JOlJRNAL OF LIGHTWAVE TECHNOLOGY, VOL. 10, NO.

1, JANUARY 1992

57

Narrow-Band Optical Channel-Dropping Filter

Hermann A. Haus, Fellow, IEEE, and Y. Lai

Abstruct- Waveguide couplers are combined with X / 4 shifted

distributed feedback resonators to produce narrow-band channel
dropping filters. With distributed feedback coupling parameters
already realized today, one may construct banks of 40 channeldroppping filters with 10-GHz bandwidths each that do not
mutually interfere. Each stage may also inject its own signal or
reinject the tapped signal so that other stages can pick it up.

hi4
4L

-,

input signal

- 5

K waveguide (211

\-

Fig. 1. Schematic of one-section structure

I. INTRODUCTION

PROMISING approach to high bit rate communication

systems is the combination of time division multiplexing
(TDM) and wavelength division multiplexing (WDM) [ 11.
The TDM would be run at rates compatible with electronic
switching rates. The WDM would be used to increase the bit
rate by utilizing the very large bandwidth of optical fibers. The
hit rates accessible to electronic switching are of the order of
5 Gb/s. If the communication is at a wavelength of 1.53 pm
and each channel is to be filtered efficiently, one needs optical
filters of the order of 0.5 A bandwidth. These channel-dropping
lilters should transfer radiation within the band to the receiver
without perturbing the radiation outside the band. It is also
desirable that each stage should be able to inject its own signal
or reinject the tapped signal so that other stages can pick it up
as well. This means, of course, that the filters be active.
The tunable filter invented by Alferness and Cross [2] is not
sufficiently narrow band to be used for that purpose. In this
letter we propose a filter that utilizes the resonant characteristic
of several X/4 shifted DFB resonators so as to achieve an
(improved) transfer characteristic. The concept originates from
thc narrow-band distributed feedback reflector design [3], [4].
The fundamental structure is illustrated in Fig. 1. An optical
waveguide is coupled to an adjacent waveguide made into a
X / 4 shifted DFB resonator. The reflection in guide (2) [the
transmission bus] is (ideally) perfect if the resonator is
lossless. It is clear that the reflection will not be perfect, that
part of the incident power will be absorbed, when there are
finite losses in the resonator, and/or when power is coupled
out of the resonator.
Near the center of the stopband, the structure of Fig. 1
can be represented by the equivalent circuit shown in Fig. 2.
Here the conductances G,, Gb represent the losses due to the
output coupling of the resonator to both sides. Their values can
be adjusted by adjusting the grating lengths. With no output
coupling, the resonator reflects perfectly at resonance, which
Manuscript received April 16, 1991. This work was supported in part by the
National Center for Integrated Photonic Technology and by the Joint Services
Electronics Program DAAL03-89-C-0001.
The authors are with the Department of Electrical Engineering and Computer Science and Research Laboratory of Electronics, Massachusetts Institute
of Tcchnology, Cambridge, MA 02139.
IEEE Log Number 9103307.

G..

Fig. 2. The equivalent circuit near center band.

occurs at the center of the stopband. In this case G, = Gb = 0.

If one shortens the length of grating on the right-hand side,
Gb # 0, power is delivered to Gb. It is clear from the
equivalent circuit that maximum power transfer occurs when
Gb = Yo/2. Then half of the incident power is absorbed
in Gb, 1/4 is reflected in the feeder waveguide, and 114 is
transmitted. This is not necessarily the limit: if another lossless
A/4-shifted DFB resonator is coupled to the waveguide as
shown in Fig. 3, the equivalent conductance in guide (2) on
the right-hand-side of resonator (1) is infinite at resonance.
Thus if Gb = Yo,full power transfer from the bus to the
output of resonator (a) is possible.
The structure in Fig. 3 can be used also to couple narrowband signals into the bus (see Fig. 4). This is simply the
time-reversed version of the operation we just mentioned.
The two structures in Figs. 3 and 4 can be combined into a
compact form as shown in Fig. 5. The central section ensures
fulI power transfer between the bus and resonators in the
first and third section. The first section takes the narrow-band
signal out of the bus while the third section couples the
signal from the transmitter into the bus.
If one wants a receiver that can receive and retransmit the
same signal at the same time, the structure in Fig. 6 can be
used. The first and third sections are now directly connected
with a X/4 shift. The signal is taken out from the central
section. Of course one needs some gain in the central section
to compensate the loss due to the output coupling.
A precise analysis of the structures requires a full analysis
of the coupled mode equations and the DFB structure. In
particular, the simple equivalent circuits of Figs. 2-4 do not

0733-8724/92$03.00 0 1992 IEEE

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 10, NO. 1, JANUARY 1991

58

to receiver

signal to recetver

from transmitter

t
Input signal
------b
output signal

___)

input signal

output signal

I
I

3a

hi4

(a)

Equivalent Circuit

3b

I
I
I

Fig. 5. Schematic of three-section structure that can be used to receive signal

and inject new signal.

+ 2 4:

T
( b)

L1a

hi4

hi4

AJ4

II

I t

II

Fig. 3. Schematic of two-section structure that couples signal out

active
layer

to receive1
from tla"sm,ttel

IS1

Fig. 6. Schematic of three-section structure that retransmit the received

signal.

hi4

"

.
.

n_h
L

I f

, ,

Fig. 4.

Schematic of two-section structure that couples signal in.

give any information about the

of the resonators on
the two sides of the stopband that cause spurious responses.
These spurious responses lie at the edge and outside the stopband of the grating. In order to avoid performance degradation
caused by the spurious, one has two options:
1) The filter is designed so as to suppress the spurious
response.
2) The stopband of each filter is made wide enough so that
all channels can be accommodated within one half of the
stopband. In this case none of the spurious need appear
within one of the other channels.
coupled waveguides of
Option 1) can be realized if one
different dispersion characteristics, similar to the AlfernessCross filter. If the coupled waveguides are sufficiently detuned
at the edge of the stopband, the spurious can be suppressed.

Option 2) is attractive because it does not require

two different waveguide dispersions. Very high ti values
have already been reported in the literature [5]. By
picking 6 large enough one can make the stopband wide;
by choosing the coupling between the two waveguides
small enough one can achieve a sufficiently narrow filter
bandwidth.
Unavoidable waveguide losses prevent high Q values and
hence narrow filter bandwidths. Therefore, any narrow band
filter of the type discussed here will have to have gain
compensating for the loss. Technically, this is not a serious
constraint, because the structure is merely a modification of a
X/4 shifted DFB laser coupled to another waveguide.
In this paper we take up all the issues discussed in the
introduction. We solve the coupled-mode equations exactly
to calculate the response of the structure. We show that,
with distributed feedback coupling parameters already realized
today, One may construct banks Of 40 channel-dropping
With 10-GHz bandwidths each that do not mUtUalIy interfere.
so that Other
Each stage may
reinject the tapped
stages can pick it up as well.

11.

COUPLING EQUATIONS AND AN

APPROXIMATE
ANALYSIS

We donote by Ai and Bi the forward and backward amplitudes of the waves in guide i(= 1,a), from which the spatial
dependences exp[fj(~/h)z]
have been removed. ti is the
coupling coefficient of the DFB structure, p is the coupling
between the two waveguides. Then the equations for the waves
are:
d
-Ai = - j S A I + KBI - j p A 2
(2.1)
dz

HAUS AND LAI: NARROW-BAND OPTICAL CHANNEL-DROPPING FILTER

59

0.6

-0.02

-0.04

0.00

0.04

0.02

8IK

where h = (w - w o ) / u g is the detuning parameter, wo is the

frequency at band center, wg is the group velocity, and 6 is real.
A real K implies that the reference planes have been picked
at a 90 distance from the peak of the index perturbation. We
use the parameter o(# 1) to allow for a difference of group
velocities in the two guides.
Before solving the equations exactly, it is useful to set up
approximate relations for the filter structure, as has been done
in [O],in order to lay out a design to zeroth order. The equivalent circuit of Fig. 2 is characterized by three parameters; the
resonance frequency of the filter, the impedance
or
the external Q , Q c , and the conductance G = Gb(G, = CO)
or the internal Q , Q0. The resonant frequency w, is the Bragg
frequency of the DFB structure.
(2.5)
, ,
where
is the phase velocity and A is the grating period.
The external Q of the lossfree resonator in Fig. 1 is easily
computed by evaluating the power escaping from the initially
excited resonator P,, and the energy W in the resonator [6]:

-0.04

-0.02

0.00

0.02

0.04

1.o

0.8

0.0

I
-3

-2

-1

81K
(c)

The inverse internal Q is given by the dissipated energy

per radian divided by the internal energy. The dissipated
power is, in fact, absorbed in the load (receiver).

Fig. 7. The resDonses of single resonator; the Darameters are: a = 1,

Thus, if the filter bandwidth Awf is given, and a certain width

of the stopband is to be achieved, one has the constraint:
Here we have assumed that Llb >> I&,
so that the power
escaping through the port on the left-hand side can be ignored.
For the structure in Fig. 3, full power transfer occurs when
G = Y,,. This gives a relation for the resonator length and the
coupling parameter p:
e-2KL2~

-41~/t4~.

(2.8)

For the structure in Fig. 3, the total Q is

and therefore the filter bandwidth is given by:

The width of the stopband, Aw, is given by

Aw, = 2v,n.

(2.11)

awf= 4

~ ~ , 1 ~ / ~ 1 ~ . (2.12)

This analysis shows clearly that there is only one degree of

freedom left in the design, the value of K . When K is chosen,
the ratio of
is fixed, and with it the resonator length. Of
course, P / K must be much less than unity to start with, in order
to satisfy the high-Q criterion of the approximate analysis on
which it is based.
The preceding simple analysis gives the order of magnitude
of the design parameters. Typically, ps of the order of
5-10 cm- are easily achievable. A K value of the order
of 300 cm- has been reported [5].Thus it appears that the
designer has sufficient freedom for a wide range of filter
characteristics. If one chooses K = 200 cm-I and p =
5 cm-, then the stopband is of the order of 4000 GHz and
the filter bandwidth is about 10 GHz. To avoid the spurious
response outside the stopband, only the central half portion of
the stopband can be used (i.e., 2000 GHz). This bandwidth can

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 10, NO. 1, JANUARY 1992

60

n
I

-0.02

-0.04

0.00

0.02

0.04

-0.02

-0.01

0.2

0.2

0.0
-0.04

-0.02

0.00

0.02

0.0
-0.02

0.04

0.01

0.02

0.01

-0.02

\/

0.00

-0.01

0.00
8/K

( b)

EIz

$ 0.6
2 0.4
c

0.4

0.2

0.0 -3

-2

-1

&/K
0

0.2

0.0

-2

&lK

(c)
Fig. 8. The responses of single resonator with different group velocities in
the two guides; the parameters are: a = 1.3, li = 200 cm-, ki = 5 cm-,
L1, = 250 pm, Llh = 150 pm.

accommodate 40 channels (each with a 10 GHz bandwidth)

spaced 5 times the channel width apart (i.e., 50 GHz). These
results show that the design of such a filter bank is realistic
In the sequel we investigate the problem exactly as
described by the differential equations.
111. RESULTSFROMEXACTANALYSIS

Equations (2.1)-(2.4) can be integrated easily by computer,

under suitable boundary conditions. Fig. 7 shows the exact
evaluation of the response of the single resonator, whose
approximate equivalent circuit is shown in Fig. 2. At and near
resonance, the equivalent circuit is a good representation of
the transmission and reflection characteristic of the structure.
Indeed, at resonance, 1/2 of the power is transferred to
the resonator when the Qs are adjusted optimally, as has
been done for the plot; 1/4 of the power is reflected, 1/4
transmitted in the transmission bus. The plot of the response
over the entire grating bandwidth shows spurious response
near the edge of the stopband that is not represented by the
equivalent circuit. If this response is considered deleterious,

(c)
Fig. 9. The filter characteristic of the structure in Fig. 3 . The parameters are:
= 1, K = 200 cm-, p = 5 cm-, L1, = Lz0 = Lst, = 250 pm. and
Lit, = 1.50 pm.

(I

one may remove it by changing the dispersion characteristics

of the two guides, by setting a # 1 in (2.3) and (2.4).
When this is done, the plot of Fig. 8 results. The spurious
response has been removed at the cost of a design complication.
This complication may be avoided, if one puts all channels
within the stopband of the grating, within which the spurious
response may be of acceptable magnitude. We shall assume
that to be the case when analyzing the more complicated
structures below that transfer (practically) all the power from
the bus to the channel dropping filter and are capable of
retransmitting the signal.
Fig. 9 shows the response functions of the two-section
structure in Fig. 3 for the case a = 1, K = 200 cm-,
p = 5 cm-, L1, = Lsa = L36 = 250pm, and Llb =
150 pm. One can see clearly the correspondence with the
prediction of the equivalent circuit of Fig. 2 although the
correspondence is not perfect. One also sees the spurious
response outside the grating stopband (Fig. 9(c)). The spu-

UAUS

AND LAI: NARROW-BAND OPTICAL CHANNEL-DROPPING FILTER

61

IV. CONCLUSIONS

0.0
-0.04

-0.02

0.00

0.02

0.04

-0.04

-0.02

0.00

0.02

0.04

u.u

We have shown the possibility of using coupled waveguides

and X/4 shifted DFB resonators to arrive at a narrow-band
channel-dropping filter. The bandwidth of the filter can be
made much narrower than the stopband of the grating. Even
if todays technology is not quite ready to realize a filter of
the ideal characteristics shown, it is only a matter of time that
it will catch up.
It is possible to remove the spurious responses of the
grating filter by appropriate dispersion characteristics for the
coupled waveguides. However, in some practical applications
it may not be necessary to do this, if all channels can be
accommodated within half the grating bandwidth of the filters.
In another paper [7], we have studied side-coupled cascades
of resonator structures. One may be able to use cascade
structures to achieve desirable filter responses.
ACKNOWLEDGMENT

H. A. Haus gratefully acknowledges stimulating discussions

with Dr. K. Utaka of KDD, Japan.
REFERENCES

-2

-1

(c)
Fig. IO. The filter characteristic of the structure in Fig. 6. The parameters are: ( I = 1.0, K = 200 cm-l, p = 5 cm-I, L1, = 300 pm,
I,,,, z L:,,, = 173 pm and L3b = 75 pm, The gain constant in the central
DFR section is 0.38 cm-I.

rious response could be suppressed without affecting the

response at band-center if the group velocity in the two
guides is different. However, for our purposes, we do not
need it because we have already had enough bandwidth to
spend.
The response curves of the structure in Fig. 6, with
the parameters as indicated, are shown in Fig. 10. One
can see that the signal is indeed successfully retransmitted.
However, due to the coupling, the resonances of the X/4shifted DFB resonators in the first and third section are
split. They are responsible for the sharp dips in Fig. 10(b).
The dips do not degrade the performances because they are
off the signal frequency. By keeping the adjacent channels
far apart (say, 50 GHz in our example), the dips will
not interfere with the adjacent channel either. The side
peaks in Fig. l0(a) are caused by the same effect. Since
they are already very small (and off the signal frequency),
one may be able to ignore them. The examples we chose
are intended for illustration, not as examples of optimum
design.

[ l ] For example, see H. Toba et al., A 100-channel optical FDM transmissionidistribution at 622 Mb/s over 50 km, J. Lightwave Technol.,
vol. 8, pp. 1396, 1990.
[2] R. C. Alferness and P. S. Cross, Filter characteristics of codirectionally coupled waveguides with weighted coupling, IEEE J. Quantum
Electron., vol. QE-14, pp. 843, 1978.
[3] R. F. Kazarinov, C. H. Henry, and N. A. Olsson, Narrow-band resonant
optical reflectors and resonant optical transforms for laser stabilization
and wavelength division multiplexing, IEEE J. Quantum Electron.,
vol. QE-23, p. 1419, 1987.
[4] For experimental demonstration, see N.A. Olsson et al., Narrow
linewidth 1.5 p m semiconductor laser with a resonant optical reflector,
Appl. Phys. Lett., vol. 15, p. 1141, 1987.
[5] H. A. Haus and Y. Lai, Narrow-band distributed feedback reflector
design, to be published in J. Lightwave Technol.
[6] H. Hillmer, S. Hansmann, and H. Burkhard, Realization of high coupling
coefficients in 1.53 p m InGaAsP/InP first order quarter-wave shifted
distributed feedback laser, Appl. Phys. Lett., vol. 57, p. 534, 1990.
[7] H.A. Haus and Y. Lai, Theory of cascaded quarter wave shifted
distributed feedback resonators, submitted to IEEE J . Quantum Electron.

Hermann A. Haus (S50-A55-SM58-F62)

was born in Ljubljana, Yugoslavia, in 1925. He
attended the Technische Hochschule, Graz, and
the Technische Hochschule, Vienna, Austria He
received the B.Sc degree from Union College,
Schenectady, NY, in 1949, the M S. degree from
Rensselaer Polytechnic Institute, Troy, NY, in 1951,
and the Sc.D. degree from the Massachusetts
Institute of Technology, Cambridge, in 1954 He
received honorary doctors degrees from Union
College and the Technical University of Vienna.
He joined the Faculty of Electrical Engineering at MIT, Cambridge, in 1954,
where he is an Institute Professor He is engaged in research in electromagnetic
theory and lasers He is the author or coauthor of five books and over 200
journal articles
Dr. Haus is a member of the American Physical Society, the National
Academy of Engineering, the National Academy of Sciences, and a Fellow
of the Optical Society of America, and the American Academy of Arts and
Sciences.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 10, NO. 1, JANUARY 1992

Y. Lai received the B.S. degree in electrical engineering from National Taiwan University, Taiwan,
in 1985 and the M.S. and Ph.D. degrees in electrical engineering from Massachusetts Institute of
Technology (MIT), Cambridge, in 1989 and 1991,
respectively. He joined the Faculty of the Institute of
Electro-Optical Engineering at the National ChiaoTung University, Taiwan, in 1991, where he is an
Associate Professor. His research interests include
nonlinear optical pulse propagation, quantum optics,
short optical pulse generation and applications, and
optoelectronics.