Towards a Scaleable 5G Fronthaul: Analog Radio-over-Fiber and Space Division Multiplexing

TL;DR: The use of analog radio-over-fiber (ARoF) is proposed and demonstrated as a viable alternative which, combined with space division multiplexing in the optical distribution network as well as photonic integration of the required transceivers, shows a path to a scaleable fronthaul solution for 5G.

Abstract: The introduction of millimeter wave (mm-wave) frequency bands for cellular communications with significantly larger bandwidths compared to their sub-6 GHz counterparts, the resulting densification of network deployments and the introduction of antenna arrays with beamforming result in major increases in fronthaul capacity required for 5G networks As a result, a radical re-design of the radio access network is required since traditional fronthaul technologies are not scaleable In this article the use of analog radio-over-fiber (ARoF) is proposed and demonstrated as a viable alternative which, combined with space division multiplexing in the optical distribution network as well as photonic integration of the required transceivers, shows a path to a scaleable fronthaul solution for 5G The trade-off between digitized and analog fronthaul is discussed and the ARoF architecture proposed by blueSPACE is introduced Two options for the generation of ARoF two-tone signals for mm-wave generation via optical heterodyning are discussed in detail, including designs for the implementation in photonic integrated circuits as well as measurements of their phase noise performance The proposed photonic integrated circuit designs include the use of both InP and SiN platforms for ARoF signal generation and optical beamforming respectively, proposing a joint design that allows for true multi-beam transmission from a single antenna array Phase noise measurements based on laboratory implementations of ARoF generation based on a Mach–Zehnder modulator with suppressed carrier and with an optical phase-locked loop are presented and the suitability of these transmitters is evaluated though phase noise simulations Finally, the viability of the proposed ARoF fronthaul architecture for the transport of high-bandwidth mm-wave 5G signals is proven with the successful implementation of a real-time transmission link based on an ARoF baseband unit with full real-time processing of extended 5G new radio signals with 800 MHz bandwidth, achieving transmission over 10 km of 7-core single-mode multi-core fiber and 9 m mm-wave wireless at 255 GHz with bit error rates below the limit for a 7% overhead hard decision forward error correction

Summary

Introduction

• I. INTRODUCTION THE introduction of fifth generation mobile networks (5G)is set to drastically expand the range of applications and use cases for mobile communications and to enable massive advancements in the capabilities and performance of mobile networks [1].
• While ARoF fronthaul holds a lot of promise for mm-wave 5G networks, the realization of such promise will strongly depend on the use of photonic integration and the implementation of ARoF transceivers in photonic integrated circuits (PICs).

A. Centralized Radio Access Networks with Analog Radioover-Fiber

• The move from distributed radio access networks (D-RANs) with baseband processing performed at every remote site to centralized radio access networks (C-RANs) with a shared and centralized baseband unit (BBU) pool has proven successful in reducing the cost of network ownership, operation and maintenance [29].
• At the same time however, the introduction of C-RAN introduced a new segment Copyright (c) 2020 IEEE.
• By moving to analog transport, not only does ARoF fronthaul maintain the centralization of the complete baseband processing, it further centralizes the digital to analog converter (DAC) and analog to digital converter (ADC) stages otherwise located at the RU, as shown in Fig. 1(d).
• The ARoF transmitter generates both the ARoF signal for optical heterodyning as well as a two-tone signal used as remote-fed LO for electrical downconversion at the RU.
• The latter includes an ARoF receiver, downlink RF front end and the antenna in downlink direction, before radiation of the mm-wave signal over the 5G NR air interface to the user equipment (UE).

B. Analog Radio-over-Fiber Transmitter Schemes

• Analog fronthaul over optical fiber can be achieved either by directly modulating the optical carrier with the RF signal or by modulating it with an IF or baseband signal and subsequent Copyright (c) 2020 IEEE.
• For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
• The goal for an ARoF transmitter is to generate the modulated mm-wave 5G RF carrier by means of optical elements, i.e, the electrical to optical (E/O) conversion of the data carrying waveform and its upconversion to mm-wave.
• Yet, integrated SLs generally have rather high intrinsic levels of phase noise, which can substantially deteriorate system performance or even make impossible to achieve signal purity requirements set in relevant standardization and regulations [19].
• This type of architecture can be very useful as it is possible to use as many lasers as desired optical tones [38], allowing the full power of one laser in each tone and consequently avoiding optical amplifiers.

C. Photonic Integration of ARoF Transmitters

• This is where integrated photonics becomes a real asset and the latter is especially true when speaking of optical beamforming integrated devices, which have proven to be a very promising field of research for aerospace, radar and telecom applications in the recent years [21], [40].
• Given commercially available possibilities a fully-integrated solution would necessarily include monolithically integrated laser sources, which is only available through indium phosphite (InP) foundries [44].
• These platforms can provide multi-project wafer (MPW) runs of standardized PICs in a commercial manner, fitting a fabless approach, and can deliver all the necessary components for 5G applications such as high-speed modulators and photodiodes, as well as optical amplifiers.
• Furthermore, this process is compatible with coupling to an SiN chip for the OBFN, as has already been demonstrated for other high performance devices [47].

A. ARoF Fronthaul Architecture with Optical Beamforming

• The blueSPACE fronthaul architecture, shown in Fig. 4, expands the basic ARoF architecture previously shown in Fig. 2 through the use of optical beamforming and optical space division multiplexing in MCF.
• Further amplification is provided by a set of power amplifiers (PAs) before radiation from the PAA.
• The received uplink signal is amplified with a bank of low-noise amplifiers (LNAs) and downconverted to IF, before entering an IFoF link with optical beamforming at the RU and MCF transport to the IFoF receiver with PD and TIA at the CO.
• In either case, optical beamformers can achieve true multi-beam transmission from a single antenna array, effectively forming a mapping matrix between M beam inputs and N outputs towards the N antenna elements (in downlink direction) in which each input is mapped to each output with a variable, progressively increasing delay or phase shift.
• Hence, with the OBFN at the CO, the fronthaul link Copyright (c) 2020 IEEE.

SOA

• Must transport N parallel signals and maintain their temporal synchronization within a narrow margin to preserve the beam patterns established by the OBFN.
• The use of MCF for transport of the fronthaul signals is key with regards to this aspect, as it allows parallel transport of multiple signals at the same wavelength and is expected to have significantly smaller differential delays between channels than solutions based on single-mode fiber (SMF) ribbons or bundles [49].
• With the OBFN at the RU on the other hand, the number of parallel signals corresponds to the number of beams M rather than the number of antenna elements and, as the transported signals are a priori independent, temporal synchronization is not required.
• A similar discussion applies in uplink direction, where placement of the OBFN at the RU again reduces the synchronization and scaling requirements compared to having the OBFN centralized at the CO.

B. blueSPACE Integrated ARoF Transmitter Designs

• Designed to directly support optical beamforming and multi-beam transmission.the authors.
• In the interest of scalability, the authors focus on using only commercially available technologies which they use to design their own circuits.
• Designs of two ARoF transmitters based on the methods described in section II-B are presented in Fig. 5. The first, shown in Fig. 5(a), uses a single on-chip distributed feedback laser (DFB) followed by a SC-MZM to generate a 22 GHz two-tone optical LO signal.
• In the SiN chip one of the LO tones is filtered out, while, in parallel, the four IF modulated channels are routed through the OBFN.
• For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

IV. PHASE NOISE IN OFDM BASED MILLMETER-WAVE ANALOG RADIO-OVER-FIBER SYSTEMS

• Phase noise plays a fundamental role in the performance of radio communication systems and limiting phase noise is one of the major concerns to be addressed before the introduction of ARoF links in fronthaul networks.
• While phase noise in ARoF systems has been studied [50]–[52], its impact on real-time processing of ARoF OFDM signals remains to be evaluated.
• In this section the measured phase noise performance of two of the aforementioned two-tone generation schemes is discussed and simulations are presented aiming at determining the maximum acceptable level of phase noise for real-time processing with the ARoF BBU.

A. Phase Noise Measurements

• The experimental evaluation of phase noise resulting from two-tone generation as required for an ARoF fronthaul focuses on the two schemes for two-tone generation discussed above, namely the SC-MZM and OPLL approaches.
• In comparison, the phase noise measured for the OPLL setup shows a significantly different behavior.
• For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
• Finally, as a reference, Fig. 6(c) further includes the phase noise measured for the IF carrier generated by the IF unit and the final 25.5 GHz mm-wave RF signal after modulation of the SC-MZM two-tone signal with the IF and optical heterodyning for upconversion (as performed in the experimental fronthaul link demonstration discussed in section V).

B. Phase Noise Simulations

• As phase noise is one of the main differentiating parameters between the two-tone generation setups, a set of OFDM simulations was conducted to study the impact of phase noise on the ARoF BBU and to estimate the required improvement in phase noise for the OPLL to become a feasible alternative.
• These simulations included a reduced version of the experimental ARoF link that comprises an ARoF BBU in a loopback configuration at the intermediate frequency, as the scope was to get an insight into the ARoF BBU real-time digital signal processing (DSP) performance in the presence of phase noise, rather than to simulate the overall link.
• The results of the simulations are presented in Fig 7(b), where the EVM and SNR of the transmitter only and the transmitter and receiver combined are plotted as a function of the phase noise level.
• For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
• With the given lasers, the 1/f2 slope of the free running beating tone should be considered as a fundamental limit and it is expected that an extension of OPLL bandwidth by one decade would lower the phase noise plateau by 20 dBc/Hz, as shown in the phase noise profiles in Fig.

V. MILLIMETER-WAVE 5G PERFORMANCE WITH ANALOG RADIO-OVER-FIBER FRONTHAUL

• In order to validate the basic ARoF setup previously shown at the top of Fig. 2 and to evaluate the achieved performance under the given phase noise levels, the experimental setup in Fig. 8 is employed [16].
• The implemented ARoF link features the blueSPACE ARoF BBU, IF unit and an ARoF transmitter based on bulk optical components at the CO, as well as commercially available RF amplifiers and antennas at the RU for mm-wave wireless transmission and electrical downconversion based on the remote-fed LO.
• The ARoF BBU generates extended 5G NR signals with a total of 4096 subcarriers, spaced at 240 kHz, of which 3136 are active, resulting in an effective signal bandwidth of 760.32 MHz.
• The received IF spectrum further has a slightly lower carrier-to-signal ratio as well as a tilt towards the higher frequencies, suggesting a non-flat system response across signal bandwidth at IF or RF.
• The resulting BER measurements as well as their averages are shown in Fig. 9(c), alongside the BER limit for a standard forward error correction (FEC) with 7 % overhead.

VI. CONCLUSIONS

• This article discussed the use of ARoF for the fronthaul of high-bandwidth mm-wave 5G NR signals, briefly introducing how ARoF fronthaul systems can address the fronthaul capacity crunch faced by traditional CPRI fronthaul, while avoiding a partial return to a distributed RAN with some signal processing at the antenna sites.
• To lay the foundation for a full implementation, the phase noise performance of two ARoF schemes was measured and the tolerance of the real-time signals processing in the ARoF BBU was investigated.
• The experimental demonstration of ARoF fronthaul over MCF with a remote-fed LO for downconversion at the RU validates the proposed basic ARoF architecture, as well as the feasibility of ARoF fronthaul with optical heterodyning for high-bandwidth 5G NR signals.
• The demonstrated link achieved BERs below the limit for a commercial FEC with 7 % overhead at a data rate of 1.4 Gbit/s.
• For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

Figures

Fig. 4. blueSPACE architecture with optical beamforming and ARoF transport over MCF: the downlink OBFN is centralized and MCF transport is employed to maintain signal synchronization during transport; the uplink is downconverted electrically and optical beamforming for the IF signal takes place at the RU.

Fig. 9. Measurement results for the real-time ARoF and mm-wave transmission: a) spectrum of the transmitted RF signal, b) spectra of the transmitted and received IF signals, c) BER over optical power of the two-tone signal, d) received constellations for eight equal size subcarrier groups.

Fig. 5. Architecture of an ARoF integrated transmitter with: a) a single on-chip laser and a SC-MZM for LO generation, b) an external OPLL providing a high power two-tone LO. SSC: spot size converter.

Fig. 8. Experimental setup of ARoF transmission with real-time signal processing, 26.5 GHz mm-wave wireless over 9 m and the remote-fed LO.

Fig. 6. Phase noise measurements for two-tone generation with SC-MZM and OPLL: a) measurement setup with SC-MZM, b) measurement setup with OPLL, c) comparison of phase noise for the two-tone generation methods and phase noise of the overall mm-wave signal of the ARoF fronthaul link (see section V).

Fig. 1. Comparison of RAN options and corresponding placement of signal processing functions: a) D-RAN with full baseband processing at the remote site, b) traditional C-RAN with DRoF CPRI fronthaul and centralized baseband processing, c) evolved C-RAN with DRoF eCPRI or NGFI fronthaul and functional split, d) C-RAN with ARoF fronthaul and full centralization of baseband processing and RF sources. SER: seralization, O/E & E/O: optical to electrical and electrical to optical conversion.

Fig. 2. General ARoF fronthaul architecture with optical heterodyne upconversion for downlink, a remote-fed LO for downconversion of the uplink signal at the RU before IF ARoF transport to the CO.

Fig. 7. Simulation results of the influence of phase noise on the OFDM transmission: a) phase noise profiles used as input of the simulation and measured OPLL phase noise, b) corresponding simulated OFDM EVM and SNR as a function of phase noise level for transmission (Tx) and transmission and reception (TxRx); insets show two of the corresponding constellations.

Fig. 3. Description of the frequency upconversion from modulated IF to mm-wave RF using optically generated two-tone signals.

Full text

Towards a Scaleable 5G Fronthaul: Analog Radio-over-Fiber
and Space Division Multiplexing
Citation for published version (APA):
Rommel, S., Dodane, D., Grivas, E., Cimoli, B., Bourderionnet, J., Feugnet, G., Morales, A., Pikasis, E.,
Roeloffzen, C. G. H., Van Dijk, P. W. L., Katsikis, M., Ntontin, K., Kritharidis, D., Spaleniak, I., Mitchell, P.,
Dubov, M., Barros Carvalho, J., & Tafur Monroy, I. (2020). Towards a Scaleable 5G Fronthaul: Analog Radio-
over-Fiber and Space Division Multiplexing.
Journal of Lightwave Technology
,
38
(19), 5412-5422. [9123569].
https://doi.org/10.1109/JLT.2020.3004416
DOI:
10.1109/JLT.2020.3004416
Document status and date:
Published: 01/10/2020
Document Version:
Accepted manuscript including changes made at the peer-review stage
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please
follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy
If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
providing details and we will investigate your claim.
Download date: 10. Aug. 2022

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. XX, NO. X, SEPTEMBER 2020 1
Towards a Scaleable 5G Fronthaul: Analog
Radio-over-Fiber and Space Division Multiplexing
Simon Rommel, Member, IEEE, Member, OSA, Delphin Dodane, Evangelos Grivas, Bruno Cimoli,
Jerome Bourderionnet, Gilles Feugnet, Alvaro Morales, Evangelos Pikasis, Chris Roeloffzen, Paul van Dijk,
Michail Katsikis, Konstantinos Ntontin, Member, IEEE, Dimitrios Kritharidis, Izabela Spaleniak, Paul Mitchell,
Mykhaylo Dubov, Member, OSA, Juliana Barros Carvalho, and Idelfonso Tafur Monroy, Senior Member, IEEE
(Invited Paper)
Abstract—The introduction of millimeter wave (mm-wave)
frequency bands for cellular communications with significantly
larger bandwidths compared to their sub-6 GHz counterparts,
the resulting densification of network deployments and the
introduction of antenna arrays with beamforming result in major
increases in fronthaul capacity required for 5G networks. As a
result, a radical re-design of the radio access network is required
since traditional fronthaul technologies are not scaleable. In this
article the use of analog radio-over-fiber (ARoF) is proposed
and demonstrated as a viable alternative which, combined with
space division multiplexing in the optical distribution network
as well as photonic integration of the required transceivers,
shows a path to a scaleable fronthaul solution for 5G. The
trade-off between digitized and analog fronthaul is discussed and
the ARoF architecture proposed by blueSPACE is introduced.
Two options for the generation of ARoF two-tone signals for
mm-wave generation via optical heterodyning are discussed in
detail, including designs for the implementation in photonic
integrated circuits as well as measurements of their phase noise
performance. The proposed photonic integrated circuit designs
include the use of both InP and SiN platforms for ARoF signal
generation and optical beamforming respectively, proposing a
joint design that allows for true multi-beam transmission from
a single antenna array. Phase noise measurements based on
laboratory implementations of ARoF generation based on a Mach-
Zehnder modulator with suppressed carrier and with an optical
phase-locked loop are presented and the suitability of these
transmitters is evaluated though phase noise simulations. Finally,
the viability of the proposed ARoF fronthaul architecture for the
transport of high-bandwidth mm-wave 5G signals is proven with
the successful implementation of a real-time transmission link
based on an ARoF baseband unit with full real-time processing
Manuscript received February 15, 2020; revised April 30, 2020, revised
June 17, 2020, accepted June 18, 2020. This work was supported by the
blueSPACE project which has received funding from the European Union’s
Horizon 2020 research and innovation programme under grant agreement No.
762055.
S. Rommel, B. Cimoli, A. Morales, J. Barros Carvalho and I. Tafur Monroy
are with the Institute for Photonic Integration, Eindhoven University of Tech-
nology, 5600 MB Eindhoven, The Netherlands (e-mail: s.rommel@tue.nl).
D. Dodane, J. Bourderionnet and G. Feugnet are with Thales Research &
Technology, 1 avenue Augustin Fresnel, 91767 Palaiseau Cedex, France.
E. Grivas and E. Pikasis are with Eulambia Advanced Technologies, 153 42
Agia Paraskevi, Athens, Greece.
C. Roeloffzen and P. van Dijk are with LioniX International, 7500 AL
Enschede, The Netherlands.
M. Katsikis, K. Ntontin and D. Kritharidis are with Intracom Telecom,
Markopoulou Avenue, 190 02 Peania, Athens, Greece.
I. Spaleniak, P. Mitchell and M. Dubov are with Optoscribe, Unit 1,
Rosebank Technology Park, Livingston EH54 7EJ, United Kingdom.
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JLT.2020.XXXXXX
of extended 5G new radio signals with 800 MHz bandwidth,
achieving transmission over 10 km of 7-core single-mode multi-
core fiber and 9 m mm-wave wireless at 25.5 GHz with bit error
rates below the limit for a 7 % overhead hard decision forward
error correction.
Index Terms—5G, fronthaul, analog radio-over-fiber, millime-
ter waves, photonic integration, real-time transmission, phase
noise.
I. INTRODUCTION
T
HE introduction of fifth generation mobile networks (5G)
is set to drastically expand the range of applications and
use cases for mobile communications and to enable massive
advancements in the capabilities and performance of mobile
networks [1]. More specifically, 5G networks are expected to
provide up to 1000× more capacity, to support a significantly
larger number of users per area, to reach multi-Gbit/s end
user data rates and to support latencies down to 1 ms – all
to support applications ranging from media consumption and
broadcasting on-the-go to autonomous driving and the control
of robots, production plants and entire factories [2]. In order to
fulfill such expectations, 5G networks introduce major changes
to the radio access network (RAN) in terms of architecture,
degrees of centralization and placement of processing func-
tionality, in order to ensure the required data rates, quality of
service (QoS) and reduced latency [3], [4]. In this context, the
introduction of novel strategies and technologies becomes a
key factor [5], [6], including the introduction of millimeter
wave (mm-wave) radio frequencies [7], [8].
The latter is considered key to achieve the required end
user data rates as at mm-wave significantly more spectrum is
available for mobile networks compared to the overcrowded
region of traditional mobile communications frequency bands
sub-6 GHz and despite the introduction of additional spectrum
in the latter. While the availability of additional spectrum
allows larger over-the-air capacities, the introduction of high-
bandwidth mm-wave signals brings significant challenges as
the viable range of signals – and thus size of the mm-wave
cells – will typically be below 500 m, requiring a significantly
more dense network deployment [9], [10]. This network densi-
fication, together with the use of larger bandwidths and large-
scale multiple-input multiple-output (MIMO) or beamforming,
further requires a major upgrade or re-design of centralized
radio access networks (C-RANs), as current solutions do not
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at http://dx.doi.org/10.1109/JLT.2020.3004416
Copyright (c) 2020 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. XX, NO. X, SEPTEMBER 2020 2
scale well especially in the fronthaul segment, i.e., between
the radio equipment or remote unit (RU) and the baseband
unit (BBU) at the central office (CO) [3], [4].
Analog radio-over-fiber (ARoF) was initially suggested for
fixed wireless access or wireless bridges between fiber net-
works [11], [12] and includes both the pure analog transport
of radio frequency (RF) signals and the use of microwave pho-
tonics for optical signal processing or optical heterodyning for
frequency conversion [13]. More recently, ARoF has received
significant interest for 5G fronthaul as a solution to solve
the fronthaul capacity issues and through optical heterodyning
assist in mm-wave generation with maximum centralization
of network resources [4], [14], [15]. ARoF at the same time
leads to new considerations for the RAN, as it must be taken
into account in resource allocation and network control [16]
and because it effectively merges the fiber optical and radio
channel and their associated impairments [17]. With optical
heterodyning, phase noise in particular may be a relevant
limitation and must be tightly controlled when designing and
implementing ARoF transmission schemes [18], [19].
While ARoF fronthaul holds a lot of promise for mm-wave
5G networks, the realization of such promise will strongly
depend on the use of photonic integration and the imple-
mentation of ARoF transceivers in photonic integrated circuits
(PICs). Only with such integration can ARoF systems achieve
the spatial and energy footprints required for ubiquitous and
densely deployed 5G mm-wave with ARoF fronthaul and also
be attractive from a cost perspective [20]. Integrated ARoF
transceivers further enable the use of optical beamforming,
where analog beamforming functionality for mm-wave sig-
nals is implemented in compact microwave photonic circuits
rather than bulky and power hungry RF electronics [21]. PICs
for microwave frequencies and ARoF transmitters have been
under study, however successful implementations of ARoF
transceivers in their full complexity are yet to emerge.
Beyond the introduction of ARoF with PIC-based
transceivers, optical space division multiplexing (SDM)
can provide a significant capacity upgrade in the RAN
and further allows sharing of a common passive optical
distribution network (ODN) for both radio access and other
fiber-based access networks, such as passive optical networks
(PONs) [22]. While initially suggested as a method to push the
ultimate capacity of fiber networks using spatial multiplexing
in different cores or modes or a combination thereof [23],
the use of SDM with multi-core fiber (MCF) in RANs has
gained significant interest, especially in connection with ARoF
transport and the use of optical beamforming [21], [24], [25].
The combination of ARoF and optical beamforming in PICs
with MCF for ARoF transport forms the central technology
proposition for 5G fronthaul for high-bandwidth mm-wave
signals of the EU H2020 5G-PPP project blueSPACE [26] and
was the topic of two workshop contributions at MWP 2019
by the authors, serving as the basis for this manuscript. This
manuscript expands upon these, briefly discussing the ARoF
fronthaul over SDM scheme of blueSPACE and introducing in
some detail its ARoF transmitter schemes and designs for their
implementation in integrated photonics. In preparation towards
a full implementation, it provides experimental phase noise
measurements for two different mm-wave ARoF transmitter
schemes and present simulations for phase noise requirements
and tolerances in systems with 5G new radio (NR) orthogonal
frequency division multiplexing (OFDM) signals and real-time
processing in an ARoF BBU. Finally, real-time transmission
of 800 MHz wide extended 5G NR signals is successfully
demonstrated for an ARoF fronthaul link with 7-core MCF
and a 9 m mm-wave link at 26.5 GHz [16], [27], i.e., in the
3GPP n258 band [28]. The demonstration is based on an ARoF
BBU implemented on an field programmable gate array (FPGA)
and intermediate frequency (IF) ARoF transport with optical
heterodyning for mm-wave upconversion of the downlink
signal and features a remote-fed RF local oscillator (LO) for
electrical downconversion of the uplink signal, maximizing
centralization of network equipment at the CO and minimizing
RU complexity. Together these contributions show the viability
of the blueSPACE architecture and establish performance
requirements for the realization of the ARoF transmitters and
optical beamforming networks (OBFNs) in PICs.
The remainder of the manuscript is structured as follows:
section II discusses ARoF fronthaul systems for mm-wave 5G
(II-A), introduces the relevant ARoF transmitter schemes (II-B)
and considers the possibilities for their integration in PICs
(II-C). Section III introduces the blueSPACE ARoF fronthaul
architecture with optical beamforming (III-A) and correspond-
ing ARoF transmitter PIC designs (III-B). Section IV discusses
phase noise in OFDM-based mm-wave ARoF systems, first,
showing the impact of phase noise on the OFDM transmission
performance and, second, presenting measurements of phase
noise from two ARoF transmitter schemes. Section V presents
the experimental transmission of extended 5G NR signals with
ARoF fronthaul over SDM and mm-wave RF transmission.
Finally, section VI summarizes and concludes the paper.
II. ANALOG RADIO-OVER-FIBER FRONTHAUL SYSTEMS
FOR MILLIMETER-WAVE 5G
This section discusses the advantages of moving from dig-
itized to analog transport in the fronthaul link and introduces
a basic architecture for analog fronthaul, before discussing
different ARoF transmission schemes and the possibilities for
their implementation in photonic integrated circuits.
A. Centralized Radio Access Networks with Analog Radio-
over-Fiber
The move from distributed radio access networks (D-RANs)
with baseband processing performed at every remote site to
centralized radio access networks (C-RANs) with a shared and
centralized baseband unit (BBU) pool has proven successful
in reducing the cost of network ownership, operation and
maintenance [29]. By centralizing all baseband processing in
a BBU pool at the central office (CO), as shown in Fig. 1(a)
and (b), C-RAN allows statistical multiplexing and sharing of
baseband processing resources and simplifies the management
and maintenance of the network by reducing the complexity
placed near the antennas at remote locations. At the same time
however, the introduction of C-RAN introduced a new segment
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at http://dx.doi.org/10.1109/JLT.2020.3004416
Copyright (c) 2020 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. XX, NO. X, SEPTEMBER 2020 3
Core
Core
Core
BBU Pool
Remote Stations
Remote Stations
Antennas
RF Amplifiers & Filters
RF upconversion
Antennas
RF Amplifiers & Filters
RF upconversion
DAC / ADC
Baseband Processing
Backhaul
Fronthaul
(CPRI)
Baseband Processing
E/O / O/E
(De)Framing+(De)SER
Backhaul
BBU Pool
Remote Stations
Antennas
RF Amplifiers & Filters
RF upconversion
Fronthaul
(eCPRI/
NGFI)
Baseband Upper
Backhaul
Baseband Lower
ARoF BBU
Pool
Remote Stations
O/E / E/O
ARoF
Fronthaul
Baseband Processing
Backhaul
Antennas
RF Amplifiers & Filters
IF/RF Upconversion
Core
a) D-RAN
b) C-RAN: DRoF legacy
c) C-RAN: DRoF evolved
d) C-RAN: ARoF
DAC / ADC
DAC / ADC
DAC / ADC
E/O / O/E
E/O / O/E
(De)Framing+(De)SER
(De)Framing+(De)SER
O/E / E/O
O/E / E/O
(De)Framing+(De)SER
Fig. 1. Comparison of RAN options and corresponding placement of signal
processing functions: a) D-RAN with full baseband processing at the remote
site, b) traditional C-RAN with DRoF CPRI fronthaul and centralized baseband
processing, c) evolved C-RAN with DRoF eCPRI or NGFI fronthaul and
functional split, d) C-RAN with ARoF fronthaul and full centralization of
baseband processing and RF sources. SER: seralization, O/E & E/O: optical
to electrical and electrical to optical conversion.
in the RAN, the fronthaul segment between BBU at the CO and
remote unit (RU) at the antenna site.
In pre-5G networks, this segment was covered by CPRI
based fronthaul – i.e., transmitting the digitized IQ samples of
the target or received OFDM waveforms – typically over fiber
optic links and hence referred to as DRoF. The transmission
of digitized samples however requires constant bitrate signals,
where the rate scales with signal bandwidth and antenna
configuration and is independent of user presence or activity
[3]. Furthermore, with multi-sector or even MIMO antennas,
these fronthaul links scale to hundreds of Gbit/s and into the
Tbit/s even for moderate signal bandwidths and MIMO config-
urations and thus, together with their stringent requirements
on latency and jitter, are clearly not a scaleable solution for
high-bandwidth 5G signals, especially at mm-wave.
One approach to mitigate the resulting fronthaul capacity
bottleneck is the use of DRoF with eCPRI or NGFI, introducing
a partial re-distribution of the baseband processing and a
functional split, as shown in Fig. 1(c) [30]. While this can
significantly reduce the required data rates and – depending
on the functional split chosen – may reduce latency and jitter
requirements [30], it requires the introduction of an additional
distributed unit (DU) between the CO and RU and thus results
in a partial loss of the centralization gain.
ARoF on the other hand solves both the fronthaul capacity
problem and maintains the centralization gain by moving
from the transport of digitized radio samples to the analog
transport of the baseband, IF or RF waveforms. By moving to
analog transport, not only does ARoF fronthaul maintain the
centralization of the complete baseband processing, it further
centralizes the digital to analog converter (DAC) and analog to
digital converter (ADC) stages otherwise located at the RU,
as shown in Fig. 1(d). The optical bandwidth or spectrum
required for ARoF transport directly relates to the IF or RF
frequency transported or targeted plus the signal bandwidth
CO
RU
ODN
5G
NR
UE
SDM (De)Mux
End User
ARoF BBU
Real-Time DSP IF IQ (de)mod
IF Unit
RF LO
ARoF
IF Rx
ARoF Tx
Laser, two tone generation
→ Remote LO
→ ARoF fronthaul signal
SDM (De)Mux
Antenna
DL
UL
ARoF
RF Rx
ARoF LO Rx
DL RF
front end
ARoF
IF Tx
UL RF
front end
Antenna
ARoF
Fronthaul
mm-
wave
Fig. 2. General ARoF fronthaul architecture with optical heterodyne upconver-
sion for downlink, a remote-fed LO for downconversion of the uplink signal
at the RU before IF ARoF transport to the CO.
and – with narrow optical filtering or multiplexing – may
even be reduced below that by interleaving optical carriers
and modulation sidebands with narrow channel spacings [15],
[22]. If combined with optical heterodyning, ARoF becomes
especially interesting for mm-wave 5G NR signal generation
as the ARoF link can not only transparently transport an IF
signal, but also perform its upconversion to mm-wave through
optical heterodyning on the photodiode (PD) at the RU – while
keeping the required RF oscillator at the CO [14], [27].
A corresponding ARoF fronthaul architecture is shown in
Fig. 2, using IF ARoF transport for the downlink and inter-
mediate frequency-over-fiber (IFoF) transport for the uplink
[15], [31]. It features the ARoF BBU and IF unit at the
CO, as well as an ARoF transmitter and IFoF receiver. The
ARoF transmitter generates both the ARoF signal for optical
heterodyning as well as a two-tone signal used as remote-fed
LO for electrical downconversion at the RU. The latter includes
an ARoF receiver, downlink RF front end and the antenna in
downlink direction, before radiation of the mm-wave signal
over the 5G NR air interface to the user equipment (UE). In
uplink direction the signal received from the UE passes the
uplink RF frontend, including downconversion to IF using the
remote-fed LO, before being transmitted as an IFoF signal back
to the CO. With this basic architecture, no RF source is required
at the RU, while employing IFoF transport for the uplink to
reduce the required bandwidth of the driver amplifier (DA)
and optical modulator.
The use of mm-wave bands further results in a significant
reduction in antenna size and brings the possibility to use
phased array antennas (PAAs) or other multi-element antenna
systems, either for massive MIMO or with analog beamforming
or a hybrid solution of both [8]. The basic ARoF link is easily
expanded to use analog optical rather than electrical beam-
forming to further reduce spatial, power and cost footprints
of the RU. The implementation of such an ARoF solution is,
however, not trivial and a number of options exist for the ARoF
transport and the generation of the two-tone signal required for
optical heterodyning, which are introduced in the following.
The ARoF fronthaul architecture with optical beamforming
and MCF transport proposed by blueSPACE is discussed in
section III-A with details on the integrated ARoF transmitter
designs given in section III-B.
B. Analog Radio-over-Fiber Transmitter Schemes
Analog fronthaul over optical fiber can be achieved either
by directly modulating the optical carrier with the RF signal or
by modulating it with an IF or baseband signal and subsequent
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at http://dx.doi.org/10.1109/JLT.2020.3004416
Copyright (c) 2020 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. XX, NO. X, SEPTEMBER 2020 4
IF + BB
f
Elect.
IF ~5 GHz
RF IF signal
Modulation
Sideband
filtering
Combining
Detection
Optical domain
Electrical domain
5G domain
Frequency
Up-conversion
IF + BB + LO
~27 GHz
RF 5G signal
f
Elect.
LO
~22 GHz
f
Opt.
Optical tone #1
Optical tone #2
f
Opt.
f
Opt.
f
Opt.
~27 GHz
f
Opt.
two-tone op�cal Local Oscillator
Fig. 3. Description of the frequency upconversion from modulated IF to
mm-wave RF using optically generated two-tone signals.
upconversion at the remote side in the electrical domain
or through optical heterodyning [13], [15], [31]. While for
traditional radio frequencies direct modulation with the RF
signal is feasible, for mm-wave signals this is limited by
modulator bandwidth and schemes based on optical hetero-
dyning are preferred as the use of optical carriers for 5G
mm-wave signal generation provides significant advantages.
First, it is relatively simple to generate optical signals that
are separated by an offset frequency in the range of the 5G
mm-wave spectrum (≈ 30 GHz) and beyond and the tunability
of telecom laser sources makes the full frequency range very
accessible. Second, the modulation bandwidth available from
optical components facilitates the use of larger signal band-
width as available at mm-wave. These aspects of microwave
photonics are crucial advantages of ARoF for 5G applications.
The goal for an ARoF transmitter is to generate the modu-
lated mm-wave 5G RF carrier by means of optical elements,
i.e, the electrical to optical (E/O) conversion of the data car-
rying waveform and its upconversion to mm-wave. A general
description of the upconversion of an IF signal to mm-wave
frequencies is shown in Fig. 3. First, the IF signal modulates
an optical carrier, creating sidebands in the optical spectrum of
the signal. The signal is then filtered so that only one sideband
remains and is combined with another optical signal (generally
referred to as LO). Finally, a fast PD performs the optical to
electrical (O/E) conversion and through optical heterodyning
generates an RF signal at a frequency equal to the sum of
the difference between the two optical tones and the IF. The
key aspect of this E/O/E link is the generation of the two
optical tones shown at the bottom of Fig. 3, as it has a major
impact on the useability of the two-tone signal and the overall
transmission performance.
The main criteria to evaluate the optical two-tone generation
are the tunability, which reflects the flexibility of the system
to cover the full 5G frequency range, the power budget, which
describes its power efficiency, and the resulting signal purity.
Several solutions exist for two-tone generation, generally
involving either several laser sources, each corresponding to
one tone, or one single laser source, used to generate multiple
tones. The former is more power efficient but suffers from
coherence issues while mixing signals which originate from
different laser sources. The latter is very commonly used
to avoid coherence issues between lasers that impact signal
purity, but at the cost of a limited power budget due to having
only one optical power source.
On one hand, solutions involving a single laser source
are mainly based on the use of interferometric modulators
and/or optical resonators. In both cases multiple optical tones
are generated from one original carrier which means that
their heterodyne recombination will not suffer from coherence
issues, given temporal alignment of the tones is maintained
during processing. This can be done using carrier-suppressed
Mach-Zehnder modulators (SC-MZMs), which is a wide-spread
method to generate such a two-tone optical signal and to
perform effective frequency doubling [13], [31], [32]. The
main limitation of this solution is linked to power efficiency
as a significant portion of the input optical power is lost due
to the suppression of the carrier. In order to compensate for
this, amplifier stages (e.g., semiconductor optical amplifiers
(SOAs) or erbium-doped fiber amplifiers (EDFAs)) are typically
required, decreasing the overall efficiency. Similarly, power
efficiency when using mode-locked laserss (MLLs) or optical
frequency combs [33], [34] is also critical as the optical power
is transferred into multiple tones while only a few are used
at any time, with filtering often required to get rid of unused
comb frequencies.
On the other hand, solutions including separate lasers are
more power efficient since power is not spread from one to
multiple tones, but suffer from coherence issues inherent to
the use of different sources. In particular, the pure heterodyne
beating of two semiconductor lasers (SLs) is very simple to
perform and strongly benefits from their tunability. Yet, inte-
grated SLs generally have rather high intrinsic levels of phase
noise, which can substantially deteriorate system performance
or even make impossible to achieve signal purity requirements
set in relevant standardization and regulations [19]. As a result,
the use of independent lasers is only feasible if the intrinsic
laser phase noise becomes very small.
An alternative solution is to make the laser sources “arti-
ficially” coherent through the use of an electronic external
control loop. This method is referred to as optical phase-
locked loop (OPLL) and its implementation in PICs has already
been demonstrated [35]–[37]. The final beating is purely
heterodyne, but the two lasers are strongly correlated to the
point of appearing to be identical, resulting in a potentially
very pure RF tone. This type of architecture can be very
useful as it is possible to use as many lasers as desired
optical tones [38], allowing the full power of one laser in
each tone and consequently avoiding optical amplifiers. It is
worth noticing that electrical PLLs have become very common
components, in particular for demodulation of signals and
frequency upconversion of low-jitter clocks [39]. However
their optical equivalents are critical to design and can prove
to be quite difficult to implement in a functional device. They
further have finite bandwidth, meaning that the two lasers have
correlated phase noises only within a certain frequency range
around their central frequency.
This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication.
The final version of record is available at http://dx.doi.org/10.1109/JLT.2020.3004416
Copyright (c) 2020 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.

Frequently asked questions

Q1. How many MHz can be extended by the inclusion of a phase advance in the loop filter?

7. While OPLL bandwidth is limited by the change of tuning mechanism of the DFB laser between thermal (low frequencies, opposite phase to frequency modulation) and carrier (high frequencies, in phase) which occurs in the MHz range, the inclusion of a phase advance in the loop filter to anticipate for this change can allow extension of the loop bandwidth by several MHz. 

Q2. What is the RF spectra of the ARoF fronthaul link?

The evaluation of the ARoF fronthaul link is based on the spectra of the transmitted RF signal and the transmitted and received IF signals, as well as on the real-time bit error rate (BER) and corresponding signal constellations. 

Q3. What is the purpose of the section IV simulations?

In section IV simulations are presented to estimate the required phase noise performance and validate the use of an OPLL in the proposed architecture, which would include the PIC described in this section. 

Q4. What is the phase noise profile with the SC-MZM?

The phase noise profile with the SC-MZM follows that of the underlying RF LO, but increased by about 6 dB, as expected due to the effective frequency doubling. 

Q5. What is the main advantage of using mm-wave bands?

The use of mm-wave bands further results in a significant reduction in antenna size and brings the possibility to use phased array antennas (PAAs) or other multi-element antenna systems, either for massive MIMO or with analog beamforming or a hybrid solution of both [8]. 

Q6. Why is the BER in the IF unit so low?

This is likely due to reduced low-frequency performance of the modulator in the IF unit as well as bandwidth limitations of the latter and the slight power reduction at higher frequencies previously observed on the received IF signal. 

Q7. What are the main advantages of a single laser source?

On one hand, solutions involving a single laser source are mainly based on the use of interferometric modulators and/or optical resonators. 

Q8. How much noise is expected to be reduced by an extension of a decade?

With the given lasers, the 1/f2 slope of the free running beating tone should be considered as a fundamental limit and it is expected that an extension of OPLL bandwidth by one decade would lower the phase noise plateau by 20 dBc/Hz, as shown in the phase noise profiles in Fig. 

Q9. What is the phase noise performance of the two ARoF transmitter schemes?

The phase noise performance of the two ARoF transmitter schemes was experimentally evaluated, showing the SC-MZM approach to match the phase noise of the underlying RF source, while the OPLL successfully suppresses part of the combined phase noise of the two lasers, but with overall higher phase noise. 

Q10. How low is the phase noise of the loop?

While the current OPLL implementation can not reach the required phase noise levels, the authors predict that with an optimized selection of lasers and design of the feedback loop, phase noise levels sufficiently low to allow operation with the ARoF BBU can be achieved. 

Q11. What is the main limitation of the pure heterodyne beating of two semiconductor lasers?

In particular, the pure heterodyne beating of two semiconductor lasers (SLs) is very simple to perform and strongly benefits from their tunability. 

Q12. What is the key step to ensuring that the final implementation is viable?

In order to ensure that the final implementation is viable, a key step is to investigate the influence of the phase noise of such an OPLL on real-time data transmission. 

Q13. What are the main advantages of a two-tone generation system?

Several solutions exist for two-tone generation, generally involving either several laser sources, each corresponding to one tone, or one single laser source, used to generate multiple tones. 

Q14. What is the compatibility of the blueSPACE fronthaul architecture with the OBFN?

this process is compatible with coupling to an SiN chip for the OBFN, as has already been demonstrated for other high performance devices [47].