вход по аккаунту



код для вставкиСкачать
signals (channel 1 and channel 2) are clearly visible in Fig. 2 at
wavelengths of 1545.0 and 1551.Onm, with peak powers of -30.94
and - 32.48dBm, respectively. The lower output power of channel
2 is probably due to a combination of the interaction of different
physical processes leading to FWM 121, and the declining gain of
the SOA at longer wavelengths.
Acknowledgments: This work was supported by Telecom Research
Laboratories, a division of Telstra Corporation Limited. The Photonics Research Laboratory is a member of the Australian Photonics Cooperative Research Centre.
0 IEE 1994
Electronics Letters Online No: 19940148
25 November I993
M. A. Summefield, I. P. R. Lacey, A. J. Lowery and R. S . Tucker
(Photonics Research Laboratory, Department of Electrical a d
Electronic Engineering University of Melbourne, Parkville 3052,
R., and RAYBON,G.: ‘All-optical demultiplexing using
ultrafast four-wave-mixingin a semiconductor laser amplifier at 20
GbiUs’, Paper ThP12.2. Proc. 19th European Conf. on Optical
Communication, ECOC ’93, 1993, (Montreux, Switzerland)
‘Observation of
2 KIKUCHI, K., KAKUI, M., ZAH,c.-E., and LEE,T.-P.:
highly nondegenerate four-wave mixing in 1.5pn traveling-wave
semcondnctor optical amplifiers and estimation of nonlinear gain
coe@icieut’, IEEE J. Quantum Electron., 1151-156992, QE28, (I),
3 AGRAWAL. G.P.: ‘Population pulsations and nondegenerate fourwave mixing in semiconductor lasers and amplifiers’, Opt. Soc. Am.
B, 1988, 5, (l), pp. 147-159
Fig. 3 Signal waveforms
a 7OOMbiUs TDM input signal at 1547.8um
b 35OMb$ls WDM s p a ‘ at 1545.0~1
c 350MblUs WDM signal at 1551.0nm
J.S. Perino, J.M. Wiesenfield and B. Glance
Indexing terms: Optical communication, Optical frequency
conversion, Optical amplifiers
Each of the two output signals was selected in turn, using
appropriately tuned filters, and the output of the optical receiver
observed using a sampling oscilloscope. Fig. 3 shows the 700MbiU
s TDM input signal, and the two 350MbiVs WDM output signals.
The first trace shows the input bit sequence (11011001110110) at
1547.8~1.The second trace shows the output bit sequence on
channel 1 (1010101) at 1545.0nm. The third trace shows the output bit sequence on channel 2 (1101110) at 1551.0~1.To measure
the bit error rate, the transmitter of the BER test set generated a
700MbiUs pseudorandom bit sequence (PRBS) of length 2’-1, and
the BER of the two 350MbiUs WDM output signals was measured using the BER test set receiver. A BER of 1W was observed
for a received optical power of -23.6dBm in the 1545.0nm channel, and for a received optical power of -2l.OdBm in the 1551.Onm
channel. A back-to-back BER measurement obtained by modulating the signal laser with a 350 MbiVs return-to-zero pulse stream
gave a BER of 10-9 for a received optical power of -25.2dBm,
indicating a power penalty for the conversion of 1.6dB in the
1545.0nm channel, and of 4.2dB in the 1551.0nm channel. These
penalties can he attributed to the amplified spontaneous emission
(ASE) noise introduced by the two EDFAs and the SOA, to intersymbol interference resulting from low frequency carrier dynamics
in the SOA, and to crosstalk caused by overlap of the transitions
of the pump lasers between the ‘on’ and ‘off states.
Conclusion: We have reported the first demonstration of all-optical TDM to WDM conversion using FWM in an SOA. A
700MbiVs TDM signal was converted to two 35OMbiUs WDM
channels spaced by 4 nm, and BERs of less than lW9 were
observed. Power penalties of 1.6 and 4.2dB were observed for the
two channels, compared to the direct transmission of a single
350MbiUs signal. Because the conversion process is all-optical,
and is not dependent on the interband carrier relaxation time of
the SOA [3], this technique is equally applicable to channels with
much higher bit rates.
Fibre transmission of lOGbit/s signals
following wavelength conversion using a
travelling-wave semiconductor optical
The wavelength of an intensity-modulated signal at IOGbiUs was
translated from 1546 to 1531nm using cross gain-compressionin a
semiconductor optical amplifier. The shifted signal was
transmitted with less than IdB penalty over a l Z 1 h span of
dispersion-shifted fibre with two in-line amplifiers. A dispersion
penalty of nearly 2dB is measured for transmission of 6GbiUs
data over 2Okm conventional fibre.
The ability to arbitrarily translate the wavelength of an optical signal at high bit rates is important to the implementation of several
types of ‘all-optical’ network architectures [I]. Frequency reuse
also increases the capacity of WDM networks. One method for
wavelength conversion is cross gain-compression in semiconductor
optical amplifiers (SOAs) [ 2 4 , in which an information carrying
signal bump) and a CW beam (probe) are both input to an SOA.
The pump significantly compresses the gain of the amplifier, and
the complementary data are thus copied onto the probe. The
speed limitation of this technique is the gain recovery time of the
SOA, usually dominated by Auger recombination. Recently, highspeed operation (> 10GbiUs) has been achieved by using input
probe powers which saturate the amplifier (several hundred microwatts) [5,6], thereby reducing the gain recovery time by probeinduced stimulated emission. For use in WDM networks, these
signals must propagate over significant fibre spans. Here, we demonstrate transmission through 121km of dispersion-shifted fibre.
The use of input pump powers of the order of ImW, however,
creates large carrier density variations, which leads to chirp and
therefore limits transmission distances through conventional fibre.
The experimental layout is shown in Fig. 1. The pump, a modelocked external cavity semiconductor laser at 1546nm, I , , is externally modulated with NRZ pseudorandom data of pattern length
223 - 1 at 2.5-10Gbith using an LiNbOp Mach-Zender modulator.
The probe is a DBR laser operating at 1531nm, h,. Each laser
output is amplified in an erbiumdoped fibre amplifier (EDFA)
before being combined in a fibre coupler. The wavelength convertor is a fibre-pigtailed, polarisation-insensitive, bulk InGaAsP
ELECTRONICS LElTERS 3rd February 1994
Vol. 30
No. 3
closed at lOGbit/s. Fig. 3 compares pump and probe bit-error rate
data taken at 6GbiVs. Data for the pump are taken at the input to
the SOA, after several metres of fibre, and then, bypassing the
SOA, through 20km of singlemode fibre. The wavelength-shifteddata are taken at the SOA output, and then again after 20km
transmission. There is a 0.2dB penalty in received power of the
probe signal, measured directly after the SOA, compared to the
input pump signal. This small penalty is caused by reduction of
the extinction ratio by wavelength shifting [5]. There is an additional dispersion penalty of 1.8dB relative to the baseline pump
after the probe is transmitted through 20km of singlemode fibre.
2 5-10 G bitls
121km DSF
20km conventional SMF
Fig. 1 Experimental layout
ECL = external cavity laser, DBR = distributed Bragg reflector,
MOD = modulator, F = filter, REC = receiver, BERT = bit error
rate test set, EDFA = erbium-doped fibre amplifier, DSF = dispersion-shifted fibre, SMF = singlemode fibre, PRBP = pseudorandom
bit pattern
The SOA output is either pro agated through 20km of conventional
fibre, 121km of dispersion-sgifted fibre, or directly input to the
-25 - 2 4
SOA, with -6dB coupling loss at each facet. The small-signal chip
gain is 24dB and the output power at 3dB gain compression is 13mW. The shifted output is filtered to remove pump and ASE
from the probe signal. The probe is then propagated without
amplification through 20km of singlemode fibre, with -16.2ps/nm
km dispersion at 1 . 5 3 ~or
; through 121km of dispersion-shifted
fibre with two in-line EDFAs. Bit error rates for the pump and
wavelength-shifted signals were measured using a pin-FET receiver
front end.
2 5 G bitls
5 Gbit I s
200 ps/divswn
-22 -21 -20 -19
received power, dBm
Fig. 3 Bit error rates against received power for conversionfrom 1546
to 1531nm at 6Gbith after 2Okm standard singlemode fhre against
back-to-back (Okm), original signal, and pump afer 20km singlemode
pump, Okm
0 pump, 20km
0 probe, okm
W probe, 20km
The use of a relatively large input pump power, necessary for
reasonable extinction ratio, creates large carrier density variations
and, hence, modulation of the device index of refraction. Values of
linewidth enhancement factor, a, for InGaAsP semiconductor
amplifiers are typically around 5, and can be much higher,
depending on bias current and wavelength, among other factors
[7l. Measurement of the 1531nm probe spectrnm after wavelength
shifting at lOGbiVs, using a scanning Fabry-Perot, revealed some
spectral broadening to I3GHz, FWHM. Thus, even a small
amount of chirp can lead to significant dispersion penalties for signals propagated through moderate distances of conventional singlemode fibre.
50 ps/dwision
Fig. 2 E y diagrams oJ navelength shi/tzd signal for conversion from
1546 to I53lnm ufter rransmusion through Mkm ofwzglemode/ibre
a 2.5GbiVs
b SGbiVs
c 8Gbitls
d l0GbiVs
n -9 10
10 -
Fig. 4 Bit error rates against receivedpower for conversionfrom 1546
to 153Inm at IOGbit/s after IZlkm dispersion-shifted fibre against
back-to-back (Okm) and original signal
A progression of eye diagrams for 2.5 to lOGbiVs transmission
of the shifted signal through 20km of standard singlemode fibre is
shown in Fig. 2. The pump and probe powers measured in the
fibre before the SOA are 0 and AOdBm, respectively. The eye is
clearly open at 2.5Gbitis. The effect of dispersion becomes more
significant as the data rate is increased, and the received eye is
3rd February 1994
Vol. 30
X baseline, pump
0 probe, okm
U probe, 121km DSF
To verify that the penalty in transmission of the shifted signal
through standard fibre arises from dispersion, and to show that
No. 3
the converted data could be transmitted undistorted, the probe
was propagated through a 121km span of dispersion-shifted fibre
with two in-line EDFAs. Fig. 4 shows bit error rate data taken at
IOGbiVs. The pump and probe powers measured in the fibre
before the optical amplifier are 2 and OdBm, respectively. There is
a 0.8dB penalty for the received probe signal at the output of the
SOA compared to the input pump. Again, this penalty is mainly
from reduced extinction ratio [SI. There is less than IdB additional
penalty after transmission through the fibre-optic lmk. The additional penalty and change in slope are caused by signal-to-noise
ratio reduction from 34dB to -27dB as the probe propagates ttuough the loss-compensated span. No error floors were observed.
6 December 1993
Q IEE 1994
Electronics Letters Online No: 19940188
J. S. Perino, J. M. Wiesenfeld and B. Glance (AT@ Bell Laboratories,
Crawford Hill Laboratory, PO Box 4W, Holmdel, NJ 07733, USA)
precompetitive consortium on wide-band alloptical networks’, J. Lightwave Technol., 1993, LT-11,pp. 714735
KIGA, M., TIKURA, N., and NAWATA, K.: ‘Gain-controlled al~-optical
inverter switch in a semiconductor laser amplifier’, Appl. Opt.,
1988,27, pp, 39W3965
and JOURDAN. A.: ‘High perfromance optical
wavelength shifter’, Electron. Lett., 1992, 28, pp. 1714-1715
STUBKJAER, K.E.: ‘4Gb/s optical wavelength conversion using
semiconductor optical amplifiers’, IEEE Photonics Technol. Left.,
1993, 5, pp. 657460
and GNAUCK, A.H.:
‘Wavelength conversion at IO Gbis using a semiconductor optical
amplifier’,IEEE Photonics Technol. Lett., November 1993,
VAA, M.,
and KLENK, M.: ‘ZOGbit/s polarisation insensitive wavelength
conversion in semiconductor optical amplifiers’. Post-deadline
paper atECOC ’93
STUBKJAER, K.E.: ‘Measurement of carrier lifetime and linewidth
enhancement factor for 1.5pm ridge-waveguide laser amplifier’,
IEEE Photonics Technol. Leff.,
1991, 3, pp. 632634
New architecturefor incoherent optical
CDMA to achieve bipolar capacity
D. Zaccarin and M. Kavehrad
Indexing terms: Oprical commuincarion, Code division multiple
such as Gold codes, cannot be used in incoherent systems, and
therefore new (0,l) codes such as Prime codes [I] were designed.
However, compared to bipolar codes, unipolar codes can usually
support fewer simultaneous active users and, perhaps more importantly in a packet-switched environment, the number of possible
subscribers is severly limited by the sue of the code set. As an
example, from a preferred pair of M-sequences of length N, N+2
Gold codes can be constructed, whereas only P codes of period Pz
are possible for Prime codes.
In this Letter, we propose an efficient method of using standard
bipolar codes in optical systems, although requiring only incoherent delay lines and direct detection receivers. In the following two
sections we discuss the code construction and propose a simple
architecture for the optical encoder/decoder. We also present theoretical performance results in terms of the probability of bit error
against the number of active users.
Unipolar composite sequences: Composite sequences are defined as
combinations of two sequences such that each chip of one
sequence is further encoded by another sequence. The first
sequence is called an outer sequence, and the other sequence is
called an inner sequence. Consider two sequences U and V constructed as follows. Let X and Y be sequences of length K, and D
and E be sequences of length M. Sequences U and V are then of
length N = KM and can be thought of as being the Kronecker
product of D with X (E with Y), and are sometimes called Kronecker sequences 121. If D = (+I, + I , -I), and X = (+I, -1, + I , +I)
then U = (+l, -1, + I , + I , + I , -1, +I, + I , -1, + I , -1, -1).
It can be shown that for composite sequences, the periodic and
odd crosscorrelation functions are given by [2]
6 u , v ( k )= ~ D , E ( ~ ) C X , Y+( ~6 D) , E ( l + 1)Cx,Y(m- K )
where k = lK + m. If D and E are chosen to he Gold codes, we
have for many values that e,&) = @ , d l+ I), so that
Suppose now that X and Y are such that @&z)
0 for x* values of m, (P< K ) . In this case, it appears that U and
V will be uncorrelated sequences for many k values. For example
let X = Y = (1, 1, -1, 1) be the Barker sequence of length 4; in this
case @,&n)= 0 for m * 0 . The advantage of using Kronecker
sequences in the context of optical CDh4A systems can be seen by
noting that all the chips of the inner sequence are encoded by the
same chip of the outer sequence.
Because we consider incoherent systems, the above bipolar
sequences must be modified. Consider a bipolar code U = (h,uI,
... uN-J for which each chip is mapped into (1,O) if U, = 1 and (0,
1) is U, = - 1. The unipolar code obtained is denoted as Up. The
following Section proposes an efficient architecture to recover the
bipolar nature of Up at the receiver without requiring optical
switches. The crosscorrelation properties of the obtained unipolar
codes are very similar to those of the original bipolar codes. As an
example, for a Gold code of period N = 127, ow,,&) = (4,-7, 9,
215, *17) whereas for the real bipolar case @ , , ( I ) is three-valued
(15, -1, -17).
A simple and efficient method for using bipolar codes in
noncoherent optical code-division multiple access systems is
proposed. The proposed system requires only incoherent optical
delay lines and direct detection receivers. Practical architectures
for the encodeddecoder are given. The Letter shows that
combining Gold codes of period M = 127 and Barker code of
period K = 4 into composite sequences of period N = 508 allows
20 simultaneous active users for P. = lW9,while requiring only
four nonprogrammable delay lines.
Introduction: Code division is a suitable multiple-access technique
for fibre optics local area networks (LANs) having bursty traffic
characteristics. Unlike time-division multiple access (TDMA), no
synchronisation between the transmitting users or reservation of
time slots for each subscriber are necessary. Incoherent optical
CDMA systems have the advantages of fewer stability requirements for optical encoderddecoders than coherent systems, and
need only sources with a small coherence time. However, because
the correlation is based on power summation, they have the limitations of positive systems. It is usually believed that bipolar codes
opt i caI
.,.,. ,
K: 1
Fig. 1 Optical encoder architecture
Architecture and performance analysis: Efficient architectures for
the implementation of the encoder and the decoder are shown in
_ _ _ ~
Vol. 30
No. 3
Без категории
Размер файла
322 Кб
Пожаловаться на содержимое документа