First demonstration of all-optical clock recovery at 40 GHz with standardcompliant jitter characteristics based on a quantum-dots self-pulsating
semiconductor laser
J. Renaudier (1,4), B. Lavigne (2), M. Jourdran (3), P. Gallion (4), F. Lelarge (1), B. Dagens (1), A. Accard (1),
O. Legouezigou (1) and G-H. Duan (1)
1-Alcatel Thales III-V Lab, Route de Nozay, 91460 Marcoussis, France
2-Alcatel Research & Innovation, Route de Nozay, 91460 Marcoussis, France
3- Aeroflex Europtest, 5 Place du Général de Gaulle, 78990 Elancourt, France
4-ENST, 46, rue Barrault, 75634 Paris cedex 13, France
E-mail : jeremie.renaudier@3-5lab.fr
Abstract We demonstrate for the first time, thanks to a quantum-dots Fabry-Perot laser, the compliance of an alloptical 40 GHz clock recovery with ITU-T standards on time jitter removal for frequencies larger than 4MHz.
Introduction
Self-pulsating (SP) semiconductor lasers based on
bulk or quantum-wells structures have been
extensively investigated for all-optical clock recovery.
However, little information has been given on the jitter
characteristics as compared with ITU-T standard for
clock recovery. Nowadays, SP laser diodes based on
quantum-dots are attracting great interests as they
provide RF oscillators with fast carrier dynamics and
also much better spectral purity [1,2]. In this paper,
we report phase noise measurements on a 40 GHz
clock recovery made on a self-pulsating quantumdots Fabry-Perot laser. We demonstrate for the first
time to our knowledge, that such a quantum-dots
(QD) laser enables a high frequency jitter suppression
that meets the ITU-T recommendation G825.1
requirements thanks to its intrinsic narrow free
running spectral linewidth, as compared with lasers
made on bulk or quantum wells.
Component and set-up descriptions
The SP laser under study is a Fabry-Perot (FP)
semiconductor laser made of a buried ridge structure
with a QD active layer on InP substrate (see [3] for
details). The QD-FP laser is 1080 µm-long, has a
central lasing wavelength around 1505 nm and selfpulsates around 39.4 GHz with a free running spectral
linewidth of 50kHz.
We display in fig.1 the experimental set-up. A pattern
generator is synchronized by an external 10GHz
clock, the jitter of which is controlled by a white noise
generator in the range 10kHz-1GHz, provided by an
Europtest jitter analyzer. It is used to modify the
amount of jitter of the incoming data signal. The
pattern generator provides a clock signal at 10 GHz
and a pseudo random sequence (PRBS) at 10 Gbit/s
that drive respectively a gain-switched laser
(λlasing=1552nm) and a LiNbO3 modulator. The
modulated signal is then launched into an optical time
division multiplexer (OTDM) that delivers the 40Gbit/s
signal. We recall that such a multiplexer does not
provide pure PRBS signal. When no PRBS data drive
the external modulator, OTDM delivers a clock that
will be considered later on as the reference 40GHz
clock (ref-ck). A polarisation controller is added due
to the polarisation sensitivity of the QD active layer.
The recovered clock signal is then amplified and
filtered by a 5nm bandwidth optical filter before
analysis.
GS-laser
Pattern Generator
Ck
Ck_in
Data
+
10GHz
Synthesizer
! = 1552 nm
Ext.
Mod.
! = 1505 nm
10Gbit/s
QD-FP laser
OTDM
Noise
10 kHz - 500 MHz
40Gbit/s
Jitter Analyzer
EDFA
Noise_out
Sampling scope
5 nm
filter
SOA
Fig 1 Experimental set-up
The clock recovery is characterized with a jitter
analyser. This last provides the phase noise (PN)
spectrum from which we deduce with high accuracy
the rms jitter in the bandwidth 100 Hz -500 MHz. In
this paper, the QD-FP laser is biased with a DC
current of 210 mA and locked at 39.4 GHz. It is
important to note that a tuning range of 1 GHz can be
achieved with a DBR structure [4] to match with ITU-T
frequencies. Besides a lauched optical power of 8
dBm is required partly due to the high coupling loss of
the chip test-bench. In addition, we measure a locking
bandwidth in the order of few MHz, which is related to
the narrow free running spectral linewidth of some
tens of kHz [2,5]. In comparison, it is typically in the
range of several hundreds of kHz and several MHz
with quantum wells and bulk respectively.
QD-FP-laser locking with unjittered OTDM signal
As first evaluation, we analysed the quality of the
recovered clock with respect to the PRBS length of
the incoming OTDM signal without jitter.
31
Fig. 2 Eye diagram of the incoming unjittered 2 -1
OTDM signal (left) and the recovered clock (right).
We report in fig.2 a temporal trace recorded when
31
using a 2 -1 PRBS. As may be seen, we observe a
high quality clock signal with high extinction ratio
(estimated to exceed 13dB).
Fig.3 shows PN curves recorded with a PRBS length
7
31
varying from 2 -1 to 2 -1. As reference curve, the PN
curve of ref-ck without jitter is also reported. Spurious
spikes appearing on QD-FP laser PN curves for
7
f>50MHz originate from PRBS contributions (2 -1).
QD-FP laser PN curves can be split up into 3 regions
as described in [1,6]. In the frequency region below
60kHz (A), the noise is dominated by the injected
signal and the clock recovery is fully transparent, ie
there is no jitter filtering effect. Conversely, for
f>4MHz (region C), we observe a clear demonstration
2
of a jitter filtering effect with a 1/f slope (-20
dB/decade): the noise is then determined by the QDFP laser. The frequency range extending from 60 kHz
to 4 MHz (region B) corresponds to the transition
region where the noises from the incoming signal and
the laser itself contribute both. The calculated rms
jitter varies from 0.16 ps (ref-ck) to 0.198 ps (with
31
OTDM 2 -1). Moreover, there is no difference on the
laser PN curves whatever the ODTM PRBS length
attesting the quality of the recovered clock.
Fig.3 Phase noise measurements performed on the
QD-FP laser clock and on ref-ck (bottom curve).
QD -FP laser locking with jittered OTDM signal
The jitter removal can be more clearly demonstrated
by injecting jittered signal into the QD-FP laser. Fig 4
31
gives the eye diagram of the incoming jittered 2 -1
OTDM signal (left) and the recovered clock (right),
showing a drastic decrease of the time jitter.
31
Fig. 4 Eye diagram of the incoming jittered 2 -1
OTDM signal (left) and the recovered clock (right).
Fig. 5 compares the PN curves of the recovered clock
with that of ref-ck with jitter. As seen, the rms jitter is
reduced from 1.37ps for the input to 0.31ps for the
31
recovered clock in the worst case (PRBS of 2 -1).
More precisely, we deduce from PN curves a 10%
rms jitter increase (from 0.28 ps to 0.31 ps) when the
7
31
PRBS length changes from 2 -1 to 2 -1. These
results undoubtedly demonstrate the high frequency
jitter suppression whatever the OTDM PRBS length.
Fig. 5 Phase noise measurements performed on the
QD-FP laser clock and on ref-ck (top curve).
Jitter transfer function
As other clear illustration of the jitter removal, we
report for the first time to our knowledge the jitter
transfer function (JTF) of an all-optical clock recovery.
This curve in decibels is obtained in the presence of
added noise by subtracting the PN curve of ref-ck to
31
the QD-FP laser PN curve (2 -1) of fig.5. As seen on
fig.6, we obtain a very good agreement between the
measured JTF and the JTF template expected from
ITU-T recommendation G825.21. It is to be noticed
that such a JTF can only be achieved with a QD laser
thanks to its narrow free running spectral linewidth.
Fig. 6 Measured Jitter transfer function
comparison with ITU-T G825.1 template
and
Conclusions
We show in this paper an all-optical clock recovery
over 40Gbit/s signal based on a quantum dots FabryPerot self-pulsating laser. From phase noise
measurements, we clearly demonstrate the high
frequency jitter removal as expected from a clock
recovery. In addition, we present for the first time to
our knowledge a jitter transfer function measured on
an all-optical device and its very good agreement with
the ITU-T recommendation G825.1. Such a result
constitutes a new step towards practical applications
of SP lasers for all-optical clock recovery.
Acknowledgement
This work is partly funded by the French RNRT
project ROTOR.
References
1 Akiyama T. et al, OFC’05, paper OWM2 (2005).
2 Renaudier J. et al, Electron. Lett., to be published.
3 Lelarge F. et al, IPRM’05, paper TP34 (2005).
4 Renaudier J. et al, CLEO’05, paper CtuV5 (2005).
5 Duan G-H. et al., Optoelectronics, Proceedings,
vol. 144, Issue 4 (1997).
6 Brox O. et al, ECOC’03, paper Tu4.5.3 (2003).