ICTON 2014 Tu.C1.

OSNR System Margin Monitoring Technique for Coherent

Transparent Optical Networks
David Dahan1,*, Senior Member, IEEE, and Uri Mahlab2, Senior Member, IEEE
1
ECI Telecom, 30 Hasivim street, 49517 Petach Tikva, Israel
2
HIT-Holon Institute of Technology, 52 Golomb street, 58102 Holon, Israel
* Tel: (+972) 3 926 6173, Fax: (+972) 3 926 6200, e-mail: david.dahan@ecitele.com
ABSTRACT
We propose a novel approach for transparent and in-service OSNR system margin monitoring for coherent
optical modulation formats. This technique is demonstrated experimentally for a 120.6 Gb/s DP-QPSK channel
in both dispersion and non-dispersion managed links.
Keywords: OSNR system margin, optical performance monitoring, coherent modulation formats.
1. INTRODUCTION
The deployment of high speed transparent and reconfigurable optical networks appeals for the development of
cost effective and in-service optical performance monitoring approaches to ensure satisfactory signal quality and
to monitor potential failures at the transmission link [1]. With the shift to advanced coherent modulation formats
and the use of Digital Signal Processing (DSP), high spectral efficiency optical networks may be designed with
almost no restriction on accumulated Chromatic Dispersion (CD) and Polarization Mode Dispersion (PMD).
Consequently, the transmission reach is mainly limited by the Amplified Spontaneous Emission (ASE) noise and
the optical fiber nonlinear effects. Therefore, efficient monitoring techniques of the Optical Signal to Noise Ratio
(OSNR) compliant to dual polarization coherent modulation formats [2-4] are required. However, monitoring the
OSNR level is not sufficient in order to evaluate the overall OSNR system margin which is defined as the
margin in terms of OSNR from the current operating OSNR level of the channel, to the reached OSNR level for
a given preFEC (Forward Error Correction) Bit Error Rate (BER) target. Link induced physical degradations,
such as received optical power, CD, PMD, Polarization Dependent Loss (PDL) and especially optical fiber
nonlinear effects, can change significantly the required OSNR level for a given BER target and therefore
challenge the estimation of the overall OSNR system margin. Monitoring the OSNR system margin is required
in different phases of the optical network operation, starting from the link commissioning, in-traffic operations,
maintenance and failure detection. OSNR system margin monitoring is particularly required when using
software defined optical coherent transceivers, in order to optimize the transceiver adaptive parameters such as
the bit rate, symbol rate modulation formats and FEC overhead. The conventional approach for OSNR system
margin monitoring is based on degrading the channel (or a tapped portion of the channel) by adding a controlled
level of ASE noise before the optical receiver and to measure the dependence of the preFEC BER on the
degraded OSNR level. Such an approach requires to bring complex and expensive equipment to the geographic
location of the network node (incurring significant operational expenses) and to find a monitor access point at
the link where the signal may be tapped out to avoid traffic disturbances during the measurement. Hence, the
systematic use of the OSNR margin monitoring is not possible. In this paper, according to the best of our
knowledge, we propose the first transparent OSNR system margin monitoring technique that does not affect the
channel quality. It is cost effective and enables potential simple remote monitoring operations. This method
relies on the estimation of the Electrical Signal to Noise Ratio (ESNR) margin of the received optical coherent
channel, scaled by a correction factor. Experimental results for a 120.6 Gb/s DP-QPSK channel show an
excellent accuracy for different physical link impairment scenarios such as nonlinear effects and optical received
power in both Dispersion Managed (DM) and Non-Dispersion Managed (NDM) links.
2. OSNR SYSTEM MARGIN MONITORING PRINCIPLES
The OSNR system margin monitoring of a coherent optical modulation format relies on the relationship between
ESNR and OSNR [5]:
A u OSNR
ESNR (1)
1  K u OSNR
where the parameter A is associated with the back to back transmission characteristics (e.g. symbol rate,
modulation format, receiver optical and electrical filter bandwidths. The parameter K refers to the saturation
effect that exists between the OSNR and the ESNR (for OSNR >> 1/K, ESNR ~ A/K) and is associated with the
overall physical impairments such as optical received power, CD, PMD, PDL and nonlinear fiber effects.
We assume that the channel OSNR level, noted OSNR1 and the A parameter are known. In addition, the
coherent receiver provides via the DSP, the estimation of the channel ESNR, noted ESNR1 Therefore, the
parameter K is evaluated by:
A 1
K  (2)
ESNR1 OSNR1

978-1-4799-5601-2/14/$31.00 ©2014 IEEE 1

ICTON 2014 Tu.C1.6

In addition, the ESNR margin to a given preFEC BER target is defined by:
ESNR1 (3)
'ESNR
ESNRref
where ESNRref is the required ESNR level in order to reach the preFEC BER target.
Therefore, using equations (1-3), the OSNR margin can be estimated by:
OSNR1 A  K u ESNRref
'OSNR 'ESNR u (4)
OSNR2 A  K u ESNR1
where OSNR2 is the required OSNR to reach the preFEC BER target.
Such an OSNR system margin technique is simple and uses the ESNR margin evaluation scaled by
a correction factor as shown in Eq. (4). It has the advantage of providing the OSNR system margin without
requiring the addition of optical or electrical noise source that would affect the signal quality.
According to the experimental results shown in the next section, we will demonstrate that the parameter A is
independent of the physical link impairments and can be derived from a simple back to back measurement.
In addition, we will show that in NDM links, ESNRref is independent of the nonlinear effects and optical
received power. Indeed, it has been demonstrated that in NDM links, after coherent detection, the NonLinear
Interference (NLI) noise exhibits a Circularly Symmetric Gaussian Distribution (CSGD) [5] and therefore cannot
be distinguished from the ASE noise. Consequently, the ESNRref values can be acquired in a back to back set up
and stored in a lookup table having entries that are function of the pre FEC BER target, symbol rate, modulation
format and filtering mode.
3. EXPERIMENTAL RESULTS
Figure 1 shows the schematic of OSNR system margin technique and the measurement set up. The coherent
receiver has a DSP block for linear link impairment compensations and a Soft Decision (SD) -FEC decoder
block. ESNR and preFEC BER levels are provided by the receiver DSP block. Before the coherent receiver, an
OSNR control stage was used in order to set the OSNR levels for the measurements. It consisted of two Erbium
Doped Fiber Amplifiers (EDFAs) in a cascade configuration with a Variable Optical Attenuator (VOA) that acts
as a span loss element located between the two amplifiers. The OSNR of the channel was estimated using an
Optical Spectral Analyser (OSA) with out-of-band estimation approach of the OSNR since no optical filters were
used after the inline amplifiers of the link spans.

Figure 1. Schematic of the measurement set up and OSNR system margin technique.
We performed the experimental verification of the proposed novel method with a real time 120.6 Gb/s
DP-QPSK transceiver over a 500 km link of G.652 (SSMF) fibers with loss parameter of 0.19 dB/km in NDM
and DM link configurations. In each scenario, we compared the OSNR system margin method based on the
ESNR margin evaluation with the correction factor to the OSNR system margin method based on the traditional
approach of inserting ASE noise to the channel before its detection. With both approaches, we considered two
preFEC BER targets (1.5×10-2 and 2×10-3) and we changed the optical impairments conditions by varying the
optical launched power per span.
3.1 Back to back measurements
We first proceeded to the back-to back performance measurement of the 120.6 Gb/s DP-QPSK channel with two
different received optical power levels (PRX = -10 dBm and PRX = -19 dBm) and a 50 GHz bandwidth optical
filter before the receiver. Figure 2a shows that the BER dependence on OSNR is affected by the received optical
power with a required OSNR of 12.4 dB and 12.7 dB for PRX = -10 dBm and PRX = -19 dBm respectively with
preFEC BER target of 1.5×10-2 and with the preFEC BER target of 2×10-3, the required OSNR is increased to
14.8 dB and 15 dB respectively. Figure 2b shows that the BER dependence on the ESNR is independent of the
optical received power with ESNRref = 8.15 dB for 1.5×10-2 and ESNRref = 10.1 dB for 2×10-3. Figure 2c shows
an excellent agreement between the measurement of the ESNR dependence on OSNR (solid lines) and the
analytic expression from Eq. (1) (dashed lines). From the fitting analytical curves for the two optical received
powers, the parameter A is found to be constant and equal to 0.41 while the K parameter is 5.5×10-3 and 7.5×10-3
respectively.

2
ICTON 2014 Tu.C1.6

0 0
10 10 20
PRX=-10 dBm
PRX=-10 dBm
-2 PRX=-10 dBm -2 PRX=-19 dBm 18
10 10 PRX=-19 dBm
PRX=-19 dBm
16
-4 -4
10 10
14

ESNR[dB]
BER
-6
BER

-6
10 10 12

-8 -8 10
10 10

8
-10 -10
10 10
6
(a) (b) (c)
-12 -12
10 10
10 12 14 16 18 20 22 24 26 28 30 6 7 8 9 10 11 12 13 14 15 16 17 18 10 12 14 16 18 20 22 24 26 28 30 32 34
OSNR [dB] ESNR [dB] OSNR [dB]

Figure 2. Experimental back to back results for 120.6 Gb/s DP-QPSK for two optical received powers (-10 and -19 dBm): (a) BER
dependence on OSNR (b) BER dependence of ESNR and (c) ESNR dependence on OSNR (dashed curves for the analytical expressions (1)).

3.2 Non-Dispersion Managed Link

In this scenario, we used 2 channels of 120.6 Gb/s DP-QPSK being co-transmitted with 2 channels of 45.8 Gb/s
DP-QPSK (see the optical channel spectrum in inset of Fig. 3a) over a NDM link consisting of 5 spans of
100 km of G.652 (SSMF) fibers. The channels are 50 GHz spaced from each other and the OSNR margin is
measured for one of the two 120.6 Gb/s channels. Figure 3a shows the ESNR dependence on the OSNR for three
launched optical channel power levels per span (1, 6 and 7 dBm) assuming PRX = -10 dBm. The fitting of the
analytic expressions from Eq. (1) exhibits an excellent agreement with the measurements. The parameter A
remains constant and equal to 0.41 while the parameter K increases with values of 9×10-3, 1.8×10-2 and 2.5×10-2,
respectively. Figure 3b shows the experimental results of the dependency between the required ESNR and
OSNR thresholds for preFEC BER targets of 1.5×10-2 and 2×10-3, on the launched optical channel power levels
per span with PRX = -10 dBm. It can be seen that for both preFEC BER targets, the required ESNR is
independent of the launched power per span and is equal to the levels found in the back to back measurement
due to the Gaussian nature of the NLI noise in NDM links [5]. In comparison, the required OSNR thresholds
increase along with the launched power per span, as a result of the nonlinear optical impairments. When the
optical launched power is increased from 1 dB to 7 dB, the required OSNR level for pre FEC BER target of
1.5×10-2 increases by 1.3 dB, while for pre FEC BER target of 2×10-3, it increases by 3.05 dB. Figure 3c shows
the comparison of the experimental OSNR system margin measurement results obtained by injection of the ASE
noise before the receiver , and by evaluating the ESNR margin with the correction factor (Eq. (4)) , as a function
of the launched optical channel power per span for the two preFEC BER targets. The channel OSNR level at the
receiver is fixed at 20 dB, the optical received power is -10 dBm and the ESNRref levels derived from back to
back measurements are used. The figure shows an excellent agreement between the two methods, with an error
below 0.15 dB for both preFEC BER targets.
20 8
19 ESNR @BER=1.5E-2
(a) ESNR @BER=2E-3 (b) (c)
18 7
OSNR @BER=1.5E-2
17 OSNR @BER=2E-3
OSNR system margin [dB]

16 6
ESNR or OSNR [dB]

15
14 5
13
-5

-10
12 4

-15
11
1
Optical Power [dBm]

-20
10 3 ESNR margin method -BER=1.5E-2
-25
9 ASE noise injection method -BER=1.5E-2
-30
8 2 ESNR margin method -BER=2E-3
ASE noise injection method -BER=2E-3
-35
1548 1548.5 1549 1549.5 1550 1550.5 1551 1551.5
7
Wavelength [nm]

6 1
1 2 3 4 5 6 7 1 2 3 4 5 6 7
Launched power per span [dBm] Launched power per span [dBm]

Figure 3. Experimental results for 120.6Gb/s DP-QPSK in 500 km NDM link with optical received power of -10 dBm: (a) ESNR
dependence on OSNR (dashed curves for the analytical expressions (1) and inset shows transmitted optical spectrum), (b) Required ESNR
and OSNR thresholds dependence on the launched optical channel power level per span for preFEC BER targets of 1.5×10-2 and 2×10-3,
(c) OSNR system margin as function of the launched optical channel power levels per span for preFEC BER targets of 1.5×10-2 and 2×10-3.
Both conventional and novel proposed methods are used.

3.3 Dispersion Managed Link

In this scenario, we used one channel of 120.6 Gb/s DP-QPSK being co-transmitted with 2 channels 45.8 Gb/s
DP-QPSK and 4 channels of 10.7 Gb/s OOK NRZ over a DM link consisting of 5 spans of 100 km of G.652
fibers. The channels are 50 GHz spaced from each other and there is a guard band of 300 GHz between the
120.6 Gb/s DP-QPSK channel and the 10.7 Gb/s channels. The optical spectrum of the channels is shown in the
inset of Fig. 4a. For each of the two first spans, a dispersion compensation fiber (DCF) was used at the span end
to compensate for the 90 km of CD, while for each of the last three spans a DCF compensating for the 95 km is
used. Figure 4a shows the ESNR dependence on the OSNR for four launched optical channel power levels per

3
 ICTON 2014 Tu.C1.6

span (0, 3, 4 and 5 dBm), assuming PRX = -10 dBm. The derivation of the analytic expression shows again an
excellent agreement with the measurements. The parameter A remains constant and equal to 0.41 while the
parameter K increases with values of 9×10-3, 1.55×10-2, 1.95×10-2 and 2.75×10-2, respectively. Figure 4b shows
the experimental results of the dependency of the required ESNR and OSNR thresholds for preBER targets of
1.5×10-2 and 2×10-3, on the launched optical channel power levels per span with PRX = -10 dBm. It can be seen
that the OSNR required for reaching the preFEC BER target increases along with the launched power per span,
as a result of the nonlinear impairments. When the optical launched channel power increases from 0 dB to 5 dB,
the required OSNR level increases by 2.25 and 5.6 dB for preFEC BER targets of 1.5×10-2 and 2×10-3
respectively. For the case of DM links, the ESNR required for reaching the preFEC BER target value increases
slightly with the launched optical channel power per span. When the launched optical power is increased from
0 dBm to 5 dBm, the required ESNR level increases by 0.25 and 0.35 dB for pre FEC BER target of 1.5×10-2
and 2×10-3 respectively. Unlike the NDM links case, the ESNR required for obtaining the pre FEC BER target is
slightly dependent on the launched optical power per span, since the NLI noise distribution deviates from the
CSGD in DM links. Figure 4c shows the comparison between the experimental results obtaining by injection of
the ASE noise before the receiver , and the results obtained by evaluating the ESNR margin with the correction
factor, according to Eq. (4), as a function of the launched optical channel power for the two pre FEC BER
targets. The channel OSNR at the receiver is fixed at 20 dB and the optical received power is -10 dBm. The
ESNRref levels derived from back to back measurements are used in order to estimate the error due to deviation
from symmetric circularity of the nonlinear noise distribution. For launched optical power levels of up to 4 dBm,
the error is below 0.1 dB and 0.35 dB for preFEC BER target of 1.5×10-2 and 2×10-3 respectively. In the case of
launched optical channel power per span of 5 dBm, the error is increased to 0.27 dB for nonlinear OSNR penalty
of 2.3 dB and preFEC BER target of 1.5×10-2 and 1.35 dB for nonlinear OSNR penalty of 5.6 dB and preFEC
BER target of 2×10-3. This error increase, especially for the BER target of 2×10-3 (when ASE noise impact is
reduced) is due to the stronger deviation of the NLI noise distribution from the CSGD.
22 8
(a) 21 ESNR @BER=1.5E-2 (b) 7 (c)
20 ESNR @BER=2E-3
19 OSNR @BER=1.5E-2
OSNR @BER=2E-3 6
18
OSNR system margin [dB]
ESNR or OSNR [dB]

17 5
16
15 4
14
3
13
0

-5

-10 12 2
ESNR margin method -BER=1.5E-2
Optical Power [dBm]

11
-15

-20

1 ASE noise injection method -BER=1.5E-2

-25 10
ESNR margin method -BER=2E-3
-30
9 ASE noise injection method -BER=2E-3
-35 0
-40
1545 1546 1547 1548 1549 1550 1551 1552
8
Wavelength [nm]

7 -1
0 1 2 3 4 5 0 1 2 3 4 5
Launched power per span [dBm] Launched power per span [dBm]

Figure 4. Experimental results for 120.6Gb/s DP-QPSK in 500 km DM link with optical received power of -10 dBm.
Same figure captions as for Fig.3 a, b, and c.

4. CONCLUSIONS
We have proposed a novel approach of in-service and cost effective OSNR system margin monitoring for
coherent optical channel based on the evaluation of the ESNR margin with a correction factor. This method is
transparent as it does not require to degrade the channel quality by injecting ASE noise into the channel and is
robust to the physical optical impairments. It only requires the preliminary knowledge of the A parameter and
ESNR reference for a given preFEC BER target (both being derived from a simple back to back measurement)
as well as the OSNR estimation of the channel. We demonstrated excellent accuracy of this method (below
0.35 dB) with 120.6 Gb/s DP-QPSK channel with nonlinear OSNR penalty up to 3.05 dB in the NDM link and
up to 2.2 dB in the DM link for both preFEC BER targets of 1.5×10-2 and 2×10-3. In the case of DM link,
a 1.35 dB error deviation is obtained with the proposed method for a large nonlinear OSNR penalty of 5.6 dB
with preFEC BER target of 2×10-3 due to deviation of the NLI noise from the CSGD.
REFERENCES
[1] D. Dahan et al.: Optical performance monitoring for translucent/transparent optical networks, IET Optoelectr., 5(1),
pp. 1-18 (2011).
[2] Z. Dong et al.: OSNR monitoring for QPSK and 16-QAM systems in presence of fiber nonlinearities for digital coherent
receivers, Opt. Express, vol. 20, no. 17, pp. 19520-19534 (2012).
[3] D. Dahan et al.: Stimulated Brillouin scattering based in-band OSNR monitoring technique for 40 Gbps and 100 Gbps
optical transparent networks, Opt. Express, 18(15), pp. 15769-15783 (2010).
[4] D. Dahan et al.: In-band optical signal-to-noise ratio monitoring technique based on Brillouin fiber ring laser, Applied
Optics, vol. 52, no.7, pp. 1487-1496 (2013).
[5] F. Vacondio et al.: On the nonlinear distortions of highly dispersive optical coherent systems, Opt. Express, vol. 20,
no. 2, pp.1022-1032 (2012).

View publication stats