Interband Short Reach Data Transmission in Ultrawide Bandwidth Hollow Core Fiber

Abstract

We report a Nested Antiresonant Nodeless hollow-core Fiber (NANF) operating in the first antiresonant passband. The fiber has an ultrawide operational bandwidth of 700 nm, spanning the 1240–1940 nm wavelength range that includes the O-, S-, C- and L- telecoms bands. It has a minimum loss of 6.6 dB/km at 1550 nm, a loss ≤ 7 dB/km between 1465–1655 nm and ≤ 10 dB/km between 1297–1860 nm. By splicing together two structurally matched fibers and by adding single mode fiber (SMF) pigtails at both ends we have produced a ~1 km long span. The concatenated and connectorized fiber has an insertion loss of approximately 10 dB all the way from 1300 nm to 1550 nm, and an effectively single mode behavior across the whole spectral range. To test its data transmission performance, we demonstrate 50-Gb/s OOK data transmission across the O- to L-bands without the need for optical amplification, with bit-error-rates (BERs) lower than the 7% forward error correction (FEC) limit. With the help of optical amplification, 100-Gb/s PAM4 transmission with BER lower than the KP4 FEC limit was also achieved in the O/E and C/L bands, with relatively uniform performance for all wavelengths. Our results confirm the excellent modal purity of the fabricated fiber across a broad spectral range, and highlight its potential for wideband, low nonlinearity, low latency data transmission.

Full text

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 38, NO. 1, JANUARY 1, 2020 159
Interband Short Reach Data Transmission in
Ultrawide Bandwidth Hollow Core Fiber
H. Sakr ,Member,OSA,Y.Hong , Member, IEEE,T.D.Bradley , Member, IEEE, Member, OSA,
G. T. Jasion ,Member,OSA,J.R.Hayes , Member, IEEE, Member, OSA,H.Kim , I. A. Davidson,
E. Numkam Fokoua, Y. Chen, Member, IEEE,K.R.H.Bottrill , N. Taengnoi ,N.V.Wheeler, Member, IEEE,
P. Petropoulos , Fellow, OSA,D.J.Richardson , Fellow, IEEE, and F. Poletti , Senior Member, IEEE
(Post-Deadline Paper)
Abstract—We report a Nested Antiresonant Nodeless hollow-
core Fiber (NANF) operating in the first antiresonant passband.
The fiber has an ultrawide operational bandwidth of 700 nm,
spanning the 1240–1940 nm wavelength range that includes the
O-, S-, C- and L- telecoms bands. It has a minimum loss of
6.6 dB/km at 1550 nm, a loss ≤7 dB/km between 1465–1655 nm
and ≤10 dB/km between 1297–1860 nm. By splicing together
two structurally matched fibers and by adding single mode fiber
(SMF) pigtails at both ends we have produced a ∼1kmlongspan.
The concatenated and connectorized fiber has an insertion loss of
approximately 10 dB all the way from 1300 nm to 1550 nm, and an
effectively single mode behavior across the whole spectral range.
To test its data transmission performance, we demonstrate 50-Gb/s
OOK data transmission across the O-to L-bands without the need
for optical amplification, with bit-error-rates(BERs) lower than the
7% forward error correction (FEC) limit. With the help of optical
amplification, 100-Gb/s PAM4 transmission with BER lower than
the KP4 FEC limit was also achieved in the O/E and C/L bands, with
relatively uniform performance for all wavelengths. Our results
confirm the excellent modal purity of the fabricated fiber across a
broad spectral range, and highlight its potential for wideband, low
nonlinearity, low latency data transmission.
Index Terms—Fiber optics communications, hollow core optical
fibers, microstructured optical fibers.
Manuscript received June 24, 2019; revised September 5,2019 and September
16, 2019; accepted September 19, 2019. Date of publication September 23,
2019; date of current version December 30, 2019. This work was supported
in part by the European Research Council project LightPipe (682724), in part
by the Engineering and Physical Sciences Research Council, U.K., under Grant
EP/P030181/1, and in part by the Royal Academy of Engineering. (Correspond-
ing author: Francesco Poletti.)
H. Sakr, Y. Hong, T. D. Bradley,G. T. Jasion, J. R. Hayes, I. A. Davidson, E. N.
Fokoua, Y. Chen, K. R. H. Bottrill, N. Taengnoi, N. V. Wheeler, P. Petropoulos,
D. J. Richardson, and F. Poletti are with the Optoelectronics Research Centre,
University of Southampton, Southampton SO17 1BJ, U.K. (e-mail: h.sakr@
soton.ac.uk; y.hong@soton.ac.uk; t.bradley@soton.ac.uk; g.jasion@soton.ac.
uk; jrh@orc.soton.ac.uk; i.a.k.Davidson@soton.ac.uk; eric.numkam-fokoua@
soton.ac.uk; yc1m12@soton.ac.uk; k.bottrill@soton.ac.uk; nt1a15@soton.ac.
uk; nvw1v10@orc.soton.ac.uk; pp@orc.soton.ac.uk; djr@orc.soton.ac.uk;
frap@orc.soton.ac.uk).
H. Kim is with the Optoelectronics Research Centre, University of Southamp-
ton, Southampton SO17 1BJ, U.K., and also with the Electrical and Electronic
Convergence Department, Hongik University, Sejong 30016, South Korea (e-
mail: h.kim@soton.ac.uk).
Color versions of one or more of the figures in this article are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JLT.2019.2943178
I. INTRODUCTION
HOLLOW core fibers (HCF) offer lower latency and nonlin-
earity than conventional fibers [1]. Major advances have
been made in HCF technology in recent years, to the point that
state-of-the-art (SoTA) fibers now show performance levels that
are compatible with a variety of short reach data transmission
applications. The first generation of HCFs (produced ∼15 years
ago) were photonic bandgap fibers (PBGFs), in which the termi-
nation of the periodic cladding to create the core-guiding defect
resulted in the formation of surface modes. This narrowed the
fibers’ usable bandwidth to 10s of nanometers, and although the
loss in these fibers was very low (∼1–2 dB/km) [2], reliable
data transmission was impractical [3]. A second generation of
PBGFs, free from surface modes (in the operating window),
offered bandwidths of ∼100 nm, losses of a few dB/km [4],
and were shown to be capable of supporting high capacity data
transmission [5].
More recently, the air-guiding bandwidth was increased to
several 100 s of nanometers through the introduction of nodeless
hollow core antiresonant fibers (ARFs). These offered octave-
spanning bandwidth, albeit at the expense of a higher loss [6],
[7]. A 7 capillary ‘tubular’ ARF with a ‘3 dB’ bandwidth
(i.e., bandwidth at which the dB/km loss doubles from its
minimum value) of 700 nm (950–1650 nm) and a minimum
loss of 25 dB/km was reported. Penalty-free data transmission
at 1, 1.5 and 2 µm was demonstrated in a single fiber, although
only at 10 Gbit/s and over just 100 m [8]. An even lower loss was
achieved in the same fiber type by optimizing the dimensions of
the tubes [7], however this work did not include fibers designed
to operate at telecoms wavelengths.
The next major evolutionary step was to add additional az-
imuthally oriented membranes to improve leakage loss. This
can be done by either adding thin glass membranes inside each
capillary [9] or, as demonstrated in this work, by adding ‘nested’
capillaries inside each antiresonant tube. Such nested antireso-
nant nodeless fibers (NANFs) promise–in theory–optical losses
lower than those of standard silica fibers [10]. The first re-
ported low-loss NANF was designed to operate in the second
antiresonant passband (which we will refer to as the second
passband) to reduce the fabrication challenges. This allowed the
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/

160 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 38, NO. 1, JANUARY 1, 2020
Fig. 1. (a) SEM images of the fabricated NANF for this work and details of
the microstructured cladding including core OD, average outer tubes diameter
and thickness. The inter-tube azimuthal gaps are between 6.1 µmto7.7µm;
(b) measured mode field image when light was launched into the fiber with a
broadband supercontinuum source.
use of relatively thick (∼1µm) tube membranes, which can be
produced using only moderate preform inflation during the fiber
draw and hence reasonably wide process tolerances. Using this
approach, a single mode, second passband NANF with a record
minimum loss for any data transmitting HCF of 1.3 dB/km was
reported [11]. Its ‘3 dB’ bandwidth of 128 nm was sufficient to
accommodate C-and L-band transmission, but not wide enough
to cover other telecoms windows/bands.
In this work, we report a NANF operating in the first antires-
onant passband (which we will refer to as first passband; Fig. 1),
which requires thinner struts than second passband NANFs and
is consequently more difficult to fabricate. Its minimum loss
of 6.6 dB/km at 1550 nm is the lowest reported amongst any
ARF operating in the first antiresonant passband. Structural
asymmetries and some relatively large inter tube gaps mean that
the loss is higher than that of ref. [11]. However, the fiber offers
a 3dB bandwidth of 700 nm (1240–1940 nm), well in excess
of the O+E+S+C+L-band range. To demonstrate its potential
for data transmission across different telecoms windows and
over km distance scales, we spliced this fiber to a similar struc-
turally matched fiber, and connectorized the whole assembly.
The excellent modal properties of these HCFs allowed us to
achieve penalty-free 50 Gbit/s OOK and 100 Gbit/s PAM4 data
transmission over ∼1 km of fiber in the 1291 to 1631 nm spectral
range. It is believed that NANFs offer a very promising solution
for future high-speed, low-latency intra/inter datacenter, 5G
and short-reach optical communications especially for mobile
fronthaul applications.
II. FABRICATION AND CHARACTERIZATION
The fiber was produced using a stack, fuse and draw process,
and is composed of 6 nested tubes defining a hollow core with
a diameter of 35 ±0.5 µm as shown in Fig. 1. Initially, 6
nested capillary elements were stacked and fused inside a jacket
tube. The assembly was then drawn into canes, which were
then jacketed and drawn down to fiber dimensions (in a second
draw stage). Control of the fiber microstructure, including the
core, inner and outer tube expansion ratio, was achieved via a
specialized pressurization system and was supported by in-line
Fig. 2. (a) Measured loss spectrum of the reported NANF; (b) loss of the
fabricated NANF (red curve) compared to previous generations of ultrawide
bandwidth antiresonant fibers operating in the 2nd and 3rd telecoms window
(purple: ’hexagram ARF with nodes [14], and green: nodeless tubular ARF [8]);
a SoTA wide bandwidth data-transmitting PBGF is also shown for comparison
(orange, dashed [4]).
fluid dynamics modelling [12]. The outer tubes of the NANF
elements shown in Fig. 1 have outer diameters (OD) ranging
from 22.2 µm to 25.2 µm and strut thicknesses (t) ranging from
405 nm to 462 nm (average 430 ±30 nm). The nested inner tubes
have ODs ranging from 8.9 µmto10µm and strut thicknesses
ranging from 457 nm to 514 nm (average 480 ±30 nm).
The inter-tube azimuthal gaps are between 6.1 µmto7.7µm,
and these values are maintained throughout the full fiber
length.
The optical loss of this fiber was measured over the full
fiber bandwidth using the cutback method. This was performed
using a tungsten lamp white light source (WLS) and two op-
tical spectrum analyzers (a Yokogawa AQ-6315A covering the
400–1750 nm wavelength range, and a Yokogawa AQ-6375A
covering the 1200–2400 nm wavelength range). A 2-nm resolu-
tion was used over the full fiber bandwidth. A 351 m length of
the fiber was spooled on a 1-m circumference bobbin and this
was cut back to 10 m. Three separate spectral transmission traces
from three different output cleaves were recorded for both the
long and the short lengths without perturbing the launch condi-
tions, and these were averaged to ensure reliable measurements.
Fig. 2(a) presents the measured cutback loss of the fabricated
NANF, showing a minimum loss of 6.6 dB/km at 1550 nm, a loss
≤7 dB/km from 1465 nm to 1655 nm and ≤10 dB/km between
1297–1860 nm [13]. To the best of our knowledge, this is the low-
est loss ever reported for an ultrawide bandwidth HCF operating
in the first passband. The measured loss is compared in Fig. 2(b)
to that of previous generations of ultrawide bandwidth HCFs
operating around telecoms wavelengths: a hexagram ARF fiber

SAKR et al.: INTERBAND SHORT REACH DATA TRANSMISSION IN ULTRAWIDE BANDWIDTH HOLLOW CORE FIBER 161
Fig. 3. (a) Measured (solid red) vs. simulated loss of the NANF reported
herein; (b) simulated fundamental mode and its radial Poynting vector at
1550 nm, showing that a considerable fraction of the loss arises from the gaps
created by one smaller than the average tube.
with nodes (∼150 dB/km) [14] and the tubular fiber mentioned
in the introduction (∼25 dB/km) [8]. For reference, the loss of a
SoTA wide bandwidth PBGF (3.5 dB/km) [4] for which several
C-band DWDM transmission experiments have been reported is
also shown. Comparing these previous wide bandwidth HCFs,
the tradeoff between loss and bandwidth/flatness can be readily
appreciated.
Despite its already good optical properties, some structural
non-uniformities, such as non-ideal inter-tube gaps or slightly
misaligned inner capillaries are present in the fabricated fiber.
These non-uniformities have a strong influence on the total
measured loss and to a lesser extent the bandwidth of operation.
To further understand where the loss arises in the fabricated fiber
and to analyze the effect of the small structural non-uniformities
that can be seen in Fig. 1 on its optical behavior, we extracted its
cross-section from an SEM image and simulated its performance
using the modelling tools described in ref. [11]. The theoretically
calculated loss includes: (a) surface scattering contributions (cf.
ref. [10]); (b) confinement loss using FEM calculations; and
(c) microbend loss contributions (cf. ref. [15]). As can be seen
in Fig. 3(a), overall, the total simulated loss (black line) is in
good agreement with the measurement (red line). It is also
clear that leakage loss dominates at longer wavelengths i.e.,
>1700 nm, whilst the main contribution at shorter wavelengths
1100–1450 nm can be attributed to microbending. Generally, for
the structure reported in this work, the contribution of surface
scattering loss (SSL) is very small and according to simulations
is <0.07 dB/km at 1550 nm. This is due to the fact that all
membranes in NANF operate in antiresonance and are therefore
effective at keeping the density electromagnetic field low at the
glass-air interfaces. The most dominant contributions to the total
loss are leakage loss (dotted blue) and microbending loss (dotted
green). On the other hand, from Fig. 3(a), the peak around 2 µm
in the simulated curve could be attributed to structural resonance,
i.e., anticrossings with modes in the glass tubes [10]. Given
the high aspect ratio between length and thickness of the glass
tubes, the overlap between air and glass modes are small but
still sufficient to create the oscillations observed in the loss plot
[16]. For the measured curve, a small but noticeable peak at
∼1.95 µmfollowedbyadipat2µm could be observed, which
could be attributed to the same phenomena. Additionally and/or
Fig. 4. Measured Loss of the NANF reported in ref. [11] (A; solid blue) vs.
current fiber (B; cf. Fig. 1; solid red). Also shown is the simulated optical loss
of an ideal geometry, allowing for regular inter-tube gaps with the same average
size as in the fabricated fiber (C; gaps of 6.5 µm; dashed purple), and with small
inter-tube gap size (D; gaps of 2 µm; dotted grey).
alternatively, an roto-vibrational overtone from CO2is also
present at ∼1.96 µm, which might give rise to some absorption.
Fig. 3(b) shows the simulated longitudinal Poynting vector of
the fundamental mode of the fiber at 1550 nm wavelength, as
well as the intensity of its radial component at the outer boundary
of the fiber (red line around the microstructure). It can be seen
that considerable leakage (a total of ∼1.5 dB/km, or ∼1/3 of
the total leakage loss at 1550 nm) occurs through two gaps at
either side of one of the tubes, which has a diameter that is
∼2µm smaller than the others. This smaller than average tube is
likely to originate from minor imperfections in the starting cane
used to draw the final fiber and can certainly be addressed as
the fabrication technology evolves. In this case we attribute
the cause of the smaller tube to a slightly misplaced in-
ner capillary (see bottommost tube in Fig. 1(b)) which has
led to local thickening of the glass layer which in turn has
led to reduced expansion under pressure compared to the
other tubes.
Fig. 4 shows a comparison between the loss of the NANF
reported here (see Fig. 1) designed to operate in the first pass-
band, and that reported in ref. [11] with a minimum loss of
1.3 dB/km and fabricated to operate in the second passband.
Although operating in the first passband offers the advantage
of greater operational bandwidth, more than double that of a
fiber operating in the second passband, the control over the
fiber’s structure during fabrication is much more challenging.
Simulations indicate that reducing the azimuthal inter-tube gap
is crucial to reduce the leakage loss contribution in NANFs [11].
However, achieving small gaps has proven to be experimentally
difficult for fibers fabricated to operate in the first passband.
This can be attributed to the thinner tube membranes (which are
almost half the thickness of those in second passband fibers) and
that makes them considerably more sensitive to subtle changes
in the applied pressure and to the initial size of the tubes within
the preform or cane. As a result of this, the azimuthal inter-tube
gaps in the fabricated fiber were of the order of ∼6.5 µmon
average, two to three times as large as those already achieved

162 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 38, NO. 1, JANUARY 1, 2020
Fig. 5. S2measurement of the reported NANF showing that no appreciable
high order mode content remains after 350 m.
in the 1.3 dB/km NANF in ref. [11], or in the 2.5 dB/km
NANF that operates at 1100 nm in ref. [17]. To reach better
understanding of the effect of the structural imperfections in the
fabricated fiber, we also simulated the optical loss of an ideal
NANF with no structural asymmetries, allowing for regular gaps
with the same average size as in the fabricated fiber (6.5 µm),
and with the same average size (2 µm) as the second passband
fiber in ref. [11]. Drawing the same fiber without asymmetries
would lead to the dashed loss curve in Fig. 4, with a minimum
loss of 3.5 dB/km. Additionally, reducing the gap size to only
2µm, which we believe should be ultimately achievable with
further improvements in fabrication control, would produce a
sub 1-dB/km fiber (dotted curve in Fig. 4).
It is worth mentioning that although producing NANFs in
the second passband could be relatively easier than in the first
passband in terms of fabrication precision and control over the
structure, it is believed that achieving similar structures with
low loss and small inter-tube gaps to that in ref. [11] could be
achieved in the very near future in NANFs produced in the first
passband. In general, in order to improve on the uniformity of
the fiber structure which results in a more controlled azimuthal
inter-tube gaps it is important to control the uniformity of the
structure in the canes to start with. This could be improved
by the initial design, stacking and manufacturing of the canes.
Additionally in order to improve on the azimuthal inter-tube
gaps in the fibers, better control and precision over the applied
pressure in the fibre draw is necessary. None of these issues
is fundamental in origin, and further progress is expected in
the very near future. Additionally, drawing NANFs in the first
passband offers multiple advantages such as a flatter ultrawide
transmission bandwidth that could be >3 times that of a second
passband fiber; and an almost doubling of the yield.
Furthermore, in HC-ARFs the higher order modes (HOMs)
are always fundamentally lossier than the fundamental mode,
LP01; this is further enhanced when utilizing the NANF struc-
ture, due to its intrinsic mode-stripping geometry that couples
core guided high order modes to modes in the cladding tubes
[10]. As a result of this, the fabricated NANF was found to be
effectively single mode after a length of 350 m, as confirmed
by an S2cutback measurement. As can be seen in Fig. 5, an
LP11 mode contribution could still be observed after only 10 m
Fig. 6. (a) Schematic of the fully spliced 960-m span comprising of two
structurally matched lengths of NANF; (b) SEM images and photograph of the
spliced fiber; (c) OTDR trace of the span measured at λ=1550 nm highlighting
reflection from the spliced joint; (d) insertion loss of the spliced span which
shows 10.15 ±0.15 dB at 1300 nm and 10.26 ±0.15 dB at 1550 nm.
of propagation, with an intensity of 17 dB below that of the
fundamental mode. As a result of the large differential loss, the
LP11 mode is heavily suppressed relative to the LP01 mode
and hence disappears in the noise floor after 350 m, with a
relative intensity of ∼−75 dB. Moreover, simulations of the
ideal geometries shown in Fig. 4(c) and 4(d) and predict that
the relative HOM extinction could reach 300x for fabricated
inter-tube gaps of 6.5 µm, and up to 1000x for smaller gaps of
2µm.
III. WIDE-BAND DATA TRANSMISSION EXPERIMENT
A. ∼1 km Band Span Preparation
Following the production and characterization of the 351 m
long NANF reported above (341 m after the cutback mea-
surement; NANF A in Fig. 6), a second, structurally matched
and 620 m long fiber (NANF B in Fig. 6), was fabricated.
Both NANFs were spliced together to form a single ∼1km
span for data transmission tests, as shown in Fig. 6(a). The
two NANFs were spliced together using a commercial splicer,
Fujikura FSM 100P. The splicer parameter were controlled to
optimize the splice in terms of transmission and to try to preserve
the microstructure. The fibers were cladding-aligned. The rota-
tional (theta axis) mismatch did not give any noticeable change
on splicing loss. Weak arc power was used to give minimum
damage to the microstructure. Moreover, the ∼1 km span was
fully-connectorized with relatively low splicing losses. Initially,
both HCFs were spliced to each other with a measured splice
loss of ∼0.1 dB (NANF A to NANF B;SEM images are shown
in Fig. 6(b)). This band was then spliced at both ends to SMF
using a combination of intermediate mode field diameter (MFD)

SAKR et al.: INTERBAND SHORT REACH DATA TRANSMISSION IN ULTRAWIDE BANDWIDTH HOLLOW CORE FIBER 163
Fig. 7. Setup of the transmission experiment using the NANF and the offline
DSP blocks.
matching solid fibers including mode field adaptors (MFA) and
single large mode area fibers (LMA). From 1300 to 1600 nm
the mode field diameter of the hollow core fiber does not change
significantly; it does however vary quite substantially in the
MFD matching fibers. We therefore decided to exploit this fact
to equalise the spectral loss of the overall ∼1 km band. We did so
by optimizing the MFD of the bridging fiber so that it produced
a minimum loss at a wavelength of 1300 nm where the fiber loss
was higher. This produced a higher splicing loss at 1550 nm,
which increases the overall insertion loss at these wavelengths
and explains the flatter overall spectral loss profile shown in
Fig. 6(d) as compared to that of NANF A (in Fig. 2(a)). Further,
to validate the full length of the spliced fiber, and to check for
longitudinal defects, the fiber span was examined using OTDR
at λ=1550 nm which also detected the spliced point at the
anticipated relevant length, and is presented in Fig. 6(c).
B. Setup of the Transmission Experiment
To evaluate the wideband data-transmission capability of the
spliced fiber, we conducted high-speed Nyquist on-off keying
(OOK) and 4-ary pulse-amplitude modulation (PAM4) data
transmission experiments in different telecom bands (O- to
L-band). The setup of the transmission experiment and the
corresponding digital signal processing (DSP) diagram is shown
in Fig. 7. At the transmitter end, two tunable lasers, covering
the 1291–1372 nm (O+E band) and 1461–1631 nm (S+C+L
band) wavelength ranges, were used as a continuous-wave (CW)
source to be modulated using a Mach-Zehnder LiNbO3modula-
tor (MZM). The offline-generated Nyquist-OOK/PAM4 signals
were fed into an arbitrary waveform generator (AWG, Keysight
M8196A) for digital-to-analogue conversion at a fixed sampling
rate of 90 GSa/s. The output of the AWG was first amplified by
an electrical amplifier and then used to modulate the MZM.
As shown in Fig. 7, two different cases were considered in the
experiments: in Case 1, the MZM was modulated by 50-Gb/s
OOK signals and its output was directly launched into the NANF
without any optical amplification. The output of the tunable laser
was fixed at 7 dBm for wavelengths shorter than 1600 nm whilst
a 2-dB increment was applied to longer wavelengths mainly
to compensate for the increased insertion loss of the MZM.
At the receiver, after the photodiode (PD, Finisar XPRV2022A),
the detected signals (105symbols) were recorded by a digital
storage oscilloscope (DSO) with a sampling rate of 80 GSa/s.
In Case 2, either a bismuth-or an erbium-doped fiber amplifier
was used to amplify the MZM’s output for the O/E and C/L
bands, respectively, before launching the signal into the NANF.
100-Gb/s PAM4 signals were adopted in this case for data
transmission, since a higher optical signal-to-noise ratio was
available. After the NANF, an optical attenuator was used to
attenuate the optical power incident to the PD. In both cases, as
shown in Fig. 7, no polarization control was adopted at the input
of the NANF, as the NANF has a low polarization dependent loss
(PDL) [18]. It is worth mentioning that similar to [19], the low
PDL ensures the feasibility of polarization division multiplexed
coherent transmission using the fabricated NANF. However,
commercially available coherent receivers will currently limit
the transmission in the NANF to around the C-band despite the
fact that other bands might also be usable for transmission.
In the offline DSP, the mapped OOK/PAM4 symbols were
first up-sampled and then filtered by a square-root raised co-
sine (SRRC) filter with a roll-off factor of 0.1. Subsequently,
down-sampling was applied to generate either 50-Gb/s OOK
or 100-Gb/s PAM4 signals. At the receiver end, the captured
signals were synchronized and re-sampled before being filtered
by a matching SRRC filter. After that, a half-symbol-spaced
(17, 7)-tap decision-feedback equalizer (DFE) with the recursive
least squares (RLS) algorithm was applied for signal recovery.
The forgetting factor of the RLS algorithm was 0.99. Finally,
the reconstructed OOK/PAM4 signals were de-mapped for bit-
error-rate (BER) calculation via error counting.
C. Transmission Results
Fig. 8 shows the results of the 50-Gb/s OOK transmission
(Case 1) across the O- to L-band. The received optical powers
(ROPs) of the tested wavelengths were around −15 dBm. The
resulting signal-to-noise ratios (SNRs) were around 11.5 dB.
Thanks to the relatively uniform ROP/SNR performance, the
corresponding BERs in all bands were lower than the 7% forward
error correction (FEC) limit, i.e., 3.8 ×10−3, therefore no
optical amplification was required. Note that the performance
degradations at wavelengths shorter than 1300 nm and around
1600 nm are principally attributed to the higher insertion loss of
the MZM at these wavelengths. Since an E-band tunable laser
was not available, only two wavelengths in the E-band, i.e.,
1361 nm and 1371 nm, were investigated in both cases. For refer-
ence, the remaining untested wavelength range is highlighted in
Fig. 8. Nevertheless, comparable BER/ROP/SNR performance
can be expected since the fiber attenuation is similar to that
of the neighboring bands, as shown in the characterization in
Section III.A. For reference, the eye diagrams of the signals
transmitted at the wavelengths of 1331 nm and 1551 nm are also
provided in Fig. 8. The corresponding BERs are 2.33 ×10−4
and 2.0 ×10−4, respectively.
The results of Case 2 (100-Gb/s PAM4 transmission) are
shown in Fig. 9. These demonstrate that comparable ROP/SNR

164 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 38, NO. 1, JANUARY 1, 2020
Fig. 8. (a) BER and ROP versus wavelengths for 50-Gb/s OOK transmission
and (b) the corresponding SNR versus wavelengths. Insets: eye diagrams of the
1311-nm and 1551-nm wavelengths.
Fig. 9. (a) BER and ROP versus wavelengths for 100-Gb/s PAM4 transmission
and (b) the corresponding SNR versus wavelengths. Insets: eye diagrams of the
1311-nm and 1551-nm wavelengths.
performance can be achieved across the investigated wave-
lengths in the O, E, C and L bands. The ROP and SNR levels
of the tested wavelengths were around −3 dBm and 18.5 dB,
respectively, and the corresponding BERs of all wavelengths are
below the KP4 FEC limit (3 ×10−4) [20]. We note that since
lower BERs were obtained for the 100-Gb/s PAM4 transmission,
the KP4 FEC limit is assumed in Fig. 9, as the KP4-FEC requires
shorter overhead (approximately 3%) and its induced latency is
also much lower [21].
For reference, the eye diagrams of the recovered PAM4 signals
at 1311 nm and 1551 nm are presented in Fig. 9 as well, and
Fig. 10. Performance comparison with back-to-back for the 1550-nm channel
with 50-Gb/s OOK: (a) BER and (b) SNR.
their BERs are 1.6 ×10−4and 1.2 ×10−4, respectively. It is
also worth noting that the tunable lasers and optical amplifiers
in Case 2 were not fully driven, that is, their output powers
were kept relatively low. Considering the ultra-low nonlinearity
of the hollow-core fibers [19], much higher launch powers
could be tolerated for data transmission. Therefore, much higher
capacities per wavelength and/or much longer transmission
distances should be easily achievable. Furthermore, since the
optical beams within the NANF are air-guided, the chromatic
dispersion (CD) effect of the bands of interest in the NANF is
negligible. It is also worth noting that when further extending
the reach to beyond just a few kilometres, it is anticipated
that the NANF will not suffer from the severe power fading
thanks to its low CD. In contrast, for a standard single-mode
fibre, the CD-induced power fading will limit the performance
of the OOK/PAM4 transmission. Therefore, a significant im-
provement in the transmission performance is expected by using
the NANF technology. This point will be further investigated in
a following study.
We further compared the performance between the scenarios
of NANF transmission and back-to-back transmission (wherein
an optical attenuator with an equivalent loss was adopted). The
comparisons of the 50-Gb/s OOK and the 100-Gb/s PAM4 cases
at 1550 nm are shown in Fig. 10 and Fig. 11, respectively.
For both cases, nearly identical BER and SNR performance
to that of the back-to-back scenario was achieved when using
the NANF, indicating that no additional penalty arose from
the transmission in the NANF. This results from the effective
single-mode guiding properties of the NANF which ensures
that any penalty induced by coupling to higher-order modes is
negligible.
It is worth noting that although most of the recently re-
ported HCF transmission experiments including this work are
demonstrated in short-reach optical interconnects, long-reach
transmission using HCF can be expected upon the realisation
of long-length and low-loss HCFs. With further refined fab-
rication process, longer yields of HCFs can be realistically
fabricated. With regard to propagation loss, the loss of the
NANF can be significantly reduced to sub-dB/km level over

SAKR et al.: INTERBAND SHORT REACH DATA TRANSMISSION IN ULTRAWIDE BANDWIDTH HOLLOW CORE FIBER 165
Fig. 11. Performance comparison with back-to-back for the 1550-nm channel
with 100-Gb/s PAM4: (a) BER and (b) SNR.
an ultrawide spectrum range when using the optimised fibre
design and refined fabrication control, as discussed in Section II.
Furthermore, in principle, NANFs can achieve lower losses than
standard silica fibers [10]. On the other hand, another associated
issue is the efficient optical amplification when longer reach is
targeted. Optical amplification solutions operating in the O-, C-,
and L-band are currently commercially available. In addition,
rare-earth-doped fibre amplifiers, such as the Bismuth-doped
fibre amplifier (BDFA) [22], [23] and the Thulium-doped fi-
bre amplifier (TDFA) [24], [25], have also demonstrated their
effectiveness of high-gain optical amplification over O-, E-,
and S-bands. These amplification techniques can be combined
together with the NANF to realise ultra-wideband long-distance
transmission.
IV. CONCLUSION
We have reported the first ultra-wide bandwidth, low loss
NANF operating in the first antiresonant passband. The fiber
has a very flat 700 nm 3-dB bandwidth that includes the O-,
S-, C- and L-bands, and the lowest loss (6.6 dB/km) ever
reported in an ultra-wide bandwidth ARF operating in the first
passband with relatively thin membrane struts. The excellent
modal purity of the fiber allows penalty-free transmission of
100-Gbit/s PAM4 signals from the O to L-band over ∼1kmof
fiber. Although already potentially interesting for some short-
reach low-latency applications, the current loss level, limited by
asymmetries and the large inter-tube gaps, can be significantly
reduced in future iterations. The predicted sub-dB/km ultra-
wide bandwidth, which appears realistically achievable, could
make NANFs a very promising solution for future high-speed,
low-latency intra/inter datacenter, 5G and short-reach optical
communications.
ACKNOWLEDGMENT
The authors would like to thank Simon Bawn, Lumenisity
Ltd., for help with splicing. All data supporting this study are
openly available from the University of Southampton repository
at: https://doi.org/10.5258/SOTON/D1083.
REFERENCES
[1] F. Poletti et al., “Towards high-capacity fibre-optic communications at the
speed of light in vacuum,” Nature Photon., vol. 7, no. 4, pp. 279–284,
2013.
[2] B. J. Mangan et al., “Low loss (1.7 dB/km ) hollow core photonic bandgap
fiber,” in Proc. Opt. Fiber Commun. Conf., 2004, Paper PDP5A.2.F.
[3] C. Peucheret, B. Zsigri, T. P. Hansen, and P. Jeppesen, “10 Gbit/s trans-
mission over air-guiding photonic bandgap fibre at 1550 nm,” Electron.
Lett., vol. 41, no. 1, pp. 27–29, 2005.
[4] N. V. Wheeler et al., “Wide-bandwidth, low-loss, 19-cell hollow core
photonic band gap fiber and its potential for low latency data transmission,”
in Proc. OFC/NFOEC, 2012, Paper PDP5A.2.F.
[5] V. A. J. M. Sleiffer et al., “High capacity mode-divisionmultiplexed optical
transmission in a novel 37-cell hollow-core photonic bandgap fiber,” J.
Lightw. Technol., vol. 32, no. 4, pp. 854–863, Feb. 2014.
[6] A.N.Kolyadin,A.F.Kosolapov,A.D.Pryamikov,A.S.Biriukov,V.G.
Plotnichenko, and E. M. Dianov, “Light transmission in negative curvature
hollow core fiber in extremely high material loss region,” Opt. Express,
vol. 21, no. 8, pp. 9514–9519, 2013.
[7] B. Debord et al., “Ultralow transmission loss in inhibited-coupling guiding
hollow fibers,” Optica, vol. 4, no. 2, pp. 209–217, 2017.
[8] J.R.Hayeset al., “Antiresonant hollow core fiber with octave spanning
bandwidth for short haul data communications,” in Proc. Opt. Fiber
Commun. Conf. Exhib., 2016, Paper PDP TH5A.3.
[9] S. Gao et al., “Hollow-core conjoined-tube negative-curvature fibre with
ultralow loss,” Nature Commun., vol. 9, no. 1, 2018, Art. no. 2828.
[10] F. Poletti, “Nested antiresonant nodeless hollow core fiber,” Opt. Express,
vol. 22, no. 20, pp. 23807–23828, 2014.
[11] T. D. Bradley et al., “Record low-loss 1.3dB/km data transmitting an-
tiresonant hollow core fibre,” in Proc. Eur. Conf. Opt. Commun., 2018,
pp. 1–3.
[12] G. T. Jasion et al., “Virtual draw of tubular hollow-core fibers,” in Proc.
Frontiers Optics /Laser Sci., 2018, Paper FW6B.3.
[13] H. Sakr et al., “Ultrawide bandwidth hollow core fiber for interband short
reach data transmission,” in Proc. Opt. Fiber Commun. Conf., 2019, Paper
PDP Th4A.1.
[14] J. R. Hayes, F. Poletti, M. S. Abokhamis, N. V. Wheeler, N. K. Baddela,
and D. J. Richardson, “Anti-resonant hexagram hollow core fibers,” Opt.
Express, vol. 23, no. 2, pp. 1289–1299, 2015.
[15] E. N. Fokoua, Y. Chen, D. J. Richardson, and F. Poletti, “Microbending
effects in hollow-core photonic bandgap fibers,” in Proc. 42nd Eur. Conf.
Opt. Commun., 2016, pp. 327–329.
[16] L. Vincetti and V. Setti, “Waveguiding mechanism in tube lattice fibers,”
Opt. Express, vol. 18, no. 22, pp. 23133–23146, 2010.
[17] H. Sakr et al., “Record low loss hollow core fiber for the 1 µm region,” in
Proc. Conf. Lasers Electro-Optics /Europe Eur. Quantum Electron. Conf.,
2019, Paper CE_5_5.
[18] Y. Hong et al., “Beyond 100-Gb/s/λdirect-detection transmission over the
S+C+L-bands in an ultra-wide bandwidth hollow core fibre,” in Proc.
ECOC, 2019, Paper Th.2.E.5.
[19] Z. Liu et al., “Nonlinearity-free coherent transmission in hollow-core
antiresonant fiber,” J. Lightw. Technol., vol. 37, no. 3, pp. 909–916, Feb.
2019.
[20] T. K. Chan and W. I. Way, “112 Gb/s PAM4 transmission over 40 km
SSMF using 1.3 µm gain-clamped semiconductor optical amplifier,” in
Proc. Opt. Fiber Commun. Conf. Exhib., 2015, Paper Th3A.4.
[21] Z. Wang and A. Ghiasi, “400GE lane configurations v.s. FEC options,” in
Proc. IEEE 802.3 Plenary Meeting, Jul. 2013. [Online]. Available: http:
//www.ieee802.org/3/400GSG/public/13_07/wang_400_01_0713.pdf
[22] E. M. Dianov, “Bismuth-doped optical fibers: A challenging active
medium for near-IR lasers and optical amplifiers,” Light Sci. Appl.,vol.1,
2012, pp. 1–7, Art. no. e12.
[23] N. K. Thipparapu, Y. Wang, A. A. Umnikov, P. Barua, D. J. Richardson,
and J. K. Sahu, “40 dB gain all fiber bismuth-doped amplifier operating in
the O-band,” Opt. Lett., vol. 44, no. 9, pp. 2248–2251, 2019.
[24] B. N. Samson et al., “Thulium-doped silicate fiber amplifier at 1460-
1520 nm,” in Proc. Opt. Amplifiers Appli., 2001, pp. 247–248.
[25] S. Aozasa and M. Yamada, “E-band thulium doped fibre amplifier,”
Electron. Lett., vol. 55, no. 6, pp. 333–334, 2019.