0733-8724 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2019.2957882, Journal of
Lightwave Technology
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
1
Abstract—Coherent technologies along with digital signal
processing (DSP) have revolutionized the optical communication
systems, significantly increasing the capacity of the fiber channel
owing to transmission of advanced modulation formats and
mitigation of propagation impairments. However, the commercial
solutions for high-capacity core networks are too complex and
costly, then hardly feasible, for access networks with high terminal
density, where cost, power budget and footprint are the main
limiting factors. This article analyses the key enabling techniques
to implement a complexity-reduced coherent transceiver (CoTRX)
by exploiting photonic integration, simplified optical modulation,
low-cost DFB lasers, consumer electronics and low-complexity
DSP. Bulk optical modulators are replaced by direct amplitude-
and-phase modulation of integrated electro-absorption modulated
laser (EML) with smaller footprint, generating up to 8-ary
modulation formats. Hardware-efficient DSP algorithms for the
coherent transmitter and receiver, including pulse-shaping for
direct phase modulation, differential detection for optical phase
recovery, and digital pre-emphasis with enhanced tolerance to
quantization noise, are investigated to face the challenges imposed
by low-cost photonic and electronic devices, such as strong phase
noise, wavelength drifts, severe bandwidth limitation, and low
resolution data converters. Through numerical simulations and
real-time experiments, the results indicate that this new class of
CoTRX enables effective implementation of wavelength-to-the-
user PON with dedicated 1.25 to 20 Gb/s per user, in an ultra-
dense 6.25 to 25 GHz spaced WDM optical grid, with >30 dB loss
budget, outperforming the current competing technologies for
access networks.
Index Terms— Coherent detection, electro-absorption
modulated laser, wavelength division multiplexing, digital signal
processing, access networks.
I. INTRODUCTION
he standardized and commercially deployed optical access
networks need to evolve facing the upcoming bandwidth-
hungry multimedia services with ultra-high definition video,
business connectivity and mobile front-haul/back-haul
(MFH/MBH) for 5G. The current passive optical networks
(PONs) attending the XG-PON standard rely on 10 Gb/s data
rate, serving the final subscribers by time division multiplexing
(TDM). To enhance the total PON capacity, the latest ITU-T
standard NG-PON2 exploits the wavelength division
multiplexing (WDM) by stacking 4/8 XG-PON wavelengths
Manuscript submitted June 28, 2019. This work was supported in part by the
Spanish Project FLIPER (TEC2015-70835), and the FPU program from the
Education Ministry of Spain (FPU12/06318).
The authors are with the Universitat Politècnica de Catalunya, Signal Theory
and Communications Department (TSC), Barcelona E-08034, Spain (email:
Fig. 1. Flexible coherent udWDM-PON with dedicated per user supporting
different types of services. DS: downstream, US: upstream, FTTH: fiber to the
home, FTTB: fiber to the business.
() achieving aggregate PON capacity up to 80 Gb/s [1].
However, scaling the actual time-and-wavelength division
multiplexing (TWDM) technology to higher data rates in a cost-
effective way might not be compatible with low-cost
commercial photonic and electronic devices, mandatory for
affordable PON implementation. Specifically, increasing the bit
rate per single constitutes a high cost and energy consuming
solution due to the large required bandwidth (BW) of the
hardware (HW) devices such as lasers/modulators, photodiodes
(PDs), and amplifiers. Moreover, transmission impairments
like chromatic dispersion (CD) become more apparent on large
BW signals, requiring digital signal processing (DSP) for
mitigation with necessary data converters that are expensive
and power-hungry if the required BW and sampling rate are
high.
The novel solutions for next generation PONs must be: (I)
affordable in cost by the final users, (II) fully compatible with
legacy access systems based on power splitting at the outside
plant, and (III) flexible in terms of user BW and channel
allocation, with reconfigurable transceivers (TRXs) enabling
convergence of residential, business and MFH/MBH services.
A key technology fulfilling these requirements is a new class of
coherent TRX (CoTRX) implementing the -to-the-user
concept, where each subscriber owns a dedicated in a cost-
effective manner by multiplexing in ultra-dense WDM
(udWDM). The bit rate per retains compatible with consumer
jtabares@tsc.upc.edu; saeed.ghasemi@tsc.upc.edu;
juavelasquez@tsc.upc.edu; jprat@tsc.upc.edu).
udWDM
OLT
DS
US
Power
splitter
Feeder
fiber
FTTH
FTTH
FTTB
MFH/MBH
Drop fiber
. . .
Coherent Ultra-Dense WDM-PON Enabled by
Complexity-Reduced Digital Transceivers
Jeison Tabares, Saeed Ghasemi, J. Camilo Velásquez, and Josep Prat, Member, IEEE
T
0733-8724 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2019.2957882, Journal of
Lightwave Technology
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
2
Fig. 2. Block diagram of a conventional CoTRX for high-capacity optical networks: TX (left) and RX (right). OA: optical amplifier; MZM: Mach-Zehnder
modulator; PBS/PBC: polarization beam splitter/combiner.
Fig. 3. Block diagram of the proposed CoTRX for udWDM-PON with simplified architecture: TX (left) and RX (right).
electronics BW, but still achieving superior aggregate PON
capacity with high spectral efficiency (SE) [2], [3] compared
with standard access systems. The CoTRX technology includes
commercial low-cost distributed feed-back (DFB) or vertical-
cavity surface-emitting laser (VCSEL) as optical sources,
flexible optical grid with limited tunability, low-speed data
converters, and low-complexity DSP. The DSP plays the
fundamental role of enhancing both the system capacity and the
SE by transmitting advanced modulation formats, as well as
compensating for all the distortions and transmission
impairments that make impact on the signal from origin to
destination.
Fig. 1 depicts the flexible udWDM-PON enabled by coherent
technologies, supporting different applications over the same
fiber infrastructure deployed for current PONs. The optical
distribution network (ODN) is totally passive and filterless,
with user terminals implementing tunable lasers with non-
preselected . Channel selection is done by -tuning of the local
oscillator (LO) laser at the CoTRX, allowing for ultra-narrow
spaced WDM grid compared with conventional selectivity by
optical filters. The link power budget is a critical factor to be
addressed in next generation PONs [1], [4]; owing to the
superior sensitivity of coherent detection, the udWDM-PON
can achieve the targeted power budget 30 dB without
increasing the launched power or extra optical amplification,
thus lowering the total energy consumption.
In this article, we discuss how the optical coherent
technologies supported by DSP, that are mature in high-
capacity and long-haul transport networks, can be migrated for
access applications or, in general, optical networks with high
terminal density. We explore photonic integration of low-cost
DFBs with electro-absorption modulator (EAM), along with
direct intensity-and-phase modulation leveraging the laser
chirp, as a key technology to reduce the complexity and energy
consumption of the coherent transmitter (TX). With this small
footprint device, up to 8-amplitude-and-phase shift keying (8-
APSK) at 7.5 Gb/s was achieved with RX sensitivity of -40
dBm. Besides, this article analyzes HW-efficient DSP
algorithms to transmit advanced modulation formats, dealing
with the impairments from low-cost photonic/electronic
devices. Low-resolution and band-limited data converters have
been considered, adapting the pre-emphasis DSP algorithm of
the TX to operate in presence of quantization noise. Simple
digital pulse-shaping at the TX DSP is proposed to generate
complex phase-modulated signals exploiting the laser chirp.
Straightforward differential detection is investigated for carrier
recovery (CR) at the receiver (RX), showing high tolerance to
strong phase noise, up to total spectral linewidth = 0.64% of
the bit rate for differential PSK (DPSK) in real-time.
The outcome is a complexity-reduced and cost-effective
digital CoTRX for udWDM-PON that is conceptually complex,
but affordable in cost by exploiting photonic integration,
simplified optical modulation, low-cost consumer electronics
and HW efficient DSP. Despite the CoTRXs simplifications,
the udWDM-PON outperforms the present PON technologies
achieving loss budget in excess of 30 dB, enabling a large
number of users in a 6.25 to 25 GHz spaced optical grid for 1.25
to 20 Gb/s -to-the-user access network, totally passive,
transparent, and fully compatible with deployed legacy PON
systems.
The article is organized as follows. Section II presents the
architecture of the simplified CoTRX for udWDM-PON,
inherited from conventional CoTRX technology along with
novel strategies reducing the complexity and cost. The DSP
algorithms for the TX and RX subsystems are analyzed in
Sections III and IV respectively, followed by the conclusions in
Section V.
II. COHERENT TRANSCEIVERS TECHNOLOGY
The actual commercially available CoTRXs are mainly
Pulse Shaping
Linear Pre-Emphasis
ASIC
Symbol Mapping
NL Pre-Distortion
DAC
DAC
DAC
DAC
PBS
PBC
MZM-I
MZM-Q
MZM-I
MZM-Q
DP – IQ Modulator
I
H
Q
H
I
V
Q
V
OA
Optical
output
PBS
90˚
Optical
Hybrid
BPD
BPD
90˚
Optical
Hybrid
BPD
BPD
PBS
Optical
input
Front-end correction
NL / CD compensation
ADC
ADC
ADC
ADC
I
H
Q
H
I
V
Q
V
Phase & Pol. diversity front-end
Equalization
CLK recovery
Carrier frequency &
phase recovery
Decision & demapping
ASIC
ASIC
Symbol Mapping
DAC
DAC
Integrated Dual-EML
EAM
Optical
output
Linear Pre-Emphasis
Pulse Shaping
PBS
PBS
Optical
input
ADC
ADC
ADC
ADC
I
H
Q
H
I
V
Q
V
Phase & Pol. diversity front-end
Equalization
CLK recovery
Carrier frequency &
phase recovery
Decision & demapping
ASIC
IQ recovery
IQ recovery
0733-8724 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2019.2957882, Journal of
Lightwave Technology
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
3
intended for high capacity core networks supporting large
volumes of data traffic. They are also gaining interest for short-
reach data center interconnect (DCI) to increase the total data
rate within the same BW. Performance evaluation of CoTRXs
for 400G DCI systems reveals that the overall complexity and
energy consumption is comparable to that the intensity-
modulation direct-detection (IM-DD) counterpart, that
dominates the DCI market segment, but with superior power
budget and SE [5].
To correctly transport such high data rates with CoTRXs,
information is mapped into advanced modulation formats with
dense complex constellations that propagate through electronic
and optical devices of the CoTRX, and through the transmission
media, i.e., the optical fiber infrastructure.
Fig. 2 depicts a block diagram of the conventional CoTRX
with all the HW components and the DSP subsystems [6]. In
general, the CoTRXs are still complex and too expensive for
optical access with high terminal density. Moreover, the
necessary digital-to-analog/analog-to-digital converters
(DACs/ADCs) are usually listed as critical elements since their
cost and power consumption increases if the required sample
rate, resolution bits, and BW are high. However, during the last
years efforts in research are being taken to develop CoTRXs
with simplified architectures and HW efficient DSP [7][11].
Yet, most of the proposed solutions still rely on external
modulators, or extra RX BW that unfavorably scales with the
bit rate. A recent alternative for CoTRXs is the Kramers-Kronig
RX [12] that substantially reduces the RX front-end to a single
PD per each state of polarization (SOP), but does not exhibit the
high sensitivity and -selectivity of coherent systems. Also, the
required BW twofold compared with coherent homodyne
detection.
In practice, the adoption of digital coherent systems for
PONs will depend on (I) reducing the footprint of the CoTRX
photonic elements; (II) lowering the requirements for ADCs
and DACs in terms of BW, resolution, and sample rate; (III)
simplifying the DSP complexity, thus reducing the energy
consumption and overall cost of the digital processor.
Fig. 3 presents the proposed realization of a CoTRX for
udWDM-PON addressing the aforementioned requirements.
First, for generation of advanced modulation formats the in-
phase and quadrature () modulators represent a severe
limitation in terms of footprint, cost, and power budget. Instead,
we successfully proposed and demonstrated in [13] a simple -
ary coherent TX based on dual electro-absorption modulated
laser (Dual-EML) for up to 8-APSK.
The Dual-EML is a monolithically integrated photonic
device where the same active layer is composed by two
sections, a DFB and an EAM. The term dual denotes that the
EML chip has dual electrodes with independent radio frequency
(RF) inputs to simultaneously phase-modulate the DFB and
intensity-modulate the EAM. Careful optimization of the bias
current/voltage is necessary to isolate the two modulations
generated by the Dual-EML. First, the DFB section is biased far
from the threshold current, minimizing the extinction ratio of
the residual IM at most, thus behaving the DFB as a pure optical
frequency modulator (FM) by the laser chirp. Similarly, the FM
contribution of the EAM by its inherent chirp can be reduced
by biasing the EAM near to the zero-chirp region, as reported
in Fig. 6 of [13]. Optical amplification is not necessary since no
external power-losing devices are attached to the Dual-EML
output, lowering the power consumption of the coherent TX.
At the RX front-end, the 90° optical hybrids in Fig. 2 are
replaced by much simpler 3 × 3 fiber couplers in Fig. 3. This
allows for phase-and-polarization diversity with six single-
ended PDs, two less than the four balanced PDs (BPD). The
signals for each SOP are recovered by linearly combining the
three photocurrents in passive HW [14]. Note that the
recovery could be implemented into the DSP; however, this
simple analog pre-processing saves two ADC channels that are
critical components of the RX.
To further reduce the complexity and cost, the CoTRX
should resort on the use of commercial low-cost lasers, such as
DFBs with statistical (i.e. non-preselected) . This in turn,
imposes extra challenges as the DFBs exhibit -instabilities,
high , and limited tunability. The DSP algorithms of the
CoTRX mitigate the first two impairments (-drifts, ), that
directly affect data making strong impact on the received
constellation, as it will be treated in Section IV.
On the other hand, the limited thermal tunability of the DFBs
is critical for the control of the TX and LO lasers in the
udWDM-PON, featuring hundreds of s in a narrow-spaced
optical grid. Typical DFBs have a tunability ratio of ~0.1 nm /
˚C [15] covering ~3 nm (375 GHz) by ±15 ˚C temperature
range, large enough to spread over many, though not all,
udWDM channels. To properly manage the activation process
and operation, a set of heuristic channel assignment algorithms
with dynamic wavelength allocation (DWA) have been
proposed to avoid collisions and efficiently organize the
udWDM spectrum, lowering the blocking probability of new
users connecting to the PON and satisfying network changes in
environmental temperature [16].
The DSP of the CoTRX can be implemented either in an
application-specific integrated circuit (ASIC) or in a field-
programmable gate array (FPGA). The digital CoTRX
reconfigures easily dynamically adjusting its parameters such
as modulation format and data BW, adapting to the type of
service and user. This is advantageous for network flexibility
compared with analog systems that require HW changes.
Some of the DSP algorithms of the conventional CoTRX in
Fig. 2 are key functionalities to correctly generate and detect
the modulated data, while others related to equalization and
impairments mitigation play a secondary role in the access
scenario with lower data rates, simpler modulation formats, and
shorter fiber spans. Next sections analyze all the proposed DSP
algorithms of the CoTRX in Fig. 3, intended for cost-effective
udWDM-PON.
III. DSP SUBSYSTEMS OF THE TRANSMITTER
In the simplified TX of Fig. 3, the bulk dual-polarization
(DP-) modulator is replaced by the EAM integrated into the
Dual-EML. The EAM has a non-linear (NL) electro-optic
response, similar to Fig. 3 of [13]. Nevertheless, the EAM bias
0733-8724 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2019.2957882, Journal of
Lightwave Technology
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
4
Fig. 4. BPSK, QPSK, and 8-APSK experimental constellation diagrams for
direct modulation with the Dual-EML.
and the extinction ratio (ER) can be adjusted to operate in the
linear region, and the NL pre-distortion by DSP might be
optional or discarded.
Therefore, the key DSP subsystems for the TX in Fig. 3 are:
symbol mapping, pulse shaping and linear pre-emphasis.
A. Symbol Mapping
In the conventional CoTRX of Fig. 2, the DP- modulator
creates a quadrature amplitude modulation (QAM) with
rectangular constellation in the complex plane by each pair of
coordinates . In contrast, the Dual-EML in Fig. 3, proposed
as a cost-effective complex optical modulator, generates a
circular constellation with polar coordinates by directly
modulating the intensity with the EAM and the phase with the
DFB laser chirp. In terms of flexibility, the TX is able to
generate a set of circular modulation formats spanning from
binary PSK (BPSK) up to at least 8-APSK, without adding extra
HW complexity. Fig. 4 shows the constellation diagram for
BPSK, quadrature PSK (QPSK) and 8-APSK, as well as the
experimental constellations obtained from direct amplitude-and
phase-modulation of the Dual-EML at 2.5 GBd. It is worth
noting the phase-widening of the constellation points for 8-
APSK, compared to that of BPSK and QPSK where the EAM
is off, due to the increment of the spectral linewidth along with
the residual chirp of the EAM under modulation.
For phase modulation only, the constellation symbols are
, with the constant radius and
the phase-
encoded data defined by
and
for
BPSK and QPSK respectively. The mean symbol power is
given by
. In the case of amplitude and phase
modulation, consecutive symbols are set to have symmetric
Euclidean distance among them. Hence, for 8-APSK the
symbols are defined by
, with
and
the amplitude and phase information
respectively. The ratio between radii for symmetric is found
to be
.
Table I summarizes our reported experiments with direct
amplitude-and-phase modulation of low-cost DFBs and Dual-
EML, with different modulation formats. Results show good
performance with high RX sensitivity, measured for a bit error
ratio (BER) of 10
-3
, fully compatible with current forward error
correction (FEC) systems. Notably, 8-APSK shows 8.5 dB
better sensitivity than 8-PSK, at the same bit rate and
comparable , due to less impact of the phase noise on the
underlying QPSK of the 8-APSK. Considering the worst Rx
sensitivity (-31.5 dBm for 8-PSK at 7.5 Gb/s), with a launched
power of 0 dBm per user, a power loss budget 30 dB is
achieved fulfilling the requirements of next generation access
[1]. In terms of channel spacing, we highlight that the 8-(A)PSK
formats, that provide the highest bit rate (7.5 Gb/s), can be fitted
in 6.25 GHz spaced udWDM channels, enhancing the SE and
throughput of the PON.
TABLE I
PERFORMANCE OF DIRECT AMPLITUDE-AND-PHASE MODULATION
Format
Bit rate [Gb/s]
Sensitivity [dBm]
Remarks
BPSK
1.25
-55
Real-time [14]
QPSK
5
-44
[13]
8-PSK
7.5
-31.5
[17]
4-APSK
5
-40
Heterodyne [18]
8-APSK
7.5
-40
[13]
This technique can be directly extended for higher order
modulation formats with more amplitude and phase levels,
depending on the total laser and the signal-to-noise ratio
(SNR) constraints. Moreover, if only phase modulation is
generated, the TX in Fig. 3 only requires a laser and one DAC
channel further reducing the complexity, cost and energy
consumption.
B. Pulse Shaping
The pulse shaping is a key enabling technology to simplify
the complexity of the optical TX, as depicted in Fig 3. It
performs two main functionalities: (I) the proper adaptation of
the modulating current for direct phase modulation with the
laser chirp, and (II) the conventional spectral shaping for
suppression of the out-band modulated signal power.
Regarding the first, it exploits the fact that when a
semiconductor laser is under direct current modulation, the
emission optical frequency varies according to variations
in the optical power , due to the chirp parameter of the
laser. This phenomena was early described in [19], [20], giving
the expression for the optical frequency variation
(1)
where is the linewidth enhancement factor or Henry
coefficient, and is the adiabatic chirp coefficient. The first
term in Eq. (1) is the adiabatic chirp, whereas the second term
is the transient chirp.
For direct phase modulation, the laser is biased at large
values while the current swing for modulation is usually small,
producing variations of the optical power that are significantly
lower than the mean emitted power. In addition, the adiabatic
chirp is dominant at mid frequencies (up to several GHz)
whereas the transient chirp is more apparent at high frequencies
during fast symbol transitions.
Under this conditions, and considering the data BWs for
0733-8724 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2019.2957882, Journal of
Lightwave Technology
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
5
(a)
(b)
(c)
Fig. 5. (a) 50% D-C RZ waveform for direct QPSK modulation; (b) 50% D-C
RZ eye diagram for direct BPSK modulation with (right) and without (left)
Nyquist filtering (roll-off = 0.25); (c) time domain waveforms of (b).
(a)
(b)
Fig. 6. Heterodyning photodetected spectra for BPSK at 2.5 Gb/s and Nyquist
filtering with (a) 0.25 and (b) 1 filter roll-off.
PON, the adiabatic chirp in Eq. (1) prevails and the
semiconductor laser can be modelled as an optical FM
modulator with residual IM, corresponding to the small-signal
approach to the laser rate equations [21]. Hence, for direct phase
modulation lasers with large are optimal to lower the residual
IM. Experimental characterization of a commercial DFB biased
at 80 mA, with modulation current swing of 11 mA
PP
, resulted
in = 10.8 GHz/mW with residual IM of about 1 dB, for direct
BPSK at 1.25 Gb/s. The average emitted power of the DFB was
2 dBm.
To translate the chirp FM modulation into phase modulation
(PM), the desired phase shift is achieved by controlling
both the frequency deviation and the time duration of
the modulating pulse, according to the expression
. As an example, for a reference phase shift of the
product equals to 0.5.
The result is a return-to-zero (RZ) pulse shape with variable
duty-cycle (D-C) that drives the direct phase modulation of the
laser [22], as represented in Fig. 5a for QPSK. Note that for data
symbols with zero phase-shift, no current swing is applied to
the laser during the symbol time , producing zero-crossing in
the eye diagram of the driving signal, as appreciated in Fig. 5b
for direct BPSK modulation.
Furthermore, Nyquist filtering can be applied to reduce the
modulated total spectral width and suppress the unwanted
spectral side-lobes that may cause severe interference to
adjacent channels in the udWDM-PON, with narrow channel
spacing. The eye diagram and time-domain waveform of the
direct-BPSK modulating current are plotted in Fig. 5b and 5c
respectively, for 50% D-C RZ with and without Nyquist
filtering (roll-off = 0.25).
An important remark is that the Nyquist filtering eliminates
the modulation side-lobes, but the filter roll-off has no impact
on the main lobe BW, because of the direct phase modulation
that relies on NL exponential modulation of the laser chirp, with
strong harmonic dynamics [21], [23]. This phenomena can be
appreciated in Fig. 6, that plots the experimental photodetected
spectra for BPSK at 2.5 Gb/s, centered at 10 GHz, for two roll-
off values of the Nyquist filter. The width of the PM modulation
main lobe retains unchanged regardless the filter roll-off, while
the width of the residual IM in the base-band is narrowed since
the IM is generated by a linear modulation.
In a real PON scenario with direct phase modulation, the
benefit of Nyquist spectral shaping becomes apparent when
maximum 15 dB differential optical path losses are allowed, as
specified in [1] for NG-PON2, due to the strong interference of
modulation side lobes on weaker adjacent channels. The results
reported in [23] indicate 25% reduction in channel spacing for
direct phase modulation with Nyquist shaping, allowing that 2.5
Gb/s BPSK users having 15 dB power difference between
adjacent channels can be fitted in a 6.25 GHz spaced optical
grid.
C. Linear Pre-Emphasis
The simplified optical TX for PONs, that employs direct
modulation of low-cost DFB, VCSEL or Dual-EML, might
present severe BW limitation and non-flat frequency response.
Furthermore, to keep the overall cost of the CoTRX down, the
low-cost data converters (DACs and ADCs) might also present
narrow BW and low-resolution, quantified through the effective
number of bits (ENOB).
The linear digital pre-emphasis (DPE) at the TX DSP jointly
mitigates the BW limitation and flattens the frequency response
of the laser, driver amplifier, and DAC. For this purpose, the
well-known zero-forcing (ZF) equalizer
, being
the frequency response of the TX HW, could cancel all
linear distortion and inter-symbol interference (ISI). However,
it presents some disadvantages since it might over-amplify the
noise greatly in the spectral region where is more
attenuated. It might also exhibit large peaks in the time domain
signal that could lead to driver amplifier saturation and
clipping. Instead, a linear DPE calculated trough the minimum
mean squared error (MMSE) between the desired and the real
system output, does not eliminate the interference completely
but minimizes the total power of the noise and the interference
components [24]. Precisely, it is optimal in presence of
significant quantization noise due to DACs/ADCs with low
ENOB, as exhibited by low-cost data converters suitable for
PONs.
The DPE calculated through the MMSE is designed in the
frequency domain as a linear filter following the
analytical derivation in [25], [26], with respect to the TX in Fig.
3, leading to