Deng et al. VOL. 2, NO. 4 /APRIL 2010/J. OPT. COMMUN. NETW. 159
Demonstration and Analysis ofAsynchronous and Survivable Optical
CDMA Ring NetworksYanhua Deng, Zhenxing Wang, Konstantin Kravtsov, John Chang, Carolyn Hartzell,
Mable P. Fok, and Paul R. Prucnal
stprnswsussCimnsdsmtpqszfthtTnitni
uottwah
Abstract—We propose an incoherent optical code-division multiple access (CDMA) ring network config-ured such that each node in the network includes all-optical addÕdrop multiplexers to provide trulyasynchronous operation and a self-healing protectionmonitoring circuit to ensure service availability dur-ing link failure. We also analyze the performance ofsuch a network considering the presence of multiple-access interference and noise from incomplete drop-ping of the desired data and accidental dropping ofthe undesired data. Furthermore, we experimentallydemonstrate the proposed design with a two-nodesystem operating at 2.5 GbitsÕs. Error-free transmis-sion is obtained and a 5.3 dB power penalty at a biterror rate of 10À9 is measured for the worst connec-tion configuration. The complete optical CDMA ringnetwork is successfully demonstrated in truly asyn-chronous operation.
Index Terms—Optical code-division multiple access;Asynchronous network; Ring network; AddÕdropmultiplexer; Nonlinear optical loop mirror.
I. INTRODUCTION
S elf-healing ring architectures provide high surviv-ability and ensure service availability [1]. Many
telecommunication infrastructures in metropolitanand local-area networks are implemented with sucharchitectures, and those networks are guaranteed tohave 60 ms or less restoration time against link fail-ure [2].
Though optical code-division multiple access(CDMA) is conventionally implemented in broadcast
Manuscript received September 4, 2009; revised January 14, 2010;accepted February 11, 2010; published March 9, 2010 �Doc. ID116374�.
tar networks [3–5], this technique’s unique proper-ies, such as large cardinality and soft blocking [6–9],rovide many advantages when implemented in aing architecture. First of all, the designed code cardi-ality is always much larger than the number of sub-cribers. With large cardinality, a survivable ring net-ork can be built such that there is no need to reserve
eparate bandwidth or a separate path for link fail-re. Compared with a WDM network that reserveseparate wavelengths and a TDM network that re-erves separate time slots for link failure, an opticalDMA ring network is more bandwidth efficient since
t can support both links in the ring network to theiraximum capacity. Also, full connectivity between
odes is possible without code switching. Second, theoft blocking property of optical CDMA allows the ad-ition of subscribers without any hard limit, and newubscribers are easily added to the network withoutodifying the existing hardware. Third, code-based
ransmission provides truly asynchronous access ca-ability. The signal is identified with a unique code se-uence, so there is no need to distribute a global clockignal in a separate channel for temporal synchroni-ation. Furthermore, full transmission is guaranteedor each node without waiting for the designatedransmission time slot. Fourth, optical CDMA allowseterogeneous data types to coexist in the same link,hus maximizing quality of service in the network.his is especially important for telecommunicationetworks where services for data, voice, and video are
ntegrated, and the quality-of-service demand for eachype differs. Last, each code is removed at the desti-ation, which increases the data security by prevent-
ng interception of data by downstream nodes.
Previous works have investigated the code-dropnit for optical CDMA ring networks using a terahertzptical asymmetric demultiplexer (TOAD) as theime-gating device [7,8]. However, such an implemen-ation requires a control signal that is synchronizedith the clock from the data source. To achieve trulysynchronous operation, a self-clocked code-drop unitas been proposed and demonstrated using a
2010 Optical Society of America
drfvwmwtbosrw
B
2cot
fmrdpTCdffe
Fn
160 J. OPT. COMMUN. NETW./VOL. 2, NO. 4 /APRIL 2010 Deng et al.
nonlinear-optical-loop-mirror-based thresholder [9].The autocorrelation peaks from the desired signal arefirst extracted from the data stream, which are thenwavelength-converted and used as the control signalfor the time-gating device [10].
In this paper, we first investigate the performanceof a complete optical CDMA ring network. The bit er-ror rate (BER) of the desired signal is calculated withGaussian approximation including multiple-access in-terference and residual noise from incomplete drop-ping of signals in the add/drop multiplexers (ADMs).This analysis is also carried out for the link failurescenario such that traffic on both links is aggregatedtemporarily. Furthermore, we experimentally demon-strate a two-node system in asynchronous operationwith a protection-monitoring circuit and an ADM thatuses only one nonlinear optical loop mirror for signaldropping and restoring.
II. OPTICAL CDMA RING NETWORK DESIGN
A. Network Architecture
There are two principle types of ring architecture.One type is a unidirectional path-switched ring(UPSR), where two copies of data are sent around thering in counterdirectional paths and the receiver se-lects the best signal from one of the paths. The othertype is a bidirectional line-switched ring (BLSR),where there is a dedicated path for normal service andanother path for backup service [2]. Each type of ar-chitecture can be configured with either a two-fiber ora four-fiber link. The two-fiber bidirectional line-switching ring architecture can be implemented moreefficiently in optical CDMA because of its soft blockingproperty.
Extensive studies have been presented on code as-signment for the fully connected bidirectional ring,where each of the nodes can communicate with everyother node [11,12]. Elrefaie has presented an algo-rithm that uses the minimum number of wavelengthsfor a ring with a fixed number of nodes. Similar algo-rithms can be applied to the optical CDMA ring net-work by assigning each node with a unique code. Theminimum number of codes F needed for fully meshedconnectivity with N nodes is shown in Eq. (1):
F = �N2 − 1
4 �. �1�
While the number of wavelengths is limited, the num-ber of codes for optical CDMA can be increased bymodifying the number of wavelengths and the numberof chips within a bit period.
Due to the soft blocking capability of optical CDMA,traffic can be aggregated with only slight performance
egradation. Thus, there is no need to reserve sepa-ate bandwidth or a separate backup path for linkailure. Both links can carry traffic during normal ser-ice. Every node can add and drop signals in bothest and east links, as shown in Fig. 1. One way toaximize the quality of service is to put the trafficith high data rates on one link and low data rates on
he other link. During a temporary link breakdown,oth west and east links are rerouted and the trafficn both links is aggregated together. Despite thelight degradation in performance, there is no inter-uption of service during link failure and the band-idth of the fiber can be utilized the whole time.
. Add/Drop Node
The design of the add/drop node is presented in Fig.. The proposed add/drop node consists of electronicircuits that monitor link connection, while the all-ptical add/drop multiplexers are used to establishransmission between nodes.
The protection-monitoring electronic circuit has twounctions: It controls the laser diodes that produce theonitoring signal, and it controls the switches that
oute the flow of traffic. To avoid interference with theata, the monitoring signal is designed to be a low-ower constant continuous-wave light at 1300 nm.his monitoring signal is filtered before the opticalDMA add/drop multiplexers and is directed to theetector in the monitoring circuit. In the case of a linkailure between any two nodes on the east link, theollowing procedure is executed (a similar procedure isxecuted for a west link failure):(1) The monitoring circuit turns on the alarm for
east link failure.(2) The laser diode for the west link is turned off to
notify the following node of the link failure.(3) The switch on the west link is connected to the
Deng et al. VOL. 2, NO. 4 /APRIL 2010/J. OPT. COMMUN. NETW. 161
together.To perform optical add/drop in the ring, the monitor-ing signal and the data are separated when the trafficcomes in to the node. The data is directed to the all-optical ADMs. The principles of each ADM are illus-trated in Fig. 3. When the data on the links enters theADM, the decoder reconstructs the signal so that thedesignated signal has autocorrelation peaks while theother signals have cross-correlation peaks. The maindevice that distinguishes the desired data from theundesired data is the nonlinear optical loop mirror(NOLM). It separates the autocorrelation peaks of thedesired data from the cross-correlation peaks of theundesired data by distinguishing the pulse intensity[13]. The NOLM consists of a dispersion-shifted fiber,an optical attenuator, and a polarization controller.Since the autocorrelation peaks from the desired datahave higher intensities than the cross-correlationpeaks from the undesired data, the desired data is di-rected to the transmitting port of the NOLM, and theundesired data is directed to the reflection port. Thedrop data from the transmitting port is sent to the re-ceiver. The thru data from the reflection port passesthrough another encoder so that the data is restoredto the same state as before the ADM. After the en-coder, the thru data is coupled with the newly addeddata to be sent onto the links.
Fig. 2. (Color online) Design of the optical CDMA and/drop nodeswith protection-monitoring circuit and optical add/drop multiplex-ers on both links. SW, switch; LD, laser diode; OADM, optical add/drop multiplexer.
. Probability of Error in the Ideal Optical CDMA Ringetwork
Incoherent optical CDMA networks implementingmplitude modulation with optical orthogonal codesupport complete asynchronous access and accommo-ate various data types. One type of optical orthogo-al code is carrier-hopping prime codes (CHPCs),hich can be described by �L�N ,� ,�a ,�c� [14], where�N is the size of the matrix, � is the weight of the
ode (the number of wavelengths for CHPC), �a is theaximum autocorrelation sidelobe, and �c is theaximum cross-correlation value. We evaluate the
erformance of such networks by calculating the prob-bility of error for the ideal system, using the assump-ions that the only source of noise is from multiple-ccess interference and each pulse has the sameower. The hit probability qi of the desired signal andth type of interfering signal is calculated in Eq. (2)15]:
qi =�a�i
2 max��a,�i�Ni, �2�
here �a is the weight of the desired signal, and �i ishe weight of the interfering signal. The averageower for a received 0 bit and 1 bit and the noise vari-nce from multiple-access interference are presentedn Eqs. (3)–(5), respectively:
P̄0 = �∀i
kiqi, �3�
P̄1 = �a + �∀i
kiqi, �4�
�02 = �1
2 = �∀i
kiqi�1 − qi�. �5�
ere, ki is the number of interfering users with theth data type. The error probability of the system ispproximated by Eq. (6) [14]:
Pe�G = Q�1
2�SNR� = Q�1
2
�a
��∀i
kiqi�1 − qi� ,
here Q�x� =1
�2�
x
�
e−y2/2dy. �6�
. Probability of Error in the Nonideal Optical CDMAing Network
We apply the same probability-of-error analysis onhe nonideal two-fiber bidirectional optical CDMA
twpnTraiodconWmsnniatr=t
Fsn
Fiet
162 J. OPT. COMMUN. NETW./VOL. 2, NO. 4 /APRIL 2010 Deng et al.
ring network, but in addition to multiple-access inter-ference, we also include the dropped residual signalfrom time-gating/thresholding devices in the add/dropmultiplexers. Before considering the probability of er-ror, first we have to investigate how the signals inter-fere with each other in two-fiber bidirectional ringnetworks. The interference depends on the connectionconfiguration. The best scenario would be one inwhich every node is transmitting to the adjacent node,so that all the codes are removed without interferingwith each other. The worst-case scenario would be theone in which every node is transmitting to the far-thest node on the link, resulting in N−2 interferingusers on every link segment.
When incomplete dropping of desired signals andaccidental dropping of partial undesired signals oc-curs, additional noise is added to the link, and we canapproximate the residual signals after the add/dropmultiplexers with a Gaussian distribution. The aver-age power of the received 0 bit and 1 bit are modifiedin Eqs. (7) and (8):
P̄0 = �∀i
qi,jrcj, �7�
P̄1 = �a�1 − ra�rcj + �
∀iqi,jrc
j, �8�
where rcj is the residual of the cross-correlation peaks
from the ith data type after jth ADMs, and ra is theresidual of the autocorrelation peak after the ADM ofthe destined receiver. The variance of multiple-accessinterference is calculated in Eq. (9):
�MAI2 = �
∀iqi,j�1 − qi,j�rc
j. �9�
The new error probability is presented in Eq. (10):
Pe�G = Q��a�1 − ra�rcj
2��MAI2 � . �10�
Figure 4 compares the performance of the ideal ringnetwork calculated using Eq. (6) and the performanceof the nonideal ring network calculated using Eq. (10)with residuals where rc=99% and ra=5%. The (12�97, 12, 0, 1) CHPC parameters used in the calcula-tion describe a 10 Gbit/s system with a 1 ps full-widthhalf-maximum pulse width. The performance of thering network is degraded accordingly to the increasein number of simultaneous transmissions on a linksegment. The calculation is done with the maximumnumber of interfering users where every node istransmitting to the farthest node, i.e., the worst-casescenario.
As mentioned previously, the heterogeneous datacharacteristic of an optical CDMA ring network allowsthe two traffic links that carry heterogeneous data
ypes to be aggregated together during link failureithout complicated network management. The errorrobabilities of the aggregated links for the ideal ringetwork calculated using Eq. (6) are shown in Fig. 5.he analysis assumes that one link carries high-data-ate traffic by implementing CHPC (12�97, 12, 0, 1)nd the other link carries low-data-rate traffic bymplementing CHPC (10�199, 10, 0, 1). The normalperation is shown with hollow square and circularata points. The calculation assumes that each link isarrying data separately without interfering with thether link, and the number of nodes would be theumber that can be supported by the individual link.e assume that both links are supporting the maxi-um number of nodes, with BER=10−12. Figure 5
hows that the high-data-rate link can support 26odes and the low-data-rate link can support 42odes. During link failure, the traffic from both links
s aggregated together. If we follow the curves for theggregated data (cross and star data points), the sys-em must accommodate 68 nodes in total, and the er-or probability during link failure becomes BER10−7 for the high-data-rate link and BER=10−5 for
he low-data-rate link.
ig. 4. (Color online) Comparing the performance of the worst-casecenario for a ring network with ideal add/drop and a nonideal ringetwork with residual as a function of the number of nodes.
ig. 5. (Color online) Performance of the ideal ring network carry-ng different types of data at each link (hollow data points), and theffect during link failure when traffic on both links are aggregatedogether (cross and star data points).
eocct
t1tiptsZs
domCcwwdwc
[ro4as
Deng et al. VOL. 2, NO. 4 /APRIL 2010/J. OPT. COMMUN. NETW. 163
This error probability calculation also shows howthe spectral efficiency can be increased in an opticalCDMA ring network. The spectral efficiency is the to-tal bit rate over the total bandwidth, and it is calcu-lated using Eq. (11) [16]:
� =N � B
� � �vBW, �11�
where N is the total number of nodes, B is the bit rate,� is the code weight, and �vBW is the bandwidth of thewavelength. For the system described above, if onlythe high-data-rate link in the ring network carriestraffic, the spectral efficiency is 43.3%, with 26 nodes,10 Gbit/s data rate, and 50 GHz bandwidth for eachof the 12 wavelengths. Now we add the low-data-ratetraffic to the other link with a 5 Gbit/s data rate. Thesystem can accommodate 68 nodes. The total band-width is still the same since the codes for opticalCDMA can have a shared spectrum. Then the spectralefficiency is increased to 78.3%.
IV. EXPERIMENTAL DEMONSTRATION
A. Experimental Network Testbed
An experimental setup of a two-node optical CDMAring network is shown in Fig. 6. The schematic showsthe worst-case connection, where each node receivesits own data, i.e., the data goes through all the nodesand undergoes both the decoding and restoring stagesat the other node. The detailed explanation of thefunctionality of the monitoring and protection circuits,and the operation of the ADMs have been discussed inthe design section. For the experimental demonstra-tion, only one link is carrying data during normal op-
Fig. 6. (Color online) Experimental setup of a two-node ring networhighly GeO -doped fiber; CDR, clock and data recovery unit; BERT
2
ration; thus one ADM is incorporated on the east linkf the node. The two nodes are sharing the same opti-al source, but they are encoded differently. The re-eived signal at one of the nodes is monitored, as thewo nodes operate similarly.
The optical source consists of four CW laser diodeshat generate wavelengths centered at 1552.92,551.72, 1550.92, and 1550.12 nm. A short pulse trainhat has 10 ps full-width half-maximum pulse widths obtained using the electroabsorption modulator as aulse carver with a 2.5 GHz clock source. The pulserain is then modulated by a 231−1 pseudorandom bitequence at 2.5 Gbits/s using a LiNbO3 Mach–ehnder modulator. The optical signal is split andent to the optical CDMA ADM’s add encoders.
The implemented ADM uses a NOLM as the code-rop device. The dispersion-shifted fiber is 2 km inne NOLM, and 2.4 km in the other. The codes imple-ented during the experiment are (4�17, 4, 0, 1)HPC, with 4 wavelengths and 17 time chips. Theode sequence for node 1’s add encoder is (0, 3, 6, 9),here the numbers indicate the chip slots for the fouravelengths from the optical source respectively. Therop decoder and the restore encoder are matchedith the add encoder. The code for node 2’s add en-
oder is (0, 7, 14, 4).
The receiver has another NOLM-based thresholder9] for suppressing the amplitude noise in the autocor-elation peaks. The thresholded signal is sent to anff-the-shelf clock data recovery unit (HP 83446A OC-8/STM 16). The optical signal and the clock signalre recovered asynchronously for bit-error-rate mea-urements.
164 J. OPT. COMMUN. NETW./VOL. 2, NO. 4 /APRIL 2010 Deng et al.
B. Experimental Results and Discussion
The dropping and passing capability of the nonlin-ear optical loop mirror is first investigated separatelywithout connecting it to the ring configuration. Thereflection port with the cross-correlation peaks and re-sidual autocorrelation peaks are shown in Fig. 7(a).After passing the thru signal from the reflection portto a matched encoder, it is restored to its original se-quence, shown in Fig 7(b), and can be directed to thenext node. The dropped signal with clean autocorrela-tion peaks is the desired signal, shown in Fig. 7(c).
Two connection configurations have been tested.The first one simulates the best scenario, where thesignal is added at one node, and it is immediatelydropped at the adjacent node. Figures 8(a) and 8(c)show the two newly added signals and the residualautocorrelation peaks. The residual autocorrelationpeaks from the NOLM are very small and have littleeffect on the thru signal. Figures 8(b) and 8(d) showthe dropped autocorrelation peaks for the two nodes.Both signals have very good eye openings and littleamplitude noise.
The second connection configuration tests the per-formance of the worst-case scenario, where the signalis received after passing through all the other nodes.In this configuration, both the multiple-access inter-ference and the residual noise from the time-gatingdevices affect the system performance. After theNOLM, noise is added onto the signal at the reflectionport. The signals on the link between the two nodesare shown in Figs. 9(a) and 9(b). The signals are acombination of the newly added signal and the re-
Fig. 7. (Color online) Waveforms after nonlinear optical loop mir-ror dropping module. (a) Reflection port of the NOLM with undes-ired thru signal before restoration (cross-correlation peaks). (b) Re-stored thru signal. (c) Dropped desired signal (autocorrelationpeaks).
tored signal after the NOLM. Figures 9(c) and 9(d)how the dropped signal at node 1 and node 2.
The performance of the complete network is charac-erized by measuring the BER as a function of the av-rage received power at the clock and data recoverynit’s photodetector, shown in Fig. 10. To demonstratehe use of the protection-monitoring circuits, we com-are the BER measurements for the system with therotection-monitoring circuit to the BER measure-ents without the circuit. For the system with protec-
ion monitoring, we investigate the normal operationhere the signal goes through the east link, as shown
n Fig. 6 and the link failure operation where the sig-al is rerouted to the west link when the east linkath is broken. There is 6 dB loss for the signal afterraveling in the west link since it goes through addi-ional couplers and switches. The BER curves for theest connection scenario, direct dropping, are indi-ated by triangular data points. Error-free transmis-ion is obtained with 3.5 dB power penalty withoutdditional signal processing at the receiver. The BERurves for the worst connection scenario, indirectropping, are indicated by circular data points. Be-ause of the residual noise from the add/drop multi-lexers, a NOLM for amplitude noise suppression isdded at the receiver. The BER curve for the indirectropping with the signal going through the west linkas a different slope from the other two paths. This isecause of the additional power needed to compensateor the loss in the protection circuits also contributesore noise to both 0 bit and 1 bit signals. For the sig-
ig. 8. (Color online) Waveforms for direct dropping. (a) Signal onhe thru link after node 1. (b) Dropped signal at node 1. (c) Signal onhe thru link after node 2. (d) Dropped signal at node 2.
ig. 9. (Color online) Waveforms for indirect dropping, after signalassing through one node. (a) Signal on the thru link after node. (b)ropped signal at node 1. (c) Signal on the thru link after node 2. (d)ropped signal at node 2.
TAg
Deng et al. VOL. 2, NO. 4 /APRIL 2010/J. OPT. COMMUN. NETW. 165
nal that goes through west link during link failure,error-free transmission is obtained with 5.3 dB powerpenalty.
V. CONCLUSION
An optical CDMA ring network with protectionmonitoring circuits and all-optical add/drop multiplex-ers for optical CDMA signals has been proposed andanalyzed. A complete two-node optical CDMA ringnetwork with error-free transmission has been experi-mentally demonstrated for a truly asynchronous op-eration, and the performance has been characterized.In optical CDMA ring networks, other than multiple-access interference, cross-correlation peak distortionand residual autocorrelation peaks from the code-dropping device also contribute to the performancedegradation. However, with additional signal process-ing at the receiver using a thresholder based on a non-linear optical loop mirror, BER=10−9 is achieved. Theexperimental demonstration shows that the proposednetwork is a viable and bandwidth-efficient designthat utilizes the unique characteristics of opticalCDMA.
ACKNOWLEDGMENTS
The authors thank Kensuke Sasaki and GyaneshwarGupta from Oki Electric Industry Co., Ltd., for provid-ing the fiber Bragg grating encoders and decoders.
Fig. 10. (Color online) Bit-error-rate measurements for the threetypes of configurations with and without a protection-monitoringcircuit. Square data points: back-to-back; triangular data points: di-rect dropping where signal is dropped immediately at the next nodein the ring network: circular data points: indirect dropping wheresignal is dropped after passing through one node. No circuit: signaldoes not go through protection circuit; east link: normal operationwhere signal goes through working link; west link failure operationwhere signal goes through backup link.
his work was supported in part by the U.S. Defensedvance Research Project Agency (DARPA) underrant MDA972-03-1-0006.
REFERENCES
[1] M. R. Wilson, “The quantitative impact of survivable networkarchitectures on service availability,” IEEE Commun. Mag.,vol. 36, no. 5, pp. 122–126, May 1998.
[2] V. Alwayn, Optical Network Design and Implementation. In-dianapolis, IN: Cisco Press, 2004.
[3] S. Ayotte, M. Rochette, J. Magne, L. A. Rusch, and S. LaRo-chelle, “Experimental demonstration and simulation results offrequency encoded optical CDMA,” in Proc. IEEE Int. Conf.Communication, June 2004, vol. 3, pp. 20–24.
[4] E. Park, A. J. Mendez, and E. M. Garmire, “Temporal/spatialoptical CDMA networks—design, demonstration, and compari-son with temporal networks,” IEEE Photon. Technol. Lett., vol.4, no. 10, pp. 1160–1162, Oct. 1992.
[5] C.-S. Brès, Y.-K. Huang, I. Glesk, and P. R. Prucnal, “Scalableand asynchronous incoherent optical CDMA,” J. Opt. Netw.,vol. 6, no. 6, pp. 599–615, 2007.
[6] P. R. Prucnal, ed., Optical Code Division Multiple Access: Fun-damentals and Applications. New York: Taylor & Francis Ltd.,2006.
[7] C.-S. Brès, I. Glesk, R. J. Runser, and P. R. Prucnal, “All-optical OCDMA code-drop unit for transparent ring networks,”IEEE Photon. Technol. Lett., vol. 17, no. 5, pp. 1088–1090, Mar.2005.
[8] B. Chen, C. Guo, J. Chen, L. Zhang, Q. Jiang, and S. He, “Add/drop multiplexing and TDM signal transmission in an opticalCDMA ring network,” J. Opt. Netw., vol. 6, no. 8, pp. 969–974,2007.
[9] K. Kravtsov, P. R. Prucnal, and M. M. Bubnov, “Simple nonlin-ear interferometer-based all-optical thresholder and its appli-cations for optical CDMA,” Opt. Express, vol. 15, no. 20, pp.13,114–13,122, 2007.
[10] K. Kravtsov, Y. Deng, and P. R. Prucnal, “Self-clocked all-optical add/drop multiplexer for asynchronous CDMA ring net-works,” IEEE J. Quantum Electron., vol. 45, no. 4, pp. 396–401, Apr. 2009.
[11] A. Elrefaie, “Multiwavelength survivable ring network archi-tecture,” in Proc. ICC, 1993, pp. 1245–1251.
[12] J. J. O. Pires, “Constraints on the design of 2-fiber bi-directional WDM rings with optical multiplex section protec-tion,” 2001 IEEE/LEOS Summer Topical Meetings (UltraLong Haul DWDM Transmission and Networking), 2001, pp.13–14.
[13] N. J. Doran and D. Wood, “Nonlinear-optical loop mirror,” Opt.Lett., vol. 13, no. 1, pp. 56–58, Jan. 1998.
[14] G.-C. Yang and W. C. Kwong, Prime Codes With Applications toCDMA Optical and Wireless Networks. Boston, MA: ArtechHouse, 2002.
[15] V. Baby, W. C. Kwong, C.-Y. Chang, G.-C. Yang, and P. R. Pruc-nal, “Performance analysis of variable-weight, multilength op-tical codes for wavelength-time O-CDMA multimedia systems,”IEEE Trans. Commun., vol. 55, no. 7, pp. 1325–1333, July2007.
[16] V. Baby, C. Bres, I. Glesk, L. Xu, and P. R. Prucnal, “Wave-length aware receiver for enhanced 2D OCDMA system perfor-mance,” in Conf. on Lasers and Electro-Optics/Int. QuantumElectronics Conf. and Photonic Applications Systems Technolo-gies, 2004, paper CWH6.
