IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 25, NO. 21, NOVEMBER 1, 2013 2133
Multiplier-Free Joint Carrier Recovery for 16-QAM
Synchronous or Agile Burst Receivers
Igor Tselniker, Alex Tolmachev, and Moshe Nazarathy, Senior Member, IEEE
Abstract— Multiplierless multi-symbol delay detector, recently
demonstrated in FPGA for QPSK and 16-QAM at 25 Gbaud for
phase recovery, is augmented here to also mitigate frequency
offset exceeding half the baudrate and chirp <80 GHz/us,
also multiplier-free (just simple additions/logic operations)—
applicable to synchronous and burst-receivers.
Index Terms—Optically coherent detection, phase estimation,
phase recovery, multi-symbol delay detection (MSDD), multi-
symbol phase estimation (MSPE), carrier frequency offset (CFO),
frequency offset estimation.
I. I NTRODUCTION
T
HE literature is replete with a plethora of carrier phase
and/or frequency recovery methods for coherent detection
(see [1] for a review). One particular method, namely Multi-
Symbol Delay Detection (MSDD) aka Multi-Symbol Phase
Estimation (MSPE) has been fruitfully pursued [1]–[6] but is
yet to receive mainstream recognition in the carrier recovery
(CR) area despite superior performance. MSDD is a decision-
directed (DD) CR, yet differing from a DD PLL. MSDD
operation is cycle-slips free, as it is used in conjunction with
modulus-preserving differential precoding (DP) [4], [6] at
the QAM Transmitter (Tx). Typically DP incurs an OSNR
penalty, however in the case of MSDD, the DP penalty may
be made negligible at least in white noise (as was shown
years ago in the wireless literature [7] but not yet assim-
ilated by our optical communication community). Recently,
we have demonstrated [5] a QPSK / 16-QAM polar-domain
MSDD variant realized without any multipliers and amenable
to “superscalar” HW parallelization, without compromising
its phase recovery performance. We constructed FPGA HW
prototypes at 25 Gbaud [5], [9] which to our knowledge are
the fastest QPSK and 16-QAM CRs to have ever been made
to work in electrical real-time. Those HW demonstrations
attested to the very low computational burden in the polar
MSDD structure. Moreover, in comparison with our prior
non-polar MSDD CR versions [4], the new polar MSDD
exhibits negligible performance penalty, in exchange for the
Manuscript received June 13, 2013; revised August 23, 2013; accepted
September 11, 2013. Date of publication September 18, 2013; date of current
version October 9, 2013. This work was supported in part by the the Chief
Scientist Office of the Israeli Ministry of Industry, in part by the Trade and
Labor - “Tera Santa” consortium under the Magnet program as well as the
Israel Science Foundation, and in part by the ASTRON FP7 European project.
The authors are with the Technion, Israel Institute of Technology,
Haifa 32000, Israel (e-mail: igortz@tx.technion.ac.il; talex@tx.technion.ac.il;
nazarat@ee.technion.ac.il).
Color versions of one or more of the figures in this letter are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/LPT.2013.2282311
multiplier-less realization major benefit. In this letter the polar-
MSDD phase-recovery system [5] is augmented by a novel
agile carrier frequency offset (CFO) estimation (EST) module
comprising just 11 adders. The resulting polar MSDD with
joint phase and frequency EST is applicable to synchronous
coherent receivers (Rx), wherein low-complexity and high
performance are at a premium. Moreover, the new CR is highly
suitable for agile wideband coherent burst receivers [9]–[11]
emerging in metro networks, wherein the optical packet bursts
must be rapidly acquired by coherent LO tuning with large
CFO laser transients [12].
II. OVERVIEW OF PROPOSED CR CHARACTERISTICS
The new extra capability of CFO mitigation is obtained
(Fig. 1) by appending here the module marked “CFO EST”
atop the polar-MSDD [5] which originally provided just phase
estimation.
Negligible Complexity: The original polar MSDD system,
was realized multiplier-less as demoed in FPGA [5]; the
CFO EST mitigation upgrade module is also implemented with
no multipliers, requiring just 11 adders total, some delays and
trivial bit-shift scalings by powers-of-two.
CFO capture range: The peak-to-peak CFO range captured
by our new joint phase and frequency EST exceeds 0.5 R
s
where R
s
is the baudrate. The precise value of the CFO capture
range (>half-baudrate) is OSNR-dependent, tending to R
s
in
asymptotically large SNR.
Negligible OSNR penalty: One should be concerned about
potential phase noises leakage through the new CFO EST
module. Fortunately, in the new system the incremental degra-
dation of the phase recovery tolerance is negligible - just
0.2 dB penalty due to the CFO EST.
Advantages for coherent burst receivers: The proposed
CR advances the state-of-the-art of coherent burst receivers,
wherein robust differential precoding and decoding methods
have been recently introduced [11]. For coherent agile burst-
receiver applications, it is critical to tolerate large and rapid
CFO swings during the burst transients of the LO. In these
prior coherent burst receivers, the CFO tolerance was achieved
by adopting simple soft differential [4], [6], [13] or doubly
differential [12] decoding (as opposed to ‘hard’ or ‘logical’
differential decoding, ahead of the slicer [14]). However
each phase soft differencing at least doubles the ASE noise,
thus substantial OSNR penalty is incurred in those current
approaches.
In contrast, in our proposed solution, robust CFO acquisition
is not accompanied by a large OSNR penalty; our OSNR
penalty due to incorporating the CFO mitigation is just 0.2 dB.
1041-1135 © 2013 IEEE