962 OPTICS LETTERS / Vol. 33, No.

9 / May 1, 2008

Repetition-rate multiplication using a single

all-pass optical cavity
Miguel A. Preciado* and Miguel A. Muriel
ETSI Telecomunicacion, Universidad Politecnica de Madrid, 28040 Madrid, Spain
*Corresponding author: ma.preciado@upm.es
Received December 20, 2007; accepted February 25, 2008;
posted March 6, 2008 (Doc. ID 91105); published April 25, 2008
We demonstrate a simple lossless method for the implementation of repetition-rate multiplication of a pe-
riodic pulse train. As it is showed, a single all-pass optical cavity (APOC) can increase the repetition rate of
the output pulse train by factors of 2, 3, and 4. Two different APOC implementations, based on a Gires-
Tournois interferometer and an all-pass ring resonator, are proposed and numerically demonstrated.
© 2008 Optical Society of America
OCIS codes: 070.676, 070.5753, 320.5550, 230.5750.

Techniques for generating ultrahigh-repetition-rate The spectral response of the APOC can be ex-
pulse trains are very attractive for future ultrahigh- pressed in base-band frequency as

冉 冉冉 冊
speed optical-communication systems. Several tech-
1+x
niques for generating periodic pulse trains at repeti- H共␻兲 = exp − j2 tan−1
tion rates beyond those achievable by mode locking 1−x

冉 冊冊冊
or direct modulation have been explored. One alter-
native is pulse-repetition-rate multiplication (PRRM) ␻
of a lower rate source by applying amplitude and/or ⫻tan + ␾offset ,
2FSR

再
phase spectral filtering [1–8].
PRRM methods based on all-pass filtering, and its r GTI
intrinsic high energy efficiency, are highly desirable x= , 共1兲
共1 − k兲 1/2
RR
[1–5]. One of these techniques is based on the propa-
gation of the pulse train in a first-order dispersive where FSR is the free spectral range of the filter, r is
medium and the temporal Talbot effect [1–3]. In prac- the field reflectivity of the partial reflecting mirror of
tice, however, the temporal Talbot condition is hard the GTI, and k is the coupling factor of the RR, and
to satisfy in a wide bandwidth. Another successful ␾offset = ␻0 / 2FSR. When this filter is applied to an in-
implementation of all-phase spectral-filtering PRRM put pulse train, the output can be written as
is based on a bulk optic pulse shaper that spatially ⬁
separates the frequency components of the input
pulse and uses amplitude and/or phase filters to pro- A2共␻兲 = A1共␻兲H共␻兲 = A0共␻兲␻T 兺
m=−⬁
H共m␻T兲␦共␻ − m␻T兲.
cess the signal [4,5].
In this Letter, a simple phase-filtering method 共2兲
based on a single all-pass optical cavity (APOC) for If we assume FSR T ⬇ N, then we can express

冉 冉冉 冊
lossless 2⫻, 3⫻, and 4⫻ PRRM, with a uniform
pulse-train envelope, is proposed and numerically 1+x
demonstrated. Optical cavities and other spectrally H共m␻T兲 = exp − j2 tan−1
1−x

冉 冊冊冊
periodic filters for PRRM have been widely studied
[6–8], but this is the first time to our knowledge that ␲m
this simple APOC-based method is proposed. Figure ⫻tan + ␾offset . 共3兲
1 shows a schematic of the system. Two kinds of N
APOC devices are proposed, based on an all-pass ring Since Eq. (3) describes a discrete and periodic se-
resonator (RR) and a Gires–Tournois interferometer quence with period N in the variable m, we can cal-
(GTI). culate the inverse Fourier transform of Eq. (2) using
The complex envelope of a periodic-input pulse the inverse discrete Fourier transform [9]:
⬁
train can be expressed as a1共t兲 = a0共t兲 丢 兺n=−⬁ ␦共t − nT兲, ⬁
where a0共t兲 represents the complex envelope of an in-
dividual pulse, 丢 is the convolution operator, and T is
a2共t兲 = 兺
n=−⬁
Cna0共t − nT/N兲, 共4兲
the temporal period of the signal. Using the Fourier
transform, we can express in spectral domain A1共␻兲 兵Cn其 = IDFTn兵H共m␻T兲其, 共5兲
⬁
= ␻TA0共␻兲兺m=−⬁ ␦共␻ − m␻T兲, where ␻T = 2␲ / T, and ␻ is
the base-band angular pulsation, i.e., ␻ = ␻opt − ␻0, where a2共t兲 represents the complex envelope of the
where ␻opt is the optical angular pulsation and ␻0 is output pulse train, IDFTn denotes the nth inverse
the central angular pulsation of the input pulse discrete Fourier transform, 兵其 denotes a sequence of
train. N elements, and Cn are complex coefficients, with
0146-9592/08/090962-3/$15.00 © 2008 Optical Society of America
 May 1, 2008 / Vol. 33, No. 9 / OPTICS LETTERS 963

Cn = Cn+N. The magnitude of the sequence, 兵兩Cn 兩 其, de-

scribes the amplitude of the output pulse train enve-
lope, which obviously is not affected by the phase of
Cn. It is worth noting that, for a given N, the se-
quence 兵兩Cn 兩 其 is a function in the variable x and a pe-
riodic function in the variable ␾offset with period ␲ / N.
If we search values of ␾offset and x that give as a re-
sult a sequence 兵兩Cn 兩 其 whose elements have similar
amplitudes, then we obtain uniform envelope PPRM
with N⫻ multiplication. Using the energy-
conservation condition, we can deduce that the opti-
mum sequence 兵兩Cn 兩 其 must have all the terms equal
to 共1 / N兲1/2. Thus, we can define a figure of merit for
PRRM with a uniform envelope as

FMN共x, ␾offset兲 =
N−1

兺
n=0
冉 兩Cn共x, ␾offset兲兩 − 冉冊冊
1
N
1/2 2
, 共6兲

where the optimum is FMN共x , ␾offset兲 = 0. In Figs.

2(a)–2(c), we represent this figure of merit in gray-
scale for N = 2, N = 3, and N = 4, respectively, where
darker regions correspond to more accurate solu-
tions. In Figs. 2(d)–2(f), we represent the magnitude
of the gradient of the figure of merit 兩ⵜFMN共x , ␾offset兲兩
in grayscale for N = 2, N = 3, and N = 4, where darker
regions correspond to a lower gradient. Both the FM
and its gradient must be taken into account for
choosing the parameter value. Note that the func-
tions represented have been limited for high values
in order to increase the contrast of the plots.
At this point, we show examples for 2⫻, 3⫻, and
4⫻ multiplication. In these examples we try to reach
a reasonable compromise of accuracy and robustness.
The solutions adopted in these examples may not be
the best solutions for all the cases, since the design
criterion depends on how these two factors are pon- Fig. 2. Figure of merit and gradient function for [(a) and
dered. (d)] 2⫻ multiplication, [(b) and (e)] 3⫻ multiplication, and
For the first example (N = 2, i.e., 2⫻ multiplica- [(c) and (f)] 4⫻ multiplication.
tion), both the FM function and the gradient magni-
tude [Figs. 2(a) and 2(d)] are optimum in multiple pa- equal to 0. The values of x, which may seem surpris-
rameter values. We choose the point x = 冑2 − 1, ␾offset ing, have been numerically obtained as the optimum
= ␲ / 4 rad, where both the FM and the gradient are of both functions for ␾offset = ␲ / 4 rad. We obtain a se-
quence 兵兩Cn 兩 其 = 2−1/2 ⫻ 兵1 , 1其 for the output-pulse-train
envelope.
For the second example (N = 3, i.e., 3⫻ multiplica-
tion), the optimum of the FM [Fig. 2(b)] is only
strictly reached for values of x near 1, but the gradi-
ent function [Fig. 2(e)] presents a sharp variation
there, so we must get a compromise between
FM and the gradient. In this case we choose
x = 0.75, ␾offset = 0.1 rad, and obtain a sequence
兵兩Cn 兩 其 = 3−1/2 ⫻ 兵1.066, 1.092, 0.8189其 for the output-
pulse-train envelope.
For the third example (N = 4, i.e., 4⫻ multiplica-
tion), again the optimum of the FM [Fig. 2(c)] is
strictly reached only for values of x near 1, but now
the gradient function [Fig. 2(f)] varies smoothly in
that region. We choose x = 0.97, ␾offset = 0 and obtain a
Fig. 1. Architecture of the system. The periodic pulse sequence 兵兩Cn 兩 其 = 0.5⫻ 兵0.9996, 1.0304, 0.9996, 0.9696其
train is processed by an APOC. Two APOC implementa- for the output-pulse-train envelope.
tions, based on ring resonator and GTI, are proposed. In these examples we assume an input periodic
964 OPTICS LETTERS / Vol. 33, No. 9 / May 1, 2008

Table 1. Summary of Parameter Values Obtained the values of the output period T / N, FSR, r, and k for
for 2Ã, 3Ã, and 4Ã Multiplication of an Input Pulse N = 2, 3, and 4. From x we can obtain the coupling fac-
Train of 100 GHz tor k for RR implementation or the reflectivity factor
r for the GTI implementation. For N = 2 and N = 3, the
T/N FSR
FSR of the APOC has a slightly different value to
N (ps) (GHz) r k
N / T in order to obtain the proper ␾offset value. Figure
2 5 200+ 2.604⫻ 10−2 0.4142 0.8284 3 shows the output-pulse-train intensity for these ex-
3 3.33 300+ 7.46⫻ 10−3 0.75 0.4375 amples, assuming an input Gaussian pulse train,
4 2.5 400 0.97 0.0591 which has been normalized by dividing it by the
maximum input intensity.
In conclusion, in this Letter we have presented a
simple method for a uniform envelope PRRM with a
theoretical energetic efficiency of 100%. We have
found that a single APOC can achieve three factors of
repetition-rate multiplication (2, 3, and 4). As has
been shown, it is specially suited for 2⫻ and 4⫻ mul-
tiplication, since 3⫻ multiplication is very restricted
by the variability of the solution. It is worth noting
that the length of the optical cavities designed in
these examples is in the order of 1 mm, which is eas-
ily feasible with current technology.
This work was supported by the Spanish Ministe-
rio de Educacion y Ciencia under Project “Plan Na-
cional de I + D + I TEC2004-04754-C03-02” and “Plan
Nacional de I + D + I TEC2007-68065-C03-02”.
References
1. S. Arahira, S. Kutsuzawa, Y. Matsui, D. Kunimatsu,
and Y. Ogawa, J. Lightwave Technol. 16, 405 (1998).
2. I. Shake, H. Takara, S. Kawanishi, and M. Saruwatari,
Electron. Lett. 34, 792 (1998).
3. J. Azaña and M. A. Muriel, Appl. Opt. 38, 6700 (1999).
4. A. M. Weiner and D. E. Leaird, Opt. Lett. 15, 51 (1990).
5. J. Caraquitena, Z. Jiang, D. E. Leaird, and A. M.
Fig. 3. Output-pulse-train intensity for (a) 2⫻ multiplica- Weiner, Opt. Lett. 32, 716 (2007).
tion, (b) 3⫻ multiplication, and (c) 4⫻ multiplication, cor- 6. K. Yiannopoulos, K. Vyrsokinos, E. Kehayas, N. Pleros,
responding to the first, second, and third examples, K. Vlachos, H. Avramopoulos, and G. Guekos, IEEE
respectively. Photonics Technol. Lett. 15, 1294 (2003).
7. D. S. Seo, D. E. Leaird, A. M. Weiner, S. Kamei, M.
pulse train with a central frequency 共␻0 / 2␲兲 Ishii, A. Sugita, and K. Okamoto, Electron. Lett. 39,
1138 (2003).
= 192 THz, and a temporal period T = 10 ps (pulse rep- 8. B. Xia and L. R. Chen, IEEE Photonics Technol. Lett.
etition rate of 100 GHz). We can totally characterize 18, 1999 (2006).
the APOC with FSR, and k for RR implementation, or 9. A. V. Oppenheim, R. W. Schafer, and J. R. Buck,
r for GTI implementation. In Table 1 we represent Discrete-Time Signal Processing (Prentice-Hall, 1999).