Receiver Design for an Optical MIMO Testbed
Heidi K¨ohnke, Robert Schwinkendorf, Sebastian Daase, Andreas Ahrens and Steffen Lochmann
Hochschule Wismar, University of Technology, Business and Design, Philipp-M¨uller-Straße 14, 23966 Wismar, Germany
Keywords:
Multiple-input Multiple-output System, Optical Fibre Transmission, Multimode Fiber (MMF), Mode Cou-
pling.
Abstract:
Within the last years multiple-input multiple-output (MIMO) systems have attracted a lot of attention in the
optical fibre community. Although the theoretical background of MIMO transmission is well understood, there
is still a need for practical investigations regarding mode combining, mode maintenance and mode splitting.
Since these components represent an essential part of an optical MIMO system, in this work a (2× 2) MIMO
testbed using fusion couplers and a multi-mode fibre (MMF) length of 1.9 km is set up for an operating
wavelength of 1326 nm. Together with the MIMO receiver-side signal processing the successful transmission
of parallel data streams is presented.
1 INTRODUCTION
The growing demand of bandwidth particularly
driven by the developing Internet has been satisfied
so far by optical fibre technologies such as Dense
Wavelength Division Multiplexing (DWDM), Polar-
ization Multiplexing (PM) and multi-level modula-
tion. These technologies have now reached a state
of maturity (Winzer, 2012). The only way to fur-
ther increase the available data rate is now seen in the
area of spatial multiplexing (Richardson et al., 2013),
which is well-established in wireless communications
(Tse and Viswanath, 2005; K¨uhn, 2006). Nowadays
several novel techniques such as Mode Group Diver-
sity Multiplexing (MGDM) (Franz and Blow, 2012)
or Multiple-Input Multiple-Output (MIMO) are in the
focus of interest (Singer et al., 2008).
Among these techniques, optical MIMO has
shown its capability for high-speed data transmission.
However, the practical implementation has to cope
with many technological obstacles such as mode mul-
tiplexing and management. This includes mode com-
bining, mode maintenance and mode splitting.
In order to investigate these effects in a whole
transmission system a MIMO testbed has been set
up. Here, fusion couplers are used for mode combin-
ing and splitting realizing a parallel data transmission
over a 1.9 km multi-mode fibre (MMF) (Ahrens and
Lochmann, 2013; Sandmann et al., 2014a; Sandmann
et al., 2014b). For the necessary implementation of
the MIMO signal processing an off-line MIMO re-
ceiver has been programmed.
Against this background the novelty of this paper
is the practical receiver implementation within a (2×
2) MIMO testbed using fusion couplers. Its proper
mode of operation is shown by the eye diagram.
The remaining part of the paper is structured as
follows: In section 2 the optical MIMO testbed and
its corresponding system model are introduced. The
further processing of the measured data, which is car-
ried out by off-line signal processing, is described in
section 3. Section 4 presents the investigated equal-
izer design. The obtained results are given in section
5. Finally, section 6 shows our concluding remarks.
2 OPTICAL MIMO SYSTEM
MODEL
The principle of optical MIMO is based on the activa-
tion of different modes or mode groups respectively,
as illustrated in Fig. 1. In order to realize a parallel
data transmission different data sources make use of
these mode groups. A possible solution is the excita-
tion of low order mode (LOM) and high order mode
(HOM) groups. These different modes travel together
in a MMF and can be separated by their spatial distri-
bution at the receiver side leading to a (2× 2) MIMO
system in this contribution.
The excitation of the different modes can be done
through various methods. Besides using Spatial Light
Modulators (SLM) (Gu et al., 2013), Long-Period
31
Köhnke H., Schwinkendorf R., Daase S., Ahrens A. and Lochmann S..
Receiver Design for an Optical MIMO Testbed.
DOI: 10.5220/0005019500310036
In Proceedings of the 5th International Conference on Optical Communication Systems (OPTICS-2014), pages 31-36
ISBN: 978-989-758-044-4
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
Data Source
Receiver
HOM LOM
Figure 1: Optical MIMO system by activating LOM and
HOM groups.
Gratings (LPG) (Giles et al., 2012) or Photonic Crys-
tals (Amphawan and Al Samman, 2013), the excita-
tion can simply be carried out by a centric or an eccen-
tric splice between a single-mode fibre (SMF) and a
MMF. Fig. 2 illustrates a typical eccentric splice used
to excite HOM groups. Unfortunately, two SMFs’
can’t easily be placed in front of standard MMF.
That’s why other solutions are in the focus of interest.
Investigations in (Ahrens and Lochmann, 2013) have
shown that fusion couplers are capable of combining
different mode groups into a MMF.
Figure 2: Eccentric single to multi-mode fibre splice (ec-
centricity 20 µm).
For mode separation at the receiver side fusion
couplers similar to the transmitter side are used. The
structure of the corresponding practical testbed is
shown in Fig. 3.
D1
D2
LOM
HOM
CH1
CH2
optical
coupler
optical
coupler
optical
channel
Figure 3: Structure of the optical (2× 2) MIMO testbed.
Two independent data sources (D1, D2) realized
by an Agilent high-speed pattern generator N4903B
produce unipolar signals, which drive the respective
laser diode modules with a pulse frequency of f
T
=
625 Mbit/s. The laser diodes can either work at 1326
nm or 1576 nm. Their light is fed to the centric
and eccentric splice with different power. The laser
diode, which has the higher power is used for activat-
ing the HOM, to compensate for the losses of higher
modes. Thereafter they are combined by the fusion
coupler. After the fibre length of 1.9 km, the trans-
mitted signals are separated by the second fusion cou-
pler followed by two broadband Agilent 81495A re-
ceivers. The obtained signals are sampled by a high-
speed sampling oscilloscope (Agilent DSO90804A)
and stored for further off-line signal processing (CH1,
CH2). Fig. 4 shows the corresponding electrical
MIMO system model.
g
11
(t)
g
12
(t)
g
21
(t)
g
22
(t)
u
s 1
(t)
u
s 2
(t)
u
k 1
(t)
u
k 2
(t)
+
+
n
1
(t)
n
2
(t)
Figure 4: (2× 2) MIMO system model.
The n
T
transmit signals u
sµ
(t) (for µ = 1,.. .,n
T
)
are mapped to the n
R
received signals u
kν
(t) (for
ν = 1,.. .,n
R
) by using the corresponding impulse re-
sponses g
νµ
(t). Additionally, white Gaussian noise
n
ν
(t) (for ν = 1,.. .,n
R
) is added at the receiver side.
Mathematically, the received signals can be described
as
u
kν
(t) =
n
T
µ=1
u
sµ
(t) g
νµ
(t) + n
ν
(t) . (1)
In this paper the number of transmitters and receivers
is limited to n
T
= n
R
= 2. The setup of the practical
(2× 2) MIMO system is shown Fig. 5.
Fig. 6 shows the measured impulse responses of
all underlying channels of this (2× 2) MIMO system
(Ahrens and Lochmann, 2013). From Fig. 6 it can
bee seen by comparing the impulse responses g
11
(t)
and g
22
(t) that the mode groups can be separated ef-
ficiently. Moreover, a typical high cross talk can be
identified from g
12
(t) and g
21
(t).
3 OFF-LINE SIGNAL
PROCESSING
In the following the off-line signal processing is de-
scribed including symbol clock recovery, frame syn-
chronisation, channel estimation and equalisation.
Referring to Fig 4, the MIMO channel can be divided
into four individual SISO channels. Fig. 7 shows such
OPTICS2014-InternationalConferenceonOpticalCommunicationSystems
32
Pattern Generator (D1, D2) E/O Converter LOM/HOM Excitation
Fusion Coupler1.9 km Channel
Fusion Coupler
Oscilloscope O/E Converter
Figure 5: Setup of the practical (2× 2) MIMO system.
0 1 2 3 4
−1
0
1
2
3
4
5
0 1 2 3 4
−0.2
0
0.2
0.4
0.6
0.8
0 1 2 3 4
−2
0
2
4
6
8
0 1 2 3 4
−0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
time (in ns) time (in ns)
time (in ns) time (in ns)
T
s
g
11
(t)
T
s
g
12
(t)
T
s
g
21
(t)
T
s
g
22
(t)
Figure 6: Measured MIMO specific impulse responses ap-
plying a pulse frequency of f
T
= 1/T
s
= 625 Mbit/s (fibre
length of 1.9 km and diode laser wavelength 1326 nm).
a SISO channel with its corresponding off-line pro-
cessing components, which have to be combined in
case of MIMO transmission.
The starting transmitter block includes a rectan-
gular transmit filter g
s
(t). It is followed by the SISO
channel described by its respective impulse response.
The receiver consists of rectangular filter g
ef
(t) and
the corresponding off-line signal processing compo-
nents.
3.1 Symbol Clock Recovery
The method of the symbol clock recovery is high-
lighted in Fig. 8. The received signals u
kν
(t) of both
MIMO-channels (i.e. for ν = 1,2) have to be squared
+
Transmitter
Receiver
Channel
d
ν
(k)
u
qµ
(t)
g
s
(t)
g
ef
(t)
g
νµ
(t)
Equalizer
T
s
n
ν
(t)
u
sµ
(t)
u
kν
(t)
Channel
estim.
Frame
sync.
Figure 7: SISO system model.
in order to obtain a frequency component at the sym-
bol pulse frequency. After bandpass filtering with the
symbol frequency the symbol clock can be found and
the sampled received signal u
ν
(k) is obtained.
downsampling
u
kν
(t)
u
ν
(k)
g
ef
(t)
IFFTFFT
Filter
Max(·)
(·)
2
Figure 8: Symbol clock recovery.
ReceiverDesignforanOpticalMIMOTestbed
33
3.2 Frame Synchronization and
Channel Estimation
The sampled data are the input data for frame syn-
chronization and channel estimation. The data struc-
ture of the transmitted signal is organized as fol-
lows: The payload data are packed into 1024-bit long
frames. The structure of this frame is shown in Tab. 1.
Each frame consists of 784 bit payload (R), a 52 bit
training sequence (T) and 188 zeros (0) as guard inter-
val to compensate the delay time difference. The 52
bit training sequence consists of two 26 bit long GSM
training sequences. Since the paper concentrates on
the prove of the optical MIMO concept, the length of
the given sequences isn’t in the focus of the testbed
optimization.
The balanced 26 bit training sequence, which re-
lates to GSM standard, is used for the synchroniza-
tion and the training of the adaptive equalizer. To find
the 52 bit sequence (T) in the data stream, the cross-
correlation function between the training sequence
and the data stream is calculated. The resulting cross-
correlation peak is now used for the frame synchro-
nisation. Furthermore, it is possible to estimate the
channel coefficients by the cross-correlation function
of the inner 16 bit of the 26 bit orthogonal GSM train-
ing sequence and the corresponding sequence of the
measured data.
4 EQUALISER DESIGN
The equalization of the filtered and sampled receive
signals u
ν
(k) refers to the principle of van Etten (van
Etten, 1975; van Etten, 1976), which is illustrated in
Fig. 9. Using the channel impulse responses h
νµ
(k),
which describe the mapping of the µth input to the νth
output, i. e.
h
νµ
(t) = g
s
(t) g
νµ
(t) g
ef
(t) , (2)
the whole MIMO system can be decomposed into a
number of SISO channels. These SISO channels in-
clude the impact of the channel as well as the transmit
and receive filtering. It is assumed that the h
νµ
(k) of
each SISO channel consist of L + 1 non-zero coeffi-
cients. Furthermore, the noise components w
ν
(k) are
obtained after receive filtering and sampling of n
ν
(t).
In order to remove the MIMO interferences, i.e.
interferences between the different input data streams
as well as intersymbol interferences, an appropriate
equalizer has to be chosen. A possible solution was
introduced in (van Etten, 1975; van Etten, 1976) for
estimating the corresponding MIMO specific equal-
izer coefficients f
νµ
(k) (with ν = 1, 2 and µ = 1, 2).
h
11
(k)
h
12
(k)
h
21
(k)
h
22
(k)
f
11
(k)
f
12
(k)
f
21
(k)
f
22
(k)
w
1
(k)
w
2
(k)
u
1
(k)
u
2
(k)
u
e1
(k)
u
e2
(k)
+
+
+
+
Figure 9: (2× 2) MIMO equalizer.
The equalizer is determined similar to the zero
forcing (ZF) T-spaced equalizer known from base-
band or single-carrier transmission (Bingham, 1988).
In order to describe the whole MIMO system,
(2 × 2) submatrices H
(for = 0,. ..,L) have to be
created taking the impact of the (2 × 2) MIMO chan-
nel at the time into account, i. e. h
νµ
[]. The (2× 2)
submatrices H
result in
H
=
h
11
[] h
12
[]
h
21
[] h
22
[]
. (3)
Taking the L + 1 non-zero coefficients of the im-
pulse responses h
νµ
(k) into account, the interferences
within the MIMO system are described by the channel
convolution matrix H, which is given by
H =
H
0
0 · · · 0
H
1
H
0
· · · 0
.
.
.
.
.
.
.
.
.
.
.
.
H
L
H
L1
· · · H
0
0 H
L
· · · H
1
.
.
.
.
.
.
.
.
.
.
.
.
0 0 · · · H
L
, (4)
where 0 is a (2× 2) zero matrix.
With the knowledge of the channel matrix H a
multidimensional equalizer can be derived similar
to the T-spaced equalizer known from the baseband
transmission. At first a 2-dimensional Nyquist vector
z = (0...0 I 0. ..0) (5)
is defined, where 0 is a (2 × 2) zero matrix and I de-
notes a (2× 2) identity matrix. The equalizer is given
according to
F =
H
T
· H
1
· H
T
· z
T
(6)
The position of the (2× 2) identity matrix I within the
matrix z is a degree of freedom to adapt the equalizer
OPTICS2014-InternationalConferenceonOpticalCommunicationSystems
34
Table 1: Frame structure consisting of training sequences (T), guard intervals (0) and data (R).
D1 (LOM) T T 0 0 0 0 0 0 R 0
D2 (HOM) 0 0 0 0 T T 0 0 R 0
bit 26 26 26 26 26 26 26 26 784 32
1024bit
to the channel conditions. Now the equalizer matrix
F can be rewritten as
F =
e
T
1
e
T
2
(7)
and the decomposition of the µth column e
T
µ
provides
the equalizer impulse response f
νµ
(k), which connect
the µth (µ = 1, 2) receive filter output to the νth equal-
izer output (ν = 1,2). With the proposedZF equalizer,
the MIMO inherent interferences are removed at the
cost of an increased noise power at the detector input.
5 RESULTS
In this contribution, the channel measurements are
carried out within a 1.9 km (2 × 2) MIMO system.
For this, the HOM groups were excited by an eccen-
tricity of 15 µm. Fig. 10 and Fig. 11 illustrate the
combination of each MIMO signal u
kν
(t) by superpo-
sition of the SISO channels u
kνµ
(t).
Taking (1) into account, the MIMO signal at the
output ν (for ν = 1,2) is obtained as
u
kν
(t) = u
kν1
(t) + u
kν2
(t) + n
ν
(t) . (8)
Therein the received signals u
kν1
(t) and u
kν2
(t) de-
scribe the influence of the transmitter 1 and 2, respec-
tively. Furthermore, there is a time delay between
the respective transmitted training sequence and the
crosstalk, caused by a different fiber length between
source and coupler of each input, which has been
compensated by the frame synchronisation module.
Based on the measured receive signals u
kν
(t) the
channel coefficients h
νµ
(k) can be estimated, which
is shown in Fig. 12. With the estimated channel co-
efficients the time-dispersive nature of the underly-
ing transmission channel is visible. After estimating
the channel coefficients h
νµ
(k), the multidimensional
equalizer can be formed in order to remove the inter-
ferences from the MIMO system. The corresponding
eye-diagram is shown in Fig. 13 for the equalized sig-
nal u
e1
(t).
0
0.02
0.04
0.06
0
0.02
0.04
0.06
0 25 50 75 100 125 150 175 200
0
0.02
0.04
0.06
u
k1 1
(t) (inV)
u
k1 2
(t) (inV)
u
k1
(t) (inV)
t/T
s
Figure 10: Measured received signal u
k1
(t) within the (2×
2) MIMO system with respect to the pulse frequency f
T
=
1/T
s
= 625 assuming a fibre length 1.9 km.
0
0.02
0.04
0.06
0
0.02
0.04
0.06
0 25 50 75 100 125 150 175 200
0
0.02
0.04
0.06
u
k2 1
(t) (inV)u
k2 2
(t) (inV)u
k2
(t) (inV)
t/T
s
Figure 11: Measured received signal u
k2
(t) within the (2×
2) MIMO system with respect to the pulse frequency f
T
=
1/T
s
= 625 MHz assuming a fibre length 1.9 km.
6 CONCLUSION
In this contribution a successful receiver design for a
(2× 2) optical MIMO system was presented. The re-
ceiver consisting of the following components: frame
synchronisation, symbol clock recovery, channel esti-
mator and equalizer was applied to a 1.9 km MIMO
testbed. The successful implementation of the MIMO
receiver was demonstrated by the open eye-diagram.
ReceiverDesignforanOpticalMIMOTestbed
35
−3 −2 −1 0 1 2 3
0
0.01
0.02
0.03
0.04
−3 −2 −1 0 1 2 3
0
0.001
0.002
0.003
0.004
−3 −2 −1 0 1 2 3
0
0.01
0.02
0.03
0.04
−3 −2 −1 0 1 2 3
0
0.01
0.02
0.03
0.04
h
11
(k)
h
22
(k)
h
12
(k)
h
21
(k)
k k
k k
Figure 12: Estimated channel coefficients of the measured
(2× 2) MIMO signal assuming a pulse frequency of f
T
=
1/T
s
= 625 MHz and a fibre length 1.9 km .
−0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5
−0.4
−0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
t/T
s
u
e1
(t) (inV)
Figure 13: Eye-diagram of received signal u
e1
(t) after
equalization.
ACKNOWLEDGEMENTS
This work has been funded by the German Ministry
of Education and Research (No. 03FH016PX3).
REFERENCES
Ahrens, A. and Lochmann, S. (2013). Optical Couplers in
Multimode MIMO Transmission Systems: Measure-
ment Results and Performance Analysis. In Interna-
tional Conference on Optical Communication Systems
(OPTICS), pages 398–403, Reykjavik (Iceland).
Amphawan, A. and Al Samman, N. M. A. (2013). Tiering
Effect of Solid-Core Photonic Crystal Fiber on con-
trolled Coupling into Multimode Fiber. In Photonic
Fiber and Crystal Devices: Advances in Materials
and Innovations in Device Applications (SPIE 8847),
pages 88470Y–88470Y–6.
Bingham, J. A. C. (1988). The Theory and Practice of Mo-
dem Design. Wiley, New York.
Franz, B. and Blow, H. (2012). Experimental Evaluation of
Principal Mode Groups as High-Speed Transmission
Channels in Spatial Multiplex Systems. IEEE Pho-
tonics Technology Letters, 24:1363–1365.
Giles, I., Obeysekara, A., C., R., Giles, D., Poletti, F., and
Richardson, D. (2012). All Fiber Components for
Multimode SDM Systems. In IEEE Summer Topical
2012: Space Division Multiplexing for optical Sys-
tems and Networks, pages 212–213.
Gu, R., IP, E., Li, M. J., Huang, Y. K., and Kahn, J. M.
(2013). Experimental Demonstration of a Spatial
Light Modulator Few-Mode Fiber Switch for Space-
Division Multiplexing. In Frontiers in Optics.
K¨uhn, V. (2006). Wireless Communications over MIMO
Channels Applications to CDMA and Multiple An-
tenna Systems. Wiley, Chichester.
Richardson, D. J., Fini, J., and Nelson, L. (2013). Space
Division Multiplexing in Optical Fibres. Nature Pho-
tonics, 7:354–362.
Sandmann, A., Ahrens, A., and Lochmann, S. (2014a).
Equalisation of Measured Optical MIMO Channels.
In International Conference on Optical Communica-
tion Systems (OPTICS), Vienna (Austria).
Sandmann, A., Ahrens, A., and Lochmann, S. (2014b). Ex-
perimental Description of Multimode MIMO Chan-
nels utilizing Optical Couplers. In 15. ITG-
Fachtagung Photonische Netze, Leipzig (Germany).
Singer, A. C., Shanbhag, N. R., and Bae, H.-M. (2008).
Electronic Dispersion Compensation– An Overwiew
of Optical Communications Systems. IEEE Signal
Processing Magazine, 25(6):110–130.
Tse, D. and Viswanath, P. (2005). Fundamentals of Wireless
Communication. Cambridge, New York.
van Etten, W. (1975). An Optimum Linear Receiver
for Multiple Channel Digital Transmission Systems.
IEEE Transactions on Communications, 23(8):828–
834.
van Etten, W. (1976). Maximum Likelihood Receiver for
Multiple Channel Transmission Systems. IEEE Trans-
actions on Communications, 24(2):276–283.
Winzer, P. (2012). Optical Networking beyond WDM.
IEEE Photonics Journal, 4:647–651.
OPTICS2014-InternationalConferenceonOpticalCommunicationSystems
36