00996589

Upload: ryandika-afdila

Post on 03-Jun-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 00996589

    1/7

    682 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 4, APRIL 2002

    Performance Analysis on Semiconductor LaserAmplifier Loop Mirrors

    Hanxing Shi, Member, IEEE

    AbstractOperating characteristics of the semiconductor laseramplifier loop mirror (SLALOM) in optical time-division multi-plexing (OTDM) applications are investigated in theory. The injec-tion current of the semiconductor laser amplifier (SLA) should beadjusted to realize a flat and high switching window. The channelcrosstalk induced by thefinite carrier lifetimeand thesignal timing

    jitter should be balanced. The optimal switching window widthshould be slightly wider than half of the OTDM signal bit interval.The control pulse should be narrower than one-third of the OTDMbit interval in order not to deteriorate the bit error rate (BER)performance. Further discussion indicates that the nonlinear gaincompression in the SLA becomes a crucial limit for the SLA-basedall-optical switches to be used in the terahertz applications.

    Index TermsDemultiplexing, fiber optical communications,nonlinear gain compression, semiconductor laser amplifier (SLA).

    I. INTRODUCTION

    ALL-OPTICAL switching is essential for high-speed op-tical time division multiplexing (OTDM) systems and net-works. A semiconductor laser amplifier loop mirror (SLALOM)

    based on the optical nonlinearity of a semiconductor laser am-

    plifier (SLA), which is also called the terahertz optical asym-

    metric demultiplexer (TOAD), is one of the most promising can-

    didates [1][4]. With an interferometric construction, it is based

    on a nonlinear phase shift induced by a change in carrier den-

    sity in the SLA, but the switching speed is not restricted by thecarrier recovery rate.

    The other fiber-based optical switches, such as soliton gates

    [5] and nonlinearoptical loop mirror (NOLM)[6], [7], require at

    least tens of meters of fiber, high control pulse energy, and strict

    wavelength matching between signal and control pulses. They

    are also sensitive to both polarization and environment. Elec-

    troabsorption modulator (EAM) switches, as electrically con-

    trolled switching, are also widely studied [8], [9]. However, they

    induce large optical loss and need the device to have a steep

    nonlinear absorption curve to realize a picosecond switching

    window.

    SLA-based all-optical switches offer the advantages of signal

    amplification, wavelength flexibility, low control pulse energy,

    and large dynamic range. They can also be polarization insensi-

    tive. Based on a similar mechanism as the SLALOM, other in-

    terferometric arrangements, such as MachZehnder interferom-

    Manuscript received December 4, 2000; revised August 15, 2001. This workwas supported in part by Beijing University of Posts and Telecommunicationsthrough National 863 Project.

    The author was with the Department of ECE, University of California, SantaBarbara, Santa Barbara, CA 93106 USA. She is now with Yotta Networks, Inc.,Plano, TX 75074 USA.

    Publisher Item Identifier S 0733-8724(02)03335-2.

    eters (MZIs) or Michelson interferometers (MIs), provide addi-

    tional flexibility and monolithic integration for stable operation

    [10][12]. So far, 250-Gb/s demultiplexing has been demon-

    strated with a SLALOM having 4-ps switching window [2].

    Therefore, SLA-based all-optical switches have a bright per-

    spective in practical applications.

    Although SLA switches have been extensively studied in

    experiments, their operating performances have not been thor-

    oughly investigated in theory. The switching characteristics,

    which are governed by many factors, such as SLAs position

    in the loop, SLAs gain dynamics, and shape and power of

    the control pulse, make the analysis complex. In [13], Zhouetal. analyzed the degradation of the signal quality induced by

    the uneven optical coupler and the spontaneous emission in

    the SLA, but they ignored several crucial issues related to the

    light pulses and the switch window configuration. In this paper,

    we will investigate these issues from the point of view of the

    system performance, especially the configurations of the SLA,

    the effect of the inevitable timing jitter in the received signal

    pulses, the optimization of the switching window width, and

    the requirement on control pulsewidth.

    In addition, Tang et al. proposed, in theory, that if a pi-

    cosecond control pulse is used, a fast gain depletion process

    will appear in the dynamics of SLA and further distort the

    switching window [14]. However, how greatly does thedistorted switching window affect the demultiplexing perfor-

    mance? In this paper, we will investigate the influence of fast

    gain depletion on the bit error rate (BER) performance of the

    demultiplexed signal and confirm the upper limitation to the

    operating speed of SLA switches.

    The remainder of the paper is organized as follows. Section II

    introduces the operating principle of the SLALOM and estasb-

    lishes the theoretical model. In Section III, the demultiplexing

    performance of the SLALOM is systematically analyzed, es-

    pecially focusing on four issues: 1) the design of the SLA; 2)

    the influence of the signal timing jitter; 3) the optimization of

    the switching window width; and 4) the limitation to the con-

    trol pulsewidth. In Section IV, the influence of nonlinear gaincompression on the BER performance and the induced speed

    limitation of SLA switches are analyzed.

    II. THEORETICALMODEL

    Fig. 1 shows the schematic diagram of the SLALOM

    device with the Sagnac interferometric configuration [1]. An

    SLA is located asymmetrically inside the fiber loop as the

    nonlinear medium. Its offset distance (where

    is the velocity of light in the fiber) from the loop midpoint

    0733-8724/02$17.00 2002 IEEE

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 8/13/2019 00996589

    2/7

    SHI: PERFORMANCE ANALYSIS ON SLALOM 683

    Fig. 1. Schematic diagram of the SLALOM device. PC: polarizationcontroller. EDFA: Erbium-doped fiber amplifier. DSF: dispersion-shifted fiber.

    approximately determines the switching window width of

    the SLALOM. The received OTDM signal with a multiplexing

    bit rate of (where is the channel number

    of the OTDM system and is the single-channel bit rate)

    is launched into port A and split into two counterpropagating

    signals with equal amplitudes and identical phases. The optical

    power of the OTDM signal is chosen small enough that it does

    not significantly affect the dynamics of the SLA. On the otherhand, intense control pulses with single-channel repetition

    rate are transmitted into the fiber loop via a wavelength

    division multiplexing (WDM) coupler in order to change the

    dynamics of SLA. In the absence of this control pulse, the

    whole configuration operates as a symmetric loop and the

    recombined signal after the loop is completely reflected back

    from the port A. However, things will be physically different

    when the refractive index in the SLA is greatly changed by the

    existence of strong control pulse. Due to SLAs offset from the

    loop midpoint, the two counterpropagating signals will arrive at

    the SLA with predicted time delay in order to experience the

    different dynamic states of the SLA. As a result, the two signals

    will experience different gains and different phase shifts. After

    they are recombined through the coupler, the corresponding

    target channel will be transmitted out from port B of the loop,

    and the other channels will be reflected back from port

    A.

    The dynamic response of SLA taking into account the intra-

    band effects can be described as [14]

    (1a)

    (1b)

    (1c)

    where and are t he o ptical p ower a nd t he p hase o f

    the input light, respectively, is the SLAs dynamic gain,

    is the small signal gain, is the mode

    confinement factor, is the differential gain coefficient, is

    thetransparentcarrier density, is thetransparentinjection cur-

    rent, is the saturation energy, is the spontaneous carrier

    lifetime, is the internal loss, is the linewidth enhancement

    factor, and is the nonlinear gain compression factor, which is

    mainly related to intraband processes, such as carrier heating,

    spectral hole burning, and two-photon absorption.

    According to the Sagnac interferometric configuration, the

    optical power of the demultiplexed signal output from port B

    of the SLALOM is [13]

    (2)

    where is the optical power of the received OTDM signal,

    is the splitting ratio of the fiber coupler, and are the

    dynamic gain experienced in the clockwise and counterclock-

    wise signal pulses, respectively, is their gain

    ratio, and is the phase differ-

    ence between the two signal pulses. When no control pulse ex-

    ists in a symmetric loop, i.e., and , no signal is

    transmitted out from port B. On the other hand, when ,

    the power of the signal pulse emitted from port B can be max-

    imized.

    In practical OTDM applications, both timing jitter and

    temporal deviation between signal and control pulses inevitably

    exist, especially after long-distance transmission. At the re-ceiving port, they will be converted into intensity fluctuations

    of the demultiplexed signal and deteriorate the system perfor-

    mance [15]. In a previous work [16], we have established a

    comprehensive theoretical model for the timing jitter of signal

    pulses in all-optical demultiplexers and analyzed its influence

    on fiber-based Sagnac interferometric switches, i.e., NOLMs.

    In this paper, the same timing jitter model, in conjunction with

    a conventional optically preamplified receiver model [17], is

    adopted to numerically analyze the SLALOMs demultiplexing

    performance.

    It is assumed that the temporal relationship between signal

    and control pulses satisfies a Gaussian probability distribution

    as follows [16]:

    (3)

    where is the shift of the received signal pulse corresponding to

    the control pulse, the mean is the synchronization deviation

    of signal pulses, and is the root mean square (rms) of the

    timing jitter of signal pulses.

    In addition to the general noise terms in the optical receiver

    [17], the additional noise induced by the amplified spontaneous

    emission (ASE) in the SLA should also be considered. In gen-

    eral, the equivalent photocurrent of ASE power in the SLA isexpressed as [18]

    (4)

    where is the spontaneous emission factor, is the electronic

    charge, and is the bandwidth of the optical filter. Thus, the

    additional noise terms induced by the SLA are mainly the ASE

    shot noise and the signal-spontaneous beat noise [18]

    (5)

    (6)

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 8/13/2019 00996589

    3/7

    684 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 4, APRIL 2002

    TABLE ICOMMON PARAMETER VALUES USED IN

    THECALCULATIONS

    where is the conversion efficiency of the optical receiver,

    is its electrical bandwidth, and is the equivalent photocurrent

    of the amplifier input power. Considering all of the noise terms,

    the BER of the demultiplexed signal is given by [ 17]

    BER erfc (7)

    where erfc is the error function, and are

    the average optical powers of 1 and 0 code of the demulti-

    plexed signal in the whole frame, respectively, and and

    are their intensity noises, respectively.

    In the next section, we study the case of a 10 10 Gb/s

    OTDM system to investigate SLALOMs demultiplexing

    performance. For comparison, we also cite the performance

    of 4 10, 16 10, and 4 40 Gb/s systems. The default

    parameters used in the calculations are summarized in Tables I

    and II. The reader is referred to [16] for the parameters of the

    optical receiver.

    III. PERFORMANCEANALYSIS

    In this section, we investigate the performance of the

    SLALOM for applications slower than 200 Gb/s. Considering

    that intraband processes become noticeable for control pulses

    narrower than 1 ps only [14], nonlinear gain compression is

    temporarily neglected in this section and will not affect the

    conclusions.

    A. SLA Configurations

    It is clear, from (2), that both the power and shape of the de-

    multiplexed signal pulses depend on the gain ratio of the two

    TABLE IIDIFFERENTPARAMETERVALUES FORDIFFERENTBITRATESYSTEMS

    Fig. 2. Dependence of the normalized demultipelxed signal power on the gain ratio of two signal branches , splitting ratio offiber coupler , and SOA linewidth enhancement factor .

    counterpropagating signals through the SLA. Therefore, opti-

    mizing the SLAs parameters is the first step to obtain the best

    demultiplexing performance.

    Fig. 2 illustrates the dependence of the normalized demul-

    tiplexed signal power on the gain ratio of two counterpropa-

    gating signals, splitting ratio of the fiber coupler and theSLAslinewidth enhancement factor . As a Sagnac interferometer,

    the symmetry of the optical coupler is a requisite to avoid the

    intrinsic crosstalk [15]. For simplification, an ideal 1:1 optical

    coupler is assumed in the following. It is also shown, in Fig. 2,

    that the normalized demultiplexed signal can be maximized with

    thegain ratio between 0.2 and 0.4. However, the optimal value

    of varies with the linewidth enhancement factor . An SLA

    with a larger requires a larger gain ratio but also produces a

    higher normalized output power. Therefore, an SLA with high

    linewidth enhancement factor can improve the demultiplexing

    performance, which was experimentally verified by Manning

    [19].

    Fig. 3 shows the dynamic gains of two signal pulses, and, and the corresponding switching window of a SLALOM

    under different injection current in the 10 10 Gb/s OTDM

    system. The width of the switching window is approximately

    equal to the time difference when the two counterpropagating

    signal pulses separately arrive at the SLA, which is twice the

    temporal offset of SLA from the fiber loop center [1]. The lower

    subfigure indicates that there is an optimal injection current to

    realize a high and flat switching window. Insufficient injec-

    tion current reduces the signal gain and further degrades the

    switching window amplitude. Conversely, an overbiased SLA

    will generate a gain ratio and a phase difference greater than

    the optimal values, which results in a dented switching window.

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 8/13/2019 00996589

    4/7

    SHI: PERFORMANCE ANALYSIS ON SLALOM 685

    Fig. 3. Dynamic gains obtained by two signal branches (upper) and thecorresponding switch window of SLALOMs (lower) in a 4 2 10 Gb/s system.

    Therefore, the injection current of the SLA must be specificallyadjusted according to the control pulse power.

    B. Optimizations of Switching Window Width and Temporal

    Parameters of Signal Pulses

    The switching window width of the SLALOM is determined

    by the position deviation of the SLA in the fiber loop. As shown

    in Fig. 3, there is a switching floor outside the main switching

    window, which means that even for the nontarget channels,

    signal leakage from port B is still inevitable. It originates from

    the finite carrier lifetime in the SLA, which results in finite

    recovery time of the dynamic gain and, finally, produces a gain

    discrepancy between the two signal pulses outside the mainwindow. This is the main contribution to the channel crosstalk

    in SLALOM operation. Widening the switching window can

    increase the gain difference outside the switching window and

    aggravate the channel crosstalk.

    However, in practice, the narrowest switching window does

    not correspond to the best system performance. If the switching

    window is too narrow, the unavoidable timing jitter in the re-

    ceived signal will be transformed into intensity noise in the de-

    multiplexedsignal and increase the biterror rate. Hence, to trade

    off the effects of the channel crosstalk and the signal timing

    jitter, the switching window width of the SLALOM has to be

    optimized to realize the best demultiplexing performance. Be-

    sides the timing jitter, any deviation of the target signal pulsesfrom the window center will also increase the error rate. There-

    fore, both the signal timing jitter and temporal deviation

    between signal and control pulses should be controlled.

    Fig. 4 shows the power penalty induced by both and

    under different switching window widths in a 10 10 Gb/s

    situation. The power penalty is defined as the excess signal

    power required for the 10 BER value in comparison to the

    ideal case having zero and optimized value of . The refer-

    ence values of the signal power received by the receiver are, re-

    spectively, 1.4 dBm, 2.9 dBm, and 2.7 dBm for , ,

    and ps. It is clear that the optimal temporal deviation should

    be slightly beyond half of the switching windowwidth , which

    Fig. 4. Power penalty induced by signal timing jitter and temporal deviationbetween signal and control pulses in a 10 2 10 Gb/s situation.

    means that the target signal pulse should be positioned around

    the switching window center. The requirement of the small de-

    viation beyond is because that the finite control pulsewidthretards the switching window. Any small departure from this

    optimum will seriously deteriorate the demultiplexing perfor-

    mance. Under the fixed , the timing jitter of the signal pulse

    will increase the power penalty, especially if the signal pulse de-

    viates from the switching window center. In addition, the wider

    switching window can tolerate larger signal timing jitter for the

    same power penalty. Comparing the received signal powers in

    the three cases, we also find that there is an optimal switching

    window width corresponding to the minimal signal power.

    Considering all of these phenomena, Fig. 5 shows the depen-

    dence of demultiplexing performance on the switching window

    width and the pulse timing jitter for different bitrate cases. Thereceived signal powers are, respectively, 5.8, 2.9, 1.0, and

    5.0 dBm for the 4 10, 10 10, 16 10, and 4 40 Gb/s sys-

    tems. In Fig. 5(b), the horizontal axis corresponds to normal-

    ized to the signal pulse interval in order

    to compare the different bit-rate cases. The temporal deviation

    between signal and control pulses are set as optimum, ac-

    cording to this analysis. Under fixed timing jitter, there is an

    optimal switching window width, which is a little bit wider than

    half of the OTDM bit interval . This phenomenon was ex-

    perimentally observed as the existence of an optimal switching

    contrast while varying the switching window width [11]. In the

    10 10 Gb/s case, the optimal value of is about 6 ps when

    ps. With a worse timing jitter, a wider switching windowis required to tolerate the timing jitter and to obtain the optimal

    BER performance. Besides, at higher speed, the SLALOM is

    more sensitive to the signal timing jitter, which asks for stricter

    requirement to the optimization of the switching window width.

    Further comparing the curves of 16 10 and 4 40 Gb/s

    cases, it is interesting to see that, in the system with fewer

    channels, i.e., relatively higher single-channel bit rate , the

    demultiplexing performance is more sensitive to the widening

    of the switching window above . This is because, for

    higher , there is less time for the SLA to recover its dynamic

    gain between two successive control pulses in order to generate

    worse crosstalk from the nontarget channels. Also, the system

    http://-/?-http://-/?-
  • 8/13/2019 00996589

    5/7

    686 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 4, APRIL 2002

    Fig. 5. BERperformanceof demultiplexed signalversus switchwindow width

    under different signal timing jitters

    . (a) 102

    10 Gb/s; (b) 42

    10, 162

    10,and 4 2 40 Gb/s.

    with a higher single-channel bit rate requires higher optical

    power of the received signal for the same BER value. Therefore,

    in order to realize the specified OTDM speed, it is beneficial to

    adopt a combination of a higher channel number and a slower

    single-channel bit rate.

    Fig. 5(a) also shows that in the 10 10 Gb/s system, the

    jitter tolerance of a SLALOM with ps in the range of

    10 BER is about 1.5 ps. Compared with a NOLM having

    the same interferometric configuration [16], the SLALOM has

    a relatively higher tolerance for timing jitter. In a fiber-based

    NOLM switch, the switching window shape is determined bythe integral effect of the walkoff between signal and control

    pulses along a few kilometers of fiber [6]. At any moment,

    the instantaneous window of the nonlinear interaction has the

    same shape as the control pulse and is very sensitive to the

    pulse timing jitter. All of the fiber-nonlinearity-based switches

    require this strict synchronization. However, the optical nonlin-

    earity of a SLALOM is concentrated in the SLA element and

    the switching window does not change with time. Therefore,

    SLA-based switches are inherently less sensitive to the signal

    timing jitter, which was proved experimentally by Sokoloff [3].

    It is undoubtedly one of the critical benefits in practical appli-

    cations.

    Fig. 6. Dependence of demultiplexing performance on normalized controlpulsewidth in different OTDM cases.

    C. Limitation on Control Pulsewidth

    As shown previously, the temporal deviation of the SLA

    from the fiber loop center, i.e., , determines the switching

    window width. In addition, the form of the control pulse is also

    important for determining the shape, especially the edges, of

    the window. A narrower control pulse can generate a flatter

    switching window, which has better tolerance for signal timing

    jitter. A narrower pulse can also achieve sharper edges of

    the switching window, which reduces the crosstalk from the

    neighboring channels.

    Fig. 6 shows the dependence of the BER performance on

    the normalized control pulsewidth under different OTDM

    bit rates. The horizontal axis is the ratio of the full-width at

    half-maximum (FWHM) of the control pulse to the signal pulse

    interval . Because the dynamic gain of the SLA is involvedwith the total optical energy, the control pulse energy is fixed for

    different cases. The optimal switching windows are set as 13,

    6, 4, and 4 ps for the 4 10, 10 10, 16 10, and 4 40 Gb/s

    systems, respectively. The received signal powers are the same

    as in Section III-B.

    It is shown that there is some limit on the control pulsewidth

    to guarantee the specified demultiplexing performance. The

    upper limit of in the scale of 10 BER value is about 1/3 of

    the single-channel bit interval. For example, in a 10 10 Gb/s

    system, the control pulse should be narrower than 3 ps to obtain

    BER .

    IV. LIMITATION OFNONLINEARGAINCOMPRESSION

    In the previous section, we only discussed the OTDM sys-

    tems slower than 200 Gb/s. In terahertz OTDM applications,

    there are mainly three factors limiting the resolution of the SLA

    switches: 1) the length of the device; 2) the control pulsewidth;

    and 3) the intraband processes. The finite length of the SLA de-

    cides the propagation time of the pulses and further induces an

    asymmetric switching window [3], [20]. Furthermore, the width

    of the switching window cannot be narrower than the control

    pulsewidth [21]. These limitations could be weakened by using

    a shorter device and a narrower control pulse. However, with

    a picosecond control pulse in a terahertz system, the intraband

    http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-
  • 8/13/2019 00996589

    6/7

    SHI: PERFORMANCE ANALYSIS ON SLALOM 687

    Fig. 7. Dynamic gain of SLA and the corresponding switch window shape ofSLALOM with in a 40 2 10 Gb/s OTDM situation.

    Fig. 8. BER performance versus control pulsewidth in the 40 2 10 Gb/sOTDM system with and without consideration of nonlinear gain compression.

    processes lead to nonlinear gain compression in SLA, which is

    no longer negligible.

    This section discusses this issue with an example of

    40 10 Gb/s system. The related parameters are that the signal

    pulsewidth is 0.6 ps, ps, ps, and the received

    signal power is 2.9 dBm. Fig. 7 shows the dynamic gain of SLA

    and the shape of the switching window under different control

    pulsewidths under gain compression factor of . There

    is apparently a gain depletion process right after the input of

    the 0.6-ps control pulse. The shape of the switching windowis correspondingly distorted, with a big cusp in the primary

    window and a small secondary window afterwards.

    Fig. 8 shows the effect of the intraband processes on the BER

    performance of the demultiplexed signal in the same system.

    Neglecting the intraband processes, the narrower control pulse

    results in the better BER performance. However, due to the exis-

    tence of the intraband processes, the over narrow control pulse

    could deteriorate the BER performance, especially when con-

    sidering the signal timing jitter.

    Fig. 9 shows the BER value versus the signal power detected

    by the receiver under different values of compression factor and

    signal timing jitter. The control pulsewidth is 0.6 ps. It confirms

    Fig. 9. BER performance versus received signal power in the same 40 2 10Gb/s situation with and without consideration of nonlinear gain compression, inwhich ps.

    that the combination of intraband processes and signal timing

    jitter can destroy the BER performance. If the signal timingjitter is negligible, the distorted switching window can only en-

    hance the channel crosstalk to a certain extent, which is shown

    as a small power penalty. However, in practice, because of the

    inevitable signal timing jitter, the secondary window will seri-

    ously deteriorate the BER performance and finally boost up the

    error floor.

    Therefore, nonlinear gain compression must be considered

    for SLA-based all-optical switches in terahertz OTDM applica-tions. Furthermore, it brings about much stricter limit on timing

    jitter of the pulses, which corresponds to stricter requirements

    on the system designs. That is one of the main reasons why

    the narrowest switching window reported with a SLALOM was

    just 4 ps, until now [2]. Even though a subpicosecond switchingwindow has been realized with a MachZehnder structure [22],

    no reports of acceptable BER performance have been published.

    V. CONCLUSION

    In this paper, we comprehensively analyze the demultiplexing

    performance of the SLALOM device in OTDM systems. Above

    all, the SLA is the key element for guaranteeing the switching

    performance. An SLA with higher linewidth enhancement

    factor is beneficial to get higher output power. The injection

    current of the SLA should be specifically adjusted to obtain a

    flat and high switching window.

    There is a compromise in switching window width betweenthe channel crosstalk induced by the finite carrier lifetime and

    the inevitable timing jitter in the received signal pulses. The op-

    timal switching window width should be slightly wider than half

    of the pulse interval of the OTDM signal. The upper limit of the

    control pulsewidth is about 1/3of the OTDM bitinterval in order

    not to restrict the BER performance. In a practical design, it is

    suggested to adopt the combination of higher channel number

    and lower single channel bitrate, which has better tolerance to

    the timing jitter of the signal pulses.

    All of these conclusions can be extended to various SLA-

    based interferometric switches and are instructive for subtera-

    hertz OTDM system design. Furthermore, compared with fiber-

    http://-/?-http://-/?-http://-/?-http://-/?-
  • 8/13/2019 00996589

    7/7

    688 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 20, NO. 4, APRIL 2002

    based switches, the SLA-based switches are inherently insensi-

    tive to the signal timing jitter, which is a merit in practical ap-

    plications.

    For terahertz OTDM applications, the nonlinear gain

    compression originating from carrier heating and spectral

    hole burning cannot be ignored. The interaction between the

    distorted switching window induced by the intraband processes

    and the signal timing jitter will seriously raise the error floorand, in the worse case, will prevent the demultiplexed signal

    from being readable. Therefore, SLA-based all-optical switches

    cannot be used in terahertz OTDM applications.

    ACKNOWLEDGMENT

    The author wishes to thank D. Cohen and P. Royo at Univerity

    of CaliforniaSanta Barbara for fruitful discussions and Prof.

    J. Lin at Beijing University of Posts and Telecommunications.

    REFERENCES

    [1] J. P. Sokoloff, P. R. Prucnal, I. Glesk, andM. Kane, A Terahertz OpticalAsymmetric Demultiplexer (TOAD),IEEE Photon. Technol. Lett., vol.15, pp. 787790, July 1993.

    [2] I. Glesk, J. P. Sokoloff, and P. R. Prucnal, Demonstration of all-opticaldemultiplexing of TDM data at 250 Gbit/s, Electron. Lett., vol. 30, pp.339340, Feb. 1994.

    [3] J. P. Sokoloff, I. Glesk, P. R. Prucnal, and P. K. Boncek, Performanceof a 50 Gbit/s optical time domain multiplexed system using a TerahertzOptical Asymmetric Demultiplexer, IEEE Photon. Technol. Lett., vol.6, pp. 98100, Jan. 1994.

    [4] K. Suzuki, K. Iwatsuki, S. Nishi, and M. Saruwatari, Error-free demul-tiplexing of 160 Gbit/s pulse signal using optical loop mirror includingsemiconductor laser amplifier,Electron. Lett., vol. 30, pp. 15011503,Sept. 1994.

    [5] K. H. Ahn, M. Vaziri, B. C. Barnett, G. R. Williams, X. D. Cao, M.N. Islam, B. Malo, K. O. Hill, and D. Q. Chowdhury, Experimentaldemonstration of a low-latency fiber soliton logic gate, J. LightwaveTechnol., vol. 14, pp. 17681775, Aug. 1996.

    [6] N. J. Doran and D. Wood, Nonlinear-optical loop mirror, Opt. Lett.,vol. 13, no. 1, pp. 5658, 1988.

    [7] M. Nakazawa, E. Yoshida, T. Yamamoto, E. Yamada, and A. Sahara,TDM single channel 640 Gbit/s transmission experiment over 60 kmusing a 400 fs pulse train and a walk-off free, dispersion-flattened non-linear optical loop mirror, Electron. Lett., vol. 34, pp. 907908, Apr.1998.

    [8] S. B. Alleston, P. Harper, I. S. Penketh, I. Bennion, and N. J. Doran,1220 km propagation of 40 Gbit/s single channel RZ data overdispersion managed standard (nondispersion shifted) fiber, in Proc.OFC/IOOC99, San Diego, CA, 1999, PD3, pp. PD3/1PD3/3.

    [9] V. Kaman, A. J. Keating, S. Z. Zhang, and J. E. Bowers, SimultaneousOTDM demultiplexing and detection using an electroabsorption modu-lator,IEEE Photon. Technol. Lett., vol. 12, pp. 711713, June 2000.

    [10] R. Hess and M. Caraccia-Grosset al., All-optical demultiplexing of 80to 10 Gb/s signals with monolithic integrated high-performance Mach-Zehnder interferometer, IEEE Photon. Technol. Letters, vol. 10, pp.165167, Jan. 1998.

    [11] S. Diez, C. Schubert, R. Ludwig, H. J. Ehrke, U. Feiste, C. Schmidt, andH. G. Weber, 160 Gbit/s all-optical demultiplexer using hybrid gain-transparent SOA Mach-Zehnder interferometer,Electron. Lett., vol.36,pp. 14841486, Aug. 2000.

    [12] A. D. Kersey, M. J. Marrone, and M. A. Davis, Polarization-insensi-tive fiber optic Michelson interferometer, Electron. Lett., vol. 27, pp.518520, Mar. 1991.

    [13] D. Zhou, K. Kang, I. Glesk, and P. Prucnal, An analysis of signal-to-noise ratio and design parameters of a Terahertz Optical Asymmetric

    Demultiplexer,J. Lightwave Technol., vol. 17, pp. 298307, Feb. 1999.[14] J. M. Tang, P. S. Spencer, and K. A. Shore, Influence of fast gain deple-tion on the dynamic response of TOADs, J. Lightwave Technol., vol.16, pp. 8691, Jan. 1998.

    [15] K. Uchiyama, T. Morioka, S. Kawanishi, H. Takara, and M. Saruwatari,Signal-to-noise ratio analysis of 100 Gbit/s demultiplexing using non-linear optical loop mirror,J. Lightwave Technol., vol. 15, pp. 194201,Feb. 1997.

    [16] H. X. Shi and J. T. Lin, Theoretical analysis on polarization deviationand switchwindow optimizationin nonlinear optical loop mirror demul-tiplexer, J. Lightwave Technol., vol. 17, pp. 25722576, Dec. 1999.

    [17] N. A. Olsson, Lightwave systems with optical amplifiers,J. LightwaveTechnol., vol. 7, pp. 10711082, July 1989.

    [18] G. P. Agrawal,Nonlinear Fiber Optics, 2nd ed. New York: Academic,1995, ch. 11, pp. 471530.

    [19] R. J. Manning, A. E. Kelly, A. J. Poustie, and K. J. Blow, Wavelengthdependence of switching contrast ratio of semiconductor optical ampli-

    fier-based nonlinear loop mirror,Electron. Lett., vol. 34, pp. 916918,Apr. 1998.[20] G. Swift, Z. Ghassemlooy, A. K. Ray, and J. R. Travis, Modeling of

    semiconductor laser amplifier for the Terahertz Optical Asymmetric De-multiplexer,IEE Proc. Circuits Devices Syst., vol. 145, no. 2,pp. 6165,1998.

    [21] K. I. Kang, T. G. Chang, I. Glesk, and P. R. Prucnal, Comparison ofSagnac and Mach-Zehnder ultrafast all-optical interferometric switchesbasedon a semiconductor resonant optical nonlinearity,Appl. Opt., vol.35, no. 1, pp. 417426, 1996.

    [22] S. Nakamura, Y. Yeno, and K. Tajima, Ultrafast (200-fs switching, 1.5Tb/s demultiplexing) and high-repetition (10 GHz) operations of a po-larization-discriminating symmetric Mach-Zehnder all-optical switch,

    IEEE Photon. Technol. Lett., vol. 10, pp. 15751577, Nov. 1998.

    Hanxing Shi (S98M01) was born in InnerMongolia, China, in 1973. She received the B.S.degree in physics from Beijing Normal University,Beijing, China, in 1994 and the Ph.D. degree inelectrical engineering from Beijing University ofPosts and Telecommunications, Beijing, China, in1999.

    From 1999 to 2001, she worked as a PostdoctoralResearcher at the University of California, SantaBarbara. She is currently with Yotta Networks, Inc.,Dallas, TX. Her research interests include photonic

    switching, optical networking, WDMOTDM technologies and optical testing.