вход по аккаунту



код для вставкиСкачать
successfully transmitted over 12960km with one lnm optical fdter
at 90% compensation ratio and with five 3nni filters at 100%
compensation ratio. Nine 3nm filters resulted in an attained distance of 12240km at 90% compensation ratio.
An equivalent bandwidth of 1nm yielded the optimum filtering
for lOGbit/s RZ signal transmission in this system with a compensation ratio of 90-100%. A compensation ratio that is too small
fails to reduce the Gordon-Haus jitter effectively, while a ratio
that is too high results in waveform distortion due partly to the
effect of higher-order dispersion.
The optimum compensation ratio slightly increases with the
equivalent fdter bandwidth. The narrow 0.3nm filter yielded the
optimum compensation ratio of 80%, but its longest distance was
only 7920km. Since the narrow filter cuts some signal spectral
components, the optimum ratio is a relatively small value, which
leads to a small amount of spectral broadening. Because of the
remaining jitter, however, the resulting transmission distance is
much smaller than the longest distances of -1300Okm. The optical
fdter with the relatively broad bandwidth of 3nm, on the contrary,
yielded the optimum compensation ratio of 100‘%,but achieved a
transmission distance of only 7920km. Too broad an optical filter
fails to suppress spectral deformation, which degrades transmission performance.
Fig. 2 shows the optical spectra observed a at the transmitter
and &j’ at the longest distance achieved for each optical filter
arrangement. Signal spectral broadening was observed in all cases.
The resulting spectral widths almost equalled the equivalent bandwidths of optical filtering, except for e, which exhibits an asymmetrical spectrum stemming from higher-order dispersion in the
transmission fibre [3].
Fig. 3 shows the attained transmission distances against amplifier output power. The compensation ratio for each filter arrangement was adjusted to the optimum value for achieving the longest
distance shown in Fig. 1. In every arrangement, the optimum
amplifier output power was -ldBm, which is higher than -5.5dSm,
the signal power satisfying the soliton condition without the DCF.
The optical filter arrangement has little effect on the optimum
optical power.
Conclusion: A single-channel 10Gbitis RZ signal was transmitted
over 12960km with periodic dispersion compensation by optimally
arranging optical filters in a 720km recirculating loop. The equivalent optimum bandwidth of the loop was lnm. Excessively narrow optical filtering degrades transmission performance by cutting
off useful spectral components, while excessively broad optical filters not only fail to control spectral broadening, but also shorten
the attainable transmission distance through waveform distortion.
These results will be useful for designing single-channel and wavelength division multiplexing (WDM) systems [4, 51.
and AKIBA, s.:
‘20GbiUs-based soliton WDM transmission over transoceanic
distances using periodic compensation of dispersion and its slope’.
ECOC’96, Oslo, September 1996, Paper ThB.3.4
Analytical estimation of measurement error
performance of multiwavelength
simultaneous monitoring circuit
M. Teshima and M. Koga
IndexlMg terms: Monitoring, Frequency measurement
The authors estimate the frequency/powermeasurement error of a
multiwavelength simultaneous monitoring circuit. The initial
frequency measurement error is dominated by an imbalance in the
bandwidths of the adjacent ports, compared with those of the
optoelectrical conversion efficiency. The polarisation dependent
power monitoring error is estimated to be a . 1 8 and +00.08dBat
the normalised-bandwidths of 0.4 and 0.6, respectively.
The optical path network offers an economical nationwide telecommunication network for the next stage B-ISDN [l]. The optical path cross-connect system (OPXC) [2], the key element of the
optical path network, will be realised by exploiting wavelengthdivision multiplexing W M ) and wavelength routing techniques.
Each pathichannel wavelength must be monitored to keep system
performance high, because the wavelength of laser diodes in the
OPXC can shift due to many causes, such as aging and temperature variation. We have demonstrated a multiwavelength simultaneous monitoring circuit (MSMC) [3, 41 that used the wavelength
crossover properties of an arrayed-waveguide grating (AWG). In
this Letter, we extend the MSMC to frequency and power monitoring, and we estimate the measurement performance of the
MSMC to give a guideline for designing and manufacturing the
Acknowledgments: The authors wish to thank I. Kobayashi for his
0 IEE 1997
Electronics Letters Online No: I9971109
18 July 1997
A. Naka, T. Matsuda and S . Saito (NTT Optical Network Systems
Laboratories, I - I Hikari-no-oka, Yokosuka, Kanagawn, 239 Japan)
:-30 dB
AKIBA, s.: ‘20Gbit/s, 8,100km straight line single channel soliton
based RZ transmission experiment using periodic dispersion
compensation’. ECOC ’95, September 1995, Brussels, Paper
2 NAKA, A., MATSUDA, T., and SAITO, s.: ‘Optical RZ signal straightline transmission experiments with dispersion compensation over
5520km at 2OGbitis and 2160km at 2 x IOGbitis’, Electron. Lett.,
1996, 32, (18), pp. 16961695
3 NAKAZAWA, M., KUBOTA, H , and TAMAURA, K.: ‘Nonlinear pulse
transmission through a optical fiber at zero-average group velocity
dispersion’, IEEE Photonics Technol. Lett., 1996, 8, (3), pp. 452454
4 MATSUDA, T., NAKA, A., and SAITO, s.: ‘4 x 10Gbit/s RZ signal
transmission over 5040 km in anomalous regime with optimally
dispersion-compensated WDM channels’, to be submitted to
Electron. Lett.
1 Ith September 1997
Each optoelectric conversion efficiency Tit equals 1
We studied the transmission characteristics of the AWG, considering crosstalk, as shown in Fig. 1. The transmission profile of
the AWG can be approximated as a Gaussian profile [4, 51. Under
the crosstalk free condition, the normalised transfer function of
the adjacent port output is
where v is optical frequency, the spacing of the AWG
transmission centre frequency, AV, IS the transmission bandwidth,
Vol. 33
No. I9
qi is optoelectric conversion efficiency including loss of the AWG
and quantum efficiency of detector, and X T is crosstalk caused by
the stray-light in the AWG or the dark current of detectors and
logarithmic amplifiers. We assumed that the crosstalk value is constant at -30dB. We evaluate the discrimination curve, the ratio of
adjacent port outputs (H,(vlAv,)lH,(vUiAv,)), for the transmission
profile of AvIAv, = 0.4 (dashed-line) and 0.6 (solid-line) shown in
Fig. 1.
Fig. 2 shows the discrimination curve for a crosstalk of -3OdB
and without crosstalk, where q1 = q2 = 1.0. The solid and dashed
lines are for AvIAv, = 0.4 and AvIAv, = 0.6, respectively. As
shown in Fig. 2, the normalised discrimination bandwidth, corresponding to a crosstalk-free line, is M.13 at AvViAv, = 0.4, and is
M.45 at AvIAv, = 0.6. Thus, the normalised bandwidth AvIAvspof
0.6 is adequate for the dynamic range (- +30dB) of logarithmic
offset adjustment of the logarithmic amplifier because the slope of
the discrimination curve is approximately linear in the discrimination bandwidth. The AV,-imbalanced condition, however, requires
software calibration using a processor and a memory holding a
frequency against output table; thus, the AV,-imbalance is inadequate for a simple hardware configuration and reasonable discrimination accuracy. When the input light polarisation state is fvced,
the frequency measurement error can be calibrated by the above
two methods. The transmission profile of the AWG itself is shifted
by the input light polarisation state. The polarisation insensitive
technique using a halfwave plate is useful [6]. The uncalibrated
error, -1.1 GHz at 125GHz spacing (v,) [4],is caused by a polarisation calibration mismatch against broad wavelength area; that
is, asymmetric polarisation mode conversion caused by the wavelength difference from the wavelength of the halfwave plate.
BW=-0 13
0 .I
b -0.1
crosstalk 4 0 d B
normalised frequency vlhvSp
normalised frequency vl~Iv,,
Fig. 4 Power measurement error against AVViAV, at transmission peak
shift of 31.5%AVrp
Fig. 2 Discrimination characteristics against normalised bandwidth
Discrimination bandwidth is area corresponding to crosstalk-free line
- 0 25
normalised frequency v/Avsp
Fig. 3 Discrimination characteristics dependent on q-imbalance or AVimbalance
Next, we discuss the measurement error against the imbalance
of the optoelectric conversion efficiency (qJ or the imbalance of
the transrmssion bandwidth (AV,). Fig. 3 shows the discrimination
curve for the AV,-balanced condition (Av,lAv, = Av21Av, = 0.6),
and AV,-imbalanced condition (AvlIAv, = 0.6, Av21Av, = 0.4). The
solid and dashed lines show the efficiencies qzlq,= 1.0 and q21ql=
0.75, respectively. In the AV,-balanced condition, the efficiencyimbalance (q2/ql= 0.75) causes the frequency measurement error
of 0.025. While in the Avv,-imbalance
condition, the zero-cross frequency is shifted as shown by (iiij (corresponding to (iii) in Fig. 1j.
The zero-cross shift value is -0.1 at AvlIAv2= 0.610.4. These zerocross shift values are 2.5 and 1OGHz for 2OOGHz spaced WDM
(AvZp= 100GHz). The zero-cross shift caused by the efficiency
imbalance in the AV,-balanced condition, can be calibrated by the
The MSMC can be also used as a power-monitor by using each
output. The monitoring range for the input wavelength is the same
as the frequency discrimination bandwidth. Fig. 4 shows the
power measurement error of port HI against the normalised frequency when the centre frequency of the transmission profile shifts
to MS%v,, i.e. a polarisation dependent shift of 1.1GwZ
(-55OOMHz) for 125GHZ spacing [4]. As shown in Fig. 4, power
monitoring error is maximised at the zero-crossing point (viv, =
0) of the discrimination characteristics, which is k0.18 and
M.08dB at the normalised-bandwidthsof 0.4 and 0.6, respectively.
In conclusion, we analytically estimated the frequency or power
measurement error performance of the MSMC. The initial zerocrossing frequency shift or discrimination slope is dominated by
an imbalance in the bandwidths of the adjacent ports compared to
an imbalance in the optoelectrical conversion efficiency. The
polarisation dependent power monitoring error is rtO.18 and
M.08dB at the transmission bandwidths of 0.4 and 0.6, respectively. The estimation of measurement error performance presented herein, will give a guideline to the design and manufacture
of the AWG.
Acknowledgments: The authors are grateful to K. Sat0 of NTT
Optical Network Systems Laboratories for his continuous encouragement.
0 IEE 1997
10 June 1997
Electronics Letters Online No: I9971091
M. Teshima and M. Koga (NTT Optical Network Systems
Laboratories, 1-1 Hikari-no-oka, Yokosuka, Kanagawa, 239 Japan)
SATO, K., OKAMOTO, s., and HADAMA, H.: 'Network performance and
integrity enhancement with optical path layer technologies', IEEE
J. Sel. Areas Commun., 1994, 12, (l), pp. 159-170
2 OKAMOTO, s., and sATO, K.: 'Optical path cross-connect systems for
photonic transport networks'. Proc. IEEE Global Telecommun.
Conf. (GLOBECOM '93), 1993, pp. 474-480
I Ish September 1997
Vol. 33
No. 79
and SATO, K.:
simultaneous monitoring circuit employing wavelength crossover
properties of arrayed-waveguide grating’, Electron. Lett,, 1995, 31,
(IS), pp. 1595-1597
4 TESHIMA, M., KOGA, M , and SATO, K.: ‘Performance of
multiwavelength simultaneous monitoring circuit employing
arrayed-waveguidegrating’, J. Lightwave Technol., 1996, 14, (lo),
pp. 2277-2285
5 TAKAHASHI, H., ODA, K., TOBA, H., and INOUE, Y . : ‘Transmission
characteristics of arrayed waveguide N X N wavelength
multiplexer’, J. Lightwave Technol., 1995, 13, (3), pp. 447455
TAKAHASHI, H.: ‘Polarization mode converter with polyimide
waveplate in silica-basedplanar lightwave circuits’, IEEE Photonics
Technol. Lett., 1994, 6, (5), pp. 626-628
Cascading of a non-blocking WDM crossconnect based on all-optical wavelength
converters for routing and wavelength slot
internal wavelengths. The routing in the cross-connect is based on
the broadcast and select principle using a coupler for broadcast
and fdters for selection. Finally, the output converters are used for
wavelength slot interchanging. In the present experiment, we consider an OXC with eight external wavelengths on each of four
fibre inlets. This size would, for example, be interesting for crossconnecting WDM rings.
A partly equipped cross-connect as shown in Fig. 1 (thick line)
is implemented to demonstrate the feasibility of the suggested
OXC. The input demultiplexer is an 8-channel SiOz-Si based
phased array demultiplexer with a 3dB bandwidth of -lnm, a
channel spacing of 2OOGHz and an insertion loss of -7dB [4]. In
the experiment, the broadcasting stage, which in reality would be a
32x32 passive coupler, is simulated by a loss of 16dB corresponding to 15dB of coupler loss and 1dB of excess loss. A tunable filter with a bandwidth of -1nm is used as channel selector. The
selected internal channel is converted to the appropriate output
wavelength by the output wavelength converter and multiplexed
by the output multiplexer which is similar to the input demultiplexer.
R.J.S. Pedersen, B.F. Jlirrgensen, B. Mikkelsen,
M. Nissov, K.E. Stubkjaer, K. Wunstel, K. Daub,
E. Lach, G. Laube, W. Idler, M. Schilling,
P. Doussiere and F. Pommerau
Indexing term: Optical communication equipment, Wavelength
division multiplexing
Fig. 2 Implementation and configuration of the two types of wavelength
converters used
The authors demonstrate experimentally the possibility of
cascading five optical cross-connects based on all-optical
wavelength converters used for routing and wavelength slot
interchanging at a channel rate of lOGbit/s. Furthermore, the
possibility of both up- and down-conversion from external to
internal wavelengths is demonstrated, as well as an input power
dynamic range of 12dB.
A cross-gain wavelength converter (XGWC) consisting of a
polarisation insensitive 120Opm long semiconductor optical amplifier (SOA) is used for input conversion. An interferometric wavelength converter (IWC) which is implemented by SOAs
monolithically integrated in a Michelson interferometer (SOA-MI)
[5] is used for output conversion. The XGWC is polarisation
insensitive and tolerant to large input power variations which is
advantageous in the input stage of the OXC. The XGWC is operated with counter propagating input signals (see Fig. 2), ensuring
that the external wavelength is sufficiently suppressed to allow a
free selection of the internal wavelengths. The main advantages of
using an IWC in the output stage are their pulse reshaping capability and excellent chirp characteristics. Thus, the extinction ratio
degradation that may occur for conversion from shorter to longer
wavelengths in the XGWC is reduced, since the IWC recovers the
extinction ratio. The IWC is operated as shown in Fig. 2 and it
has previously been shown that this IWC configuration is well
suited for cascading [6]. The polarisation state of the output signal
from the XGWC is fured which is an advantage in, for example,
realisation of integrated fast tunable fdters. Since the output of the
XGWC is inverted, the IWC is operated in an inverting state leading to a non-inverting OXC.
Introduction: WDM in transport networks will offer advantages
such as higher transmission capacity and increased cross-connect
throughput [l, 21. Non-blocking operation in the space and wavelength dimensions, as well as short re-configuration time, are
essential features of the transport network, so the performance of
the optical cross-connects ( 0 x 0 ) will be of major importance.
The performance evaluation includes sensitivity to cascading, since
a channel will generally pass several OXCs. Furthermore, low
cross-talk and low sensitivity to variations in polarisation and
power levels are essential. Using wavelength converters, an OXCarchitecture could be based on simple cross-gain wavelength converters (XGWCs) at the input for routing, and interferometric
wavelength converters (IWCs) at the output for wavelength slot
interchanging [3].
In this Letter, we demonstrate experimentally the possibility of
cascading up to five OXCs based on the above mentioned principle at a channel rate of 10Gbit/s. Moreover, we show that the
polarisation insensitive OXC can handle large input power variations of individual channels and demonstrate both up- and downconversion from external to internal wavelengths. The demonstration is performed in a re-circulating loop experiment.
= a,
s --
splrtter (1EdB
arrayed wav&uxde WWM
Fig. 1 Configuration of all-optical cross-connect
matically in Fig. 1. Each of n fibre inlets carrying m external wavelengths is demultiplexed and wavelength converted to one of nxm
17th September 1997
Fig. 3 BER curves for both up- and down-conversion
Configuration of cross-connect: The cross-connect is shown sche-
received optical power, dBm
Up conversion, 1559nm to 1562nm
0 1 oxc; 2 0x0; n 3 oxcs; 0 4 oxcs; 05 oxcs
Down conversion, 1559nm to 1556nm
1 OXC; V 2 OXCs; A 3 OXCs; 0 4 OXCs; 5 OXCs
VGd. 33
No. 19
Без категории
Размер файла
438 Кб
Пожаловаться на содержимое документа