Wideband Tapped Delay Line Channel Model at 3.5GHz for Broadband Fixed Wireless Access system as function of Subscriber
Antenna height in Suburban Environment
Chia Leong Hong 1
, Ian J. Wasll 1
, Georgia E. Athanasiadou 1, Steve Greaves 2, Malcolm Sellars 2关于郑州
1
Laboratory for Communication Engineering, University of Cambridge
2
Cambridge Broadband Limited, United Kingdom
Abstract
This paper prents the results of measurements made to characteri the 3.5GHz Broadband Fixed Wireless Access channel in a suburban environment, using a ctored ba station antenna and a directional subscriber antenna. A tapped delay line channel impul respon model of the Single-Input Single-Output (SISO) channel is derived from the measurements. Previously, it has been reported that the delay spread decreas with an increa in the height of the Subscriber Unit (SU) antenna [9]. In this paper, it is reported that the multipath tap gain also decreas with an increa in the SU antenna height. Furthermore, it is reported that a 3-tap channel model with tap spacing of 200ns, and maximum tap delay at 400ns can adequately describe the radio channel under investigation. The narrowband Ricean K -factor and wideband root mean squared delay spread are obrved to correlate strongly with excess path loss.
1. Introduction
The fundamental limit that is impod on any wireless system is due to the radio channel. Propagation and channel models are esntial for the analysis and simulation of wireless systems. Knowledge of the radio channel is esntial to the development and deployment of a wireless system. Radio channel models are ud to support rearch into methods for mitigating channel impairments, such as the design of the equalir and the choice of single-carrier or multi-carrier syst
ems [1]. In addition, they also form an integral part of wireless system planning and deployment.
There is a large body of literature concerning the mobile radio channel, but publications concerning the Fixed Wireless Access (FWA) channel are still very limited [14],[15]. Previous works have investigated the narrowband channel characteristics of FWA systems, e.g. Crosby et al [2]. As the data rate of FWA systems increas, the effect of the wideband channel plays an increasingly important role on system performance. It is therefore necessary to characteri the wideband channel effects. Investigation into the effects of wideband channels on FWA systems have been conducted in the 2.5 GHz frequency band by Porter et al [3] and Gans et al [4], in the 1.9 GHz frequency band by Erceg et al [5]; and at 3.5 GHz by Siaud et al [6]. Models for path loss, Ricean K -factor and tapped delay line
channel impul respon for Single-Input Single-Output (SISO) fixed wireless channel were reported in [7]. Models for Ricean K -factor, Cross Polarisation Discrimination and path loss at 2.5GHz for a 2x2 Multiple-Input Multiple-Output (MIMO) fixed wireless system were reported in [8].
Previous literature has reported that the directional subscriber antennas usually employed in FWA systems significantly reduce the delay spread of the channel compared with that when an omni-direc
tional antenna is ud [3]. However the influence of SU antenna height has not previously been thoroughly investigated. This paper prents the results of propagation measurements at 3.5 GHz for a SISO Broadband Fixed Wireless Access (BFWA) system with a directional subscriber antenna and a ctored ba station antenna. In the previous paper [9], we have reported the statistics of the path loss and Root Mean Squared (RMS) delay spread with respect to SU antenna height. In this paper, we report on Ricean K -factor, excess path loss and the wideband tapped delay line channel impul respon model with respect to SU antenna height. First we describe the equipment and data processing methods ud for extracting the channel impul respon. This is followed by the results of wideband channel measurements, prented in the form of the wideband tapped delay line channel models, with respect to SU antenna height.
2. Measurement Methods
The measurements were conducted in the northern suburbs of Cambridge, England, during the summer months from June to September 2002 with trees in full foliage. This area has a relatively flat terrain and few high ri buildings, and covers an area of 9 km 2. The ba station (BS) is located at a height of 15 m above ground level. The SU antenna is positioned on top of a retractable vehicle-mounted mast. Two ts of readings are taken at each location parated by a horizontal distance of
about 2m. At each location, the bearing of the antenna is adjusted to point in the direction with highest received power. Once the bearing is fixed, the height of the SU antenna is subquently varied between 4 to 10 m in steps of 1 m. At each height, a total of 100 delay profiles and path loss measurement were collected over a period of 30 conds. A total of 540 ts of measurement have been collected, from 65 subscriber locations with BS to SU distances ranging from 730 m to 3.1 km. The preci
塞博
1B7.4
0-7803-8185-8/03/$17.00 © 2003 IEEE
ICICS-PCM 200315-18 December 2003Singapore
locations of the measurement sites are identified with a GPS receiver.
The BS antenna and the SU antenna have half power beamwidths of 90° and 20° respectively, and are both vertically polarid. The gains of the BS and SU antennas are 12.5dBi and 15.6dBi respectively. The bandwidth of the system is 5 MHz, giving ri to a delay spread resolution of 200ns. The length of the pudo-random sounding quence ud is 128 symbols and the maximu
m excess delay that can be measured with the system is 12.8 µs. The sounding quence is modulated using QPSK. At the subscriber, the in-pha and quadrature components of the received sounding burst are captured on the hard disk of a computer for later processing. A post-processing method bad on correlation processing extracts the power delay profiles and the RMS delay spread from the stored data [9].
3. Data Processing Methods
The processing method involves Fourier transformation of the correlation between sounding quence and received signal. The radio channel is often modelled as a linear time-variant filter with impul respon h(t,τ) or equivalently by its frequency respon H (f,t ), where h (t,τ) and H (f,t ) are a Fourier transform pair, [10]. Without loss of generality, consider a linear time-invariant system that is characterid by its impul respon h (τ), as shown in Figure 1.
Figure
1. Linear single input/single output system The complex envelope of the channel output y (t ) is the
convolution of the impul respon h (τ) with the complex envelope of the channel input x (t ), i.e.
∞
∞
−−=⊗=ττττd t x h h t x t y )()()()()( (1)
The time-invariant impul respon h (τ) is a special ca of the time-variant impul respon h (t,τ) if the unit impul respon function is independent of the time an input is applied, i.e.,
∞<<∞=t -h t h for )(),(ττ
(2)
To estimate the channel impul respon h (τ), the first step is to cross-correlate the input of the channel with its output, assuming that the input to the channel, i.e. the transmitted signal, is known. For jointly stationary stochastic process x (t ) and y (t ), it can be shown [11] that their cross-correlation φxy (τ) is related to the autocorrelation of the input φxx (τ) as
αατφατφd h xx xy )()()(−= ∞
∞
− (3)
which is a convolution integral. Since convolution in time domain is equivalent to multiplication in frequency domain, the relation (3) becomes
)()()(f H f f xx xy Φ=Φ (4)
where Φxy (f ) denotes the Fourier transform of φxy (τ), Φxx (f ) denotes the Fourier transform of φxx (τ) and H (f ) is the frequency respon of the channel. Hence, the channel impul respon h (τ) is found via the inver Fourier transform of its frequency respon H (f ). This forms the basis of the technique ud in the post-processing algorithm.
The power delay profile is the expected value of the magnitude squared of h (τ), i.e.,
2
]|)(| [)(2ττh E P =
(5)
The RMS delay spread of the channel is calculated according to
=−=
n
i i i T
RMS P P 1
2
21ττ
节奏感强的歌τ (6)
where
==
n
i i
i T
P P 1
01
τ
τ (7)
A Blackman window is applied to the signals before Discrete Fourier transformation. This enhances the signal to noi ratio of the power delay profile to more than 30dB. Since RMS delay spread is known to be nsitive to noi components on the power delay profile having large excess delays [12], a noi exclusion threshold is applied on the power delay profile so that any components more than 30d
B below the peak respon are excluded before calculating the RMS delay spread. Using this criteria, 330 ts of measurements out of a total of 540 ts were procesd and the results are prented as follows.
The measurements are parated into groups having a range of 1m according to the SU antenna height. Thereafter, the tap gains of the tap delay line channel models for each height group are computed by taking the average of tap gains at the same tap position for all power delay profiles in each group.
The Ricean K -factor is computed using the moment-method [13]. The Ricean K -factor for each tap of the wideband tapped delay line in each height group is the average of all K -factors of the power delay profiles at the same tap position.
4. Results
The narrowband Ricean K -factor is plotted against the excess path loss (path loss in excess of free space loss) in Figure 2. The result shows that Ricean K -factor decreas, (approaching Rayleigh fading) as excess path loss increas (due to heavy shadow fading).
Excess Path Loss (dB)
R i c e a n K -f a c t o r (d B )
Figure 2 Ricean K-factor vs. Excess Path loss
Figure 3 shows the RMS delay spread as a function of excess path loss. The delay spread is obrved to be highly correlated to excess path loss. However, Figure 4 shows that RMS delay spread does not significantly depend on distance between SU and BS.
The wideband tapped delay line channel model is summarid in Table 1. The tapped delay line channel model at a 5m SU height is shown in Figure 5. The results show that the tap gain at delays of 200ns (2nd tap) and 400ns (3rd tap) diminish as the SU antenna height is incread. The Ricean K -factors of the cond tap are very small, whilst the value for the third taps are very clo to zero, showing that the amplitude of the multipath echoes is clo to Rayleigh distributed.
excess loss (dB)
T r m s (n s )
Figure 3 RMS delay spread vs. Excess Path loss
distance (km)
T r m s (n s )
Figure 4 RMS delay spread vs. distance between subscriber
antenna and ba station
SU Height (m) 4.5<h<5.5 31 samples 5.5<h<6.5 48 samples 6.5<h<7.5 65 samples 7.5<h<8.5 43 samples 8.5<h<9.5 78 samples 9.5<h<10.5 57 samples Tap delay (ns) Tap gain (dB) K-factor (dB) Tap gain (dB) K-factor (dB) Tap gain (dB) K-factor (dB) Tap gain (dB) K-factor (dB) Tap gain (dB) K-factor (dB) Tap gain (dB) K-factor (dB) 0 0 43.8 0 44.1 0 44.6 0 45.6 0 46.0 0 46.7 200 -19.8 5.5 -21.0 8.6 -21.6 7.05 -22.5 9.0 -23.2 7.7 -24.3 9.6 400
-25.5
心愿清单-0.34
-26.0
1.54
-26.9
1.9
-25.5
2.8
-28.4
3.6
-28.5
3.5
Table 1 Summary of Wideband Tapped delay line channel model
Excess Delay (ns)M a g n i t u d e (d B )
Figure 5 Wideband tapped delay line channel model at 5m SU
antenna height
单国栋5. Conclusions
This paper has shown the influence of SU antenna height
on a BWA system employing directional SU antennas and
ctored BS antennas at 3.5 GHz. It has been previously
reported that the average RMS delay spread is obrved to decrea with increasing SU antenna height [9]. This paper further elaborates on the wideband characteristics of
outdoor SISO BFWA radio channel at 3.5GHz, by showing
the average tapped delay line channel impul respon.
Similarly to [7], it is confirmed that a 3-tap model can
adequately describe the channel. However, due to the
relatively few high ri building and flat terrain of the
environment and the narrower SU antenna beamwidth (20o
cf. 30o in [7]), the maximum average channel tap delay obrved is 400ns. The individual tap gains of the multipath components are obrved to decrea with an increa in
工程验收报告模板the SU antenna height. This is consistent with [9] which
reported that the RMS delay spread decreas with an
increa in the SU antenna height.
The narrowband Ricean K -factor is obrved to correlate
strongly with excess path loss, which is similar to the finding in [8]. In summary, the influence of distance between SU and BS on the delay spread is less significant
历史常识
than is the height of the SU antenna.
References
[1] Cambridge Broadband Limited, “Single carrier and
OFDM modulation: Their suitability for broadband
fixed wireless access systems”, 2001
/
[2] Crosby D., Greaves S. and Hopper A., “A Theoretical
Analysis of Multiple Diffraction in Urban
Environments for Wireless Local Loop Systems”,
Proceedings 9th Virginia Tech/MPRG Symposium on
Wireless Personal Comms, 1999 Blacksburg, Virginia.
[3] Porter J. W., Thweatt J. A., “Microwave propagation
characteristics in the MMDS frequency band”, ICC2000 (International Conference on
Communication) Conference Proceedings, 2000, pp.1578-1582.
[4] Gans M. J., Amitay N., Yeh Y. S., Damen T. C.,
Valenzuela R. A., Cheon C., Lee J., “Propagation
Measurements for Fixed Wireless Lopps (FWL) in a
Suburban Region With Foliage and Terrain
Blockages”, IEEE Trans. on Wireless落花生说课稿
Communications, 2002 Vol. 1, No. 2, pp. 302-310.
[5] Erceg V., Michelson D. G., Ghasmzadeh S. S., Greenstein L. J., Rustako A. J., Guerlain P. B., Dennison M. K., Roman R. S., Barnickel D. J., Wang
S. C., Miller R. R., “A Model for the Multipath Delay Profile of Fixed Wireless Channels”, IEEE JSAC, 1999
Vol. 17, No. 3, pp.399-410.
[6] Siaud I., Morin B., “Investigation on radio propagation channel measurements at 2.2GHz and 3.5GHz for the fixed wireless access in an urban area” Ann. Telecommun., 1999 Vol 54, No. 9-10, pp.464-478.
[7] V. Erceg, K.V.S. Hari, M.S. Smith, D.S. Baum, K.P.
Sheikh, C. Tappenden, J.M. Costa, C. Bushue, A. Sarajedini, R. Schwartz, D. Branlund. “Channel models for fixed wireless Applications (final IEEE 802.16 TG ad hoc version)”, IEEE 802.16 Broadband Wireless Access Working Group , <ieee802/16/tg3/index.html> report number IEEE 802.16.3c-01/29r4, 17 July 2001.
[8] P. Soma, D.S. Baum, V. Erceg, R. Krishnamoorthy,
A.J. Pauraj, “Analysis and Modeling of multiple-input multiple-output (MIMO) radio channel bad on outdoor measurements conducted at 2.5GHz for Fixed BWA Applications”, International Conference on Communications, (ICC 2002), Vol 1, pp 272-276, May 2002.
[9] C. L. Hong, I. J. Wasll, G. E. Athanasiadou, S.
Greaves, M. Sellars, “Wideband channel measurements and characterisation for broadband wireless
access,” Twelfth International Conference on Antennas and Propagation, 2003 (ICAP 2003), Volume 1, pp 429-432, April 2003.
[10] Bello P. A., 1963, “Characterization of Randomly
Time-Variant Linear Channels”, IEEE Trans on
Communication Systems, Vol. CS-11, No. 1, 360-393
[11] Proakis, J. G., 1995, “Digital Communications”, Third
Edition, McGraw-Hill Book Company, New York, 68-72
[12] Cullen P. J., Fannin P. C., Molina A., “Wide-band
Measurements and Analysis Techniques for the Mobile
Radio Channel”, IEEE Trans. On Vehicular Technology, 1993 Vol. 42, No. 4, 589-602.
[13] L.J. Greenstein, D.G. Michelson, V. Erceg, “Moment-
method estimation of the Ricean K-factor”. IEEE Comms Letter, Vol 3 No 6 June 1999.
[14] D. Crosby, S. Greaves, and A. Hopper. The Effect of
Building Height Variation on the Multiple Diffraction Loss Component of the Walfisch-Bertoni Model. IEEE Personal, Indoor andMobile Radio Communications Conference (PIMRC), September 2003 (Beijing, China).
[15] M.P. Sellars, G.E. Athanasiadou, B. Ziolko, S.D.
Greaves, and A. Hopper. Simulation of Broadband FWA Networks in High-ri Cities with Linear Antenna Polarisation. IEEE Personal, Indoor andMobile Radio Communications Conference (PIMRC), September 2003 (Beijing, China).
