See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/313031353

Toward a low-cost, low-power, low-complexity DAC-based multilevel (M-ary

QAM) coherent transmitter using compact linear optical field modulator

Conference Paper · January 2017

DOI: 10.1117/12.2257421

CITATIONS READS

0 1,125

1 author:

Benjamin Dingel

78 PUBLICATIONS 580 CITATIONS

SEE PROFILE

All content following this page was uploaded by Benjamin Dingel on 09 August 2018.

The user has requested enhancement of the downloaded file.

 Invited Paper

Toward a Low-cost, Low-power, Low-complexity DAC-based

Multilevel (M-ary QAM) Coherent Transmitter using
Compact Linear Optical Field Modulator
Benjamin Dingel

Nasfine Photonics Inc., Painted Post, NY, 14870, USA. ([email protected]).

ABSTRACT
In this invited paper, we summarize the current developments in linear optical field modulators (LOFMs) for
coherent multilevel optical transmitters. Our focus is the presentation of a new, novel LOFM design that provides
beneficial and necessary features such as lowest hardware component counts, lowered insertion loss, smaller RF power
consumption, smaller footprint, simple structure, and lowered cost. We refer to this modulator as called Double-Pass
LOFM (DP-LOFM) that becomes the building block for high-performance, linear Dual-Polarization, In-Phase-
Quadrature-Phase (DP-IQ) modulator. We analyze its performance in term of slope linearity, and present one of its
unique feature --- a built-in compensation functionality that no other linear modulators possessed till now.

Keywords: Linear optical field modulator (LOFM), multilevel coherent transmitters, Mach Zehnder Interferometer
(MZI) modulator, Sin-Sin2 Linearization

1. INTRODUCTION
In the emerging optical transport networks operating at 100/400 Gbit/s and beyond, there are growing demands to
network providers and data center providers to meet the never-ending network challenge of “increased-performance-
at-lower-cost”! Performance is defined as getting the highest spectral efficiency (SE) at the longest possible
transmission reach [1-3] per WDM wavelength channel across an established fiber infrastructure. To meet the
performance side of the challenge, there is a preferred choice to use advanced modulation formats in conjunction with
digital signal processing (DSP) and coherent detection technologies.
Advanced modulation formats are realized with the use of different multilevel coherent transmitters, especially
quadrature amplitude modulation (M-ary QAM)-driven modulators. They support the emerging needs for adaptable
rate, variable-modulation format, and selectable bandwidth [1-3]. However, since the Mary QAM modulator requires
multilevel input signal, this leads to (i) higher hardware complexity, (ii) difficult fabrication, (iii) higher power
consumption, and (iv) sophisticated circuit designs. These difficulties are further compounded by the emerging
standards requiring (a) smaller form factors, (b) lower total system power consumption, and (c) increase port density
than the existing Optical networking Forum (OIF) [4]. All these technical and standard issues are increasing the overall
cost compared with typical binary driven optical transmitters. Thus, these are weighing against the expected gains in
the system performance.
This implies big challenge toward fulfilling the cost side of the challenge. There are three elements to this cost
challenge, namely: (1) fabrication, (2) design, and (3) linearization issue. As a key step and still ongoing development,
the component suppliers, network providers, and data center providers are lowering their equipment power
consumptions and shrinking their footprint by using innovative monolithic or hybrid fabrication technologies such as
InP, Silicon Photonic Integrated Circuits (PICs), etc. [5-6]. These fabrication approaches are very important strategies
to increase the overall port density, reduce chip size, and lower the total cost.
The second element to cost reduction calls for innovative multilevel modulator design. Presently, the current
designs of coherent optical modulators are complex [7-8] especially for electrical driving signal level greater than 4
(or 4 bits are mapped to one symbol). Even if the current modulator designs can be miniaturized and fabricated with
PICs, they still involve too many active components that eventually increase the RF power consumption and occupy
a larger footprint. Thus, the design of multilevel modulators needs serious “complexity reduction”. At present, the
search for efficient and cost effective coherent QAM modulator design is still on-going.
The third element to cost reduction is the search for creative solutions to address the issue of linearization of
optical multilevel modulators which was not a problem before in the binary-level driven modulator [9-12].
Linearization of these modulators has become necessary, especially for 100/400 Gbit/s optical communication system.

Next-Generation Optical Communication: Components, Sub-Systems, and Systems VI,

edited by Guifang Li, Xiang Zhou, Proc. of SPIE Vol. 10130, 101300O · © 2017 SPIE
CCC code: 0277-786X/17/$18 · doi: 10.1117/12.2257421

Proc. of SPIE Vol. 10130 101300O-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Without linearization, the modulator’s resolution and effective dynamic range would be reduced. On the other hand,
current linearization approaches lead to increased hardware component counts, higher insertion loss, and greater RF
power consumption which clearly have negative impact on meeting the stricter standard requirements.

1.1 Goal and Organization

In this invited paper, we present a low-cost, low-power, low-complexity DAC-based Multilevel (Mary QAM)
coherent transmitter/modulator using a novel linear optical field modulator (LOFM) [13]. It promises to provide a
balanced approach to the issues of modest hardware component counts, lowered insertion loss, and smaller RF power
consumption that optical component standards demand.
This paper is organized as follows. In section II, we briefly review the different schemes to generate multi-level
coherent optical transmitters. In section III, we focus on one type of coherent multilevel transmitter/modulator called
Coherent Multilevel Dual-Polarization IQ (DP-IQ) Modulators, discuss its nonlinearity issue, and summarize the
different current approaches to mitigate this problem. In section IV, we present our Double-Pass DP-IQ linear optical
field modulator (LOFM) [13], analyze its performance in term of slope linearity, and present one of its unique feature
--- built-in compensation functionality before giving our conclusion in section V.

2. Different Types of Coherent Multilevel Modulators/Transmitters

In this section, we summarize the different coherent multilevel transmitters that have been proposed or
demonstrated [14]. These transmitters consist of CW laser combined with some external modulator arrangements
consisting of mix of intensity modulator (IM), phase modulator (PM), or In-Phase-Quadrature-Phase (IQ) modulator.
These modulators are summarized in Fig. 1. These modulators can be grouped into two general classes, namely; (1)
Binary-Drive Cascaded/Nested Multilevel Modulators, and (2) Alternative Multilevel Modulators. The second
class can be divided further into (2a) Direct-drive Segmented Modulators, and (2b) Digital-Analog-Converter (DAC)-
based Multilevel Modulators.

2.1 Binary-Drive Cascaded/Nested Multilevel Modulators

Binary-Drive Cascaded/Nested Multilevel Modulators class has been intensely researched and demonstrated
[15-27]. This class of modulators is still the most commonly used when driving signal level requirement is less than
or equal to 4 or 4 bits are mapped to one symbol. In theory, higher order M-QAM signal can be synthesized from
binary electric signals by using multiple nested MZ modulators or cascade of MZMs. Structurally speaking, it
combines two or more conventional binary-drive optical modulators under proper power scaling to generate the
corresponding multilevel optical signal. There are numerous variations of the basic implementation which can be
grouped depending on particular optical configurations or combinations involving PM, IM, and IQ modulators.
Types of Coherent Multilevel Optical Modulators

Binary-Drive Cascaded/ Alternative Multilevel

Nested Multilevel Modulators Modulators

DAC-based
or Direct Drive Multilevel DAC-based Multilevel
Not? Modulators MZ Modulators

IQ Modulator + PM Segmented MZ DP-IQ Modulators

IQ Mod. + PM + PM Modulator
What are the
Segmented MRR Single DDMZM
actual optical MZM + IQ Mod
implementations? Modulator
Serial MZM plus
Quad IQ Modulator AWG-based PM
others Generalized MZM

Is linearization
required ? No YES YES

Component Few
Many (for level N >4) Moderate
Counts?

Technology Present solution for Near- to-Mid-

Long Term solution Term solution
solution? signal level N < 4
DAC – Digital Analog Converter MZ –Mach Zehnder AWG—Array Waveguide Grating MRR—Microring Resonator

Fig. 1: Different types of optical coherent multilevel modulators

Proc. of SPIE Vol. 10130 101300O-2

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fig. 2 shows some modulator configurations that generate 16-QAM constellations. These configurations can
either be in discrete or integrated form. More detailed information on these types of modulators can be found in [15-
27]. Clearly, the tradeoff is between using matured, easy-to-handle electrical binary input signals in exchange for more
complicated and “bulky” optical modulator design. This approach has many advantages. First, it can generate, operate,
and add-up, high-speed, electrical binary-level modulations comparatively easier than high-speed multilevel
electronics modulation. Second, the electronic device and components dealing with generation and modulation of
binary data streams are well established. Third, the driving-signal level of each symbol is more stable (or has lower
fluctuation) due to the rectification characteristics of the binary-driven MZM. Lastly, it does not have linearization
issues so there is no need for additional component like electrical DACs and linear modulator drivers.
However, this class of multilevel modulators becomes troublesome when the multilevel signal requirement
becomes higher than 4. This leads to the following disadvantages, namely: (1) more complicated optical circuitry due
to increased component counts, (2) challenging fabrication difficulties, especially for the integrated modulators, (3)
larger device footprint, (4) higher electrical/RF power consumption, and (5) increased overall cost.
(a) U1 (c) U1
U3 (f) I1
MZM U3 U4
In MZM PM
PM In MZM
Out In
MZM π/2 Out Q1 I2
MZM
Out
PM π/2
U2 MZM π/2
U4
U2
Q2

U1
(b) (d) U3
I1

MZM (g) MZM

U3 U4 Out
In U1 MZM
PM PM In Out
MZM MZM π/2

MZM π/2 In Q1 Out

U2 MZM π/2 I2

U2 U4
MZM 6 dB

MZM π/2

Q2

(e) U1 U3

MZM MZM
In

Out
MZM π/2 MZM π/2

U2 U4

Fig. 2: Different configurations of Binary-level driven multiple modulators. (a) Conventional single-drive IQ modulator plus serial PM, (b)
IQ modulator plus two serial PMs, (c) Modified IQ modulator with two internal PMs, (d) MZM plus IQ modulator, (e) Cascaded IQ modulators,
(f) Double-drive IQ modulator, and (g) Quad-Parallel MZMs. Modulators in the gray boxes imply integrated modulators

2.2 Alternative Multilevel Modulators

As shown in Fig. 1, alternative multilevel modulator has two types namely; (1) Direct drive Segmented
modulators, and (2) DAC-based multilevel modulator. Both types of modulators require some form of linearization
to correct the nonlinearity in their respective output signals.

2.2.1 Direct-drive Segmented Modulator

Direct-drive segmented modulator modifies the design of the conventional optical modulator in which the
typically long, RF electrode in the arms of conventional Mach-Zehnder modulator (MZM) is replaced by multiple
shorter RF electrodes. Then, these segmented electrodes are driven directly by the parallel data signals (with binary
level) to generate the multilevel optical signal. Fig. 3 shows three different demonstrated implementations of this type
of modulators namely; (1) multi-electrodes MZM [28], (b) multi-electrode microring resonator (MRR) modulator [29],
and (3) multi-electrode coupling coefficient MRR [30]. One immediate benefit of this general approach is the absence
of DAC components. These modulators are promising candidates for future multi-level coherent modulators since
they combine the (i) low complexity binary driving circuits, and (ii) low complexity optical modulator design.

(a) (b) (c)

N-bit Datain
N-bit Datain
Electrodes Electrodes
Electrodes
In Out
Electrodes
Electrodes

In Electrodes Electrodes Out

In Out
N-bit Datain
N-bit Datain

Fig. 3: Different configurations of Binary-level Direct-driven Segmented Modulator. (a) multi-electrodes, segmented Mach-Zehnder
Modulator, (b) multi-electrodes, segmented Ring Resonator Modulator, and (c) multi-electrode coupling coefficient microring resonator
(MRR).

Proc. of SPIE Vol. 10130 101300O-3

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 However, this technology is still in the early stage of development and at present, is considered as a long term
solution. It also require linearization to correct the inherent nonlinearity in the main MZM or MRR structures.

2.2.2 DAC-based multilevel modulators

It is reasonable to expect that a compromised solution in many cases can be found somewhere between the current
Binary-Drive Multiple Modulators, and the future-bound Direct-driven Segmented Modulators. The challenge is to
find the right mixture of complexity between the electronic drive circuitry and the optical circuit design.
DAC-based multilevel modulators class is an excellent compromised solution at this point in time. The basic idea
of the design approach is to first convert and synthesize the RF multilevel signal from serial or parallel binary data
sequence signals into a single analog multilevel electrical signal using Digital Analog Converter (DAC) or Arbitrary
Waveform Generator (AWG) before it is injected into the optical MZ modulator as an RF drive signal. Thus, the role
of the optical modulator is to insure that the electrical-to-optical conversion process is as linear as possible [9-13].
Binary data Binary data
Binary data Binary data

DAC DAC
IV1 I
DAC
RFin DAC
RFin
MZM
In In In
Out Out
MZM PM MZM π/2
V2
DAC Q
DAC

Binary data Binary data

(a) DD-MZM (b) Serial MZM plus PM (c) Conventional IQ Transmitter
Fig. 4: Three different types of multilevel coherent optical modulators capable of generating Square 16-QAM constellation, (a) Dual-Drive
Mach-Zehnder Modulator (DDMZM) with arbitrary M-level, (b) Serial MZM plus PM that generates 3-level amplitude modulation (MZM),
and 12-level phase modulation (PM), and (c) Conventional IQ Modulator with standard 4-level for each of the arms of the main MZM.

Fig. 4 shows some proposed/demonstrated configurations [34-35] that can generate 16-QAM constellations based
on this approach. These configurations range from (i) simple, single device (Dual-Drive, Mach-Zehnder Modulator,
DD-MZI) having 16 electrical driving levels as shown in Fig. 4a, to standard IQ modulator comprising of two sub-
MZMs each having 4 electrical driving signal levels, as shown in Fig. 4c. Compared with dual-drive MZM
configuration (Fig. 4a) and Serial MZM and PM (Fig. 4b), the IQ modulator (Fig. 4c) can be used with lower number
of electrical driving signals which reduces greatly the complexity of electrical driving circuits.

3. Coherent Multilevel Dual-Polarization IQ (DP-IQ) Modulators

Given today’s technology, a good compromised multilevel coherent modulator possesses the right mixture of
complexity between the electronic drive circuitry and the fabrication of nested optical modulators. The Dual-
Polarization, In-Phase-Quadrature-Phase (DP-IQ) modulator (under the DAC-based multilevel modulators sub-
group) appears to be the most attractive configuration that offers excellent balanced features [36]. At present, DP-IQ
modulators with 4 to 8 sub-MZMs have been demonstrated using monolithic or hybrid technology. Although higher
level of integration (>8 sub-MZMs) is possible, it appears that the “sweat spot” is with 4 sub-MZMs.
Binary data Binary data

DAC DAC
V1 V1
I1
Table I: Advantages and Challlenges of DP-IQ Modulator
MZM
Advantages Disadvantages

P-Polarization MZM π/2 Actual integrated modulator is available DAC or AWG is currently limited to
In Q1 Out and its fabrication is also well established < 20Gbaud/sec.
I2
S-Polarization MZM 6 dB Uses matured DACs as electrical signal Linearization in DP-IQ modulator
(g)
aggregator or needs to be addressed
MZM π/2 Uese very flexible electronic AWG to
generate various modulation formats.
Q2
Has smaller chirp value compared with
DAC DAC
other configurations
Binary data Binary data

Fig. 5: Schematics of a conventional IQ optical modulators together with a Table showing its advantages and challenges

Proc. of SPIE Vol. 10130 101300O-4

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 A schematic of an optical DP-IQ modulator is shown in Fig. 5. It consists of 4 sub-MZMs, where each of the sub-
MZMs corresponds to I and Q components of the two orthogonal polarizations, respectively. DP-IQ modulator has
been used extensively in Binary-Driven Multiple Modulators class, although the signal input level is restricted to
binary level. Now, each sub-MZMs is driven with a multilevel electronic signal generated by a currently available
DAC. Table I summarizes the advantages and the technical challenges related to DP-IQ modulators.
3.1 Nonlinearity Problem with DP-IQ Modulators
A significant challenge to coherent multilevel modulator-based on DP-IQ is the linearization of its output signal.
Linearization is all about achieving low-distortion mapping of the electronic multilevel signal into the optical
multilevel signal [9-13]. The critical source of the distortion is due to the nonlinear sinusoidal optical field response
of each of the 4 sub-MZMs. The important focus and new aspect of the problem is not the usual linearization of
intensity response of MZM that has been investigated for decades [37-46] but the linearization of its optical field
response!
Many different approaches have been proposed in literature to mitigate this nonlinear distortion effects, but with only
moderate successes. The first approach limits the peak-to-peak range of the electric signals to a small interval (around
the null or quadrature points) of the transfer sinusoidal function to achieve the linear E/O conversion. However, the
penalty of this approach is increased optical losses and limited use of the entire dynamic range of the modulator. The
second approach develops an inverse electrical arcsine circuit to pre-distort the input RF electrical signal [47-49].
However, its effectiveness is limited by the electrical bandwidth of the resultant arcsine circuit. It also reduced the
effective resolution of the DAC. The third approach replaces each of the 4 standard sub-MZMs with linear optical
field modulators (LOFMs) [9-13].
3.2 Optical Linearization Approach: Sin-Sin2 linearization
Fig. 6a shows the basic principle of “Sin-Sin2 linearization” to produce the LOFM. It requires the generation and
subsequent combination of an electric field signal, sin[θ]with an attenuated electric field signal, α∗sin[bθ] under
appropriate amplitude scaling (1-r: r). The role of the second term, -α sin[bθ], is to compensate for the nonlinearity
in the scaled sin[θ] signal [9-13].
DATAin AWG 1.0
Divider Inverter Electric Field Amplitude
or DAC Parameter:
r=0.12 √r Sin[2θ]
(a) Vbias V(t)
Vbias -V(t) (b) 0.5

Εin Εout Cos[x] Sin[2x]

Out 0.0
E©out

Sin(θ) In
Vbias V(t)
-Sin(2θ) Vbias -V(t) Sin[x] Sin[x] - 0.5
√(1-r) Sin[θ]
Inverter Divider - 1.0
- 1.5 - 1.0 - 0.5 0.0 0.5 1.0 1.5
Driving Voltage, V(t)/Vπ

(a) (b) (c)

Fig. 6: Generic model of the optical field linearization scheme (a), standard implementation 2 MZM in series with Y-branch combiner [x] (b),
and the electric field response (c).

Fig. 6b shows its standard implementation [9-12] using two-stage lattice configuration. It consists of (i) a dual-
output, dual-drive MZM, (ii) a single-output, dual-drive MZM, (iii) an asymmetric coupler, and (iv) two RF signal
sources (+V and −V, or one RF signal with RF divider and inverter). The two MZMs function in push-pull operations
for chirp-free performance. The input signal, Ein is injected onto input port#1 of MZM (with appropriate biasing
condition) to generate the output signals E3 = Ein∗cos[θ], and E4 = Ein ∗sin[θ] at its output ports, respectively. Then
signal E3 is injected to the second MZM to generate an output signal, E11 = Ein cos[θ]sin[θ ]=Ein sin[2θ ]/2 while
signal E4 is injected into a delayed waveguide (DW) combined with E11 by the Y-branch combiner so that the resultant
electric field output signal, Eout is given as

E ( A) OUT r
= 1 − r Sin[θ ] − Sin[ 2θ ] (1)
E in 2

Proc. of SPIE Vol. 10130 101300O-5

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 where Ein is the incident electric field input signal, and (1-r: r) is the power split ratio of the asymmetric coupler. The
term θ is the phase of MZM which is equal to πV(t)/(2Vπ) with V(t)= Vm cos[ωmt], where Vm is the amplitude voltage
of the radio frequency (RF) or Data driving signal, ωm is the angular frequency, Vπ is the half-wave voltage of the
MZM. The excellent linearity results of the electric field as function of the RF voltage is depicted in Fig. 6c. More
detailed discussion is given in [9-12].

3.3 Problems with Standard Implementation

Although the standard implementation of “Sin-Sin2 linearization” is straightforward and produces excellent
linearity, it unfortunately has clear drawbacks. First, it increases the device’s complexity by doubling the 4 sub-MZM
units into 8 units for an optical signal with two orthogonal polarizations. Second, it doubles its footprint. Third, it
increases the overall RF power consumption due to the additional RF/electrical powers needed to drive and bias the
added sub-MZMs. Fourth, it requires potentially more RF components such as an RF signal divider and an inverter
that contribute to the overall RF losses that might lead to the use of an external RF amplifier. Fifth, it has no inherent
compensation functionality that would be essential, especially to mitigate unavoidable fabrication error.
Eventually, all these important factors increase the overall cost of coherent transmitters and go against the current
trends imposed by the new more, stricter OIF standards.

4. DOUBLE-PASS LINEAR OPTICAL FIELD MODULATOR (DP-LOFM)

4.1 Novel Implementation
Recently, we reported a different implementation [13] of Sin-Sin2 linearization that achieved the required linearity
but without its disadvantages. The new implementation is shown in Fig. 7a. It is based on double-pass modulation
(DPM) to generate the required -α*sin[2θ] signal. Thus, we will refer to it as Double-Pass LOFM (DP-LOFM).
Unlike the standard LOFM [9-12], our design employs only (i) a single, dual-drive MZM configured as dual-
input/dual-output-port using two directional couplers (DC). The input/output ports of the two DC are labelled as 1, 2,
3, and 4. It also includes (ii) an Optical Delay Module, (ODM), (iii) an asymmetric coupler, and (iv) single pair of
RF signal source.

DATAin AWG 1.5

Divider
Electric Field Amplitude

or DAC Best Linear Fit

(b) 1.0
(a)
Vbias Cos[x] M 0.5 Proposed
Ein 1 V(t)
3
LOFM
0.0 Parameter:
2 4 Delay Line, L
EOut E(P)out r=0.113
Sin[2x] - 0.5
Vbias -V(t) Sin[x]
Sin[θ],
- 1.0 Typical MZM
Sin[x]
Inverter - 1.5
- 1.5 - 1.0 - 0.5 0.0 0.5 1.0 1.5
Driving Voltage, V(t)/Vπ
Figure 7: Design of (a) DP-LOFM , and (b) Electric field transfer function

Briefly speaking, the input signal, Ein is injected onto input port#1 of MZM to generate the output signals E3=
Ein*cos[θ], and E4= Ein*sin[θ] at port #3 and port #4, respectively. At end-surface of port#3, a delay module (DM),
(comprising of a “mirror” M and delay line L) is positioned to reflect the E3 signal back into the MZM to generate an
output signal, E2 = Ein α1 cos[θ]sin[b1θ + ϕ] exiting at port#2. Then signal E4=Ein*sin[θ] is looped back toward port#2
using an extended waveguide to combine with E2 using the Y-combiner with a split ratio (1-r:r) to give an output
signal, E(P)out as

E ( B ) OUT
= 1 − r Sin[θ ] + a1 r Cos[θ ]Sin[ θ + ϕ ] . (2)
E in

where a1 is the power reflectance of mirror M, b1(ω) is a frequency-dependent ratio between the backward to forward
MZM modulation indexes, and ϕ is the phase delay of the modulating signal as seen by the reflected E2 signal [50-
53]. The parameters b1 is set to 1. The physical argument and arrangement related to this assumption is discussed in

Proc. of SPIE Vol. 10130 101300O-6

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 ref [52-53]. On the other hand, parameter ϕ is set to +/− π by positioning mirror M at the end-surface of port #3 and
using a flat mirror so that the resultant signal E(P)out becomes

E ( P ) OUT r
(3)
= 1 − r Sin[θ ] − a1 Sin[ 2θ ]
Ein 2

which is similar to Eq. (1) except for the parameter a1.

4.2 Immediate Benefits

The immediate benefits of DP-LOFM is the reduction of power consumption by one-half compared with standard
LOFM. It needs only one pair of V(t) and −V(t) signals. Compared with Fig. 6b, DP-LOFM has fewer RF components,
lower RF power consumption, single MZM, and simple device structure that all lead to compact modulator chip, and
simplified fabrication.

4.3 Linearity Analysis of the Slope

In ref [9-12], the linearity of a standard implementation of Sin-Sin2 linearization was analyzed using SFDR. To
complement this SFDR evaluation, we analyze our reported implementation in terms of slope linearity of the electric
field response. We express the electric field given by Eq. (3) as series of the higher-order harmonic terms of
∆θ = θ − θ0 using a Taylor expansion around point θ0 =0, such that

(θ ) 2 n +1 ( 2θ ) 2 n +1 (4)
E OUT # 3 (θ ) = 1 − r ∑ (−1)
n =0
n

( 2 n + 1)!
− a1 r ∑ n =0
( −1) n
( 2 n + 1)!

Note that there are no even-order harmonic distortion terms since (2n +1) is always odd number. Neglecting 7th and
higher-order distortions, we get the first three normalized coefficient terms as:

T1= (√(1-r) – a1√r), (5)

T2= (-√(1-r) + 8a1√r)/3! and (6)

T3=(√(1-r) -32a1√r)/5! (7)

The difference between field Eout#3(θ) and its best linear fit line, f3(θ) is quantified by the normalized deviation, (ND)
which is defined as:

ND = Eout #3 (θ ) − f 3 (θ ) /( E max − E min ) (8)

where Emax and Emin are the respective maximum and minimum amplitude of the output electric fields when θ=θmax
and θ=θmin respectively. The term (Emax - Emin) is approximately around 2. In general, optical field response with less
than 1% peak-to-peak ND value, (NDp-p) is considered to have excellent linearity [54-55].

1.5 (b) 10
Norm. Deviation, ND (in %)

(a) Best 0.06

Electric Field Amplitude

Linear Fit ND = 2.1% (c)

ND = 0.73%
Eout (θ)- f3(θ)

1.0
0.04
(r=0.113) (r=0.1) 8
(rsub-opt =0.0588)
0.5 0.02
6
0.0 0.00
- 0.02
4
- 0.5 (ropt =0.113)
- 0.04 ND = 7.25% 2
- 1.0 (r=0.0588)
- 0.06
0
- 1.5
- 1.0 - 0.5 0.0 0.5 1.0
0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
- 1.5 - 1.0 - 0.5 0.0 0.5 1.0 1.5
Driving Voltage, V(t)/Vπ Driving Voltage, V(t)/Vπ Power Split Ratio, (r)

Fig. 8. The field response, best linear fit line, (a) and corresponding normalized deviation, ND (b) when r= 0.0588 and when r=0.113.

Proc. of SPIE Vol. 10130 101300O-7

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 Fig. 8a shows the electric field response of Eout#3(θ) for two values of r with a1=1, and best linear fit line f3(θ). On
the other hand, Fig. 8b shows the actual Eout – f3(θ) electric field difference for (1) r=0.113 (optimum value), (2)
r=0.0588 (sub-optimum value) and r=0.1. Finally, Fig. 8c depicted the calculated NDp-p value as a function of r.
There are two notable observations from Fig. 8. First, there is mild agreement between Eout3 (θ) and f3(θ) for rs−opt
=0.0588 as seen in Fig. 8b. The rs−opt is obtained when the 3rd-order harmonic distortion coefficient, T3 is equated to
zero. This is a common procedure in current intensity linearization approaches. However, this criteria is not
satisfactory for optical field linearization as reflected by the resultant NDp−p value of 7.25% shown in Fig. 8b and 8c.
The remaining higher-order harmonic terms (like the 5th-order and 7th-order) cannot be neglected since they contribute
largely to the distortion.
Second, there is an excellent agreement between Eout3 (θ) and f3(θ) for ropt = 0.113 which is also verified by the
associated excellent NDp−p value shown in Fig. 8b. The optimum value is obtained when the 3rd-order distortion
term is negated by the 5th order distortion (and other higher-orders) since they have opposite signs. This process
eventually minimizes the overall higher-order harmonic coefficients and gives lowest NDp−p value. As shown in Fig.
8b the calculated NDp−p value is reduced from 7.25% to superior 0.73% and is the lowest attainable value as depicted
in Fig. 8c.

4.4 Effect of Non-Optimum value of r and its MITIGATION

Fig. 8c reminds us the importance of parameter r in getting the best linearity performance for both LOPM and
DP-LOFM. However, we also know that fabrication error is unavoidable and the initial value of r is hard to precisely
set, and may not be precisely fabricated at its optimum value. Thus, it is highly desirable that any LOFM should have
a built-in compensation feature.
0.10
C: (r=0.2, a1=1)
NDp-p = 9.2%
B: (r=0.15, a1=1)
NDp-p = 3.8%
Eout (θ)- f3(θ)

0.05
C’: (r=0.2, a1=0.6)
NDp-p = 0.6%

0.00

A: (r=0.113, a1=1)
NDp-p = 0.73%
- 0.05 B’: (r=0.15, a1=0.81)
NDp-p = 0.8%

- 0.10
- 1.0 - 0.5 0.0 0.5 1.0
Driving Voltage, V(t)/Vπ
Fig.9. The positive effect of parameter a1 as a mitigating factor when r is not optimized.

One unique feature of DP-LOFM is its inherent compensation functionality. Fig. 9 depicts the sensitivity of the
calculated NDp−p on parameter a1 for different values of r (ropt = 0.113, r = 0.15 and r = 0.2) under lossless (a1 = 1)
and lossy (a1 = 1) conditions. It shows the positive effect of parameter a1. When there is no loss, the calculated NDp−p
value increases to 3.8 % when r = 0.15 (Curve B; green solid line), and to 9.2 % when r = 0.2 (Curve C; blue solid
line), compared to the optimum value of 0.73 % when r = 0.113 (Curve A, red solid line). To mitigate this increased
distortion, we can simply introduce loss into the mirror M that lowers its value from a1 = 1 to a1 = 0.81 for curve B’
(r=0.15, green dotted line) and a1 = 0.6 for curve C’ (r=0.2, blue dotted line).
Clearly, with this approach, we maintain the effective NDp−p value to be less than 0.8%. Basically, this can be
understood easily if we look at the second term in Eq. (3). Here, the decrease in the value of parameter a1 compensates
for the increase in the value of √r when r is greater than its optimum value. This positive feature on parameter a1 is
very important since fabrication error is unavoidable. It functions as an externally controlled compensating feature to
optimize the function of the device.

Proc. of SPIE Vol. 10130 101300O-8

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 4. CONCLUSION
We presented a new linear optical field modulator (LOFM) for Digital-to-Analog (DAC)-based multi-level
coherent transmitter. It main benefits are reduced complexity of the modulator design, lower power consumption, and
smaller form factor which are all not found in current implementation of LOFM. These requirement are also very
important especially with the stricter optical component OIF standards. The new modulator uses only one MZM which
is configured for double-pass modulation instead of two cascaded MZM units. It has excellent linearity with a
normalized deviation value of less than 0.8% when r=0.113. Another unique feature of this modulator is its built-in
linearity compensating feature which is essential, especially when parameter r cannot be precisely set during
fabrication. Externally adjusted parameter a1 allows the nonlinear distortion to be mitigated.
REFERENCES
[1] Seimetz, Matthias. "Multi-format transmitters for coherent optical M-PSK and M-QAM transmission." Transparent Optical
Networks, 2005, Proceedings of 2005 7th International Conference. Vol. 2. IEEE, 2005.
[2] Geisler, David J., et al. "Bandwidth scalable, coherent transmitter based on the parallel synthesis of multiple spectral slices
using optical arbitrary waveform generation." Optics express 19.9 (2011): 8242-8253.
[3] Nakazawa, Masataka, Kazuro Kikuchi, and Tetsuya Miyazaki, eds. High Spectral Density Optical Communication
Technologies. Vol. 6. Springer Science & Business Media, 2010.H.
[4] Optical networking Forum (OIF), Technology Options for 400G Implementation, OIF-Tech-Options-400G-01.0 July 2015.
[5] Rang-Chen (Ryan) Yu, “Cost Efficient 100G/400G Optical Transceiver Solutions”, OFC’2014 Market Watch March 13, 2014
[6] Chen, L., et al. "Silicon photonics for 100G-and-beyond coherent transmissions." Optical Fiber Communications Conference
and Exhibition (OFC), 2016. IEEE, 2016.
[7] S. Mino, H. Yamazaki, T. Goh, and T. Saida, “Transmitter Synthesis for Advanced Modulation Formats Utilizing Silica-
LiNbO3 Hybrid Integration Technology”, (OFC/NFOEC), 2013, OM3C.3, pp. 1 – 3.
[8] T. Sakamoto, and A. Chiba, “Coherent Synthesis of Optical Multilevel Signals by Electro optic Digital-to-Analog Conversion
Using Multi parallel Modulator”, IEEE J. OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 16, NO. 5, 2010, pp.
1140-1149.
[9] Yamazaki, Hiroshi, et al. "Linear optical IQ modulator for high-order multilevel coherent transmission." Optical Fiber
Communication Conference. Optical Society of America, 2013.
[10] A. Kaneko, H. Yamazaki, Y., “Linear Optical Modulator”, Optical Fiber Communication Conference (OFC/NFOEC), 2014,
W3K.5, pp. 1 – 3.
[11] Yamazaki, Hiroshi, et al. "Optical Modulator With a Near-Linear Field Response." Journal of Lightwave Technology 34.16
(2016): 3796-3802.
[12] Yamazaki, Hiroshi. "Linear optical modulator for DAC-based coherent fiber communications systems." 2014 IEEE Compound
Semiconductor Integrated Circuit Symposium (CSICS). IEEE, 2014.
[13] Dingel, Benjamin. "Reduced Complexity, Low-Power Linear Modulator for DAC-Based Multilevel Coherent Transmitters."
IEEE Photonics Technology Letters 28.7 (2016): 717-720.
[14] Al-Bermani, Ali, and Reinhold Noé. "Comparison of cost and complexity for various 16-QAM transmitter structures in
coherent optical systems." SPIE OPTO. International Society for Optics and Photonics, 2016.
[15] M. Seimetz, "Multi-format transmitters for coherent optical M-PSK and M-QAM transmission," in Proc. ICTON,Th.B1.5,
Barcelona, 2005.
[16] G.-W. Lu, T. Sakamoto, A. Chiba, T. Kawanishi, T. Miyazaki, K. Higuma, M. Sudo, and J. Ichikawa, "Monolithically
Integrated Quad Mach-Zehnder IQ Modulator for Optical 16-QAM Generation," in Conference on Lasers and Electro-Optics,
OSA Technical Digest (CD), paper CPDA7, 2010.
[17] L. Zhang, X. Hu, Y. Wu, Q. Liu, T. Wang, and Y. Su, "Multiple 16QAM signals generation at 40Gbit/s using a novel
transmitter," in Proc. 15th Asia-Pacific Conference on Communications (APCC), 2009.
[18] G.-W. Lu, M. Sköld, P. Johannisson, J. Zhao, M. Sjödin, H. Sunnerud, M. Westlund, A. Ellis, and P. Andrekson, "40-Gbaud
16-QAM transmitter using tandem IQ modulators with binary driving electronic signals," Optics Express, vol. 18, no. 22, pp.
23062-23069, 2010.
[19] M. Seimetz, "Transmitter design," in High-order modulation for optical fiber transmission, Springer, 2009, p. 28.
[20] X. Zhou, and J. Yu, "200-Gb/s PDM-16QAM generation using a new synthesizing method," in Proc. ECOC 2009, paper
10.3.5, 2009.
[21] H. Y. Choi, T. Tsuritani, and I. Morita, "Optical transmitter for 320-Gb/s PDM-RZ-16QAM generation using electrical
binary drive signals," Optics Express, vol. 20, no. 27, pp. 28772-28778, 2012.
[22] M. Secondini, E. Forestieri, and F. Cavaliere, "Novel optical modulation scheme for 16-QAM format with quadrant
differential encoding," in Proc. Photonics in Switching, 2009.
[23] S. Yan, D. Wang, Y. Gao, C. Lu, A. Lau, Y. Ji, L. Liu, and X. Xu , "Generation of square or hexagonal 16-QAM signals
using a single dual drive IQ modulator driven by binary signals," in Proc. OFC/FOEC, paper OW3H.3, 2012.
[24] A. Malacarne, F. Fresi, J. Klamkin, and L. Poti, "Versatile offset-free 16-QAM single dual-drive IQ modulator driven by
binary signals," Optics Letters, vol. 37, no. 19, pp. 4149-4151, 2012.

Proc. of SPIE Vol. 10130 101300O-9

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx

 [25] T. Sakamoto, A. Chiba, and T. Kawanishi, "50-Gb/s 16 QAM by a quad-parallel Mach-Zehnder modulator," in Proc.
European Conference on Optical Communication (ECOC), PDP2.8, 2007.
[26] T. Sakamoto, A. Chiba, and T. Kawanishi, "50-km SMF transmission of 50-Gb/s 16 QAM generated by quad-parallel
MZM," in Proc. of European Conference on Optical Communication (ECOC), Tu.1.E.3, 2008.
[27] L. Lei, Y. Yu, F. Lou, Z. Zhang, L. Xiang, and X. Zhang, "A simple experimental scheme for M-QAM optical signals
generation," Frontiers of Optoelectronics, vol. 5, no. 2, pp. 200-207, 2012.
[28] Ehrlichman, Yossef, Ofer Amrani, and Shlomo Ruschin. "Improved digital-to-analog conversion using multi-electrode Mach–
Zehnder interferometer." Journal of Lightwave Technology 26.21 (2008): 3567-3575.
[29] Ehrlichman, Yossef, Ofer Amrani, and Shlomo Ruschin. "Direct-Digital-Drive Microring Modulator." arXiv preprint
arXiv:1603.03802 (2016).
[30] Karimelahi, Samira, and Ali Sheikholeslami. "Quadrature amplitude modulation (QAM) using binary-driven coupling-
modulated rings." Optics Communications 366 (2016): 354-361.
[31] Yamazaki, Hiroshi, et al. "Optical Modulator With a Near-Linear Field Response." Journal of Lightwave Technology 34.16
(2016): 3796-3802.
[32] Kaneko, Akimasa, Hiroshi Yamazaki, and Yutaka Miyamoto. "Linear optical modulator." Optical Fiber Communication
Conference. Optical Society of America, 2014.
[33] Yamazaki, Hiroshi. "Linear optical modulator for DAC-based coherent fiber communications systems." 2014 IEEE Compound
Semiconductor Integrated Circuit Symposium (CSICS). IEEE, 2014.
[34] Ho, K.P., Cuei, H.W, "Generation of arbitrary quadrature signals using one dual-drive modulator," IEEE Journal of
Lightwave Technology , vol. 23, no. 2, p. 764–770, 2005.
[35] M. Seimetz, "Multi-format transmitters for coherent optical M-PSK and M-QAM transmission," in Proc. ICTON, Th.B1.5,
Barcelona, 2005.
[36] S. Yan, D. Wang, Y. Gao, C. Lu, A. Lau, Y. Ji, L. Liu, and X. Xu , "Generation of square or hexagonal 16-QAM signals
using a single dual drive IQ modulator driven by binary signals," in Proc. OFC/FOEC, paper OW3H.3, 2012.
[37] Djupsjobacka, A., “A Linearization Concept for Integrated-Optic Modulators,” IEEE PTL, 4(8), 869-872 (1992).
[38] L. M. Johnson and H. V. Roussell, “Reduction of intermodulation distortion in interferometric optical modulators,” Opt. Lett.,
vol. 13, no.10, pp. 928–930, 1988.
[39] S. K. Korotky and R. M. Ridder, “Dual parallel modulation schemes for low-distortion analog optical transmission,” IEEE J.
Sel. Areas Commun., vol. 8, no. 7, pp. 1377–1381, Sep. 1990.
[40] B. B. Dingel, “Ultra-linear broadband optical modulator for high performance analog fiber link system,” in Proc. IEEE Int.
Topical Meeting Microw. Photon., Oct. 2004, pp. 241–244.
[41] Dingel B., Prescod, A., Madamopoulos, N., and Madabhushi, R., “Performance of Ring Resonator-based Linear Optical
Modulator (IMPACC) under Critical Coupling (CC), Over Coupling (OC), and Under Coupling (UC) Conditions,” IEEE Photonics
Annual Society Meeting (2011).
[42] Dingel, B.B.,“Ultra-high Linearized Optical Modulator,” US patent 6,943,931, (2005).
[43] Prescod, A., Dingel, B.B., Madamopoulos, N., “Traveling wave optical modulator with high dynamic range (>130 dB) at large
modulation frequencies,” Sarnoff Symposium, 2009. SARNOFF '09. IEEE, 1-5, (2009)
[44] Prescod, A., Dingel, B.B., Madamopoulos, N., and Madabhushi, R., “Effect of Ring Resonator Waveguide Loss on SFDR
Performance of Highly Linear Optical Modulators under Sub-Octave Operation,” IEEE Photonics Techn. Letters, 22, 17 (2010).
[45] Prescod, A. J., Dingel, B.B., Madamopoulos, N., “An Ultra-Linear Modulator with Inherent SFDR Compensation Capability,”
in Integrated Photonics Research, Silicon and Nanophotonics and Photonics in Switching, OSA Technical Digest (CD) (Optical
Society of America, 2010), paper JTuB22, (2010)
[46] Dingel, B. Madamopoulos, N., Prescod A. and R. Madabhushi, “Analytical Model, Analysis and Parameter Optimization of a
Super Linear Electro-Optic Modulator (SFDR>130 dB),” Optics Communications, 284(24), 5578-5587 (2011)
[47] Chiu, Y., Jalali, B., Garner, S. and Steier, W., “Broadband electronic linearizer for externally modulated analog fiber-optic
link,” IEEE Photonics Technology Letters, 11, 48-50 (1999).
[48] Sadhwani, R. and Jalali, B., “Adaptive CMOS pre-distortion linearizer for fiber-optic link,” Journal of Lightwave Technology,
21, 3180-3193 (2003).
[49] Sadwhani, R., Basak, J. and Jalali, B., “Adaptive electronic linearization for fiber,” Proceedings OFC, 2, 477-479 (2003).
[50] Ou, Haiyan, et al. "All-optical transversal filter with tap increasing and negative coefficients based on double-pass
modulation." Photonics Technology Letters, IEEE 23.14 (2011): 938-940.
[51] E. H. W. Chan and R. A. Minasian, “Novel All-Optical RF Notch Filters Equivalent Negative Tap Response”, IEEE Photonics
Technology Letters, Vol. 16, No. 5, May, 2004, pp. 1370-1372.
[52] P. C. Schindler et al., “Ultra-short silicon-organic hybrid (SOH) modulator for bidirectional polarization-independent
operation,” in Proc. Eur. Conf. Opt. Commun., Cannes, France, Sep. 2014, pp. 1–3.
[53] P. C. Schindler et al., “Ultra-dense, single-wavelength DFT-spread OFDMA PON with laserless 1.2 Gb/s ONU ready for
silicon photonics integration,” J. Lightw. Technol., vol. 33, no. 8, pp. 1650–1659, Apr. 15, 2015.
[54] E. Zolotov, R. Tavlykaev, “Integrated optical Mach-Zehnder modulator with a Linearized modulation characteristic,” Sov. J.
Quantum Electron. 18, 401-402 (1988)
[55] R. Tavlykaev, R. Ramaswamy, “Highly linear Y-fed directional coupler modulator with low intermodulation distortion,” J.
Lightwave Technol. 17 282 – 291 (1999)

Proc. of SPIE Vol. 10130 101300O-10

DownloadedViewFrom:
publicationhttp://proceedings.spiedigitallibrary.org/
stats on 02/05/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx