IEICE Communications Express, Vol.4, No.5, 136 142 Time ... · 3 Diode switching characteristics and conducted emissions for the CCM boost converter Fig. 2(a) depicts the CCM boost

Time and frequencymixed-domain analysis ofconducted emissions fortypes of diode

Takaaki Ibuchia) and Tsuyoshi FunakiOsaka University, Division of Electrical, Electronic and Information Engineering,

Graduate School of Engineering, Suita, Osaka 565–0871, Japan

a) [email protected]

Abstract: Electromagnetic interference (EMI) noise in time and frequency

mixed domains is analyzed to understand the influence of noise source

behavior on the conducted emissions in boost converters. The switching

characteristics of a sillicon PiN diode and a sillicon carbide Schottky barrier

diode in a boost converter are compared and evaluated as EMI noise sources,

and the influence of diode switching operation on the generation and

attenuation of conducted emissions are discussed on the basis of spectrogram

analysis.

Keywords: switching converter, conducted emissions, silicon carbide

Schottky barrier diode, spectrogram

Classification: Electromagnetic Compatibility (EMC)

References

[1] A. R. Hefner, R. Singh, J. Lai, D. Berning, S. Bouche, and C. Chapuy, “SiCpower diodes provide breakthrough performance for a wide range of applica-tions,” IEEE Trans. Power Electron., vol. 16, no. 2, pp. 273–280, Mar. 2001.DOI:10.1109/63.911152

[2] N. Oswald, P. Anthony, N. McNeill, and B. H. Stark, “An experimentalinvestigation of the tradeoff between switching losses and EMI generation withhard-switched all-Si, Si-SiC, and all-SiC Device Combinations,” IEEE Trans.Power Electron., vol. 29, no. 5, pp. 2393–2407, May 2014. DOI:10.1109/TPEL.2013.2278919

[3] T. Ibuchi and T. Funaki, “Effect of diode operating temperature on conductednoise spectrum for CCM DC–DC boost converter,” IEICE Commun. Express,vol. 3, no. 9, pp. 269–274, Sept. 2014. DOI:10.1587/comex.3.269

[4] J. Lutz, A. Schlangenotto, U. Scheuermann, and R. Doncker, SemiconductorPower Devices—Physics, Characteristics, Reliability, Springer, 2011.

[5] G. Spiazzi, S. Buso, M. Citron, M. Corradin, and R. Pierobon, “PerformanceEvaluation of a Schottky SiC Power Diode in a Boost PFC Application,” IEEETrans. Power Electron., vol. 18, no. 6, pp. 1249–1253, Nov. 2003. DOI:10.1109/TPEL.2003.818821

[6] T. Lobos, J. Rezmer, and P. Schegner, “Parameter estimation of distorted signalsusing Prony method,” Proc. IEEE Bologna Power Tech. Conf., vol. 4, pp. 23–26, June 2003. DOI:10.1109/PTC.2003.1304801

[7] T. Ibuchi and T. Funaki, “A study on modeling of dynamic characteristics of

© IEICE 2015DOI: 10.1587/comex.4.136Received March 28, 2015Accepted April 14, 2015Published May 15, 2015

136

IEICE Communications Express, Vol.4, No.5, 136–142

http://dx.doi.org/10.1109/63.911152

http://dx.doi.org/10.1109/63.911152

http://dx.doi.org/10.1109/63.911152

http://dx.doi.org/10.1109/TPEL.2013.2278919




http://dx.doi.org/10.1587/comex.3.269








http://dx.doi.org/10.1109/PTC.2003.1304801




circuit component in TDR measurement based on Prony analysis,” IEICEElectron. Express, vol. 8, no. 18, pp. 1534–1540, Sept. 2011. DOI:10.1587/elex.8.1534

[8] M. Kuisma and P. Silventoinen, “Using spectrograms in EMI-analysis— anoverview,” 20th Annual IEEE Applied Power Electronics Conference andExposition (APEC ’05), vol. 3, pp. 1953–1958, Mar. 2005. DOI:10.1109/APEC.2005.1453323

[9] S. Braun, T. Donauer, and P. Russer, “A real-time time-domain EMI meas-urement system for full-compliance measurements according to CISPR 16-1-1,”IEEE Trans. Electromagn. Compat., vol. 50, no. 2, pp. 259–267, May 2008.DOI:10.1109/TEMC.2008.918980

1 Introduction

Silicon carbide (SiC) semiconductors’ material properties are more attractive than

conventional Si semiconductors for high-power, high-frequency, and high-temper-

ature operating applications [1, 2]. The fast switching of power devices with high

voltages and large currents results in high dv/dt and di/dt and leads to high-

frequency electromagnetic interference (EMI) noise by interacting with circuit

parasitic components [2]. Solving this EMI noise problem in power converters is

very important to meet standard electromagnetic compatibility (EMC) require-

ments.

The conventional approach for evaluating conducted and radiated emissions is

to analyze the noise spectrum amplitude in the frequency domain. However, the

EMI noise source in a switching power converter is the intermittent, transient

voltages and currents caused by the switching operations of power semiconductor

devices. Therefore, understanding the influence of switching voltages and currents

in the device on the EMI noise is necessary to clarify EMI noise generation

mechanism and to design a circuit with fast and high-frequency switching of high

voltages and large currents. Previous work [3] compared the temperature dependen-

cy of the reverse recovery behaviors of the Si PiN diode (PiND) and the SiC

Schottky barrier diode (SBD). The SiC SBD demonstrated an invariant switching

behavior and conducted the power converter’s emission levels to the operating

temperature. However, its fast turn-off causes lower damping of oscillation than Si

PiND, which results from having a lower Q in the resonance between the diode

terminal capacitance and circuit parasitic inductance. This study focuses on the

switching characteristics of Si PiND and SiC SBD as EMI noise sources in a

continuous-current-mode (CCM) DC–DC boost converter, and EMI noise gener-

ation will be discussed on the basis of a time and frequency mixed-domain analysis

of the conducted emissions.

An STTH8L06 (STMicrosemiconductors, 600V, 8A) Si PiND and a

TRS8E65C (Toshiba, 650V, 8A) and IDH08SG60C (Infineon, 600V, 8A) SiC

SBDs, with comparable voltages and current ratings, are used. These devices were

packaged in TO-220. Section 2 discusses these diodes’ static characteristics.

Section 3 evaluates their switching characteristics and investigates the influence

of those characteristics on conducted emissions in the tested converter on the basis© IEICE 2015DOI: 10.1587/comex.4.136Received March 28, 2015Accepted April 14, 2015Published May 15, 2015

137


http://dx.doi.org/10.1587/elex.8.1534





http://dx.doi.org/10.1109/APEC.2005.1453323





http://dx.doi.org/10.1109/TEMC.2008.918980




of a time and frequency mixed-domain analysis. The study’s conclusions are

presented in Section 4.

2 Temperature dependence in the static characteristics of diodes

Fig. 1(a) shows the measured forward current–voltage (IF–VF) characteristics of

the studied diodes at room temperature (25 °C) using a curve tracer (Agilent

B1505). Both Si PiND and SiC SBD have knee voltage and their forward

characteristics show almost the same tendency. The two SiC SBDs have almost

the same knee voltage; however, the on-resistance of IDH08SG60C is approxi-

mately 1.7 times larger than that of TRS8E65C.

Fig. 1(b) shows the measured capacitance–voltage (Cj–VR) characteristics of

the studied diodes in a blocking condition with a 100mV, 1MHz AC measurement

signal. The SiC SBDs have larger junction capacitances than Si PiND because of

higher impurity concentrations in the drift region [4]. The capacitance for

TRS8E65C is twice as large as that for IDH08SG60C in the lower reverse bias

voltage range, which corresponds to the difference in on-resistance.

Fig. 1(c) shows the measured frequency characteristics of the studied diodes’

impedance jZj in a blocking condition (VR ¼ 20V) using an impedance analyzer

(Keysight E4991B). All samples’ frequency responses show as RLC series equiv-

alent circuits. The equivalent series inductances are almost identical for the same

TO-220 package. The series resonance frequency of 100–300MHz corresponds

with the diodes’ junction capacitances. STTH8L06 shows a larger equivalent series

resistance (ESR) than that of the SiC SBDs, which are identified from minimum

impedance at the series resonance frequency. IDH08SG60C’s ESR is approxi-

mately 1.3 times larger than that of TRS8E65C.

3 Diode switching characteristics and conducted emissions for the

CCM boost converter

Fig. 2(a) depicts the CCM boost converter’s circuit diagram [5]. The MOSFET

(Infineon, IPP60R099CP) was operated at a 100 kHz switching frequency with

50% duty cycle for a 100V DC input voltage and 75Ω load resistance. Output

voltage was 200V DC. Si PiND induces large reverse recovery currents during

turn-off operations because of stored minority carriers, as shown in Fig. 2(a), which

results in large switching losses and noise in CCM converters. In contrast, SiC SBD

is free from the reverse recovery phenomena because it operates with the majority

carrier; thus, only the depletion charge is observed during turn-off operations.

Hence, SiC SBDs have less switching losses and their lower peak reverse currents

induce less surge voltages. However, their fast switching and small terminal

capacitances could cause high frequency oscillation and noticeable EMI noise in

high-frequency ranges. This section focuses on the turn-off operations of the

studied Si PiND and SiC SBDs and evaluates the switching characteristics of

diode currents during turn-off operations.


138


3.1 Diode switching characteristics

Figs. 2(b) and (c) show the measured diode current iak for turn-on and turn-off

switching operations at room temperature (25°C), respectively. Diode types are

given as parameters. During the turn-on operation, current behavior is almost

unaffected by the diode type, as shown in Fig. 2(b). STTH8L06’s measured current

in turn-off operations (Fig. 2(c)) shows the characteristic of bipolar devices with

large peak reverse currents. However, though they lead to ringing oscillation, SiC

SBDs show lower peak reverse currents and shorter recovery times than those of Si

PiND. TRS8E65C exhibits larger reverse current peaks and less damping in ringing

oscillation than those of IDH08SG60C.

Prony analysis is applied to the measured time-domain data to evaluate the

damping factor of the ringing oscillation mode [6, 7]. The damping factor of the

ringing oscillation for SiC SBD is calculated as 2:43 � 107/s and 3:53 � 107/s for

TRS8E65C and IDH08SG60C, respectively; the difference corresponds to the ESR

in diode blocking condition as shown in Fig. 1(c).

Fig. 2(d) shows the Fourier spectrum of measured diode current for each diode.

The spectrum will decrease with 20 dB/decade up to the break frequency for the

periodic trapezoidal waveform with rise and fall times. The large peak of the

reverse recovery current for Si PiND at 1–30MHz results in a 5–15 dB larger

spectrum level than occurs for SiC SBDs. The spectrum levels at 50MHz show no

significant difference among diode types. SiC SBDs exhibit spectrum peaks at

60–70MHz, which are caused by ringing oscillations during turn-off operations.

Prony analysis extracted ringing oscillation frequencies of 62.6 and 70.7MHz in

Fig. 1. The diodes’ measured static characteristics


139


the diode turn-off currents for TRS8E65C and IDH08SG60C, respectively. The

larger capacitance of TRS8E65C results in lower frequency spectrum peaks than

those in IDH08SG60C.

3.2 Measured spectrogram of the conducted emissions

This section presents the measured frequency spectrogram of the conducted

emissions at terminal disturbance voltage va. A mixed-domain oscilloscope

(Tektronix MDO4104-3) and line impedance stabilization network (LISN, 9117-

5-PJ-50-N, Solar Electronics Co., Ltd.) were used for measurement, as shown in

Fig. 2(a).

Fig. 3(a) shows the spectrum of va at 1–100MHz with a peak detection mode

spectrum analyzer. The spectrum levels above several MHz differ with the type of

diode used. At up to 30MHz, STTH8L06 exhibits spectrum levels 5–10 dB higher

than those of the SiC SBDs because of its larger reverse recovery current peak.

Not much difference shows in the spectrum levels and distribution of the two SiC

SBDs in the conducted emission regulation frequency range (CISPR22, 150 kHz–

30MHz). STTH8L06 exhibits the lowest spectrum peak level above 50MHz.

TRS8E65C and IDH08SG60C exhibit spectrum peaks around 60 and 70MHz,

respectively. The differences in the extracted ringing oscillation frequencies of

diode currents during turn-off operations agree with the Prony analysis results.

EMI noise is conventionally evaluated in the frequency domain. However, this

evaluation method assumes a signal’s time-invariant frequency distribution. This

study applies a time and frequency mixed-domain analysis to measure and

characterize EMI noise. The short-time fast Fourier transform (STFT) is used [8,

9]. The resulting STFT (called a “spectrogram”) can visualize both the frequency

and time behaviors of the signal [8, 9]. A mixed-domain oscilloscope was triggered

Fig. 2. Tested CCM boost converter configuration and diode currents


140


on the diode current and simultaneously captured a frequency spectrum of va

during the diode switching operation. Fig. 3(b) depicts the spectrogram of va up

to 100MHz. The horizontal linear axis represents frequency and the vertical axis

time; the color scale represents noise emission level. Fig. 3(b) also shows the

corresponding diode currents in the time domain for one switching cycle. The

emission level and spectrum distribution for diode turn-on operations do not differ

Fig. 3. The measured frequency spectrum of conducted emissions


141


with the type of diode. STTH8L06 exhibits the highest spectrum level, stemming

from its larger reverse current peak, at 1–30MHz; it also exhibits the lowest

spectrum levels above 50MHz during diode turn-off operations (Fig. 3(b)). The

ringing oscillations during SiC SBD turn-offs cause conducted emissions at

60–80MHz. Fig. 3(c) is the spectrogram of va for diode turn-off operations

magnified from 50–80MHz. TRS8E65C results in higher levels and less damping

noise spectrums than IDH08SG60C at 60–70MHz. The damping time constant of

the noise spectrum for IDH08SG60C is approximately 1.3 times shorter than that

for TRS8E65C. These differences agree with the extracted ESR shown in Fig. 1(c)

and the extracted damping of ringing oscillations obtained through Prony analysis.

Thus, diode turn-off characteristics have significant influence on the conducted

emissions level and its dynamic behavior in frequency ranges of >50MHz.

4 Conclusion

This study elucidates the static and switching characteristics of diodes that could

affect the conducted emissions of CCM boost converters. Because of their smaller

recovery currents, SiC SBDs give less line-conducted emissions levels at up to

30MHz than does Si PiND. The ringing oscillation of diode currents during turn-

off operations and EMI noise levels at frequency ranges of >50MHz depend on the

ESR and junction capacitance of SiC SBDs. Prony analysis could evaluate the

switching characteristics of diode currents during turn-off operations. The extracted

damping factor and ringing oscillation frequency are in good agreement with the

results of the spectrogram for conducted emissions. The spectrogram demonstrates

the transient characterization of the noise spectrum. Analysis results show that

diode turn-off characteristics influence conducted emissions in CCM DC–DC boost

converters. Thus, the analysis of time and frequency mixed domains can character-

ize both noise sources and EMI noise emissions.

Acknowledgments

This work was partially supported by Grant-in-Aid for Japan Society for the

Promotion of Science (JSPS) Fellows (DC2), and by Council for Science, Tech-

nology and Innovation (CSTI), Cross-ministerial Strategic Innovation Promotion

Program (SIP), “Next-generation power electronics” (funding agency: NEDO).


142


Configuration of MIMOsystem using single leakycoaxial cable for linear cellenvironments

Yafei Hou1,2a), Satoshi Tsukamoto2, Takahiro Maeda2,Masayuki Ariyoshi2, Kiyoshi Kobayashi2,Tomoaki Kumagai2, and Minoru Okada11 Graduate School of Information Science, Nara Institute of Science and Technology,

8916–5 Takayama, Ikoma, Nara 630–0195, Japan2 ATR Wave Engineering Laboratories,

2–2–2 Hikaridai, Keihanna Science City, Kyoto 619–0288, Japan


Abstract: It has been conventionally assumed that a single leaky coaxial

(LCX) cable only is utilized as one antenna to configure a multiple-input

multiple-output (MIMO) system. In this letter, we show one single LCX

cable can be used as two antennas to configure a MIMO system owing to the

designed intersection angle between the different radiation directivities of

both input signals. The measurement results of 2 × 2 LCX-MIMO channel

quality confirm that our proposed LCX-MIMO can realize a promising

channel condition over 5GHz frequency band. It also points out that the

intersection angle between radiation directivities of LCX cable needs to be

considered carefully to reduce the channel degradation at the both edge

portions of the LCX cable.

Keywords: leaky coaxial cable, multiple-input multiple-output, channel

quality

Classification: Wireless Communication Technologies

References

[1] J. Medbo and A. Nilsson, “Leaky coaxial cable MIMO performance in an indooroffice environment,” Proc. 23th IEEE international Symposium on Personal,Indoor and Mobile Radio Communications (PIMRC2012), Sydney, Australia,pp. 2061–2066, Sept. 2012. DOI:10.1109/PIMRC.2012.6362694

[2] Y. Hou, S. Tsukamoto, M. Okada, et al., “2 by 2 MIMO system using singleleaky coaxial cable for linear-cells,” Proc. 25th IEEE international Symposiumon Personal, Indoor and Mobile Radio Communications (PIMRC 2014),Washington, USA, pp. 408–412, Sept. 2014.

[3] Y. Hou, S. Tsukamoto, M. Ariyoshi, K. Kobayashi, T. Kumagai, and M. Okada,“Performance comparison for 2 by 2 MIMO system using single leaky coaxialcable over WLAN frequency band,” Proc. 25th Asia-Pacific Signal and Informa-tion Processing Association Annual Summit and Conference (APSIPA ASC2014), Siem Reap, Cambodia, pp. 1–6, Dec. 2014. DOI:10.1109/APSIPA.2014.7041735


143


http://dx.doi.org/10.1109/PIMRC.2012.6362694




http://dx.doi.org/10.1109/APSIPA.2014.7041735




[4] S. Tsukamoto, T. Maeda, M. Ariyoshi, Y. Hou, K. Kobayashi, and T. Kumagai,“An experimental evaluation of 2 × 2 MIMO system using closely-spacedleaky coaxial cables,” Proc. 25th Asia-Pacific Signal and Information ProcessingAssociation Annual Summit and Conference (APSIPA ASC 2014), Siem Reap,Cambodia, pp. 1–6, Dec. 2014. DOI:10.1109/APSIPA.2014.7041676

[5] Y. Hou, S. Tsukamoto, M. Ariyoshi, K. Kobayashi, and M. Okada, “4-by-4MIMO channel using two leaky coaxial cables (LCXs) for wireless applicationsover linear-cell,” Proc. 3rd IEEE Global Conference on Consumer Electronics(GCCE 2014), Tokyo, Japan, pp. 125–126, Oct. 2014. DOI:10.1109/GCCE.2014.7031177

[6] Y. Hou, S. Tsukamoto, M. Okada, et al., “Realization of 4-by-4 MIMO channelusing one composite leaky coaxial cable,” Proc. 12th Annual IEEE ConsumerCommunications and Networking Conference (CCNC 2015), Las Vegas, USA,pp. 103–108, Jan. 2015.

[7] J. W. Huang and K. K. Mei, “Theory and analysis of leaky coaxial cables withperiodic slots,” IEEE Trans. Antennas Propag., vol. 49, no. 12, pp. 1723–1732,Dec. 2001. DOI:10.1109/8.982452

1 Introduction

One of the hottest topics among the global telecoms industry is of how to improve

the spectrum efficiency for specific wireless network scenarios. There are many

scenarios known as linear-cell environments where the wireless transmission is

operated over a long and shallow place such as tunnels, along railways, under-

ground shopping malls, and so on. For these environments, a wireless system using

leaky coaxial (LCX) cable as an antenna for radio communication is widely used

because LCX has many potential advantages compared with conventional Omni-

directional antenna. For example, its coverage is uniform and the antenna place-

ments to deduce an interference might be simpler. In addition, the handover process

of conventional cells can be avoided when users are moving.

Recently thin LCX has been developed to be utilized for higher frequency

such as 2.4GHz band, 5GHz band for WLAN systems. To improve the spectral

efficiency of WLAN system over linear-cell environments, LCX cable can be

utilized as antenna to configure the multiple-input multiple-output (MIMO) chan-

nel. In paper [1], two independent LCX cables were applied to configure a 2 � 2

indoor MIMO system for corridor scenario and office landscape scenario, which

can realize an channel quality close to an independent and identically distributed

(i:i:d:) one. However, utill now, these researches have assumed that one LCX cable

is singlely treated as one antenna. Therefore it needs more than one LCX cable

to configure an MIMO system. This requires more cost for configuration of

MIMO, and large spacing between LCX cables is also necessary to reduce channel

correlation.

We have recently proposed a method that one single LCX cable can be used as

two antennas [2, 3, 4, 5, 6]. When different RF transmit signals are fed to each end

of the cable, the single LCX can work as two antennas. The results in [2] confirm

that, over 2.4GHz frequency band, the proposed MIMO channel using a single

LCX cable can realize good 2 � 2 channel condition even within a highly correlated© IEICE 2015DOI: 10.1587/comex.4.143Received March 26, 2015Accepted April 16, 2015Published May 19, 2015

144






http://dx.doi.org/10.1109/GCCE.2014.7031177




http://dx.doi.org/10.1109/8.982452

http://dx.doi.org/10.1109/8.982452

http://dx.doi.org/10.1109/8.982452

propagation condition. The 2 � 2 MIMO performance comparison between that

using the LCX cable and using Omni-direction antenna has been reported with

simulated results in [3] and with experimental evaluation in [4]. It also showed that

by adjusting the period of slot, we can use one composite cable, which consists of a

pair of LCX cables with different radiation directivities, to configure a 4 � 4MIMO

channel. The measurement results using the proposed LCXs over 5GHz band

confirm that the proposed composite LCX cable can realize a good MIMO channel

condition even if the spacing between the pair of LCXs is just 2 cm [5, 6].

Therefore the proposed LCX cable can reduce the space requirement for MIMO

deployment over linear-cell environments.

In this letter, we show the measurement results about the 2 � 2 MIMO channel

quality using single LCX with different radiation directivities over 5GHz frequency

band. The results confirm that our proposed LCX-MIMO can realize better channel

quality than the i:i:d: channel. It also shows that the intersection angle between

dominant radiation directivities of the LCX cable needs to be designed carefully to

reduce the channel degradation at the both edge portions of LCX cable.

2 LCX cable and its propagation direction

Fig. 1(a) shows an example of LCX cable structure. Radio wave is radiated and

received from slots which are periodically located along the outer conductor in the

cable. Signal strength of radio waves at the far field region from LCX is related to

the sum of the radiated radio waves from the slots. Radiation angles with peak

directivity are

�m ¼ sin�1ffiffiffiffi�r

p þ m�RFP

� �; ðm ¼ �1;�2; . . .Þ ð1Þ

where m is an order of harmonic, �r is relative permittivity of insulator in LCX, �RFis the wavelength of input radio frequency (RF) signal, and P is the period of slots

[7]. �m are propagation angles. To avoid radiation of harmonics, LCX is typically

Fig. 1. LCX cable structure and its propagation direction


145


designed as taking value m ¼ �1. By changing the period of slot P, the dominant

propagation angle of LCX ��1 is accordingly changed. Therefore, we can design

��1 as any nonzero value. As an example shown in Fig. 1(b), the dominant

propagation angles of two different input RF signal propagations from port 1

and 2 can have an intersection angle as 2��1. We can use this property to adjust the

value 2��1 to achieve a promising MIMO channel condition.

3 Proposed 2-by-2 LCX-MIMO system

3.1 Proposed 2-by-2 LCX-MIMO structure

On the basis of the previous explanation, we designed a 2 � 2 LCX MIMO system

using single LCX cable as shown in Fig. 2. Here station (STA) has two conven-

tional antennas. Since the LCX have dominant wave directions that is not 0 degree,

the access point (AP) exploits the MIMO technique using both ends of LCX cable

as the antenna inputs shown as port 1 and 2. Two different radiation wave

directivities are shown in the figure as A and B. The signal at one receiving

antenna can have low correlation compared with that at the other. If two receiving

antennas have appropriate spacing and keep an appropriate distance from center of

the LCX, two-stream MIMO can be achieved.

3.2 Measurement parameters

For the confirmation of the proposed LCX-MIMO performance over 5GHz

frequency band, 2 � 2 LCX-MIMO channel matrices between the LCX cable

and two receiving antennas were measured using a multi-ports vector network

analyzer in an anechoic chamber. The configuration of the measurement is also

shown with measurement points within the area in Fig. 2. Here X axis is the

direction along the LCX cable and Yaxis is the direction from the LCX cable to the

receiving antenna. An origin point is defined at the center of the LCX cable of 10

meters length. Each receiving antenna has omnidirectional radiation pattern. The

separation between the two receiving antennas is one wavelength or 6mm. Since

there are reduced reflection paths and the channel propagation is static in an

anechoic chamber, a shape of a cell that is formed by the LCX is assumed to be

symmetric against X axis and Y axis. We measure the channel with the equal

spacing of 25.4mm when STA is moving along X axis from 0 meter to 5 meter

which has 200 positions for each fixed Y.

Fig. 2. 2-by-2 MIMO system with proposed LCX cable


146


To show the channel condition of LCX with different radiation directivity, 2��1,we choose two types of vertically polarized (V-type) LCX for comparison. By

designing the slot period P using Eq. (1), the difference of an angle of dominant

radiation resulting from different input ports j2��1j can be set as 36Deg. (��1 ¼�18 degree) and 110Deg. (��1 ¼ �55 degree), respectively. Antenna gain of the

monopole antenna for STA is 1 dBi. Measurement bandwidth is 500MHz which is

from 5.15GHz to 5.65GHz. The frequency resolution is 1.25MHz. 401 samples

were obtained in each measurement position.

The slot period P for realizing angles of 36Deg. and 110Deg. are 40mm and

30mm respectively. The coupling loss and cable loss are 60 � 5 dB/1.5m and

0.6 dB/m. The inner copper wire diameter and insulator diameter are set as 2mm

and 5mm respectively. The outer sheathe thickness of LCX is 1mm. The VSWR

(voltage standing wave ratio) for the LCX cables is smaller than 1.2.

4 Channel condition comparison of proposed 2� 2 LCX MIMO

4.1 Metrics of channel quality

We use condition number (CN) γ as a metric of MIMO channel quality. First of all,

the measured 2 � 2 matrix H is decomposed using singular value decomposition

(SVD) as H ¼ U�V�. Here U and V are unitary matrices. � ¼ diagf�1; �2g where

�i is the ith singular value of H. The condition number γ [dB] is computed as

� ¼ 20 � log10ð�1=�2Þ. A matrix with a low condition number is said to be a well-

conditioned matrix which means that the MIMO channel has good condition for the

multiplexing capacity increase or large diversity order.

4.2 Channel measurement results

Fig. 3 shows the cumulative distribution function (CDF) of CN value γ for two

kinds of LCX cables. Fig. 3(a) shows the CDF of condition number for that using

LCX with the angle of 36 degree between two dominant radiation directivities. The

Y axis of measurement points is increased from 0.5m to 2.5m to evaluate the effect

on the channel condition from the distance between the STA and LCX cable. As

shown in Fig. 3(a), the LCX with 36 degrees between two dominant radiation

directivities can realize a promising channel condition for MIMO transmission.

About 82% and 90% CN values are smaller than 10 dB and 20 dB which shows a

better channel condition than that of i:i:d: MIMO channel. On the other hand, when

distance between the STA and LCX is increased from 0.5m to 2.5m, channel

condition is slightly improved.

Fig. 3(b1) shows the CDF of CN for that using LCX with the angle as 110

degrees between two dominant radiation directivities. In this case, when the

distance between the STA and LCX is increased from 0.5m to 2.5m, the MIMO

channel condition is largely degraded. The reason is that, due to the large angle

between two signals, almost only one propagation signal appears at the both edge

portions of LCX cable. The lack of the other signal reduces the probability of

realizing efficient MIMO channel. When the value of Y axis is increased, such edge

portion is enlarged which degrades the overall channel condition. We use Fig. 3(b2)

to demonstrate the reason. In this case, the measurement positions are just selected© IEICE 2015DOI: 10.1587/comex.4.143Received March 26, 2015Accepted April 16, 2015Published May 19, 2015

147


from 0 to 2m. We find that the proposed LCX-MIMO can achieve a better channel

condition than that of i:i:d: MIMO channel. Therefore it needs to be considered

how to improve the MIMO channel condition of the LCX edge area by selecting an

appropriate ��1.

5 Conclusions

This letter has showed the measurement results about the 2 � 2 MIMO channel

quality using a single LCX with different radiation directivities over 5GHz

frequency band. The results confirm that the proposed LCX-MIMO can realize a

better channel condition than that of the i:i:d channel. It also showed that its

coverage area needs to be considered carefully to reduce the channel condition

degradation at the both edge portions of LCX cable.

Acknowledgments

This work is supported by the Ministry of Internal Affairs and Communications

SCOPE (Strategic Information and Communications R&D Promotion Programme)

2014 with grant number as 135007001.

Fig. 3. CDF of condition number of 2 � 2 LCX-MIMO


148


Path loss model for the 2to 37GHz band in streetmicrocell environments

Minoru Inomata1a), Wataru Yamada1, Motoharu Sasaki1,Masato Mizoguchi1, Koshiro Kitao2, and Tetsuro Imai21 NTT Access Network Service Systems Laboratories, NTT Corporation,

1–1 Hikarinooka, Yokosuka, Kanagawa 239–0847, Japan2 NTT DOCOMO, INC., 3–6 Hikarino-oka, Yokosuka, Kanagawa 239–8536, Japan


Abstract: A path loss model based on the Rec. ITU-R P.1411 model, which

can cover the frequency range from microwave to millimeter-wave bands, is

presented. The path loss characteristics are analyzed on the basis of measure-

ment results obtained using the 2 to 37GHz band in street microcell

environments. It is clarified that the characteristics depend on distance from

transmitter to intersection and frequency dependency. By taking these

dependencies into account, the proposed model can decrease the root mean

square error of prediction results to within about 5 dB in the 2 to 37GHz

band.

Keywords: propagation, millimeter wave, street microcell environment

Classification: Antennas and Propagation

References

[1] NTT DOCOMO, INC., “DOCOMO 5G White Paper, 5G Radio Access:Requirements, Concept and Technologies”, July 2014.

[2] METIS, https://www.metis2020.com/.[3] Rep. ITU-R M.2135-1, “Guidelines for evaluation of radio interface technol-

ogies for IMT-Advanced,” ITU-R Report, vol. 1, M Series, ITU, Geneva, 2009.[4] Rec. ITU-R P.1411-7, “Propagation data and prediction methods for the

planning of short-range outdoor radiocommunication systems and radio localarea networks in the frequency range 300MHz to 100GHz,” ITU-R Report,vol. 7, P Series, ITU, Geneva, 2013.

[5] X. Zhao, T. Rautiainen, K. Kalliola, and P. Vainikainen, “Path-loss models forurban microcells at 5.3GHz,” IEEE Antennas Wireless Propag. Lett., vol. 5,pp. 152–154, Dec. 2006. DOI:10.1109/LAWP.2006.873950

[6] K. Sakawa, H. Masui, M. Ishii, H. Shimizu, and T. Kobayashi, “Microwavepath-loss characteristics in an urban area with base station antenna on top of atall building,” International Zurich Seminar on Broadband Communications,pp. 31-1–31-4, 2002. DOI:10.1109/IZSBC.2002.991774


149


https://www.metis2020.com/



http://dx.doi.org/10.1109/LAWP.2006.873950




http://dx.doi.org/10.1109/IZSBC.2002.991774




1 Introduction

Traffic in wireless communication systems has been rapidly increasing in recent

years and is assumed to reach 1000 times higher than the current traffic amount in

the next 10 years [1]. To cope with this growth, the application of both microwave

and millimeter-wave bands for the next generation mobile systems is being

examined since millimeter-wave bands can use wider frequency bandwidth that

can provide attractive higher throughput.

It is assumed that one of the possible service areas of mobile systems using

millimeter-wave bands is urban downtown areas, well known as street microcell

environments [2]. A lot of path loss models for street microcell environments have

been proposed; most of them can be categorized into two types. The first type is

two-slope models, of which the Rep. ITU-R M.2135 model [3] is the representative

example. However, the applicable frequency range of this model is only from 2 to

6GHz. The other is three-slope models, of which the Rec. ITU-R P.1411 model [4]

is the representative example. The applicable frequency range of this model is from

2 to 16GHz. However, no models covering millimeter-wave bands have yet been

reported.

A single path loss model that can cover the frequency range from microwave to

millimeter-wave bands is preferable in terms of measuring the consistency of

frequency characteristics, though to the authors’ knowledge no such models have

been reported. To address this deficiency, we propose a model that is based on the

Rec. ITU-R P.1411 model and can cover the frequency range from microwave to

millimeter-wave bands. Since the P.1411 model is constructed with parameters it

can extend the frequency range so as to cover bands from microwave to millimeter-

wave. However, since each of the coefficients of Rep. ITU-R M.2135 model are

optimized in the target frequency range, it will be necessary to reconstruct the

coefficients to extend the frequency range [5, 6].

2 Measurement parameters and environments

The path loss measurements were taken in the environment shown in Fig. 1(a). The

measurements were taken in four frequency bands 2, 4.7, 26, and 37GHz in order

to ascertain the frequency dependence of the measurements. The transmitter (Tx)

antenna was set to a height of either 6 or 10m and used to transmit continuous

waves. A receiver (Rx) antenna was fixed on the roof of a measurement car whose

height was 2.5m. The antenna directivity was omni-directional. The received

power level was measured while driving the measurement car along the line-of-

sight (LOS), non-line-of-sight (NLOS) 1, NLOS 2, and NLOS 3 routes as shown

in Fig. 1(b). The measurement procedure was conducted in Tokyo, Japan. The

distance from Tx antenna to intersection was about 242m for NLOS 1, 57m for

NLOS 2, and 169m for NLOS 3. Tall buildings of about ten stories about 40m

height surrounded the street in the measurement site.


150


3 Path loss characteristics in 2 to 37GHz band in NLOS street

microcell environments

3.1 Introduction of ITU-R P.1411 model and measurement results

In this subsection we discuss the path loss characteristics in the 2 to 37GHz band at

the LOS and NLOS 3 routes at hBS ¼ 6m as shown in Fig. 2(a). The path loss

predicted by the P.1411 model is obtained by using the following equations.

(i) The free space loss is obtained by using equations (1), (2), (3).

LLOS;m ¼ Lbp þ 6 þ20 log10

d1Rbp

� �for d1 � Rbp

40 log10d1Rbp

� �for d1 > Rbp

8>>><>>>:

ð1Þ

Lbp ¼ 20 log10�2

8�h1h2

� �� ð2Þ

Rbp ¼ 4ðhBS � hsÞðhMS � hsÞ

�ð3Þ

where Lbp is the value for the basic transmission loss at the breakpoint and Rbp is

the breakpoint distance, hBS is Tx antenna height, hMS is Rx antenna height, hs is

the effective road height, and d1 is the traveling distance calculated from equa-

tion (4).

d1 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix21 þ ðhBS � hMSÞ2

qð4Þ

where x1 is distance from Tx antenna to intersection. We assumed that hs ¼ 0.

(ii) The pass loss LC in the corner loss region is obtained by using equation (5), (6).

Lc ¼Lcorner

log10ð1þdcornerÞ log10ðd2 � w1=2Þ for w1=2 þ 1 < d2 < w1=2 þ 1 þ dcorner

Lcorner for d2 > w1=2 þ 1 þ dcorner

(

ð5Þd2 ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix22 þ ðhBS � hMSÞ2

qð6Þ

where the corner loss Lcorner is expressed as the additional attenuation over the

distance dcorner. Distance from intersection to Rx antenna is denoted by X2 and LOS

street width is denoted by w1. For an urban environment, the recommended Lcorner

value is 20 dB and the recommended dcorner value is 30m.

(a) LOS route from Tx antenna. (b) Measurement routes.

Fig. 1. Measurement environment.


151


(iii) The path loss Latt in an NLOS region is obtained from equation (7). The NLOS

region lies beyond the corner loss region, where a coefficient parameter β

applies. Total path loss LNLOS is calculated from equation (8).

Latt ¼10� log10

d1þd2d1þw1=2þdcorner

� �for d2 > w1=2 þ 1 þ dcorner

0 for d2 � w1=2 þ 1 þ dcorner

8<: ð7Þ

LNLOS ¼ LLOS;m þ Lc þ Latt ð8ÞFrom Fig. 2(a), it can be seen that from 0 to 200m the P.1411 model obtains the

good fit with measured results for the LOS and corner loss regions. Up to 200m the

root mean square error (RMSE) is about 3 dB, but after that it increases to about

19 dB at 2GHz and about 12 dB at 37GHz in the NLOS region. It is therefore

important to analyze the path loss characteristics in the NLOS region.

3.2 Analysis of the attenuation coefficient

We mainly analyzed the attenuation coefficient β in equation (7), which is used as a

fitting parameter in the NLOS region, in terms of its dependence on frequency, Tx

antenna height, and distance from Tx antenna to intersection. The coefficient is

obtained by calculating its maximum likelihood value such that the RMSE for

LNLOS and that for the measurement path loss are minimized. Fig. 2(b) shows the

dependency between the coefficient and the distance from Tx antenna to inter-

section. The coefficient is obtained by using the median value of each frequency

band. From Fig. 2(b) it can be seen that the coefficient decreases as distance from

Tx antenna to intersection increases. Fig. 2(c) shows the dependency between

the coefficient and the frequency. It shows that the coefficient increases as the

frequency increases in each route and for each antenna height. The coefficient has

frequency dependency because scattered and diffracted waves are dominant. It is

also affected by the distance from Tx antenna to intersection because the slope

values of approximation vary. However, it does not depend on the Tx antenna

height because the value of 6m Tx antenna height approximately corresponds to

the value of 10m Tx antenna height. Accordingly, its dependency on both

frequency and distance from Tx antenna to intersection need to be taken into

account.

3.3 Proposed path loss model

As the measurement results in 3.2 indicate, in order to predict the path loss in bands

from microwave to millimeter-wave it is important to take into account the

dependency on frequency and distance from Tx antenna to intersection. We there-

fore propose a path loss model that takes these into account for extending the

applicable frequency band. The model is shown below in (9). We assumed that the

effects of both the Tx-antenna-to-intersection distance and the frequency could be

formulated by multiplying logarithmic functions of frequency and distance from Tx

antenna to intersection. Here, the fitting parameters for the proposed model are

carrying out multiple regression analysis of the measurement results for both 6m

and 10m Tx antenna heights. The median value of fitting parameters is calculated.

The path loss calculation for the attenuation coefficient is substituted into equa-© IEICE 2015DOI: 10.1587/comex.4.149Received March 31, 2015Accepted April 14, 2015Published May 19, 2015

152


tion (9). The fitting parameters are calculated for �const ¼ 4:15, �f ¼ 1:41, Cf ¼�7:82, �x ¼ 0:79, and Cx ¼ �1:00. The unit of frequency is MHz and unit of

distance is m.

� ¼ �const þ ð�f log10 f þ CfÞð�x log10 x1 þ CxÞ ð9Þ

(a) Comparison of measurement and prediction results with P.1411 model.

(b) Dependency on distance from Tx to intersection.

(c) Frequency dependency.

Fig. 2. Measurement results.


153


3.4 Verifying validity of proposed path loss model

Fig. 3(a) shows path loss measurement results, prediction results obtained with the

P.1411 model, and prediction results obtained with the proposed model for the LOS

and NLOS 3 routes at hBS ¼ 6m. Quantitative evaluation results are depicted in

Fig. 3(b), where RMSE values for each frequency band are calculated. Black dots

indicate NLOS 1 with hBS ¼ 6m, blue dots NLOS2 with hBS ¼ 6m, green dots

NLOS 3 with hBS ¼ 6m, red dots NLOS 1 with hBS ¼ 10m, purple dots NLOS 2

with hBS ¼ 10m, and brown dots NLOS 3 with hBS ¼ 10m. From Fig. 3(a) it can

be seen that better fitting between predicted and measured values is obtained with

the proposed model than with the P.1411 model, with the former reducing the mean

RMSE value to about 11 dB. Fig. 3(b) shows that with the P.1411 model the RMSE

values were about 6.6 dB at 37GHz and 23.9 dB at 2GHz. The values decrease as

the frequency increases because the recommended attenuation coefficient value is

six, which is a large value to apply to the high frequency band. Fig. 3(b) shows that

with the proposed model the RMSE values are about 2.2 dB at 2GHz and 4.7 dB at

37GHz. This shows that the proposed model can cover the frequency range from 2

to 37GHz.

4 Conclusion

We analyzed path loss characteristics in the 2 to 37GHz band in street microcell

environments with the aim of developing a path loss model that can cover the

frequency range from microwave to millimeter-wave bands. The measurement

results clarified that path loss depends on distance from Tx antenna to intersection

and on frequency. We proposed a frequency extension model that takes account of

these dependencies and confirmed its validity by evaluating the RMSE values of

prediction results obtained with it. The evaluations confirmed it can predict path

loss with RMSE of less than 5 dB in the 2 to 37GHz band.

(a) Path loss in the 2 to 37 GHz band. (b) RMSE in the 2 to 37 GHz band.

Fig. 3. Comparison of proposed and conventional model results.


154


Performance improvement ofZF-precoded MU-MIMOtransmission by collaborativeinterference cancellation

Hidekazu Murataa) and Ryo ShinoharaGraduate School of Informatics, Kyoto University,

Yoshida-hommachi, Sakyo-ku, Kyoto 606–8501, Japan


Abstract: Multi-user multiple-input and multiple-output (MU-MIMO)

transmission has been extensively studied to enhance the spectral efficiency

of wireless communication systems. The performance of MU-MIMO suffers

due to the presence of residual multi-user interference. Linear interference

cancellation can offer excellent performance when a large number of

antennas are available. This, however, is impractical in a mobile station. In

this paper, we apply a collaborative interference cancellation scheme to a

precoded MU-MIMO system. In order to effect collaboration, we implement

received signal sharing among mobile stations using wireless local area

network connections. We carried out transmission experiments in an indoor

environment to test the performance of our precoded MU-MIMO system

with collaborative interference cancellation.

Keywords: multi-user MIMO, user cooperation, collaborative MIMO,

transmission experiment, interference canceller


References

[1] H. Murata, “Collaborative interference cancellation for multi-user MIMOcommunication systems,” IEICE Technical Report, RCS2013-201, pp. 159–164,Nov. 2013 (in Japanese).

[2] H. Murata, “Collaborative interference cancellation for future wirelesscommunications,” Proc. 2015 Vietnam–Japan International Symposium onAntennas and Propagation, Ho Chi Minh City, Vietnam, pp. 35–38, Jan. 2015.

[3] H. Kwon and J. M. Cioffi, “Multi-user MISO broadcast channel with user-cooperating decoder,” Proc. IEEE Vehicular Technology Conference, Sept.2008. DOI:10.1109/VETECF.2008.335

[4] J. Lee, S. Kim, H. Suman, T. Kwon, Y. Choi, J. Shin, and A. Park, “Downlink nodecooperation with node selection diversity,” Proc. IEEE Vehicular TechnologyConference, pp. 1494–1498, May 2005. DOI:10.1109/VETECS.2005.1543568

[5] M. Taniguchi, H. Murata, S. Yoshida, K. Yamamoto, D. Umehara, S. Denno,and M. Morikura, “Indoor experiment of multi-user MIMO user selection algo-rithm based on chordal distance,” Proc. Global Telecommunications Conference(GLOBECOM2013), Atlanta, GA, USA, Dec. 2013. DOI:10.1109/GLOCOM.2013.6831692


155


http://dx.doi.org/10.1109/VETECF.2008.335




http://dx.doi.org/10.1109/VETECS.2005.1543568




http://dx.doi.org/10.1109/GLOCOM.2013.6831692




1 Introduction

Multi-user multiple-input and multiple-output (MU-MIMO) transmission has been

extensively studied in order to improve the spectral efficiency of wireless commu-

nication systems. However, precoding techniques are inherently sensitive to chan-

nel variations. In other words, the performance of MU-MIMO is significantly

affected by the presence of residual multi-user interference. It is well-known that a

linear interference cancellation scheme with a large number of receive antennas can

offer satisfactory performance with reasonable computational complexity. How-

ever, it is impractical to expect such a number of antennas in mobile stations due to

size limitations.

In this paper, we apply collaborative interference cancellation (CIC) [1, 2] to a

precoded MU-MIMO system and test its performance through an experiment. We

employ received signal sharing among mobile stations [3, 4] in the proposed

scheme in order to enhance its interference cancellation capability. We conducted

transmission experiments using wireless local area network (LAN) links for

collaboration in an indoor environment in order to test the performance of our

collaborative interference cancellation scheme implemented on a MU-MIMO

testbed.

2 System model

2.1 Precoding

Consider downlink communication in a MU-MIMO system. For the sake of

simplicity, let us consider a single-carrier frequency in a single-coverage environ-

ment. The base station (BS) equipped with K antennas is serving M mobile stations

(MSs), each with a single antenna. We introduce a linear zero-forcing (ZF) precoder

by exploiting the downlink channel state information (CSI). The transmitted signal

to each MS is collected in the vector s 2 CM�1. In the context of our study, the

received signal of the MSs can be expressed as

y ¼ HWs þ n ¼ s þ n: ð1Þwhere H 2 C

M�K is a channel matrix and n 2 CM�1 is an additive white Gaussian

noise vector. The precoding matrix W 2 CK�M is given by

W ¼ Hy ¼ HHðHHHÞ�1: ð2ÞIf there exist channel variations �H 2 C

M�K , the off-diagonal elements of

ðI þ �HWÞ cause multi-user interference.

y ¼ ðH þ �HÞWs þ n ¼ ðI þ �HWÞs þ n ð3Þ

2.2 User selection algorithm

In order for the linear precoder to suppress interference, we need to select MSs (also

referred to as users) to serve simultaneously. The number of users cannot be greater

than the number of BS antennas. As a low-complexity user selection method

exploiting the CSI, we introduce the chordal distance-based user selection (CDUS)

algorithm [5].© IEICE 2015DOI: 10.1587/comex.4.155Received March 31, 2015Accepted April 16, 2015Published May 19, 2015

156


The chordal distance is given by

dcdðHcand;HselÞ ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiXJj¼1

sin2 �j

vuut

¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiJ � trð ~Hcand

~HHsel

~Hsel~HHcandÞ

qð4Þ

where Hcand 2 CJ�K is the downlink CSIs of candidate users, Hsel 2 C

NselectedJ�K is

the downlink CSIs of the selected users, �j is the principal angle between the two

subspaces spanned by row j of Hcand and the rows of Hsel, Nselected is the number of

selected users, J is the number of receive antennas of MS, ~½�� represents the Gram–

Schmidt orthonormalization of the rows.

Therefore, by selecting users with greater channel chordal distance among one

another, we can bring the channels of the selected users closer to the orthogonal,

hence improving the performance of linear precoding.

2.3 Collaborative interference cancellation

In order to suppress residual multi-user interference, we apply a collaborative

interference cancellation scheme by utilizing the short-range wireless interface of a

mobile station. In this paper, we employed a simple minimum mean squared error

(MMSE) weight WMMSE 2 CM�M. The weight is given by

WMMSE ¼ ðHWÞHRyy�1

¼ ðHWÞHðHWWHHH þ �Þ�1; ð5Þwhere Ryy 2 C

M�M and � 2 CM�M are autocorrelation matrices of the received

signals and the noise, respectively.

3 Transmission experiments

3.1 Equipment

The testbed used in this research consisted of a BS with four antennas (K ¼ 4) and

six MSs, both implemented using software-defined radio features. We employed

amplify-and-forward (AF) based channel estimation, CDUS, and linear spatial

precoding [5]. The major parameters of the experiment are shown in Table I.

The BS consisted of two instrument chassis and a radio frequency (RF) front

end that included RF amplifiers and RF switches. Power amplifiers (output power at

1 dB compression P1dB ¼ 34:6 dBm, typical) and low-noise amplifiers (noise figure

1.9 dB, typical) were connected to signal generators (SGs) and signal analyzers,

respectively. RF switches were used for duplexing. The control signals of the RF

switches were generated by a field-programmable gate array (FPGA). The trigger

signal was sent from an SG to the FPGA in order to synchronize the transmission

timing of the SGs with the control signals of the RF switches.

A universal software radio peripheral (USRP) was used as a mobile station. We

connected a general-purpose PC to a USRP N210, and carried out baseband signal

processing on the host PC. The motherboard of USRP N210 was equipped with 14-

bit analog-to-digital converters, 16-bit digital-to-analog converters, and an FPGA.

Digital signal processing, such as frequency translation, decimation, and interpo-

lation, was performed on the FPGA.


157


3.2 Signaling format

A transmission frame was composed of three time slots. The BS first transmitted the

training sequences (TS1s) for round-trip channel estimation at time t ¼ 0. The TS1s

from the BS were orthogonal sequences, and were transmitted at the same time and

frequency. Each MS sent back the received round-trip TS1 using amplify-and-

forward relaying along with another training sequence (TS2) for uplink estimation

at t ¼ 2ms. The BS estimated the downlink channel using the two TSs, and then

transmitted the precoded data packet at t ¼ 10ms. The data packet consisted of the

synchronization word (SW) used for timing synchronization, the training sequence

(TS3) used for demodulation, the control signal used for measurement automation,

and the data sequence along with its cyclic redundancy check.

We established timing synchronization using the SW, which was transmitted by

the BS in the third timeslot. Each MS detected the timing of the SW by means of a

sliding correlator. Channel estimation, user selection, and precoding weight gen-

eration were carried out at every transmission frame. The duration of a transmission

frame was set to 50ms in order to ensure that the MSs had sufficient time to

calculate and display the measurement results.

3.3 Experimental setup

We tested the performance of our proposed collaborative interference cancellation

scheme through a transmission experiment in an indoor environment. The experi-

ment was conducted in the foyer of the faculty of engineering building no. 3. The

foyer and the plan for it are shown in Fig. 1. The four mobile stations MS1∼MS4

moved at 40 cm/s. The other mobile stations MS5 and MS6 remained stationary.

Table I. Major parameters of experiment.

System parameters Value

Number of BS antennas 4

Number of MSs 6

Carrier frequency 5.11GHz

Modulation QPSK

Symbol rate 312.5 k symbols/s

Tx/Rx filter Square root Nyquist

(roll-off factor = 0.4)

BS parameters Value

Channel estimation AF-based two-way method

Precoding Linear (ZF)

Transmission power of TS1 7 dBm

Antenna height 2m

Antenna gain 5 dBi

MS parameters Value

Transmission power of TS2 7 dBm

Antenna height 88 cm

Antenna gain 3 dBi


158


Wireless LAN modules of the host PCs were employed as collaboration links

among mobile stations. The independent BSS mode channel 48 was used. The

received training signal (TS3) and the received data signal were sampled and

expressed in 20-bit I/Q data streams with 16-bit gain information. Due to possible

packet losses, the number of received signals available for interference cancellation

was not always six. Each MS independently demodulated the available signals. The

performance of the system was evaluated in terms of bit error rate (BER), and data

throughput is beyond the scope of this study.

3.4 Experimental results

In our experiments, we examined the CDUS and round robin (RR) schemes frame

by frame. In the RR scheme, ð 64Þ ¼ 15 groups of four MSs (M ¼ 4) were scheduled

once for transmission. The target received signal power of the precoded data signal

was set to −90 dBm. The performance was evaluated in terms of the BER of BS–

MS links, the number of MSs selected in nine frames, and the number of gathered

quantized received signals through Wi-Fi links. The BER was also averaged over

nine packets of each user selection scheme.

Fig. 2(a) shows the performances of MS1 against the packet index. As can be

seen, in the case not involving CIC (w/o CIC), CDUS offered better BER than RR

at all times, except for the packet index from 10 to 18 and from 28 to 36. In the case

involving the use of CIC (w/ CIC), there was no obvious difference between the

performance of CDUS and RR.

Fig. 1. Experimental equipment installation.

Fig. 2. Measured performance.


159


Fig. 2(b) shows the cumulative distribution function (CDF) of the measured

BER of MS1∼MS4. As can be seen from this figure, the collaborative interference

cancellation scheme clearly outperformed independent reception (w/o CIC) for

both the cases involving CDUS and RR. This also confirmed that CDUS is superior

to RR. Moreover, we confirmed that CIC was effective even when a user selection

scheme took orthogonality among users’ channels into account in a user mobility

scenario.

4 Conclusion

In this paper, we applied a collaborative interference cancellation scheme to a ZF-

precoded MU-MIMO system to eliminate residual multi-user interference. Through

a transmission experiment, we confirmed that received signal sharing among

mobile stations can enhance transmission performance, especially among dynamic

mobile stations.

Acknowledgments

This work was supported by the Strategic Information and Communications R&D

Promotion Programme (SCOPE) of the Ministry of Internal Affairs and Commu-

nications, Japan.


160


Dual polarized open-endedwaveguide using 7-layerPTFE board

Takashi Maruyamaa), Satoshi Yamaguchi, Tomohiro Takahashi,Masataka Otsuka, and Hiroaki MiyashitaInformation Technology R&D Center, Mitsubishi Electric Corporation,

5–1–1 Ofuna, Kamakura, Kanagawa 247–8501, Japan


Abstract: We propose a wideband and dual polarized open-ended wave-

guide for planar array antennas. The proposal forms most of the antenna

structures including waveguide and feeding networks in the dielectric sub-

strates. This yields two advantages. First, it reduces the antenna dimensions

because the waveguide is filled with dielectric; this yields 30% bandwidth

and short element spacing without grating lobes. Second, it simplifies the

machining process, especially for the case of large scale arrays consisting of

several hundred elements. We design antenna elements and a four-element

array, and then fabricate the four-element array using a 7-layer board of

PTFE. Though the calculated and measured reflection characteristics diverge

slightly, the radiation patterns well match.

Keywords: dual polarization, open-ended waveguide, planar array, multi-

layer board

Classification: Antennas and Propagation

References

[1] S. Vaccaro, D. Llorens del Rio, J. Padilla, and R. Baggen, “Low cost Ku-bandelectronic steerable array antenna for mobile satellite communications,” Proc.5th EUCAP, pp. 2362–2366, Apr. 2011.

[2] J. M. Inclán-Alonso, A. García-Aguilar, L. Vigil-Herrero, J. M. FernandezGonzalez, J. SanMartín-Jara, and M. Sierra-Perez, “Portable low profile antennaat X band,” Proc. 5th EUCAP, pp. 2043–2047, Apr. 2011.

[3] H. Arai and T. Izumi, “Wideband dual-polarized small tapered slot antenna forMIMO emulator,” IEEE AP-S, pp. 438–439, 2013. DOI:10.1109/APS.2013.6710880

[4] T. Maruyama, S. Yamaguchi, T. Takahashi, and H. Miyashita, “Design andFabrication of Wideband and Dual Polarized Open-Ended Waveguide,” IEEETrans. Antennas Propag., vol. 62, no. 9, pp. 4872–4876, Sept. 2014. DOI:10.1109/TAP.2014.2333093


161


http://dx.doi.org/10.1109/APS.2013.6710880




http://dx.doi.org/10.1109/TAP.2014.2333093




1 Introduction

Active phased array antennas (APAA) are attractive for vehicular wireless commu-

nications because of their low profile and lack of moving parts [1]. However, gain

loss for beam scanning is unavoidable and the many RF components raise costs

excessively. Our solution is a mechanically driven planar array where all elements

are excited in phase. Our goal is to realize a mechanically driven planar array

yielding dual polarization and 30% frequency bandwidth. Because the array

antenna with several hundred elements is required for high gain, the antenna

structure must be easy to make. Because all elements are connected to a passive

beam forming network, low loss antenna elements are also required. This paper

proposes solutions to meet all of these requirements.

The first candidate for an array element with wideband and dual polarization is

the stacked patch antenna [2]. It has simple structure and it may achieve 30%

bandwidth. Unfortunately, relatively large conductor loss and dielectric loss are

unavoidable. The second candidate is the tapered slot antenna [3]. It is suitable for

array antennas because it can be fabricated on the dielectric substrate. Unfortu-

nately, its dimension on the boresight axis is relatively long. Moreover, large scale

arrays are hard to manufacture because the many substrates must be aligned in grid

pattern.

We focused on the open-ended waveguide in earlier work [4]. Our design

places strip lines in the substrate to feed the multiple elements. We designed the

antenna and confirmed that the calculated and measured results agree well.

However, it has two issues. The first one is antenna size. Waveguide size must

be 0:5�l or wider to pass the dominant mode, where �l is the wavelength at the

lower limit frequency. On the other hand, antenna spacing must be 1:0�h or shorter

to avoid grating lobes, where �h is the wavelength at the upper limit frequency. Our

design has difficulty in satisfying both constraints due to waveguide size. The

second issue is manufacturing complexity. The design stacks multiple metal parts

and dielectric substrates which raises machining and assembly costs. Additionally,

it was difficult to connect the signal line from the antenna bottom to the strip line

because the line must pass through metal parts and dielectric substrates. In this

paper, we drastically change the antenna structure. Most antenna structures,

including the waveguide and feeding circuits, are formed in the multi-layer board.

We fabricate 7-layer board of PTFE substrates to reduce dielectric loss. Though

PTFE multi-layer boards are not common due to their viscosity, this structure has

significant advantages. The antenna dimensions are reduced because the waveguide

is filled with dielectric. This achieves the 30% frequency bandwidth and the

antenna spacing of 1:0�h or shorter. Additionally, when several hundred elements

are required, the machining process can be reduced because etching is used. The

vertical signal line from the antenna bottom to the strip line is also easily realized

by through holes formed when fabricating the multi-layer board. We designed the

antenna element and four-element array with antenna spacing of 1:0�h or shorter.

We fabricated the four-element array. Though there was some discrepancy between

the calculated and measured reflection characteristics, the radiation patterns agreed

well.© IEICE 2015DOI: 10.1587/comex.4.161Received March 17, 2015Accepted April 21, 2015Published May 19, 2015

162


2 Antenna structure

The proposed element antenna structure is shown in Fig. 1(a). The antenna consists

of a multi-layer board together with a metal aperture. ‘L1’ to ‘L8’ are copper layers.

The 8-copper layers and 7-dielectric substrates are united to form one multi-layer

board. L2 is the strip line and excitation probe layer for V polarization. L1 and L3

are its ground layers. L4 is the strip line and excitation probe layer for H polar-

ization. L3 and L5 are its ground layers. Each polarization has opposite two

excitation probes in a symmetrical structure. They yield small coupling between

the orthogonal polarizations and low sidelobes. The strip lines of both polarizations

can not be co-layered when opposite two excitation probes are connected to

branched strip line, explained in Section 3, so the two feeding networks are

stacked. However, this structure creates a characteristic difference between the

polarizations. L7 is a slits layer for characteristic equalization. The slits act as

reflectors for V polarization because the slits are parallel to the electric field. The

electric field of H polarization, which is perpendicular to the slits, passes through

the slits and is reflected at the antenna ground L8. This equalizes the characteristic

difference between the polarizations.

(a) Exploded perspective view (b) Outline of copper layers

(c) Active reflection

Fig. 1. Element antenna structure and its active reflection


163


We designed the antenna element using HFSS, an electromagnetic field

simulator. The dielectric substrate is PTFE composite with the relative permittivity

of around 2.1. The calculation model has four feeding points at the edge of strip

lines. The aperture size is 0:79�, where λ is center frequency wavelength. This

corresponds to 0:91�h. The waveguide size, the white square in L1, 3, 5, 6, is 0:58�.

It corresponds to 0:49�l. In the proposed antenna, the waveguide size is mini-

aturized because the waveguide is filled with PTFE; the remaining space holds the

strip line. Its design is shown in Section 3. Through holes are placed to form the

conducting walls of the waveguide. The probe length is adjusted so as to decrease

reflection. The length was 0:29� for V polarization and 0:31� for H polarization.

Fig. 1(c) shows the active reflection plots. The plot of V-pol (H-pol) means the

active reflection yielded when the two signal lines of L2 (L4) are excited in

reversed phase. The two plots are similar due to L7 slits. The active reflection

value is −10 dB or lower for both polarizations in the 30% bandwidth.

3 Design of four-element array

We designed and fabricated a four-element array; the four elements were strip line

fed. A picture of the prototype is shown in Fig. 2(a). Its side view is in (b). Though

the prototype has six apertures, each polarization uses four apertures. Due to the

flange size of the SMA connector at the antenna bottom, the two connectors could

not be closely placed, so the four apertures used are shifted by one column for the

two polarizations. This is a limitation of just this prototype. When the coaxial line is

directly connected to the feeding network for a large scale array, no SMA connector

is used and this shift is not required. Fig. 2(c) shows L2 feeding circuit pattern. The

antenna spacing is 0:97�h even if the strip line is placed between waveguides.

Because the antenna spacing does not exceed 1:0�h, no grating lobe is radiated. The

strip line length is adjusted so that the opposite probes are excited in reversed

phase. The cable core of the coaxial line passes through the dielectric substrate and

is connected to the strip line. The outer conductor of the coaxial line in the substrate

section is formed by copper layers and through holes. When the SMA connector is

not required for a large scale array, the cable core of the coaxial line is formed by

using a through hole. This simplifies connection between the strip line and the

antenna bottom. This is an additional advantage of the multi-layer board. L4

feeding circuit pattern is 90 degree rotated version of Fig. 2(c).

Reflection coefficient plots are shown in Fig. 3(a) and (b). Measured and

calculated pattern shapes agree. However the measured frequency characteristics

are shifted by about 5% and the measured level is higher. This discrepancy is

probably due to the difference in the relative permittivity of PTFE between the

calculation and the actual value. This antenna is sensitive to the relative permittivity

because the waveguide between L1 to L8 in Fig. 1(a) is filled with PTFE. The

actual permittivity can be estimated by comparing measured and calculated values.

The discrepancy will be reduced if the design is modified by using the actual

permittivity. Sum of dielectric loss and conductor loss does not exceed 0.4 dB

within the frequency bandwidth in the calculation. Most losses occur in the strip

line. This small loss may be acceptable because the subsequent beam forming© IEICE 2015DOI: 10.1587/comex.4.161Received March 17, 2015Accepted April 21, 2015Published May 19, 2015

164


network is simplified even for large scale arrays. Fig. 3(c) and (d) are radiation

patterns at the center frequency when the V polarization feeding circuit is excited.

Fig. 3(e) and (f ) are radiation patterns at the upper limit frequency; they show that

no grating lobe is observed. Calculated and measured patterns approximately

agreed including the cross polarization pattern. The worst cross polarization value

on the boresight was −34 dB. Radiation patterns of H polarization were similar to

(but rotated) those of V polarization.

(a) Picture of four-element array (b) Side view

(Each polarization uses four apertures)

(c) Feeding circuit

Fig. 2. Four-element array


165


4 Conclusion

We proposed a wideband and dual polarized open-ended waveguide for array

antennas. Most antenna structures including strip lines and waveguide cavity are

formed in a multi-layer board. A slits structure to equalize the frequency character-

istics of polarizations is also included. This structure enables both 30% bandwidth

and small antenna spacing because the waveguide is filled with dielectric. We

designed and fabricated a four-element array with no grating lobes within the 30%

bandwidth. Though the measured frequency characteristics of the reflection differed

by about 5% due to a difference in the relative permittivity value assumed, the

calculated and measured radiation patterns approximately agreed.

(a) Reflection of V polarization (b) Reflection of H polarization

(c) Radiation pattern (φ=0deg, Fc)

(e) Radiation pattern (φ=0deg, Fh)

(d) Radiation pattern (φ=90deg, Fc)

(f) Radiation pattern (φ=90deg, Fh)

Fig. 3. Measured and calculated results


166


Feasibility of RSSI based60GHz WLAN discovery formulti-band WLAN

Sho Wadaa), Masahiro Umehirab), Shigeki Takedac),Teruyuki Miyajimad), and Kenichi Kagoshimae)

College of Engineering, Ibaraki University,

4–12–1 Nakanarusawa-cho, Hitachi-shi, Ibaraki, Japan


b) [email protected]

c) [email protected]

d) [email protected]

e) [email protected]

Abstract: This letter proposes RSSI (Received Signal Strength Indicator)

based 60GHz WLAN discovery for multiband WLAN (Wireless LAN),

which uses both 2.4/5GHz band and 60GHz band to achieve high trans-

mission speed, sufficient reliability and power-saving of a multiband WLAN

device. The proposed 60GHz WLAN discovery detects 60GHz WLAN

coverage by using RSSI of 2.4/5GHz WLAN signals. Space/time diversity

is employed to improve the false detection probability of 60GHz WLAN

discovery. Ray-tracing simulation results confirm that 60GHz WLAN dis-

covery is feasible and space/time diversity is effective to improve the

reliability of 60GHz WLAN discovery.

Keywords: RSSI, multiband WLAN, 60GHz, space/time diversity


References

[1] I. Toyoda, F. Nuno, Y. Shimizu, and M. Umehira, “Proposal of 5/25-GHz dualband OFDM-based wireless LAN for high-capacity broadband communica-tions,” PIMRC 2005, vol. 3, pp. 2104–2108, Sept. 2005. DOI:10.1109/PIMRC.2005.1651810

[2] M. de Courville, S. Zeisberg, M. Muck, and J. Schönthier, “IST Project:BroadWay—The way to broadband access at 60GHz,” PWC 2003, vol. 2775,pp. 219–221, Sept. 2003. DOI:10.1007/978-3-540-39867-7_24

[3] H. Singh, J. Hsu, L. Verma, S. S. Lee, and C. Ngo, “Green operation of multi-band wireless LAN in 60GHz and 2.4/5GHz,” CCNC 2011, pp. 787–792, Jan.2011. DOI:10.1109/CCNC.2011.5766599

[4] M. Umehira, G. Saito, S. Wada, S. Takeda, T. Miyajima, and K. Kagoshima,“Feasibility of RSSI based access network detection for multi-band WLANusing 2.4/5GHz and 60GHz,” WPMC 2014, pp. 243–248, Sept. 2014. DOI:10.1109/WPMC.2014.7014824

[5] EEM Inc., “EEM-RTM,” http://www.e-em.co.jp/rtm/eem_rtm.htm (2014/2/1access) (in Japanese).© IEICE 2015

DOI: 10.1587/comex.4.167Received April 3, 2015Accepted April 14, 2015Published May 28, 2015

167






http://dx.doi.org/10.1007/978-3-540-39867-7_24

http://dx.doi.org/10.1007/978-3-540-39867-7_24

http://dx.doi.org/10.1007/978-3-540-39867-7_24

http://dx.doi.org/10.1109/CCNC.2011.5766599




http://dx.doi.org/10.1109/WPMC.2014.7014824




http://www.e-em.co.jp/rtm/eem_rtm.htm







1 Introduction

Multiband WLAN using both 2.4/5GHz and 60GHz or other higher frequency

band has been proposed to achieve both high transmission speed and sufficient

reliability [1, 2, 3]. In the multiband WLAN, 60GHz is used to achieve high

transmission speed and 2.4/5GHz is used if 60GHz is unavailable due to human

body shadowing for example. Power-saving is an important issue for a battery-

operated multiband WLAN device because power consumption of the multiband

WLAN device can be very large if both 2.4/5GHz and 60GHz radio parts are

always turned on. Therefore, a novel power saving technique is needed to turn on

60GHz radio only if the multiband WLAN device is within 60GHz WLAN

coverage. This is a spectrum efficient approach if 60GHz is used instead of

2.4/5GHz as far as 60GHz can be used. This usage scenario requires a multiband

WLAN device to detect 60GHz WLAN coverage using 2.4/5GHz WLAN radio

instead of 60GHz WLAN radio [4].

This letter proposes RSSI based 60GHz WLAN discovery for multi-band

WLAN, by using 2.4/5GHz WLAN signals without using 60GHz WLAN radio.

Assuming that a line-of-sight (LOS) path is available in 60GHz WLAN, we can

expect LOS path is also available in 2.4/5GHz, thus, 60GHz RSSI is in proportion

to 2.4/5GHz RSSI. However, multi-path propagation in 2.4/5GHz WLAN can

cause 2.4/5GHz RSSI variation, and degrade miss- and false-detection of 60GHz

WLAN coverage. To solve this problem, we introduce space/time diversity to

improve the reliability of 60GHz WLAN discovery. Ray-tracing simulation results

confirm that 60GHz WLAN discovery is feasible and space/time diversity is

effective to improve the reliability of 60GHz WLAN discovery.

2 Operation of multiband WLAN

We assume that multiband WLAN AP (Access Point) and STA (Station) have both

2.4/5GHz and 60GHz radio parts. As 2.4/5GHz WLAN has wider coverage than

60GHz WLAN, 2.4/5GHz WLAN can be used whenever 60GHz WLAN can be

used. A multiband WLAN AP always uses both 2.4/5GHz and 60GHz radio parts,

however 60GHz radio part needs to be turned-off or in a sleep mode when a

multiband WLAN STA is out of 60GHz WLAN coverage. 60GHz radio part is

turned on or waked up as soon as 60GHz WLAN coverage is detected and 60GHz

link is used to provide high transmission rate as far as 60GHz link is available. If

60GHz link is blocked due to moving human body for example, wireless con-

nection is immediately switched from 60GHz to 2.4/5GHz to maintain the

wireless connection. This operation strategy of multiband WLAN enables to

efficiently utilize valuable 2.4/5GHz spectrum in combination with 60GHz.

3 Principle of RSSI based 60GHz WLAN discovery

Since a directional antenna is used and LOS is maintained in 60GHz WLAN, we

can assume that propagation of 60GHz WLAN is almost free space propagation,

where a direct wave is dominant and received level of other reflected waves are

much smaller than that of a direct wave. Friis formula can be used to calculate

received signal power of 60GHz, Pr60G, which is given by:

© IEICE 2015DOI: 10.1587/comex.4.167Received April 3, 2015Accepted April 14, 2015Published May 28, 2015

168


Pr60G ¼ �60G4�r

� �2

Pt60GGt60GGr60G; ð1Þ

where r is the distance between a transmitter and a receiver, �60G is wavelength of

60GHz and Pt60G is transmission power. Gt60G and Gr60G are antenna gain of a

transmitter and a receiver.

In a multiband WLAN STA using 2.4GHz and 60GHz, LOS is maintained and

a direct wave is dominant in 2.4GHz if STA is within 60GHz WLAN coverage.

However, as omni-directional antenna is generally used in 2.4GHz WLAN,

2.4GHz RSSI can vary more than 60GHz RSSI due to multi-path propagation.

In this case, the instantaneous received power at 2.4GHz, P is given by:

P ¼ 1

2ErðtÞErðtÞ� ¼ 1

2

Xkm¼0

Amej!ðt��mÞ

! Xkm¼0

Amej!ðt��mÞ

!�

¼Xkm¼0

Am2

2þXk�1m0¼0

Xkm¼m0

AmAm0 cos!ð�m0 � �mÞ;ð2Þ

where ErðtÞ is electric field strength at the receiver, Am is amplitude of the m-th

path, �m is delay time of the m-th path, k is the number of multi-paths, and m ¼ 0

means a direct wave. In this equation, the first term means averaged RSSI and the

second term means RSSI variation due to multi-path propagation. The second term

becomes zero if averaging is performed with sufficiently large number of samples.

Averaged received signal power of 2.4GHz, Pr2:4G is approximated by:

Pr2:4G ¼ Pt2:4GGt2:4GGr2:4G

Xkm¼0

�2:4G4�rm

� �2

R2m

!ffi Pt2:4GGt2:4GGr2:4G

�2:4G4�r0

� �2

; ð3Þ

where �2:4G is wavelength of 2.4GHz and Pt2:4G is transmission power. Gt2:4G and

Gr2:4G are antenna gain of a transmitter and a receiver, and Rm is reflection

coefficient of the m-th path. In LOS propagation condition in 2.4GHz WLAN,

received level of a direct wave is assumed to be larger than that of other reflected

waves. As Pt60G, Gt60G, Gr60G, Pt2:4G, Gt2:4G, and Gr2:4G are known design parame-

ters, we obtain the following equation from the Eq. (1) and (3):

Pr60G ¼ Pr2:4G�260GPt60GGt60GGr60G

�22:4GPt2:4GGt2:4GGr2:4Gð4Þ

This equation indicates that 60GHz RSSI can be estimated by averaging multiple

samples of 2.4GHz RSSI in LOS condition. However, we need various 2.4GHz

RSSI samples in various propagation conditions for averaging. We propose to apply

diversity techniques for this purpose. Supposing that N is the number of space

diversity branches and M is the number of time diversity branches, the total number

of diversity branches, L is given by N �M and L samples of 2.4GHz RSSI can be

averaged to estimate 60GHz RSSI. Based on Eq. (4), the estimated received power

Pr2:4G is given by:

Pr2:4G ¼ 10 log101

L

XLi¼1

10PR2:4GðiÞ

10

!½dB� ð5Þ

where PR2:4GðiÞ is the measured RSSI at each diversity branch i.© IEICE 2015DOI: 10.1587/comex.4.167Received April 3, 2015Accepted April 14, 2015Published May 28, 2015

169


4 Performance evaluation

4.1 Simulation conditions

Ray-tracing simulation was conducted to evaluate 60GHz RSSI estimation error

according to 2.4GHz RSSI when space/time diversity is employed. Three branch

space diversity (N ¼ 3) is employed at AP, where each antenna is vertically located

with antenna separation of 20 cm [2]. Regarding time diversity, we assume STA

moves to 9 directions by about 1m, thus number of time diversity branch, M is 9.

We assume a large conference room of 11m � 15m � 2:8m and an auditorium

of 22m � 30m � 5:0m for ray-tracing simulations. Material of two rooms is

assumed to be concrete. In the large conference room, AP is set at ðx; yÞ ¼ð12:5; 10:7Þ, AP antenna height is 1.8m, 2.0m, 2.2m and STA antenna height is

1m. The simulated number of STA positions is 19 � 27 ¼ 513 where STA position

spacing is 0.5m in both x- and y-axis. In the case of auditorium, AP is set at

ðx; yÞ ¼ ð27:5; 21:7Þ and the number of STA positions is 21 � 29 ¼ 609 where STA

position spacing is 1.0m.

4.2 Simulation results

60GHz WLAN parameters are based on IEEE802.11ad, and 2.4GHz WLAN

parameters are based on IEEE802.11n for multiband WLAN. Ray tracing simu-

lation software, EEM-RTM using a ray-launching method is employed for simu-

lations and the maximum numbers of reflection and diffraction are three [5].

2.4GHz RSSI is calculated by summing up the power of 56 subcarriers with

subcarrier spacing of 312.5 kHz, considering frequency selective fading. 60GHz

RSSI is calculated for 2GHz bandwidth signals. Fig. 1 shows a correlation diagram

of 2.4GHz RSSI and 60GHz RSSI at a conference room. As shown here, rms (root

mean square) 60GHz RSSI estimation error for the prediction line is 3.39 dB in (a),

2.84 dB in (b), and 1.92 dB in (c). The results confirm that space/time diversity is

effective to reduce RSSI estimation error. In the auditorium, space/time diversity

with L ¼ 3 � 9 can reduce rms RSSI estimation error from 3.73 dB to 2.44 dB.

4.3 False-detection of 60GHz coverage

False-detection probability of 60GHz coverage is evaluated using the correlation

diagram of 2.4GHz and 60GHz RSSI. Fig. 3 illustrates principle of 60GHz

(a) Without diversity (c) Space/time diversity (L =3×9)(b) Space diversity (L =3)

Fig. 1. Correlation diagram of 2.4GHz and 60GHz RSSI at a largeconference room.


170


coverage estimation by using 2.4GHz RSSI. Supposing that 60GHz coverage

needs the received power more than −45 dBm, the correlation diagram is divided

into four areas. Area (2) and (3) mean correct detections, i.e. “within 60GHz

coverage” and “out of 60GHz coverage”. Area (1) means miss detection, i.e.

“60GHz cannot be used though it can be used”, and area (4) means false-detection,

i.e. “60GHz can be used though it cannot be used”. If false detection happens,

60GHz RF part is turned on though 60GHz WLAN is unavailable, thus battery

power is wasted. False detection probability can be reduced by introducing

detection threshold margin as shown in Fig. 3. Large detection margin can reduce

false detection but increase miss detection. Miss detection and false detection is a

trade-off issue and larger detection margin makes 60GHz coverage effectively

smaller. Therefore, detection margin should be as small as possible.

Fig. 3 shows false detection probability as a function of detection threshold

margin. The results show that space/time diversity decreases the detection thresh-

old margin by 2 dB in the conference room and by 1.5 dB in the auditorium. This

means that 60GHz coverage can be increased by 41%∼58% while the false-

detection probability is maintained at 0.05 or 0.02.

Fig. 2. Principle of 60GHz coverage estimation by 2.4GHz RSSI.

Fig. 3. False detection probability as a function of detection thresholdmargin.


171


5 Conclusion

This letter proposes RSSI based 60GHz WLAN discovery for multi-band WLAN.

Ray-tracing simulation results confirm that RSSI based 60GHz WLAN discovery

is feasible and space/time diversity is effective to improve detection reliability.

Acknowledgments

This work is supported by the Ministry of Internal Affairs and Communications of

Japan, under the grant of research and development for radio resource enhance-

ment, “multi-band/multi-mode sensor wireless communications technologies”.


172


Documents

IEICE Communications Express, Vol.4, No.5, 136 142 Time ... · 3 Diode switching characteristics and conducted emissions for the CCM boost converter Fig. 2(a) depicts the CCM boost