of 8
Enabling high spectral efficiency coherent superchannel transmission
with soliton microcombs
Mikael Mazur
1
, Myoung-Gyun Suh
2
, Attila F ̈ul ̈op
1
, Jochen Schr ̈oder
1
, Victor
Torres-Company
1
, Magnus Karlsson
1
, Kerry J. Vahala
2
and Peter A. Andrekson
1
1
Photonics Laboratory, Fibre Optic Communication Research Centre (FORCE), Department of
Microtechnology and Nanoscience, Chalmers University of Technology, Gothenburg SE-412 96, Sweden
2
T. J. Watson Laboratory of Applied Physics, California Institute of Technology, Pasadena, CA 91125, USA
These authors contributed equally to this work
Optical communication systems have come
through five orders of magnitude improvement in
data rate over the last three decades
1
. The in-
creased demand in data traffic and the limited
optoelectronic component bandwidths
2
have led
to state-of-the-art systems employing hundreds
of separate lasers in each transmitter.
Given
the limited optical amplifier bandwidths, focus is
now shifting to maximize the
spectral efficiency
,
SE
1,3
. However, the frequency jitter from neigh-
bouring lasers results in uncertainties of the exact
channel wavelength, requiring large guardbands
to avoid catastrophic channel overlap
4
. Optical
frequency combs with optimal line spacings (typ-
ically around 10-50 GHz) can overcome these lim-
itations and maximize the SE
5
. Recent develop-
ments in microresonator-based soliton frequency
combs (hereafter
microcombs
)
6–11
promise a com-
pact, power efficient multi-wavelength and phase-
locked light source for optical communications
12
.
Here we demonstrate a microcomb-based commu-
nication link achieving state-of-the-art spectral
efficiency that has previously only been possible
with bulk-optics systems. Compared to previous
microcomb works in optical communications
12,13
,
our microcomb features a narrow line spacing of
22.1 GHz. In addition, it provides a four order-of-
magnitude more stable line spacing compared to
free-running lasers
4
. The optical signal-to-noise
ratio (OSNR) is sufficient for information encod-
ing using state-of-the-art high-order modulation
formats. This enables us to demonstrate trans-
mission of a 12 Tb/s superchannel over distances
ranging from a single 82 km span with an SE
exceeding 10 bits/s/Hz, to 2000 km with an SE
higher than 6 bits/s/Hz. These results demon-
strate that microcombs can attain the SE that will
spearhead future optical networks.
Combined
with further advances in hybrid integration
14
,
high-SE microcomb-based transmitters could en-
able novel transmission schemes with lower en-
ergy consumption
5
while continuing the decades
of exponential growth in optical communications.
Early optical communication systems encoded infor-
mation in a binary-like fashion by switching the laser
on and off. The invention of all-optical amplification
using erbium-doped fibre amplifiers (EDFAs) provided
a paradigm shift in the 90s by enabling simultaneous
amplification over about 10 THz bandwidth, the C+L-
bands. Throughput could be massively increased using
wavelength-division multiplexing transmission (WDM).
Today, high-speed digital-to-analogue and analogue-to-
digital converters (DACs/ADCs)
2
combined with digital
signal processing (DSP) have enabled the adaptation of
wireless coherent technologies in optical systems
15
. Clas-
sical on-off keying systems are being replaced with co-
herent solutions using advanced modulation formats
1,15
.
These formats exploit both amplitude and phase together
with polarization multiplexing (PM) to transmit inde-
pendent information. Moreover, DSP has replaced opti-
cal compensation of fibre dispersion
15
.
The throughput demand of single transceivers is
rapidly approaching 10 Tb/s
16
. At the same time the
1
0
1
0
WSS
TX
2
RX
2
TX
3
RX
3
TX
1
to RX
3
>10 Tb/s dense
superchannel
λ
λ
λ
λ
λ
λ
0
WSS
WSS
01011
λ
Site 1
Site 2
Site 3
λ
01011
TX
2
to RX
1
10100
FIG. 1:
Concept of microcomb-based superchan-
nels in future networks.
In an all-optically routed
network, superchannels can be routed arbitrarily across
the entire system. They enable multi-Tb/s throughput
with high spectral efficiency by avoiding excising guard-
bands between channels which are needed to separate
individual channels using wavelength selective switches
(WSS). The sketch shows an example of such a scenario
using a chip-scale microcomb-based superchannel trans-
mitter (Tx). The superchannel is transmitted from site
1 to a receiver (Rx) in site 3 (red) while a second su-
perchannel goes from site 2 to site 1 (blue). High spec-
tral line density microcombs allow the superchannels
to be generated on a chip while maintaining state-of-
the-art data spectral effiencies, thereby maximizing the
throughput of the entire network.
arXiv:1812.11046v1 [physics.app-ph] 22 Dec 2018
2
Silica Disk
Silicon Substrate
a)
d)
1550.80 nm
1554.37 nm
1559.74 nm
16QAM
64QAM
256QAM
1550
1552
1554
1556
1558
1560
10 dB / div
Wavelength [nm]
kHz + 22.10985 GHz
10 dB / div
-250
0
250
1520
1530
1540
1550
1560
1570
1580
159
0
-70
-60
-50
-40
-30
Power [dBm]
Wavelength [nm]
RBW
9 kHz
Superchannel (52 lines)
5 cm
RBW
0.01 nm
RBW 0.01 nm
x
N
Loading
EDFA
Fibre span
Odd
Even
c)
b)
f)
e)
Pump
Superchannel transmitter
Link
Coherent
receiver
sech
2
fitting
(~ 140 fs)
Pump
Microresonator
FIG. 2:
Coherent superchannel transceiver using soliton microcombs.
a. Schematic experimental setup
used to perform the superchannel transmission. A single wavelength continuous-wave (CW) laser was amplified via
an erbium-doped fibre amplifier (EDFA) and coupled into a microresonator to generate a soliton microcomb. The
high-Q silica microresonator consists of a whispering-gallery-mode silica wedge disk resonator (blue) on a silicon
substrate (gray). After suppressing the pump laser line using a fibre Bragg grating, the soliton microcomb was am-
plified, flattened, and demultiplexed. A test-loading-band approach with 5 test channels was used to emulate indi-
vidual modulators for each channel. After modulation, the channels are combined again via a multiplexer. The op-
tical signal was launched into a single span fibre loop or a recirculating loop. Finally, the signal was demultiplexed
and detected with a coherent receiver. b. Picture of the packaged microresonator module. Inset: Zoomed-in view of
22.1 GHz soliton microcomb device. c. Optical spectrum of the soliton microcomb. The squared hyperbolic secant
envelope (dashed red curve) gives a soliton pulse width of 140 fs. d. Zoom-in of the optical spectrum showing the
superchannel (52 comb lines) used in the transmission experiment. e. Electrical spectrum (measured for
>
3 hours
in maxhold mode) showing the long-term stability of the 22.1 GHz repetition rate. f. Constellation diagrams of 16,
64, 256QAM at three different wavelengths after 82 km of superchannel transmission (X-polarization).
electrical bandwidth, determining the maximum symbol
rate and the performance of the transceiver electronics,
is increasing at a much lower pace. Previously, sufficient
excess fibre bandwidth, limited by the EDFA gain, has
been available to use standard WDM transmission with
poor spectral efficiency to scale beyond the limitations of
a single transceiver. As this bandwidth is rapidly filling
up and optical networks are changing from simple point-
to-point to optically routed networks supporting flexible
channel bandwidths, the growth from WDM has stag-
nated. One promising path to enable continued growth
is to introduce superchannels, as conceptually illustrated
in Fig. 1. A superchannel is a high-throughput chan-
nel formed by optical multiplexing of multiple classical
channels. It can therefore reach orders of magnitude
higher bandwidths than the individual channels. Super-
channels are crucial to maximize the efficiency of future
optical networks supporting flexible channel widths and
use all-optical routing
1
. Achieving similar throughput
using classical WDM channels would lead to poor per-
formance in flexible optically routed networks. This fol-
lows from the limitations of optical routers which require
guardbands exceeding 10 GHz between 50 GHz-spaced
channels to avoid substantial performance penalty from
channel-selective filtering using optical wavelength selec-
tive switches (WSSes)
17
. As the exact guard-band re-
quired depends on the number of nodes, the penalty in-
crease with network flexibility. However, since a super-
channel is viewed as a single broadband channel, it is
always routed as a unit and there is no need to have any
routing guardbands between each channel. The mini-
mum guard-band between the channels forming the su-
perchannel is therefore mainly dictated by the uncer-
tainty in absolute laser frequency
5
.
Using independent lasers, the minimum required
guard-band is a few GHz
4,18
. This limitation can lead
to loss in efficiency of about 10% for systems target-
ing symbol rates around 30 Gbaud
18
. The guard-band
penalty is thus the strongest factor limiting system data
rates. High symbol rates help to mitigate this problem,
but at the expense of increasing the penalty associated
with DAC and ADC imperfections
2
. To maximize the
spectral efficiency, the gaps must be minimized which can
only be done using mutually frequency locked optical car-