REDUCTION OF MULTIPLE-ACCESS
INTERFERENCE FOR OPTICAL
CDMA SYSTEMS
Maw-Yang Liu1and Hen-Wai Tsao1 1Department of Electrical Engineering
and Graduate Institute of Communication Engineering National Taiwan University
Taipei, Taiwan 107, R.O.C.
Recei¨ed 18 January 2001
ABSTRACT: Interference reduction¨ia a cancellation scheme for syn
-chronous optical CDMA systems is discussed. Based on integrating o¨er one bit duration of the recei¨ed optical signal, the interfering signals can be estimated and cancelled by the desired correlator. Irrespecti¨e of high or low load, our proposed method can precisely estimate the interference. Furthermore, thermal noise, APD noise, and interference are included in the analysis.䊚 2001 John Wiley & Sons, Inc. Microwave Opt Technol
Lett 30:1᎐3, 2001.
Key words: optical CDMA; interference cancellation scheme INTRODUCTION
A synchronous optical CDMA system can provide a greater number of simultaneous users than an asynchronous system w x1 . However, in an optical CDMA system, multiple-access
Ž .
interference MAI will cause a penalty on system perfor-mance. To reduce the adverse impact of multiple-access interference, some alleviating methods such as using optical
Ž .
hard limiter s can effectively enhance system performance, but only in the case of a lightly loaded traffic situation.
Interference cancellation is an effective way to alleviate
w x
the adverse impact of multiple-access interference 2, 3, 6 . In w x2 , a simple scheme utilizes averaging interference cancella-tion to estimate multiple-access interference from the re-ceived optical signal. However, such a method can precisely estimate the interfering signals only in a high-load traffic situation.
Ž .
Schemes using reference correlator s have been
investi-w x w x
gated to reduce the multiple-access interference 3, 6 . In 6 , some reference correlators are necessary in the interference cancellation schemes, thus reducing the cardinality of the signature sequence. To prevent a reduction of the cardinality of the signature sequence, all of the subscribers can use the same reference correlator based on a padded modified prime
w x
sequence 3 , whereas using the reference correlator method will cause additional power splitting loss due to the optical
w x
tapped delay-line decoder. As in 3 , we use the padding method so that the entire spreading sequence family has a unique cross-correlation value of ‘‘1.’’ Combined with esti-mating the received optical signal and cancellation scheme, this approach can completely eliminate the multiple-access interference irrespective of high- or low-load traffic. Also, this scheme can cause an increase of about 3 dB power in
w x sensitivity over that using the reference correlator in 3 .
SYSTEM DESCRIPTION
w x 2
The prime sequence 1 , whose length is p , can be con-structed from the modulo-p multiplication of the Galois field
Ž . 4
GF p s 0, 1, . . . , p y 1 , where p is a prime number. The padding subsequence to the signature sequence has been investigated to improve the cross-correlation property. For Contract grant sponsor:Ministry of Education
Contract grant number:89-E-FA06-2-4
Ž .
example, a prime code with padding py 1 zeros in each
w x subblock of the codeword has been discussed 5 .
Modified prime sequences are generated from time-shifted versions of the prime code, which can be applied to the
w x
synchronous systems 1 . The modified prime code has p groups, and each contains p mutually orthogonal code
se-2 Ž
quences. Each code sequence has p chips i.e., code length 2.
s p with weight p. Sequences between any different groups have a cross-correlation value of ‘‘1.’’
w x
In 3 , padding the same subblock of length p to each group of the modified prime sequence can make the whole code family have a unique cross-correlation value of ‘‘1.’’ Table 1 illustrates the padded modified prime sequences for
w x
ps 5 3 . This padded spreading code has the same
cardinal-ity as the original one, but the code weight and spreading
Ž . 2
length become ws p q 1 and F s p q p, respectively. The proposed interference cancellation scheme is shown in Figure 1. The received optical signal is split into two arms, with ratios ␣ and 1 y ␣, respectively. As the optical CDMA decoder employs optical tapped delay lines, this will cause a
Ž . Ž
1r p q 1 power-splitting loss in accordance with the code .
weight . We denote the desired signal as S and the interfer-i ing signal as S . After parallel cancellation, the output Z canj be represented as N ␣ Ž . Zs Si⭈ p q 1 q
Ý
Sj pq 1 js1; j/i N Ž . Ž . Ž . y 1 y␣ S ⭈ p q 1 qiÝ
pq 1 Sj js1; j/iTABLE 1 Padded Modified Prime Sequence Code for p = 5
Group Code Sequences
0 C0, 0s 10000 10000 10000 10000 10000 10000 C0, 1s 00001 00001 00001 00001 00001 10000 C0, 2s 00010 00010 00010 00010 00010 10000 C0, 3s 00100 00100 00100 00100 00100 10000 C0, 4s 01000 01000 01000 01000 01000 10000 1 C1, 0s 10000 01000 00100 00010 00001 01000 C1, 1s 01000 00100 00010 00001 10000 01000 C1, 2s 00100 00010 00001 10000 01000 01000 C1, 3s 00010 00001 10000 01000 00100 01000 C1, 4s 00001 10000 01000 00100 00010 01000 2 C2, 0s 10000 00100 00001 01000 00010 00100 C2, 1s 00100 00001 01000 00010 10000 00100 C2, 2s 00001 01000 00010 10000 00100 00100 C2, 3s 01000 00010 10000 00100 00001 00100 C2, 4s 00010 10000 00100 00001 01000 00100 3 C3, 0s 10000 00010 01000 00001 00100 00010 C3, 1s 00010 01000 00001 00100 10000 00010 C3, 2s 01000 00001 00100 10000 00010 00010 C3, 3s 00001 00100 10000 00010 01000 00010 C3, 4s 00100 10000 00010 01000 00001 00010 4 C4, 0s 10000 00001 00010 00100 01000 00001 C4, 1s 00001 00010 00100 01000 10000 00001 C4, 2s 00010 00100 01000 10000 00001 00001 C4, 3s 00100 01000 10000 00001 00010 00001 C4, 4s 01000 10000 00001 00010 00100 00001 Padded subsequence
Figure 1 Receiver using interference cancellation scheme via inte-gration over one bit interval T from received optical signalb
␣ Ž . Ž . s S ⭈ p q 1i
ž
y 1 y␣/
pq 1 2 py1 ␣ Ž .Ž . q y 1 y␣ p q 1 ⭈Ý
Sj pq 1 js1; j/iwhere N is the number of active users. Ž
To suppress the interfering signals i.e., to make the .
output signal irrelevant to interference , the value of ␣ should be 2 Žpq 1. Ž . ␣ s 2. 1 Ž . 1q p q 1 Ž
Remark:Without padding i.e., only a modified prime .
sequence , the desired correlator cannot extract the impact of Ž
interfering signals from the same group because of orthogo-.
nality . Hence, under low- to medium-load traffic, this scheme w x
cannot operate effectively 2 .
SYSTEM PERFORMANCE ANALYSIS
In this section, we take account of thermal noise, APD noise, and interference to analyze the system performance. The APD output can be modeled approximately as a Gaussian
w x
distribution 4 . The received optical signal intensity over a chip interval T is modeled as a Poisson point process. Thec average number of absorbed photons is T , where is thes c s arrival rate of incident photons due to chip ‘‘1’’ transmission, which can be represented as s P rhf. Here, P is thes w w optical power incident upon APD, is the APD quantum
Ž y34 2.
efficiency, h is Planck’s constant 6.624= 10 W⭈ s , and
f is the optical carrier frequency.
We use two Gaussian random variables X and Y to represent the APD outputs of the upper arm and lower arm, respectively, and their mean and variance are given by
w x Ž . s GT P ⭈ q I re q T ⭈ I rex c w x s b c s 2 2 2 w x 2 Ž . s G F T P ⭈ q I re q T ⭈ I re q x e c w x s b c s th 3 w x Ž . s GT P ⭈ q F ⭈ I re q F ⭈ T ⭈ I rey c w y s b c s 4 2 2 w x Ž 2. Ž . s G F T P ⭈ q F ⭈ I re q F ⭈ T ⭈ I re q y e c w y s b c s th 5 where G is the average APD gain, I is the APD surfaces leakage current, I is the APD bulk leakage current, e is theb electron charge, Pw x and Pw y are the optical power incident upon the APD in the upper and lower arms, respectively, and
Fe is the excess noise factor given by
Ž .Ž . Ž .
Fes K G q 2 y 1rG 1 y Keff eff 6
where K is the APD effective ionization ratio and 2 is
eff th
the variance of thermal noise, given by
2 Ž 2 . Ž .
s 2 K T T r e Rth B r c L 7
where T is the receiver noise temperature, K is Boltzmann’sr B constant and R is the receiver load resistance.L
After cancellation, the decision variable Zs X y Y is also Gaussian, and since X and Y are independent, its mean and variance are given by
Ž .
s y z x y 8
2 2 2 Ž .
s q .z x y 9
Assuming that transmitted data are equally likely, the condi-tional probability density function of k users transmitting bit
Ny 1 yŽ Ny1.
w x
‘‘0’’ or ‘‘1’’ is Pr krN y 1 s
ž /
k ⭈ 2 .Therefore, the bit-error probability can be represented as 1 w Ž . Pes Pr Z)rS s 0, N y 1i 2 Ž .x qPr Z FrS s 1, N y 1i Ny1 1 Ž . s
Ý
Pr Z)rS s 0, k, N y 1i 2 ks0 Ž . 4 w x Ž . qPr Z FrS s 1, k, N y 1 ⭈ Pr krN y 1i 10 where is the OOK decision threshold, which is usually setw x to be half of the mean of the decision variable Z 3 .
NUMERICAL RESULT AND DISCUSSION
w x
The numerical parameters used here are the same as in 3, 4 . Figure 2 shows the bit-error probability comparisons for the
Ž . w x
reference correlator Ref-Cor 3 and the proposed
interfer-Ž . Ž 2.
ence estimation IE scheme with full-load subscribers p
for ps 5, 7, and 11, respectively. In the present scheme, the performance can be improved by about 3 dB in sensitivity
w x
over that using the reference correlator 3 because using the reference correlator in the lower arm of the interference
Ž .
canceller will lead to a power-splitting loss 1rp q 1 caused by the optical tapped delay-line decoder. To obtain a larger
Figure 2 Bit-error probability comparisons for reference correlator w x3 and interference estimation scheme with full-load subscribers for
ps 5, 7, and 11, respectively
MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 30, No. 1, July 5 2001
aggregate capacity, increasing the optical power and using a larger prime number p are the necessary costs.
Due to the unique cross-correlation value of the padded modified prime sequence, our proposed scheme can precisely estimate the multiple-access interference from the received optical signal. Consequently, via parallel cancellation, the desired signal can be exactly detected.
REFERENCES
1. W.C. Kwong, P.A. Perrier, and P.R. Prucnal, Performance com-parison of asynchronous and synchronous code division multiple access techniques for fiber-optic local area networks, IEEE Trans
Ž .
Commun 39 1991 , 1625᎐1634.
2. H. Walle and U. Killat, Averaging interference cancellation for optical DS-CDMA systems, Proc IEEE 4th Int Symp Spread Spectrum Techniques and Appl, 1996, vol. 2, pp. 609᎐613. 3. M.Y. Liu and H.W. Tsao, Cochannel interference cancellation via
employing a reference correlator for synchronous optical CDMA
Ž .
systems, Microwave Opt Technol Lett 25 2000 .
4. H.M. Kwon, Optical orthogonal code division multiple access systemᎏPart I:APD noise and thermal noise, IEEE Trans
Com-Ž .
mun 42 1994 , 2470᎐2479.
Ž .
5. G.C. Yang and W.C. Kwong, Prime codes, Electron Lett 31 1995 . 6. H.M.M. Shalaby, Synchronous fiber-optic CDMA systems with
Ž .
interference estimators, J Lightwave Technol 17 1999 , 2268᎐2275. 䊚 2001 John Wiley & Sons, Inc.
MINIATURIZED 20 GHz CPW
QUADRATURE COUPLER USING
CAPACITIVE LOADING
R. Baliram Singh1and T. M. Weller1 1Department of Electrical Engineering University of South Florida
Tampa, Florida 33620
Recei¨ed 21 January 2001
( )
ABSTRACT: This paper describes coplanar wa¨eguide CPW quadra
-( )
ture couplers that utilize metal᎐insulator᎐metal MIM capacitors at each corner, eliminating the need for air bridges and significantly reduc
-ing the coupler footprint. A 20 GHz coupler on high-resisti¨ity silicon ( )
with branch lengths of 40⬚ 660m is presentedᎏthe area is 80% smaller than a con¨entional design. A measured insertion loss of ;4.5
" 0.5 dB was achie¨ed at 20 GHz.䊚 2001 John Wiley & Sons, Inc.
Microwave Opt Technol Lett 30:3᎐5, 2001.
Key words: coplanar wa¨eguide; quadrature coupler; capaciti¨e loading; thin-film capacitor
I. INTRODUCTION
Directional couplers are passive microwave components that may be used for power division andror combining in mi-crowave circuits such as mixers, balanced amplifiers, data modulators, phase shifters, and feed networks in antenna arrays. Conventional quadrature hybrid couplers utilize r4 Ž90⬚ transmission lines. However, as demonstrated in 1 , the. w x lengths of the r4 lines can be reduced by increasing the characteristic impedance and introducing lumped capacitance at the ends of the lines. Additional examples of capacitive
w x
loading in coupler design can be found in 2᎐3 . Contract grant sponsor:Raytheon C3 Systems, St. Petersburg, FL
In this paper, miniaturized CPW quadrature couplers using thin-film MIM capacitors are described. The work is distinguished from prior publications in that a CPW coupler is designed using three capacitors located at each corner in place of the conventional air bridges, thus providing ground equalization along with size reduction. Previous works were done mostly on microstrip, while those done with the CPW topology included only one capacitor located at each corner, and still utilized air bridges. New guidelines for coupler design are also given herein.
A summary of the design approach is presented in the following section. The analysis provides for the determination of necessary characteristic impedance and capacitive loading for a given electrical length of the coupler branch arms. The simulation and optimization of the couplers are presented in Section III, followed by experimental results in Section IV. The designs studied here used;1-m-thick Cr᎐Ag᎐Cr᎐Au layers for the MIM and CPW metal, and a 0.58-m-thick SiO
Ž . w x
layer s 6 for the capacitor dielectric 4 .r
II. DESIGN APPROACH
The geometry parameters associated with a capacitively loaded hybrid coupler can be simplified as shown in Figure 1. To minimize the number of variables, the down and across
Ž .
branches are taken to be equal in length s s , and1 2 the capacitance at each corner is also assumed to be equal. Thus, there are four quantities to be determined:the
electri-Ž .
cal length , the impedance of the down and across branches ŽZ1, Z , and the capacitance Cap .2. Ž .
The approach taken here is to solve for Z , Z , and Cap1 2 as a function of such that the even- and odd-mode condi-w x tions for a conventional quadrature hybrid are satisfied 5 . For both the even and odd modes, we want
Ž .
S11s 0. 1
For the even mode, we require
Ž .
y 1 q j
Ž .
S21s
'
22 and for the odd mode,
Ž1y j.
Ž .
S21s
'
. 32
Figure 1 Geometry parameters associated with the 90⬚ hybrid coupler