factor of 3.6.

In addition, correlation time is set to 64 chips during describes local and global spatial variations of a signal envelope,
the code acquisition, with a chip rate of 4.096Mcps, and assuming
20 traffic channels/cell.
Figs. 1 and 2 show the probability of the locations at which the
cell search time from macro to micro, or vice versa, is below a cer-
tain value with the power ratio a = 3, 6dB of pilot against traffic.
It is noted that the cell search time becomes shorter with a = 6dB
because the interference from traMic channels decreases. The cell
search time of the extended Gold codes is comparable to that of
the orthogonal Gold codes or Gold codes because the correlation
time of 64 chips is shorter than the pilot code length. The cell
search time in the macrocells using eight group codes is shorter
than in the microcells with 32 cell-specific codes.
Conclusions: An I/Q multiplexed code assignment has been pro-
posed to facilitate fast cell search under HCS with asynchronous
cell site operation. During the inter-frequency handoff, the three-
step cell search algorithm uses a shared pilot code between a mac-
rocell and the overlaid microcells, allowing efficient combining of
and/or short-term and long-term temporal fluctuations of a signal
envelope in fading-shadowing wireless channels [2]. The lack of
general acceptance of this distribution may have been caused by
its complicated integral form. Note thatfdu) is very similar to the
Suzuki distribution [3]. However, in the Suzuki distribution,
instead of v, dv has the lognormal distribution. Although both def-
initions have already been used as mixtures of Rayleigh and log-
normal distributions [2], we focus on the former. However, it
should be mentioned that the Suzuki distribution can also be
approximated by a K distribution.
The K distribution, extensively used for modelling diverse scat-
tering phenomena such as tropospheric propagation of radio
waves, various types of radar clutter, optical scintillation from the
atmosphere etc., is a mixture of the Rayleigh and gamma distribu-
tions [4]:
fx(x)= 1 00

fXlY

2
(XlY = Y ) f Y ( Y W Y
5 P+1
the I and Q channels and hence reduced cell search time. It was
shown that the cell search time in the case of the handoff from
- ar(p + 1) (GI
2 2 0 a>O
KO(:)
p>-1 (3)
macrocell to microcell, or vice versa, can be reduced to -50ms at
90% of the locations. where fxl(XI Y=y) is the Rayleigh distribution with mode J Jf,& ~ )
is the gamma distribution with parameters a and p:
Acknowledgment: This work was supported by the Korea Tele-
com, Wireless Communications Research Lab.

0 IEE 1998 28 January 1998 r(.)is the gamma function, and I@(.) is the modified Bessel func-
Electronics Letters Online No: 19980608 tion of the second kind and order p.
Dong In Kim (University of Seoul, School of Electrical Engineering,
Seoul 130-743, Korea)

References

1 GILHOUSEN, K.S., JACOBS, I.M., PADOVANI, R., VITERBI, A.J.,

and WHEATLEY, c.E.: On the capacity of a cellular
WEAVER, L.A., Jr.,
CDMA system, IEEE Trans. Veh. Techno[., 1981, VT-40, (5), pp.
303-312
2 FIEBIG, U,-c.G.: Auto- and crosscorrelation properties for extended
m-sequences and related sequences. IEEE Proc. ISSSTA 94, 1994,
2, (7), pp. 406410
3 BAIER, A., FIEBIG, U,-c., GRANZOW, w., KOCH, w., TEDER, P., and
THIELECKE, J.: Design study for a CDMA-based third-generation
mobile radio system, IEEE J. Sel. Areas Commun., 1994, SAC-12,
(5), pp. 733-143
Fig. 1f u ( u ) and fx(x) for a = 2 and different values of p
(i) p = -0.2
(ii) p = 0.3
(iii) p = 0.9
- - _ _ fdu)
fA4
Kdistribution: an appropriate substitute for
Rayleigh-lognormal distribution in fading-
shadowing wireless channels Comparison of the moment generatingfunctions of 1nU and 1nX It
is proven that lognormal and gamma distributions can closely
approximate each other [5, 61. This fact, together with the mixture
A. Abdi and M. Kaveh representations in eqns. 1 and 3, motivated us to investigate how
The Rayleigh-lognormal distribution, which has proved useful for similar the Rayleigh-lognormal and K distributions are. The
modelling fading-shadowing wireless channels, has a complicated moment generating function of In U, +,l,dt)=flexp(tlnU)], can be
integral form. The authors have accurately approximated it by the obtained using auk/
V=v] = (2v)T(l+k/2) and aP]=exp(np
K distribution. This distribution is simpler and thus more +n2h2/2)[6]:
appropriate for analysis and design of wireless communication
systems.

Introduction: The Rayleigh-lognormal distribution is a mixture of Since 4x] = ( 2 a ) ~1(+k/2)r(1+p+k/2)/r( 1+p) [4], +,At) can be
the Rayleigh and lognormal distributions [l]: written as:

fv(u)= 1 CO

fujv(ulV = v)fv(v)dv U 20 (1)

Clearly ln@,,,dt) = lnr(1+t/2)+(ln2+p)t/2+h2tz/8,
while substitution
where fqdul V = v) is the Rayleigh distribution with mode v and
A(v) is the lognormal distribution with parameters p and h: of power expansion of Inr(l+p+t/2) into the logarithm of eqn. 6
gives:

The Rayleigh-lognormal distribution in eqn. 1 appropriately

ELECTRONICS LETTERS 30th April 1998 Vol. 34 No. 9 851
where Y(.), U(.), U(.), and Y(.) are the psi function and its
derivatives, respectively [7]. By neglecting t3, P, ... in eqn. 7 and
then comparing it with ln$,,,dt), we obtain the required relation-
ship between @,A) and (a$) that yields approximate equivalence
of U and X
p = ln(2a2) + P ( l + p) X2 = (1 + p) (8)
Discussion: According to eqn. 7, a has no effect on the accuracy
of approximation. In fact, on application of the relations in eqn.
8, E[cF]/E[x] will be independent of a. So the accuracy of
approximation depends on p. In eqn. 7 IY(m)(s)l, m = 1, 2, ..., is a
decreasing function of s such that lY(m)(s)l + as s + 0 and
00
iY(m)(s)l = (m-l)!/r for large positive s. Therefore,f d u ) and fJx)
increasingly coincide as p increases. This fact is depicted in Fig. 1.
Moreover, we observe that the tails of both distributions are
approximately the same, independent of the value of p.
Conclusions: Based on our results, the Rayleigh-lognormal and K
distributions are similar, but the latter has a simpler form. In addi-
tion, the existence of numerous analytic results regarding Bessel
functions makes it possible to obtain closed-form solutions in the
calculation of bit error rates, diversity effects etc. using the K dis-
tribution. We can also take advantage of numerous available
methods, especially in radar literature, for the simulation and
parameter estimation of the K distribution. Another important
supporting fact for the K distribution is its theoretical explanation:
it is a limit solution to a general scattering problem modelled as
two-dimensional random walks [SI.
0 IEE 1998 9 March 1998
Electronics Letters Online No: 19980625
A. Abdi and M. Kaveh (Department of Electrical and Computer
Engineering, University of Minnesota, Minneapolis, M N 55455, USA)
E-mail: abdi@ece.umn.edu
References
1 HANSEN, F., and MENO, F.I.: Mobile fading-Rayleigh and lognormal
superimposed, IEEE Trans. Vehic. Technol., 1977, VT-26, (4), pp.
332-335
2 STUBER, G.L : Principles of mobile communication (Kluwer,
Boston, Massachusetts, 1996)
3 SUZUKI, H.: A statistical model for urban radio propagation, IEEE
Trans. Commun., 1977, COM-25, (7), pp. 673-680
4 RAGHAVAN, R.s.: A model for spatially correlated radar clutter,
IEEE Trans. Aerosp. Electron. Syst., 1991, AES-27, (2), pp. 268-
275
5 CLARK, J.R., and KARP, s.: Approximations for lognormally fading
optical signals, Proc. IEEE, 1970, 58, (12), pp. 19661965
6 JOHNSON, N.L., and KOTZ, s.: Distributions in statistics: continuous
univariate distributions (Wiley, New York, 1970)
7 MAGNUS, w., OBERHETTINCER, F., and SONI, R.P.: Formulas and
theorems for the special functions of mathematical physics
(Springer, New York, 1966), 3rd edn.
8 JAKEMAN, E., and PUSEY, P.N.: Significance of K distribution in
scattering experiments, Phys. Rev. Lett., 1978, 40, (9), pp. 546-550
Method for eliminating narrowband
shortwave interference in upstream channel
of HFC
Suming Ju and Guangguo Bi
A scheme is presented which uses a finite impulse response digital
filter (FIRDF) to repress narrowband shortwave interference for
increased utilisation of the upstream bandwidth of hybrid fibre
coax. A precoding scheme is described to eliminate intersymbol
interference caused by the FIRDF.
Introduction: Because the hybrid fibre-coax (HFC) network has a
rich frequency resource, which can meet the requirements of multi-
media communications, and because subscribers share the HFC
plant, with a relatively low cost per home, the HFC network is
852
recognised as being one of the types of broadband user access net-
works. Providing a reliable solution for bidirectional HFC com-
munications will require a thorough understanding of the channel
characteristics of the HFC plant. The HFC downstream channel
has a bandwidth of up to 750MHz and a low noise level, which
meets the requirements of large capacity communications and
highly bandwidth efficient transmission techniques (e.g. 64 QAM).
However, because the HFC plant comprises tree and branch net-
works, emanating from a headend or central office and terminat-
ing at the subscriber residences, noise funnelling occurs, i.e. the
summation of the noise and ingress from both the subscribers and
the cable plant in the upstream channel [I]. At present, most cable
modems use quadrature phase-shift keying (QPSK) which has
high noise immunity, but low transmission efficiency (< 2biUHd
s). The upstream channel has narrow bandwidth (500 users share
5 4 2 MHz of bandwidth). The noise further decreases utilisation
of the upstream bandwidth. There are five types of noise and
ingress in the upstream channel [l].Narrow shortwave interference
(NSI) and additive white Gaussian noise (AWGN), which both
result in a funnelling effect at the receiver, are the main elements
that reduce the transmission capacity of the upstream channel.
AWGN limits the high efficiency modulation method of the
upstream channel. Because NSI signals from subscribers are corre-
lated and are in phase, its funnelling factor can be up to 20 logn
where n is the number of subscribers. Fig. 2 of [I] shows that the
high power level of NSI is only 7dBmV lower than that of the
data signals. The FIRDF method which is described below can
not only eliminate NSI, but also increase the utilisation of
upstream bandwidth.
r _ _ _ _ _ _ - _ _ _ _ _ _ _ _ _ _ _ _ _
I
I
d[kI
I
I ;I[hfin[ij;
FIRDF method: The bandwidth of the upstream channel can be
divided into a number of subbands with small bandwidth 1.5-
6MH2, called upstream paths. The bandwidth of NSI is generally
several tens of kilohertz, with relatively stable frequency. This
means that the receiver can utilise a type of FIRDF to repress
NSI. The cable channel shown in Fig. 1 is known to be nearly
ideal [I] and only has NSI and AWGN. First, the receiver (head-
end) measures the frequency, bandwidth and power level of NSI in
the upstream channel. If an upstream path has NSI, with fre-
quencyf,, the computer designs an FIRDF H(z) with a stop band
corresponding tofr, H(Z) = Cf=, h(k)zk.For easy realisation, it is
required that h)(k)(k = 0 - p ) are real and h(0) = 1. Thus
H ( 2 ) = (1 - e 7 w I z - l ) ( I - e - 3 w I z - l 1
= 1 - 2 cosw,z-1 + 2-2 (1)
where a, = 2.r~AIL, fr is the centre frequency of NSI and is the
sampling frequency of the receiver (here,L = 2MHz). As shown in
Fig. 2, the decrease in H(z) in the rangef, f 20kHz is > 40dB,
which can completely eliminate the effect of NSI on upstream
transmission. In D-transform notation r(D) = x ( D ) + I ( 0 ) + n(D)
where x(D) is a transmission symbol sequence, I(D) is a discrete
NSI sequence corresponding tof, and n(D) is an independent iden-
tically distributed Gaussian sequence of zero mean and variance
02.The output sequence of H(z) is
+ +
r(D) = H ( D ) z ( D ) H ( D ) I ( D ) H ( D ) n ( D )
+
= w(D) I(D) d ( D ) + (2)
where I(D) = H ( D )I ( D ) = 0. Thus:
ELECTRONICS LETTERS 30th April 1998 Vol. 34 No. 9