Professional Documents
Culture Documents
Abstract—Millimeter wave (mmWave) communication has been cular arrays. Path loss, shadowing, cell coverage, polarization
a key technology for the fifth generation (5G) wireless communica- properties, and RMS DS were obtained. In [10], 15 GHz
tions. There have been various mmWave channel measurements. channel measurements were conducted using a VNA and a
However, many measurements in the literature are conducted with
different configurations, which may have large impacts on the spectrum analyzer with 1 GHz bandwidth in two indoor cor-
propagation channel characteristics, and make the comparison of ridor environments. Parameters like large-scale fading, Ricean
propagation channel characteristics for different mmWave bands K-factor, and RMS DS were obtained, as well as their statistical
impossible. In this paper, we carry out channel measurements at properties. In [11], a sliding correlator based direction-scan-
11, 16, 28, and 38 GHz bands in an indoor environment using a sounding method was used to measure the 28 GHz indoor
vector network analyzer (VNA). The space-alternating generalized
expectation-maximization (SAGE) algorithm is used to obtain office environment with 500 MHz bandwidth. Both line-of-
the multipath component (MPC) parameters including three sight (LOS) and non-line-of-sight (NLOS) scenarios were
dimensional (3D) angular domain information. The propagation considered, and the SAGE algorithm was applied to estimate
characteristics like average power delay profile (APDP), power delay and angular parameters of MPCs. In [12], the authors
azimuth profile (PAP), power elevation profile (PEP), root mean conducted 28 GHz measurements in an urban environment
square (RMS) delay spread (DS), and azimuth and elevation
angular spread (AS) are shown and compared for the four with 250 MHz bandwidth. The transmitter (Tx) antenna was
frequency bands. The results show similar properties for different scanned in both azimuth and elevation angle domains. After
bands and indicate the possibility of the derivation of a unified post-processing the angular power delay profiles (PDPs) and
channel model framework for 10–40 GHz bands. omni-directional measurements were synthesized. In [13], a
Index Terms—Millimeter wave, 5G, SAGE, parameter estima- commercial backhaul equipment was used to conduct channel
tion.
measurements at 38 GHz in urban outdoor and outdoor-to-
indoor environments. The LOS propagation, reflection, scat-
I. I NTRODUCTION
tering, diffraction, transmission, as well as polarization effects
MmWave communication has been considered as a key tech- were studied. Authors in [14] measured the 60 GHz channel
nology for the 5G wireless communications [1], e.g., for wide- with 4 GHz bandwidth in indoor conference room and office
band cellular communication (hotspot and small cell), wireless environments. At both ends of Tx and receiver (Rx), virtual
backhaul, indoor and device-to-device (D2D) communications planar array was utilized to investigate the influence of array
[2]. MmWave generally corresponds to the 30–300 GHz bands, size on beamforming performance. Prof. T. S. Rappaport et
but sometimes the 10–30 GHz bands are also included as they al. from New York University had also conducted various
share some similar propagation characteristics. Compared to the channel measurements at 28, 38, 60, and 73 GHz bands in
traditional sub 6 GHz frequency bands, mmWave has larger campus, base station-mobile access, peer-to-peer, and vehicular
available bandwidth but suffers from additional higher path scenarios [15]–[17].
loss. Recently, many groups and standards including METIS However, most of the measurements in the literature were
[3], MIWEBA [4], mmMAGIC [5], 5GCM [6], 3GPP 38.900 conducted with different channel sounders, measurement en-
[7], and NYU Wireless [8] have tried to develop mmWave vironments, antenna configurations, and even post-processing
channel models for the frequency range from 6 GHz to 100 methods, which may have large impacts on the propagation
GHz. channel characteristics, and make the comparison of prop-
Various channel measurements have been conducted at some agation channel characteristics for different mmWave bands
popular mmWave frequency bands, such as 11, 15, 28, 38, impossible. Thus it is unable to have a fair comparison of the
60, and 73 GHz bands. In [9], 11 GHz outdoor urban cellular available mmWave channel measurement results, though the
channel measurements were conducted using a 24×24 multiple- comparison is very important for the development of a unified
input multiple-output (MIMO) channel sounder with 400 MHz channel model framework for the whole mmWave bands.
signal bandwidth and dual-polarized 12-element uniform cir- To fill this gap, we carry out channel measurements at
A. The SAGE Algorithm 2) PAP, PEP, and RMS AS: The PAP and PEP can be
obtained as
For VNA-based channel measurements, due to the quasi-
static nature of the environment and relatively long measure- 𝐿
∑ 2
ment time, a large number of snapshots are not available. In 𝑃 𝐴𝑃 = ∣𝛼𝑙 ∣ 𝛿(𝜙 − 𝜙𝑙 ) (5)
reality, there is only a single snapshot. Thus, conventional 𝑙=1
subspace based high resolution algorithms like multiple signal 𝐿
∑ 2
classification (MUSIC) and estimation of signal parameters 𝑃 𝐸𝑃 = ∣𝛼𝑙 ∣ 𝛿(𝜃 − 𝜃𝑙 ) (6)
via rotational invariance technique (ESPRIT) can not be used, 𝑙=1
because these algorithms rely on an estimation of the noise where 𝛿() is the Dirac delta function.
covariance matrix which is rank deficient in the single-snapshot Note that the power of estimated MPCs with the same
scenario. The maximum likelihood (ML) based SAGE algorith- azimuth or elevation angles are summed up to get unique
m [19] is applied to the data post-processing instead. values.
In the SAGE algorithm, the number of MPCs 𝐿 is usually Similar to the calculation of RMS DS, the RMS AS is
predefined large enough to capture all the significant paths. calculated as
Some values for the numbers of MPCs have been investigated.
∑𝐿 ∑𝐿
2 2 2
We found that when the number of MPCs equals 100, a 1 ∣𝛼𝑙 ∣ 𝜓𝑙 1 ∣𝛼𝑙 ∣ 𝜓𝑙 2
𝐴𝑆 = ⎷ ∑ 𝐿 2
− ( ∑𝐿 2
) (7)
good trade-off between accuracy and computational complexity
1 ∣𝛼𝑙 ∣ 1 ∣𝛼𝑙 ∣
can be achieved. The outputs of the SAGE algorithm are the
parameter sets Θ𝑙 = [𝛼𝑙 , 𝜏𝑙 , 𝜙𝑙 , 𝜃𝑙 ], 𝑙 = 1, ..., 𝐿, where 𝛼𝑙 , 𝜏𝑙 , where 𝜓𝑙 denotes the azimuth angle 𝜙𝑙 or elevation angle 𝜃𝑙 .
𝜙𝑙 , and 𝜃𝑙 denote the complex amplitude, delay, azimuth angle, IV. R ESULTS AND A NALYSIS
and elevation angle for the 𝑙th MPC. These spatial-temporal
MPC parameters are used to obtain some important propagation There are some important channel propagation characteristics
channel characteristics. for channel modeling, among which are the APDP, PAP, PEP,
RMS DS, and azimuth and elevation AS. At this section, these
properties will be compared and analyzed for the four mmWave
B. Propagation Channel Characteristics frequency bands at the four Tx locations.
1) APDP and RMS DS: The channel transfer function H ∈ A. APDP
2
C𝑁 ×𝐾 can be obtained by synthesizing the estimated MPCs The measured and estimated APDPs are compared for the
four frequency bands at Tx1 location, as shown in Fig. 2. For
𝐿
∑ each frequency band, the estimated results match well with the
H(𝑛, 𝑘) = 𝛼𝑙 exp(j2𝜋𝜆−1 ⟨e(Ωl ), rn ⟩)exp(−j2𝜋fk 𝜏l ) (1)
measured data. Most of the MPCs are extracted accurately. A
𝑙=1
strong LOS path is seen with the same delay but different power
gains. When the frequency increases, the attenuation tends to
⟨𝑒(Ω𝑙 ), 𝑟𝑛 ⟩ =[𝑥𝑛 , 𝑦𝑛 , 𝑧𝑛 ] be larger. The reflected and scattered MPCs also show some
(2)
× [cos(𝜙𝑙 )sin(𝜃𝑙 ), sin(𝜙𝑙 )sin(𝜃𝑙 ), cos(𝜃𝑙 )]T different characteristics. The maximum excess delay is within
80 ns. Apart from the LOS path, an average of five clusters of
where 𝑁 2 is the number of array elements, 𝐾 is the sweeping MPCs are seen for the four bands.
points, 𝑓𝑘 denotes the 𝑘th frequency point, and 𝑥𝑛 , 𝑦𝑛 , and 𝑧𝑛
B. PAP and PEP
denote the coordinates of the 𝑛th antenna array element.
The PAP and PEP are related with the measurement envi-
The channel impulse response can be obtained by the inverse
ronment. Fig. 3 shows the PAPs at the four Tx locations for
Fourier transform. For the measurement and estimated data, the
the four frequency bands. The azimuth angles show almost
APDP is obtained by averaging over the array elements
the same properties for different frequency bands at the same
2
Tx location. An average of five clusters in azimuth domain
𝑁
1 ∑ 2 are seen. The PAPs validation is shown in Fig. 4. For each
𝐴𝑃 𝐷𝑃 = 2 ∣𝐼𝐹 𝐹 𝑇 (𝐻(𝑛, 1 : 𝐾))∣ . (3) Tx location, the LOS direction is pointed to the Rx antenna,
𝑁 𝑛=1
while other reflection MPCs can be related to scatterers in
The RMS DS is calculated as the environment according to the delay and angle information.
The PEPs at Tx1 are shown in Fig. 5. Because the Tx and
Rx antenna height differences are the same for the four Tx
∑𝐿 ∑𝐿
2 2 2 locations, the elevation angles also show similar properties. The
⎷ 1 ∣𝛼𝑙 ∣ 𝜏𝑙 1 ∣𝛼𝑙 ∣ 𝜏𝑙 2
𝐷𝑆 = ∑𝐿 2
− ( ∑ 𝐿 2
) . (4) azimuth and elevation angles are sparse in space, and the beams
1 ∣𝛼𝑙 ∣ 1 ∣𝛼𝑙 ∣ are very narrow.
-50 -60
Measured, 11 GHz Measured, 16 GHz
-60 Estimated, 11 GHz Estimated, 16 GHz
-70
-70
-80
-80
Power gain (dB)
-100 -100
-110
-110
-120
-120
-130
-140 -130
0 50 100 150 200 0 50 100 150 200
Delay (ns) Delay (ns)
(a) Measured and estimated APDPs at 11 GHz. (b) Measured and estimated APDPs at 16 GHz.
-70 -70
Measured, 28 GHz Measured, 38 GHz
-80 Estimated, 28 GHz -80 Estimated, 38 GHz
-90
-90
-100
Power gain (dB)
-140 -150
-150 -160
0 50 100 150 200 0 50 100 150 200
Delay (ns) Delay (ns)
(c) Measured and estimated APDPs at 28 GHz. (d) Measured and estimated APDPs at 38 GHz.
Fig. 2. (a) Measured and estimated APDPs at 11 GHz; (b) Measured and estimated APDPs at 16 GHz; (c) Measured and estimated APDPs at 28 GHz; (d)
Measured and estimated APDPs at 38 GHz.
TABLE I
M EASUREMENT SYSTEM SETUP.
R EFERENCES
[1] C.-X. Wang, F. Haider, X. Gao, X.-H. You, Y. Yang, D. Yuan, H.
C. RMS DS and AS Aggoune, H. Haas, S. Fletcher, and E. Hepsaydir, “Cellular architecture
and key technologies for 5G wireless communication networks,” IEEE
Table III shows the RMS DS and AS for the four bands Commun. Mag., vol. 52, no. 2, pp. 122–130, Feb. 2014.
averaging over the four Tx locations. Generally, the RMS DS, [2] T. S. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, G.
N. Wong, J. K. Schulz, M. Samimi, and F. Gutierrez, “Millimeter wave
azimuth AS, and elevation AS show similar values for different mobile communications for 5G cellular: It will work!” IEEE Access, vol.
bands, which indicate that there are no large differences for the 1, pp. 335–349, May 2013.
0 0
-50 -50
-30 30 -30 30
-60 -60
-70 -70
-60 60 -60 60
-80 -80
-90 -90
11 GHz 11 GHz
16 GHz 16 GHz
-90 90 -90 90
28 GHz 28 GHz
38 GHz 38 GHz
-70 -70
-60 60 -60 60
-80 -80
-90 -90
-90 90 -90 90
11 GHz
-120 11 GHz 120 -120 120
16 GHz 16 GHz
28 GHz 28 GHz
38 GHz 38 GHz
-150 150 -150 150
180 180
7
[8] NYU Wireless Website, [Online]. Available: http://wireless.engineering.
Rx
nyu.edu/nyusim/
-90° 90°
6 [9] M. Kim, J. Takada, Y. Chang, J. Shen, and Y. Oda, “Large scale
characteristics of urban cellular wideband channels at 11 GHz,” in Proc.
5 0° EuCAP’15, Lisbon, Portugal, Apr. 2015, pp. 1–4.
Room width, R (m)
Tx4
[10] X. Zhou, Z. Zhong, B. Zhang, R. He, K. Guan, Q. Wang, and D. Ma-
W
Tx2
4
tolak, “Experimental characterization and correlation analysis of indoor
Tx1
3 channels at 15 GHz,” Int. J. Antennas Propag., Article ID 601835, 2015.
Tx3 [11] X. Yin, C. Ling, and M.-D. Kim, “Experimental multipath cluster
2 characteristics of 28 GHz propagation channel in office environments,”
IEEE Access, vol. 3, pp. 3138–3150, 2015.
1
[12] S. Hur, Y.-J. Cho, T. Kim, J. Park, A. F. Molisch, K. Haneda, and M.
0 Peter, “Wideband spatial channel model in an urban cellular environments
0 1 2 3 4 5 6 7
at 28 GHz,” in Proc. EuCAP’15, Lisbon, Portugal, Apr. 2015, pp. 1–5.
Room length, RL (m)
[13] I. Rodriguez, H. C. Nguyen, T. B. Sorensen, J. Elling, J. A. Holm, P.
Mogensen, P. Mogensen, and B. Vejlgaard, “Analysis of 38 GHz mmWave
Fig. 4. PAPs validation for the measurement environment. propagation characteristics of urban scenarios,” in Proc. European Wire-
less, Budapest, Hungary, May 2015, pp. 1–8.
0
-55
[14] S. Wyne, K. Haneda, S. Ranvier, F. Tufvesson, and A. F. Molisch,
30 30
“Beamforming effects on measured mm-Wave channel characteristics,”
-70
IEEE Trans. Wireless Commun., vol. 10, no. 10, pp. 3553–3559, Nov.
60 -85 60
2011.
-100 [15] T. S. Rappaport, E. Ben-Dor, J. N. Murdock, and Y. Qiao, “38 GHz and
11
16
GHz
GHz
60 GHz angle-dependent propagation for cellular & peer-to-peer wireless
90 90
28
38
GHz
GHz communications,” in Proc. IEEE ICC’12, Ottawa, Canada, Jun. 2012,
pp.4568–4573.
120 120 [16] Y. Azar, G. N. Wong, K. Wang, R. Mayzus, J. K. Schulz, H. Zhao,
F. Gutierrez, D. Hwang, and T. S. Rappaport, “28 GHz propagation
150 150 measurements for outdoor cellular communications using steerable beam
180 antennas in New York city,” in Proc. IEEE ICC’13, Budapest, Hungary,
Jun. 2013, pp.5143–5147.
[17] G. R. MacCartney and T. S. Rappaport, “73 GHz millimeter wave
Fig. 5. PEPs for Tx1.
propagation measurements for outdoor urban mobile and backhaul com-
munications in New York city,” in Proc. IEEE ICC’14, Jun. 2014, Sydney,
NSW, pp. 4862–4867.
[18] X. Wu, C.-X. Wang, J. Sun, J. Huang, R. Feng, Y. Yang, and X. Ge, “60
[3] METIS Website, [Online]. Available: https://www.metis2020.com/
GHz millimeter-wave channel measurements and modeling for indoor
[4] MIWEBA Website, [Online]. Available: http://www.miweba.eu/#Start
office environments,” IEEE Trans. Antennas Propag., 2017, in press.
[5] mmMAGIC Website, [Online]. Available: https://5g-mmmagic.eu/ [19] B. H. Fleury, M. Tschudin, R. Heddergott, D. Dahlhaus, and K. I.
[6] 5GCM Website, [Online]. Available: http://www.5gworkshops.com/ Pedersen, “Channel parameter estimation in mobile radio environments
5GCM.html using the SAGE algorithm,” IEEE J. Sel. Areas Commun., vol. 17, no. 3,
[7] 3GPP Website, [Online]. Available: http://www.3gpp.org/DynaReport/ pp. 434–450, Mar. 1999.
38900.htm