10ms. The minimum sleep period


618


Ts;min, smoothing factor α, and the cut-off frequency gdis were set to 10ms, 0.2, and

100 rad/s, respectively. In the PID-based controller, Kp, Ki, and Kd were set to

8:0 � 10�7, 1:0 � 10�12, and 1:9 � 10�1, respectively. In the DOB-based controller,

Kp, gdis, Ts0, and Ta were set to 1:0 � 105, 9:0 � 105, 0, and 10ms, respectively. The

simulations were performed under the following three conditions: in cases 1, 2, and

3, the parameters ðTs;max; TqdÞ were set to (20ms, 10ms), (40ms, 20ms), and

(80ms, 40ms), respectively. We assumed that there was only downstream traffic

whose arrival rate followed the Poisson distribution. The frame size was set to 1250

bytes.

The simulation results are shown in Fig. 3. The average downstream queueing

delay in the OLT is shown in Fig. 3(a). In all the cases, the PID-based controller

maintained the average queueing delay at a constant level regardless of the amount

of downstream traffic. However, the PID-based controller generated an error

between the target queueing delay and actual delay. The DOB-based controller

reduced the error compared with the PID-based controller. The time occupancy of

active periods of an ONU is shown in Fig. 3(b). It was confirmed that the DOB-

based controller provided power-saving performance that was similar to that of the

PID-based controller.

(a) Control system implemented in OLT

(b) Internal structure of DOB

Fig. 2. DOB-based controller.


619


6 Conclusion

This letter proposed a feedback controller with a DOB that includes a linearized

cyclic sleep model to improve the QoS in terms of the average downstream

queueing delay. The simulation results showed that the proposed DOB-based

controller reduced errors between the target queueing delay and the actual delay

compared with a conventional PID-based controller.

Acknowledgments

This research was supported in part by the National Institute of Information and

Communications Technology (NICT), Japan.

(a) Average queueing delay

(b) Time occupancy of active period

Fig. 3. Simulation results.


620


Near-metal-insensitiveantenna for closed spacewireless communications

Yuta Nakagawa1a), Shingo Tanaka1, Tatsuo Toba1,Kenji Matsushita1, and Hisashi Morishita21 Transmission Technology Research Dept., Yazaki Research and Technology Center,

Yazaki Corporation, 3–1 Hikari-no-oka, Yokosuka, Kanagawa 239–0847, Japan2 Graduate School of Science and Engineering, National Defense Academy,

1–10–20 Hashirimizu, Yokosuka, Kanagawa 239–8686, Japan


Abstract: In order to obtain near-metal-insensitive antenna for closed-

space wireless communications, the impedance characteristics of U-shaped

folded monopole antenna is investigated in detail. In this Letter, the near-

metal-insensitive means antenna VSWRs are hardly influenced by the near

object, especially metal. The simulated and measured results show that the

proposed higher impedance model has stronger near-metal-insensitiveness

than the conventional middle impedance model. The simulated and measured

results show that the antenna gains of higher impedance models are 3 dB

greater at maximum than those of middle impedance models, when metal

plane approaches.

Keywords: antennas, U-shaped folded monopole antenna, closed space

wireless communications, near-metal-insensitive antenna, robust antenna

Classification: Antennas and Propagation

References

[1] T. Kobayashi, “Measurements and characterization of ultra widebandpropagation channels in a passenger-car compartment,” IEICE Trans.Fundamentals, vol. E89-A, no. 11, pp. 3089–3094, Nov. 2006. DOI:10.1093/ietfec/e89-a.11.3089

[2] S. Horiuchi, K. Yamada, S. Tanaka, Y. Yamada, and N. Michishita,“Comparisons of simulated and measured electric field distributions in a cabinof simplified scale car model,” IEICE Trans. Commun., vol. E90-B, no. 9,pp. 2408–2415, Sep. 2007. DOI:10.1093/ietcom/e90-b.9.2408

[3] M. Ohira, T. Umaba, S. Kitazawa, H. Ban, and M. Ueba, “Experimentalcharacterization of microwave radio propagation in ICT equipment for wirelessharness communications,” IEEE Trans. Antennas Propag., vol. 59, no. 12,pp. 4757–4765, 2011. DOI:10.1109/TAP.2011.2165494

[4] H. Hatamoto, N. Nakamoto, N. Kikuchi, and S. Shimizu, “A study on patchantenna design for wireless networks in ICT equipment,” IEICE General Conf.,B-1-134, Mar. 2012 (in Japanese).

[5] M. Ohira, “A bandwidth-enhanced low-profile antenna for in-machine wirelessharness communications and its clearance distance evaluation,” Proc. IEEEAP-S Int. Symp., Orland, Florida, pp. 978–979, Jun. 2013. DOI:10.1109/APS.

© IEICE 2017DOI: 10.1587/comex.2017XBL0117Received July 24, 2017Accepted August 8, 2017Publicized August 29, 2017Copyedited November 1, 2017

621


http://dx.doi.org/10.1093/ietfec/e89-a.11.3089




http://dx.doi.org/10.1093/ietcom/e90-b.9.2408




http://dx.doi.org/10.1109/TAP.2011.2165494




http://dx.doi.org/10.1109/APS.2013.6711148


2013.6711148[6] T. Ito, M. Nagatoshi, S. Tanaka, and H. Morishita, “Fundamental character-

istics of folded monopole antennas with parasitic element for triple bandMIMO antenna,” Int. Rev. Prog. Appl. Computational ElectromagneticsSociety., pp. 24–28, Mar. 2013.

[7] R. W. Lampe, “Design formulas for an asymmetric coplanar strip foldeddipole,” IEEE Trans. Antennas Propag., vol. 33, no. 9, pp. 1028–1031, Sep.1985. DOI:10.1109/TAP.1985.1143698

[8] C. A. Balanis, Antenna Theory: Analysis and Design, 3rd ed., John Wiley &Sons, Chap. 9, 2005.

[9] T. Ito, M. Nagatoshi, S. Tanaka, and H. Morishita, “Characteristics of widebandfolded dipole antenna with feed line for mounting to small terminal,” IEICETrans. Commun. (Japanese Edition), vol. J96-B, no. 2, pp. 124–132, Feb.2013.

[10] Y. Nakagawa, S. Tanaka, T. Toba, T. Kimura, K. Shirasu, T. Oki, N. Nishiyama,and H. Morishita, “A study on metal-insensitive antenna for closed spacewireless communications,” Proc. IEEE Int. Symp. Antennas and Propag.,Hobart, Australia, pp. 1–4, Nov. 2015.

1 Introduction

The research and development activities on wireless communication technologies

are rapidly growing, not only for open space applications, but also for closed space

applications [1, 2, 3]. The propagations and distributions of electric fields in closed

space are more complicated than those in open space, but it is expected that

those research and development will pioneer novel application fields on wireless

communications.

However, the problems for closed space wireless communications are not only

electric field propagations, but also antenna characteristics. It is known that the

antenna impedances are strongly influenced by the near object (especially metal),

so the near-metal-insensitive antennas are strongly required for closed space wire-

less communications [4, 5].

In this Letter, therefore, we introduce fundamental study on near-metal-insen-

sitive antenna, by employing U-shaped folded monopole antenna (UFMA) with

ground plane (GP) [6]. UFMA is a modified folded monopole antenna, so the basic

impedance characteristics are depends on step up ratio [7, 8, 9], that will be defined

in Section 3.

The impedance characteristics of UFMAs are investigated in detail when metal

object approaches and step up ratio is controlled in order to enhance near-metal-

insensitive characteristics. In addition to simulated results [10], measured results

will be shown and compared.

2 Antenna structures

The structure of investigated UFMA is shown in Fig. 1(a). The antenna element is

composed of two parallel metal strips with widths wa1 and wa2. Two strips are

short-connected by metal strips with width wa3 and length sa, at one side.© IEICE 2017DOI: 10.1587/comex.2017XBL0117Received July 24, 2017Accepted August 8, 2017Publicized August 29, 2017Copyedited November 1, 2017

622








At the other side, one strip with width wa1 is connected to GP and the other

strip with width wa2 is connected to feed point. The total length of two metal strips

is la þ h, and they are folded to keep low profile antenna with height h. The antenna

impedance can be changed by step up ratio. The parameter values shown in

Fig. 1(a) are selected in order to obtain VSWR � 3 at 2.4GHz with bandwidth

10MHz, when there is no infinite plane.

The antenna element is placed on the GP (50mm � 80mm), which models

printed circuit board of electronic control unit. Therefore, considering the mounting

space and the layout flexibility of electronic components, the placement of the

antenna elements is preferably the edge of GP.

We assumed installation image of UFMA is shown in Fig. 1(b). The infinite

plane (perfect conductor) of Fig. 1(a), which models metal wall of closed space,

is located in parallel to GP. The distance between GP and infinite plane is defined

as hgp, and changed in order to investigate the near-metal-insensitive characteristics

of the antenna.

3 Impedance characteristic [7]

The input impedance characteristics of the UFMA are investigated in this chapter,

when hgp values are changed. The simulated results using FEKO simulation, based

on method of moment and measured results are shown and compared. For the

measurement, Cu plane with size 400mm � 635mm is placed near the antenna,

instead of infinite plane for simulation. The input impedance of the folded

monopole antenna Z is expressed by equation (1).

Z ¼ ð1 þ aÞ2Zm ð1ÞWhere Zm is impedance of the monopole antenna (Zm ¼ 9:2Ω, in case of Fig. 1(a)),

ð1 þ aÞ2 is step up ratio. The a value is obtained from (2) and (3).

(a) Antenna mounted on ECU. (b) Assumed installation image.

Fig. 1. UFMA.


623


a ¼ln

(4c þ 2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið2cÞ2 �

�wa2

2

�2s )

� lnðwa2Þ

ln

(4c þ 2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið2cÞ2 �

�wa1

2

�2s )

� lnðwa1Þð2Þ

c ¼

wa1 þ wa2

2þ sa

2ð3Þ

Where the wa1, wa2, and sa are indicated in Fig. 1(a). The step up ratio is

determined by wa1, wa2, and sa.

3.1 Middle impedance model

The UFMA with parameter values shown in Fig. 1(a) is referred to as middle

impedance model (MIM), as it is designed to obtain 37Ω at 2.4GHz when there is

no infinite plane (hgp ¼ 1). In MIM, the step up ratio value calculated from

equations (1)–(3) is 4 [7]. The simulated and measured VSWRs of MIM are shown

in Fig. 2(a) and (b). The relevant Smith Charts are also shown inset. Fig. 2(a)

shows that VSWR � 3 is satisfied at 2.4GHz when hgp ¼ 16mm. However, as are

shown Fig. 2(b), VSWR � 3 cannot be satisfied at 2.4GHz when hgp ¼ 4mm. We

can say that the measured VSWR values agree well with simulated VSWR values.

The Smith Charts in Fig. 2(a) and (b) indicate that the larger VSWRs with

hgp ¼ 4mm is caused by lower antenna impedances that those with hgp ¼ 16mm.

So we can say that this is the influence of the infinite plane coming close to parallel

with the horizontal element of length la and the radiation resistance is reduced due

to the generation of the image current. Therefore, we expect that, if we can obtain

higher antenna impedance by adjusting step up ratio, the reduced impedance will be

cancelled and the antenna will be more near-metal-insensitive than MIM.

3.2 High impedance model

In order to obtain higher antenna impedance, we changed parameters shown in

Fig. 1(a). wa1 ¼ 2mm and wa2 ¼ 0:2mm are selected to obtain higher step up

ratio. la ¼ 26mm is selected for frequency adjustment, but other parameters are not

changed from Fig. 1(a), and the antenna is referred to as higher impedance model

(HIM). In HIM, the step up ratio value calculated from equations (1)–(3) is 13.8

[7], and the impedance value is 126Ω at 2.4GHz when there is no infinite plane

(hgp ¼ 1). Fig. 2(c) and (d) show the VSWRs and Smith Charts of simulated and

measured of HIM. Fig. 2(c) shows that the antenna impedance of HIM is higher

than that of MIM, but VSWR � 3 is satisfied when hgp ¼ 16mm. Fig. 2(d) shows

that antenna impedance with hgp ¼ 4mm is lower than that with hgp ¼ 16mm,

as are observed in Fig. 2(a) and (b). However, due to higher impedance with

hgp ¼ 1, VSWR � 3 is satisfied even when hgp ¼ 4mm. Regarding the band-

width, HIM became narrower than that of MIM.

The simulated and measured VSWR values vs. hgp values at 2.4GHz are

summarized in Fig. 2(e). We can say that, in order to obtain VSWR � 3, hgp �© IEICE 2017DOI: 10.1587/comex.2017XBL0117Received July 24, 2017Accepted August 8, 2017Publicized August 29, 2017Copyedited November 1, 2017

624


5mm is necessary for MIM. However, hgp � 2mm is acceptable for HIM. HIM is

more robust against nearby metal than MIM.

4 Radiation characteristics

The simulated and measured radiation patterns of UFMAs in H-plane at 2.4GHz,

when hgp ¼ 4mm, 16mm are shown in Fig. 3. Fig. 3(a) and (b) is xz-plane

radiation patterns for MIM and Fig. 3(c) and (d) is xz-plane radiation patterns

for HIM, respectively. The definitions of x/y/z axes are shown inset, and maximum

radiation is observed toward 0 or +30 degree.

Fig. 3(e) shows the simulated and measured antenna gains of maximum

radiation direction vs. hgp values at 2.4GHz. We can say that the antenna gain

of HIM is greater than that of MIM in hgp � 6mm. The difference of antenna gain

between MIM and HIM is the 3 dB at maximum in hgp � 6mm.

(a) MIM when hgp=16mm

(c) HIM when hgp=16mm

(b) MIM when hgp=4mm

(d) HIM when hgp=4mm

(e) VSWR values vs. hgp values at 2.4GHz

Fig. 2. Simulated and measured VSWRs.


625


5 Conclusion

We investigated impedance characteristics of UFMA, in order to obtain near-metal-

insensitive antenna for closed-space wireless communications. The simulated and

measured results show that the antenna impedance changes to small value when

metal plane approaches. To encounter this problem, we proposed higher impedance

model by selecting step up ratio. The limitation value in order to obtain VSWR � 3

is 5mm for the conventional MIM and 2mm for the proposed HIM, respectively.

The simulated and measured results show that the antenna gains of HIM are 3 dB

greater at maximum than those of MIM.

(a) MIM when hgp=16mm

(c) HIM when hgp=16mm

(b) MIM when hgp=4mm

(d) HIM when hgp=4mm

(e) Antenna gains of maximum radiation direction vs. hgp values.

Fig. 3. Simulated and measured radiation characteristics of UFMAs at2.4GHz.


626


Experimental study ofincoming waves separatingadaptive array for ISDB-Tmobile reception

Takanobu Tabata1a), Mitoshi Fujimoto2, Satoshi Hori1,Tomohisa Wada3,4, and Hirokazu Asato41 Kojima Industries Corporation,

1099–2 Marune, Kurozasa-cho, Miyoshi, Aichi 470–0201, Japan2 Graduate School of Engineering, University of Fukui,

3–9–1 Bunkyo, Fukui, Fukui 910–8507, Japan3 University of the Ryukyus,

1 Senbaru, Nishihara-cho, Nakagami-gun, Okinawa 903–0213, Japan4 Magna Design Net, Inc.,

Daido Seimei Naha Building 4F 3–1–15 Maejima, Naha, Okinawa 900–0026, Japan


Abstract: In terrestrial digital TV broadcasting in Japan, OFDM is utilized

as a modulation scheme. However, the communication quality could be

seriously deteriorated at the high speed mobile reception (Ex. vehicle).

Furthermore, most of the mobile reception system, 4 antennas are mounted

on a vehicle (Front: 2 antennas, Rear: 2 antennas).

The authors proposed a new combing method (Incoming waves separating

system by adaptive array antenna). In this paper, the performances in case of

4 directional antennas are evaluated. Furthermore the performance in more

complicated environment is evaluated. From experimental results, it is

clarified that the BER performances of the proposed system are superior to

the conventional system.

Keywords: ISDB-T, OFDM, adaptive array antenna, vehicle, directivity,

BER

Classification: Antennas and Propagation

References

[1] T. Tabata, H. Asato, D. H. Pham, M. Fujimoto, N. Kikuma, S. Hori, and T.Wada, “Experiment study on adaptive array antenna system for ISDB-T highspeed mobile reception,” IEEE Antennas and Propagation Society InternationalSymposium (APS2007), Hawaii, USA, June 2007. DOI:10.1109/APS.2007.4395840

[2] T. Tabata, N. Kikuma, T. Wada, S. Hori, M. Fujimoto, and H. Asato,“Experimental study of adaptive array antenna system for ISDB-T mobilereception,” 2009 International Symposium Antenna and Propagation(ISAP2009), vol. 1, pp. 719–722, Oct. 2009.

[3] T. Tabata, M. Fujimoto, S. Hori, T. Wada, and H. Asato, “Incoming waves

© IEICE 2017DOI: 10.1587/comex.2017XBL0118Received July 27, 2017Accepted August 22, 2017Publicized September 7, 2017Copyedited November 1, 2017

627






separating adaptive array for ISDB-T mobile reception,” InternationalSymposium on Antennas and Propagation (ISAP16), Vol. 1, pp. 1056–1057,2016.

[4] S. Hori, N. Kikuma, and N. Inagaki, “MMSE adaptive array utilizing guardinterval in the OFDM systems,” Electron. Commun. Jpn. Pt. I, vol. 86, no. 10,pp. 1–9, 2003. DOI:10.1002/ecja.10121

[5] S. Hori, N. Kikuma, T. Wada, and M. Fujimoto, “Experimental study on arraybeam forming utilizing the guard interval in OFDM,” International Symposiumon Antennas and Propagation (ISAP05), Vol. 1, pp. 257–260, 2005.

1 Introduction

Orthogonal Frequency Division Multiplexing (OFDM) is adopted as a modulation

scheme in the terrestrial digital TV broadcasting in Japan (ISDB-T). However, the

communication quality is seriously deteriorated at the high speed mobile reception

(Ex. vehicle), running in a radio environment that the delayed wave exceeding

guard interval or running extremely high speed. Furthermore, most of the mobile

reception system, 4 antennas are mounted on a vehicle (Front: 2 antennas, Rear: 2

antennas) and array signal processing techniques is adopted to improve the

reception quality.

The authors proposed a combining system and proved that our combing method

has better performances than the conventional system [1, 2]. To improve the

performance further, the authors proposed a new combing method (Incoming

waves separating system by adaptive array antenna) [3]. In ref. [3], the performance

of the new combining method is evaluated in case of Omni-directional antennas,

however, the directional pattern of antennas mounted on the vehicles are different

in either front or rear side. In this paper, the performances in case of 4 directional

antennas are evaluated. Furthermore the performance in more complicated environ-

ment is evaluated. From experimental results, it is clarified that the BER perform-

ances of the proposed system are superior to the conventional system.

2 System configurations

2.1 Antenna mounted on vehicles

In this paper, we suppose that 4 antennas are mounted on a vehicle where two

antennas are mounted in front side and rear side, respectively. The top view of

the antenna location on the vehicle is shown in Fig. 1(a), and the directional

patterns of the front side and rear side antennas are shown in Fig. 1(b). F/B (Front/

Back) ratio for each antenna is defined as shown in Fig. 1(b).

2.2 Proposed combining system

Fig. 1(c) shows the block diagram of the proposed system. The signals received by

the antennas are combined independently in the front side and the rear side of the

vehicle. At Main process, received signal by both antenna elements is combined

based on Minimum Mean Square Error (MMSE), and then largest arrival signal

(usually the 1st arrival signal) is remained. At Sub process, the largest arrival signal© IEICE 2017DOI: 10.1587/comex.2017XBL0118Received July 27, 2017Accepted August 22, 2017Publicized September 7, 2017Copyedited November 1, 2017

628


http://dx.doi.org/10.1002/ecja.10121



is suppressed firstly by Power Inversion (PI) algorithm. Therefore 2nd largest signal

(usually 2nd arrival signal) is remained by MMSE.

Similar process is conducted for rear side antennas.

(a) Antennas position (b) Directivities

(c) Proposed System. (Pre-FFT type)

(d) 4FFT System. (e) Prototype.

Fig. 1. System configurations© IEICE 2017DOI: 10.1587/comex.2017XBL0118Received July 27, 2017Accepted August 22, 2017Publicized September 7, 2017Copyedited November 1, 2017

629


The feature of the proposed system is that the largest signal is combined by

Main process as desired signal and the 2nd largest signal is combined by Sub

process as also desired signal. Subsequently, these signals in front side and in rear

side are combined again and this combined signal is demodulated. Fig. 1(d) shows

4FFT system (Post-FFT type) which is conventional configuration. Both systems

have been implemented in the prototype which is shown in Fig. 1(e).

2.3 Algorithms for incoming wave separating

It is supposed that the array is composed of K-element. The received signals by the

antenna element xk and weight coefficients wk are expressed as follows,

XðtÞ ¼ ½x1ðtÞ; x2ðtÞ; . . . ; xKðtÞ�T ð1ÞXðtÞ ¼ ½x1ðtÞ; x2ðtÞ; . . . ; xKðtÞ�T ð2Þ

Then, the array output is given by

yðtÞ ¼ WHXðtÞ ð3ÞHere, the superscript T and H indicates the transpose and the conjugate transpose,

respectively.

The OFDM signal in time domain consists of the guard time GI and the

effective symbol. Let xhkðtÞ (k ¼ 1; 2; . . . ; K ) express the extracted signals from the

received signal during the GI. In a similar manner, xtkðtÞ (k ¼ 1; 2; . . . ; K ) express

the extracted signals from the received signal during the last part of the effective

symbol. They are expressed in a vector as

XhðtÞ ¼ ½xh1ðtÞ; xh2ðtÞ; . . . ; xhKðtÞ�T ð4ÞXtðtÞ ¼ ½xt1ðtÞ; xt2ðtÞ; . . . ; xtKðtÞ�T ð5Þ

Hence the extracted signals from guard interval (GI) and last part of effective

symbol are expressed as

yhðtÞ ¼ WHXhðtÞ; ytðtÞ ¼ WHXtðtÞ ð6ÞIn this paper, we adopt an algorithm, Minimum Mean Square Error (MMSE), and

the weight coefficient vector WMMSE is expressed as follows [4, 5],

WMMSE ¼ R�1XhXhWAMBF ð7Þ

where RXhXh ¼ E½XhðtÞXhHðtÞ�, WAMBF ¼ E½XhðtÞy�tðtÞ�

An algorithm separating the largest arrival signal and 2nd largest signal is PI, and

the weight coefficient vector WPI is expressed as follows

WPI ¼ R�1XhXhS ð8Þ

S ¼ ½1; 0; � � �; 0�T ð9Þ

3 Validations through experiment

3.1 Condition

The BER performances are evaluated using the fading simulator. The frequency in

the experiments is channel No. 35 (605.143MHz). The OFDM signal conditions

are given below. Number of carriers, Effective symbol length, GI length (1/8) and

Modulation scheme is 5617, 1008µs, 126µs and 64QAM, respectively. The


630


performance is evaluated in a 2-arrival wave environment as shown in Fig. 2.

Distance of 4 antennas mounted on the vehicle are dx ¼ dy ¼ 0:5� as shown in

Fig. 1(a). The F/B ratios of antennas are 0 dB and 10 dB. The desired signal and

undesired signal power ratio (D/U) is 3 dB. The delay time between the desired

signal to the undesired signal is 21 ð1=6 � GIÞ � 399 ð19=6 � GIÞ [µs].

3.2 Experimental results

Fig. 3 shows the BER performances of the proposed system and the 4FFT system.

Fig. 3(a) shows the BER performances when F/B ratio = 0 dB (Omni-directional),

Fig. 3(b) shows the BER performances when F/B ratio = 10 dB (antennas are

mounted on a vehicle). In this experiment, we evaluated the Pre-Viterbi BER

performances, therefore the criteria of BER performances is 2:0 � 10�2 (Error Freeat Post-RS coding).

(1) Comparison of combining system

When the delay time is large, the BER performance of the proposed system is

superior to that of the 4FFT system. Especially, the BER performance degradation

of the proposed system is very small even if the delay time is large. The reasons are

that the proposed system perform adaptive array (Beam-forming for only desired

signal, and null-steering to undesired signals.) independently in the front side and

the rear side of the vehicle, and the 2nd largest signal is combined by Sub process

as also desired signal.

(2) Comparison of F/B ratio

When F/B ratio = 0 dB and 10 dB, the BER performances has the same tendency.

Therefore it can be considered that there is no effect from the directional patterns of

the 4 mounted antennas.

(3) Comparison of environment

At the 2 waves environment like this paper, these BER performances are the

similar.

Fig. 2. Radio environment.


631


4 Conclusion

The BER performances of proposed system (Incoming waves separating system)

was compared with that of 4FFT system (Post-FFT carrier diversity) as a conven-

tional system. From the experimental result, it was proved that the BER perform-

ance of the proposed system was superior when the OFDM signals arrive with

delay exceeding the Guard Interval length even if the 4 directional antennas are

mounted on a vehicle. The reason is that the proposed system perform adaptive

array independently in the front side and the rear side of the vehicle.

(a) F/B ratio =0dB

(b) F/B ratio =10dB

Fig. 3. BER performances


632


IEICE Communications Express, Vol.6, No.11, 615 Delay ...

Documents

Transcript of IEICE Communications Express, Vol.6, No.11, 615 Delay ...