Professional Documents
Culture Documents
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 8, SEPTEMBER 2011
I. I NTRODUCTION
N RECENT years, with the explosive growth of wireless
communication industry in terms of network infrastructures, network users, and various new applications, the energy
consumption of wireless networks and devices is experiencing
a dramatic increase. Because of ubiquity of wireless applications, such an increasing energy consumption not only
results in a high operational cost and an urgent demand for
battery/energy capacity to wireless communications operators,
but also causes a more severe electromagnetic (EM) pollution
to the global environment. Therefore, an emerging concept of
Green Communications has received considerable attention
in hope of finding novel solutions to improve energy efficiency,
relieve/reduce radio pollution to unintended users, and maintain/improve performance metrics.
c 2011 IEEE
0733-8716/11/$25.00
1699
L1
hl [k l],
(1)
l=0
Fig. 1.
lTS
T
(2)
|h[l]| , k = 0, 1, . . . , L 1.
g[k] = h [L 1 k]
l=0
(3)
In the above equation, we ignore the noise term to simplify
derivation1. Thanks to the channel reciprocity, the multi-path
channel forms a natural matched filter to the basic waveform
g[k], k = 0, 1, . . . , L 1, and hence a peak is expected at the
receiver.
The base station loads the data stream on the basic waveform, and transmits the signal into the wireless channel.
Usually the baud rate is much lower than the sampling rate,
and the ratio of the sampling rate to the baud rate is also
known as the rate back-off factor D [10]. Mathematically, if a
sequence of information symbols are denoted by {X[k]} and
assumed to be i.i.d. complex random variables with zero mean
1 By sending a large number of channel training sequences from the receiver,
the noise term is diminishing asymptotically.
1700
Fig. 2.
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 8, SEPTEMBER 2011
(5)
(h g)[Dl]X[k l] + an[k],
(2L2)/D
+a
(7)
1
]
PSig = EX
(h g)[L 1]X[k
D
L1
2
2
= P |(h g)[L 1]| = P
|h[l]| , (11)
l=0
PISI = EX
(h g)[Dl]X[k l]
l=0
l=(L1)/D
=P
l=0
(ISI)
(N oise) (10)
(2L2)/D
where
(h g)[k] =
(h g)[Dl]X[k l]
+ an[k]
l=0
L1
(Signal)
l=(L1)/D
Y [k] = a
L1
]
D
l=0
[D]
|(h g)[Dl]| ,
(12)
l=0
h[l]h [L 1 k + l]
,
L1
2
|h[l]|
l=(L1)/D
(8)
l=0
L1
l=0
=P
DT
L
R 1
|h(l) |2 ,
(15)
l=0
A. Power Reduction
Note that in (8), when k = L 1, it corresponds to the
maximum-power central peak of the autocorrelation function
of the CIR, i.e.
L1
2
|h[l]| .
(9)
(h g)[L 1] =
l=0
LR 1
|h(l) |2
P
rP = DT =
l=0
,
(17)
L1
2
P
l=0 |h[l]|
and the ratio of the expected transmission power needed for
TR and direct transmission can be expressed as
LR 1
2
E
|h
|
(l)
l=0
E[P ]
=
L1
.
(18)
P =
2
E[P DT ]
l=0 E[|h[l]| ]
In order to derive the numerator of (18), one needs to
analyze the order statistics of the |h[l]|2 s. However, since the
|h[l]|2 s are not identically distributed and it is also unknown
which LR out of all the |h[l]|2 s are the LR largest channel
taps, it is very difficult to obtain the closed-form expression of
the numerator in (18). Therefore, we will first assume that the
|h[l]|2 s are identically and independently distributed (i.i.d.),
and derive the numerator of (18). Then we will calibrate the
results for non-identically distributed |h[l]|2 s.
Before we start our analysis, let us first introduce the
concept of quantile [15] in order statistics. Denote F (z) as
the distribution function for a continuous random variable.
Definition 1: Suppose that F (z) is continuous and strictly
increasing when 0 < F (z) < 1. For 0 < q < 1, the q-quantile
of F (z) is a number zq such that F (zq ) = q. If F 1 represent
the inverse of F (z), then zq = F 1 (q).
Now, let us suppose that the h[l]s are i.i.d random variables;
then, |h[l]|2 are also i.i.d. Denote Zl = |h[l]|2 for short, and the
numerator in (18) can be approximated by the sample mean,
i.e.,
L 1
L 1
n
R
R
1
i
Z(l) lim
z(l)
,
(19)
E
n n
i=1
l=0
l=0
l=0
1701
LR 1
2
E
|h
|
can
be obtained by substituting the largest
(l)
l=0
quantile in (21), i.e.,
L 1
R
E
Z(l) LR E Z(0) |Z(0) z(0),q ,
(22)
l=0
(23)
l=0
l ,z 0,
1
e
FZl (z) =
(24)
0,
z < 0.
Therefore, Zl is also exponentially distributed, with mean
lTs
E[Zl ] = l2 = e T . Solving the inverse function of FZl (z)
R
yields the q-quantile of Zl
and substituting q = LL
L
zl,q = l2 ln(1 q) = e
s
lT
ln(
L
).
LR
(25)
Considering the approximation in (23), z(l),q is the (l + 1)th largest q-quantile corresponding to zl,q in (25), and Z(l)
corresponds to Zl Exp(1/l2 ), we can get
L
lTs
(26)
E Z(l) |Z(l) z(l),q = 1 + ln
e T .
LR
Then, the numerator of (18) can approximated as
L 1
L
R
R 1
L
lTs
2
E
|h(l) | 1 + ln
e T ,
LR
l=0
(27)
l=0
(28)
l=0
1702
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 8, SEPTEMBER 2011
l=0
(30)
L1
(31)
with h1 [l] being the l-th tap of the CIR and L the length of
the CIR. Each h1 [l] has the same distribution as h[l], i.e., a
circular symmetric complex Gaussian random variable with a
lTs
zero mean and a variance e T , but they are assumed to be
independent due to the location difference.
Then, we can express the received signal from the transmitter at the victim receiver with the TR-based transmission
as
Y1 [k]= a(h1 g)[L 1]X[k
L1
]
D
(Signal)
(2L2)/D
(h1 g)[Dl]X[k l]
(ISI)
l=0
l=(L1)/D
+ an1 [k]
L1
(N oise)(32)
h1 [l]X[k l] + n1 [k].
l=0
(34)
Then, the interference at the un-intended receiver can be
expressed as
2
L1
L1
I DT = EX
h1 [l]X[k l]
= P DT
|h1 [l]|2 ,
l=0
l=0
(35)
and we can obtain the ratio of the interference caused by the
two schemes as
IT R
rI = DT .
(36)
I
Define
E[I T R ]
I =
(37)
E[I DT ]
as the ratio of the expected interference caused by TR and
direct transmission. Substituting (33) and (35) into (37) and
taking expectation with respect to h and h1 , we can approximate I as
2
L1
E
h1 [l]h [l]
l=0
I P L1
L1
2
2
E
|h[l]|
|h1 [l]|
l=0
L1
l=0
!2
E |h[l]|2
L1
=P L1 l=0
2
2
E
|h[l]| E
|h1 [l]|
l=0
h1,l [k l],
l=0
+a
h
[l]h
[l]
1
I T R = P |(h1 g)[L 1]|2 = P l=0
.
(33)
L1
2
|h[l]|
L1
=P
l=0
L1
l=0
=P
1+e
(38)
l=0
e
e
s
2lT
s
lT
T
s
LT
1e
Ts
T
,
1e T 1+e T
where the second equality holds since h[l] and h1 [l] are i.i.d.
random variables, and h[l] and h1 [k] are independent for l =
k. Note that the ratio of the expected transmission power P
should be chosen according to (18) in order to maintain the
same performance.
In general, the observation time LTs satisfies LTs T ,
and the sampling period TS is much smaller than the delay
spread T , and thus we know that I is much less than 1.
According to our simulation results with typical parameters,
under the ideal assumption that channel responses of two
different locations are completely independent, interference
s
LT
2
Ts
C. Multi-Path Gain of TR
Since TR can utilize the multi-paths as virtual multiantennas, the multiple paths can provide spatial diversity.
In this part, we briefly talk about the maximum achievable
diversity order of TR transmission.
We first consider a binary
phase-shift keying
(BPSK) sig
naling with amplitude P , i.e., X[k] = P . By omitting
the ISI term, the error probability of detecting X is
L1
PSig
|h[l]|2 SNR ,
(39)
= Q 2
Q
2 /2
E[rP] vs. LR
0.5
0.4
0.1
SNR e
l=0
s
lT
1
22
lTs
T
E[rI] (dB)
26
28
30
32
simulation (L=100)
theory (L=100)
simulation (L=200)
theory (L=200)
L(L1)Ts
2T
38
5
10
15
Number of fingers of the Rake receiver
20
(SNR)L
l=0
=e
24
36
(40)
20
E[rI] vs. LR
5
10
15
Number of fingers of the Rake receiver
20
34
l=0
L1
&
L1
&
l=0
0.3
0.2
simulation (L=100)
theory (L=100)
simulation (L=200)
theory (L=200)
simulation (L=300)
theory (L=300)
0.6
E[rP]
1703
(SNR)
1704
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 8, SEPTEMBER 2011
E[r ] vs. L
P
0.8
0.7
0.6
E[rP]
PTR / 2 (dB)
0.9
equal
D=5
D=10
D=15
0.5
indoor
outdoor
0.4
0.3
0.2
1
0
0.1
0
DT
/ (dB)
Fig. 5. Expected transmit power needed for a TR-based system vs. an LR finger Rake receiver (ISI non-negligible).
5
10
15
Number of fingers of the Rake receiver
20
Fig. 7. Expected ratio of energy needed for a TR-based system vs. direct
transmission (IEEE 802.15.4a channel model).
E[r ] vs. L
I
20
15
IDT
22
ITR
10
24
E[r ] (dB)
I/ (dB)
26
28
10
indoor
outdoor
30
15
10
12
14
16
18
20
22
Fig. 6. Expected interference alleviation of a TR-based system vs. an LR finger Rake receiver (ISI non-negligible).
32
5
10
15
Number of fingers of the Rake receiver
20
V. E XPERIMENTAL M EASUREMENTS
10
theory
simulation
In this part, we demonstrate some experimental measurements taken in practical multi-path channels. The tested signal
bandwidth spans from 490 MHz to 870 MHz, centered at
the carrier frequency 680 MHz. Two measurement sites are
considered, an office room and a corridor, both of which are
located on the second floor of the J. H. Kim Engineering
Building at the University of Maryland. The layout of the
two sites are given in Fig. 10, where transceiver A transmits
time-reversed signals to transceiver B, and electromagnetic
waves are reflected by walls, ceiling/floor, and other objects
in the surrounding area. We fixed the location of transceiver
A, whereas moving transceiver B in a rectangular area (the
length is about four wave lengths) in the experiment.
10
10
10
10
10
10
10
1705
10
12
P/ (dB)
5 Although the real channel length is generally much longer than the chosen
parameter, computers cannot afford the simulation using real channel length
that requires 10L channel realizations to get an error bit. Therefore, we choose
a much shorter multi-path channel just for illustration purpose.
1706
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 8, SEPTEMBER 2011
1707
0.5
0.8
0.45
0.6
0.4
0.4
0.35
Amplitude
0.2
0.2
0.4
0.3
0.25
0.2
0.15
0.6
0.1
0.8
0.05
100
200
300
Time (ns)
400
500
600
6
8
10
12
14
The number of fingers in the Rake receiver
16
18
20
16
18
20
(a) rP (office)
0.3
0.15
0.1
Amplitude
0.05
0
0.05
0.1
0.25
0.2
0.15
0.2
0.15
0.25
100
200
300
Time (ns)
400
500
600
0.1
6
8
10
12
14
The number of fingers in the Rake receiver
(b) rP (corridor)
Fig. 12. Power reduction by the TR-based transmission obtained by the
experiment measurements.
1
0.9
Normalized Magnitude
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
500
1000
1500
2000
2500
Taps of Preequalized Channels
3000
1708
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 8, SEPTEMBER 2011
0.8
2
dB
0.6
0.4
3
4
5
0.2
6
10
0
10
8
6
6
4
4
2
2
0
Multiples of /2
Multiples of /2
Multiples of /2
Multiples of /2
(a) rI (office)
1
1
0.8
2
dB
0.6
3
4
0.4
0.2
6
8
0
8
4
2
Multiples of /2
Multiples of /2
2
0
Multiples of /2
4
2
0
Multiples of /2
(b) rI (corridor)
Fig. 14. Interference alleviation by the TR-based transmission obtained by
the experiment measurements.
40
35
0.95
30
25
0.9
20
15
0.85
1709
10
VII. C ONCLUSION
0.8
Fig. 15.
10
15
20
25
30
35
40
1710
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 29, NO. 8, SEPTEMBER 2011
Yu-Han Yang (S06) received his B.S. in electrical engineering in 2004, and two M.S. degrees in
computer science and communication engineering
in 2007, from National Taiwan University, Taipei,
Taiwan. He is currently pursuing the Ph.D. degree
at the University of Maryland, College Park. His
research interests include wireless communication
and signal processing. He received Class A Scholarship from National Taiwan University in Fall 2005
and Spring 2006. He is a recipient of Study Abroad
Scholarship from Taiwan (R.O.C.) Government in
2009 and 2010.