Making OSNR Measurements

In a Modulated DWDM
Signal Environment

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Jack Dupre
Jim Stimple

Making OSNR measurements

in a modulated DWDM
signal environment
May 2001

In a DWDM spectrum, it is desirable to measure the optical signal-to-noise

ratio (OSNR) of each channel in order to gauge signal quality. A
standardized definition and test method is described.
With higher modulation rates and narrower channel spacing, the
modulation sidebands from adjacent channels interfere and limit the ability
to measure the noise level between channels. Increasing the resolution of
the optical spectrum analyzer does not affect this limitation. This paper
provides simulations and actual measurements demonstrating the OSNR
limit for 50-GHz channel spacing and modulation rates of 2.5 and 10 Gb/s.

2-1

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Signal impairments in a DWDM transport

network can come from a variety of sources
Single-channel interface
multichannel interface

MUX

OA

OA

OADM
OA

OA

OA

DEMUX

T
r

...

...

Fiber loss
Optical amplifier ASE Can be deduced from optical measurements
Optical crosstalk
Fiber nonlinearities
Chromatic dispersion
Polarization-mode dispersion
Waveshape distortion

Require per-channel signal measurements

In the unrelenting move toward all-optical networking, there are increasingly larger
segments of the network in which there is no optical- to-electrical conversion. Here
is shown an example of such a network section. Interfaces to this network section
are standardized single-channel interfaces. As the DWDM signals propagate
through the network, signal impairments that impact overall system quality can
occur.
At the multichannel optical interfaces within such a network segment, it is
desirable to measure parameters that provide information about signal quality.
Such parameters are necessary in the manufacture of network elements, the
installation of the network, and in performance monitoring of the operating
network.
Ideally, such parameters would directly correspond to the bit error ratio (BER) of
each channel of a multichannel carrier at the particular optical interface. Related
parameters such as Q-factor or optical eye patterns would provide similar
information and could be correlated to per-channel BER. However, it is difficult to
obtain access to these parameters at a multichannel interface point. It is necessary
to demultiplex the potentially large number of channels and make BER, Q-factor, or
eye-diagram measurements on a per-channel basis.
In contrast, useful information about the optical properties of the multichannel
carrier is readily obtained by measuring the optical spectrum. Wavelength-resolved
signal and noise levels provide information on signal level, signal wavelength, and
amplified spontaneous emission (ASE) for each channel.

2-2

Making OSNR Measurements in a

Modulated DWDM Signal Environment

OSNR in a transport network is dependent

upon optical amplifier noise figure
single span
Span loss
L (dB)

Signal-spontaneous noise figure*

F (dB)
Pout
ASE(
)

OA

N spans
L (dB)

F (dB)

h is Plancks constant
is optical frequency
r is the reference optical bandwidth

where

F (dB)

F (dB)
L (dB)

L (dB)
OA

OSNR = Pout - L - F - 10 Log (h

r)

OA

OA

Pout
ASE()

OSNR = Pout - L - F - 10 Log (N) - 10 Log (h

r )
* Signal-spontaneous noise figure is defined in IEC Publication 61290-3, Basic specification for
optical fibre amplifiers test methods Part 3: Test methods for noise figure parameters

For a transport network, the OSNR can be calculated for given system
parameters including span loss and optical amplifier gain and noise figure.
It is therefore a useful parameter to determine if a system is performing to
its designed performance.

2-3

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Optical signal-to-noise ratio (OSNR)

Useful to predict system impairment due to optical power and
optical amplifier gain and ASE

Optical power

missing channels

noise

Wavelength

A typical optical spectrum at a multichannel interface is shown here.

Important characteristics are:
(a) The channels are placed nominally on the grid defined by ITU G.692.
(b) Individual channels may be non-existent because it is a network
designed with optical add/drop demultiplexers or because particular
channels are out of service.
(c) Both channel power and noise power are a function of wavelength.
For calculating OSNR, the most appropriate noise power value is that at the
channel wavelength. However, with a direct spectral measurement, the
noise power at the channel wavelength is included in signal power and is
difficult to extract. An estimate of the channel noise power can be made by
interpolating the noise power value between channels.

2-4

Making OSNR Measurements in a

Modulated DWDM Signal Environment

OSNR Standardized Definition*

With an optical spectrum
analyzer, noise power is
interpolated at each channel
Ni =

P i + Ni

N( i- )+N( i+)
2
N ( i - )

OSNR = 10Log

Ni

N ( i + )

Pi
+ 10Log Bm
Ni
Br

* TIA/EIA-526-19, OFSTP-19, Optical Signal-to-Noise Ratio Measurement Procedures for

Dense Wavelength-Division Multiplexed Systems

In the figure:
Pi is the optical signal power in watts at the i-th channel.
Ni is the interpolated value of noise power in watts measured in noise
equivalent bandwidth, Bm, at the i-th channel.
Dl is the interpolation offset equal to or less than one-half the ITU grid
spacing,
Br is the reference optical bandwidth. (The units for Bm and Br may be in
frequency or wavelength but must be consistent.) Typically, the reference
optical bandwidth is 0.1 nm.
Note that in the equation for OSNR, there is an adjustment to convert the
OSA measured value of noise to that in the reference bandwidth. Also, the
signal power is is obtained by subtracting the the measured noise value.

2-5

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Modulated source spectral width can

cause an error in signal measurement
Power Envelope for
10 Gb/s 27-1 PRBS

0.06 nm
Resolution
Bandwidth
Shape

OSA resolution
bandwidth set
too narrow will
not capture all
source power

The spectral width of each channel is broadened from that of the CW laser
due to several causes:
(a) laser chirp
(b) intensity modulation for signal transmission
(c) modulation to suppress stimulated Broullion scattering (SBS).
(d) self-phase modulation (SPM)
For dense WDM systems in which external modulation is generally used,
laser chirp is not a factor. Broadening due to SBS suppression and SPM are
typically small compared to the broadening due to the signal modulation at
2.5-Gb/s and higher rates.
If the OSA resolution bandwidth is too narrow, the measured signal power
will be lower than the true signal power.

2-6

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Wide RBW required to avoid errors in

signal measurement
10 Gbit/s NRZ Modulation 2
Amplitude Error dB

Amplitude Error dB

2.5 Gbit/s NRZ Modulation 1

~0.5 dB error
with 50 pm RBW
Resolution Bandwidth nm

~2.5 dB error
with 50 pm RBW
Resolution Bandwidth nm

1 Assumes 7.5-GHz

electrical and 0.36-nm

optical filtering with zero chirp.

2 Assumes 30-GHz

electrical and 0.64-nm

optical filtering with zero chirp.

Because a portion of the signal power is not captured by the OSA, an error
in the measured signal power occurs. These charts show the magnitude of
the error for 2.5-Gb/s and 10-Gb/s data rates respectively.
To minimize the error in the signal power measurement, a resolution
bandwidth of sufficient width should be chosen. For less than 0.1-dB error
and the modulated signal condition shown on the graphs, the RBW should
be wider than 0.09 nm for 2.5 Gb/s and 0.2 nm for 10 Gb/s.

2-7

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Amplitude (dBm)

High dynamic range required to measure

noise between closely spaced channels
-10
-20
-30
-40
-50
-60
-70
-80

noise measured
here

dynamic
range

1549.5

1550

1550.5

1551

Wavelength (nm)

In order to accurately measure the noise between channels, the optical spectrum
analyzers dynamic range must be sufficiently high.
The dynamic range of an OSA is a measure of the OSA's ability to make measurements of
low-level signals and noise that are close in wavelength to large signals. It is important to
note that narrowing the RBW does not necessarily correlate to better dynamic range.
RBW is a measure of the 3-dB bandwidth or noise equivalent bandwidth of its filter
characteristic. Dynamic range, on the other hand, is a measure of the steepness of the
filter characteristic and the OSA noise floor. Dynamic rage is defined as the ratio, in dB,
of the filter transmission characteristic at the center wavelength, li, and at one-half a grid
spacing away, li D i.
The OSAs dynamic range must exceed the expected noise level by at least 10 dB. For
example, if the grid spacing is 50 GHz, and the expected OSNR is 25-dB, the dynamic
range must be at least 35-dB at 25 GHz for a 0.1-nm RBW setting. If the RBW is reduced
to 0.05 nm, the OSNR requirement increases to 38 dB.

2-8

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Combination of wide RBW for signal and narrow

RBW for noise provides the best accuracy
Wide RBW includes all relevant sidebands

Signal
Measurement
(1st sweep)
1550

1550.4

Spectra with
10 Gbit/sec
data rate

Narrow RBW to include noise only

Noise Power
Density
Measurement
(2nd sweep)

1550

1550.4

For narrow channel spacing, finding a single RBW setting that is optimum
for both signal and noise measurement is difficult. As discussed, the signal
measurement requires a setting that is sufficiently wide to accommodate the
chirped modulated signal. The noise measurement requires high dynamic
range that usually means a narrower RBW setting.
The solution is to take two consecutive sweeps of the OSA with different
RBW settings. As shown, the first sweep measures the signal power at each
channel and the second sweep measures the noise power between channels.

2-9

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Agilent OSAs automatically measure

OSNR in a DWDM signal environment
Each signal is marked and
numbered
Easy to read table of:
Channel number
Channel Wavelength
Channel Power
Channel OSNR
Span Tilt
Peak to Peak Deviation

Features of Agilents WDM Channel Analysis Application:

Uses two sweeps to accurately measure both signal power and noise floor power
density
Uses noise markers that are corrected for OSA filter shape to more accurately
measure noise floor power density
Utilizes OSAs 10 pm wavelength accuracy for an accurate center wavelength
measurement
Measures OSNR, center wavelength, peak to peak deviation, span tilt for signals
spaced down to and including 50 GHz
Remote control capable
Three choices for the position of the noise measurement are available:
1. Between channels - user inputs channel spacing and the application will
measure noise halfway in between the channels (standardized definition).
2. Pit - the application will find the pit (lowest point) in between channels and use
this as the noise measurement.
3. Specified offset- the user inputs an offset, noise is then measured at this offset
away from the channel peak.

2 - 10

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Modulation sidebands limit the

measurement of the ASE level
SIGNAL

ASE

True OSNR cannot be measured by conventional means

As modulation rates have increased and channel spaces have narrowed a

limitation is imposed on the measurement of the actual noise level. As
shown here, the optical spectrum between channels is dominated by
modulation sideband power. The ASE level is too low to measured by
conventional means.
An OSNR value can still be measured but its meaning has changed. It
becomes a signal-to-modulation sideband ratio. While it may still be a
useful parameter, it does contain information about the build-up of ASE in a
system.

2 - 11

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Signal simulation methodology

1. Calculate spectral envelope for
PRBS modulation

P(f) = sin

2. Bandpass filter the spectral

envelope

(f - fo)
2 fm

(f - fo)
2 fm

where fm = bit rate/2 for NRZ format

= bit rate for RZ format

3. Convolve with simulated OSA

resolution bandwidth filter

Gaussian filter shape

P(f)

modulated signal spectral envelope

filtered signal spectral envelope
OSA simulated spectrum

fo fo+ f m

optical frequency, f

To determine noise measurement limitations imposed by modulation

sidebands, a simple simulation is used. The spectral envelope for PRBS
modulation is the familiar sin(x)/x function.
Because the signal envelope is filtered electrically in the modulation
circuitry and optically in the demultiplexing filter, the simulation performs
a bandpass filter function on the modulation envelope. Typical filter
bandwidths are two to 5 times the bit rate.
An optical spectrum analyzers resolution bandwidth filter characteristic is
simulated by a Gaussian function

2 - 12

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Measured spectra with 50-GHz

channel spacing
Two modulated channels

Two Unmodulated Channels

0

-10

-10

RBW, nm-20
0.06
0.1
0.2

-30

dB

dB

-20

-40

-30
-40

-50

-50

-60

-60

-70
193.095

193.145

193.195

-70
193.095

193.245

Optical Frequency, THz

193.145

193.195

193.245

Optical Frequency, THz

To validate the simulation, the spectrum of two laser sources separated by

50-GHz was observed on an OSA. The OSAs resolution was set to 0.06, 0.1,
and 0.2 nm. Note that depth of the null between the signals is significantly
less with the modulation turned on. In this case, the modulation was 2.5Gb/s NRZ.

2 - 13

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Comparison of measured and simulated

spectra 2.5-Gb/s NRZ, 50-GHz spacing
2.5 GB/s NRZ Simulated Spectrum
0

Relative Amplitude (dB)

-10
RBW, nm

-20

0.20
0.10

-30

-40

measured
values

0.06

-50

-60
192.9

192.91 192.92 192.93 192.94

192.95 192.96 192.97 192.98 192.99

193

Optical Frequency (THz)

Here is the simulated data for the modulated spectra shown in the previous
slide. The measured value of the minimum between signals is reasonable
close to the simulation (< 2 dB) for the 0.06 and 0.1 nm resolution
bandwidth settings. The 0.2-nm RBW, however, is considerably different.
This can be explained by the fact that the OSAs wider RBW settings have
filter characteristics that are more rectangular than Gaussian. An
assumption of Gaussian shape was used in the simulation. The narrower
RBW values tend to be more Gaussian in shape and give better agreement
with the simulation.
A bandwidth of four times the bit rate (10-GHz) was used which
approximates the electrical bandwidth of the lasers modulation circuitry.

2 - 14

Making OSNR Measurements in a

Modulated DWDM Signal Environment

10-Gb/s NRZ simulation

Multichannel Simulated Spectrum

10-Gb/s NRZ

Relative Amplitude (dB)

0
-10
-20

0.1 nm RBW

-30

0.06 nm RBW

-40

0.01 nm RBW

-50
-60
-70
-80
193.1500

193.2000

193.2500

193.3000

193.3500

193.4000

Optical Frequency (THz)

Now, we will look at simulations of 10-GB/s modulation with three channels

separated by 50 GHz. For the NRZ format a bandwidth of 2.5 times the
clock rate or 25-GHz is assumed. Shown are the calculated spectra as would
be observed on a optical spectrum analyzer for resolution bandwidth
settings of 0.1, 0.06, and 0.01 nm. With the widest RBW, 0.1 nm, the signal
peaks are identifiable and the modulation depth is only 22 dB. As the RBW
is decreased to 0.06 and 0.01 nm, the depth of the modulation between
signal peaks is increased. At 0.01-nm RBW, the spectral detail of the
modulated carrier can be observed.

2 - 15

Making OSNR Measurements in a

Modulated DWDM Signal Environment

10-Gb/s RZ simulation

Multi-channel Measured Spectrum

10-Gb/s RZ
0
modulation sideband ratio

Relative Amplitude (dB)

-10
-20

0.1 nm RBW
0.06 nm RBW

-30

0.01 nm RBW

-40
-50
-60
-70
-80
193.1400

193.1900

193.2400

193.2900

193.3400

193.3900

Optical Frequency (THz)

This is the same simulation but for RZ modulation. The bandwidth, still at
2.5 times the clock rate, has been increased to 50 GHz. We will define the
ratio of the spectral amplitude at the signal peaks to the amplitude halfway
between the signal peaks as the modulation sideband ratio. The
modulation sideband ratio is less for RZ modulation due to the increased
bandwidth requirement.

2 - 16

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Modulation sideband ratio is relatively

constant for OSA RBW less than 0.1 nm
Modulation Sideband Ratio

0.1 nm
0.06 nm
0.01 nm

Modulation Sideband ratio,

dB

for 10 Gb/s data rate

RBW =

30
25
20
NRZ
RZ

15
10
5
0
zero

0.01 nm

0.06 nm

0.1-nm

0.2-nm

Resolution Bandwidth

The modulation sideband ratio is charted here for resolution bandwidth

setting of zero to 0.2 nm for NRZ and NRZ modulation formats. The
important thing to note is that decreasing RBW below 0.06 nm does not
change the measured value of the modulation sideband ratio. This has some
interesting implications on the measurement of OSNR.

2 - 17

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Implications on OSNR measurements

Procedure to obtain OSNR from 10-Gb/s data rate simulations:
1. Obtain the signal value from a wide (0.2 nm) RBW
2. Obtain the modulation-limited power spectral data halfway between channels for
each RBW
3. Adjust the value in step 2 to convert to a 0.1-nm reference bandwidth.
OSNR Limit Imposed by 10-Gb/s Modulation, dB
0.2-nm RBW*

NRZ format
RZ format

0.1-nm RBW*

0.06-nm RBW*

0.01-nm RBW*

(11.1)

23.3

26.6

25.5

(8.3)

14.8

16.1

15.3

Decreasing the OSA RBW below 0.06-nm does not

significantly change OSNR value !
*For the noise measurement

As discussed before, it is best to measure the OSNR by taking to OSA

sweeps: one to obtain the signal power and the second the noise power.
Because OSNR is defined for a particular r4eference optical bandwidth,
usually 0.1 nm, the noise value needs to be corrected for the reference
optical bandwidth.
Using this algorithm for the 10-Gb/s modulated signal as discussed, the
OSNR limit due to modulation sidebands is calculated and summarized in
the table.
A perhaps surprising result is that decreasing the OSAs RBW below 0.06
nm has minimal effect on the OSNR limit.

2 - 18

Making OSNR Measurements in a

Modulated DWDM Signal Environment

An alternative to OSNR when modulationsideband-limited values are not sufficient to

guarantee system performance

1540

nm

1550

Filter a selected channel and route

to signal measurement apparatus

Earlier, it was mentioned that OSNR may be a useful parameter to measure

the ASE build-up in a system. Clearly, with 10-Gb/s data rate and 50-GHz
channel spacing, OSNR has limited utility. To achieve a better measure of
signal quality, it may be necessary to filter each channel and route the
selected channel to single channel measurement apparatus.

2 - 19

Making OSNR Measurements in a

Modulated DWDM Signal Environment

ITU identifies test issue at

multichannel interface*
Optical Network Element
1
3R
3R

3R

Optical Network Element

1
3R
MPI-S M

MPI-R M

O
M

O
D

Domain A

3R

3R

Domain B

Measurement 1
(Eye diagram)

Reference
Bandpass
Filter

Measurement 2
(BER)

Reference
Optical
Receiver

*G.959.1, Optical Transport Network Physical Layer Interfaces (draft)

The ITU, in draft recommendation G.959.1, suggests that a meaningful

multichannel interface performance measures require channel selection.
Depending on whether the interface is at the transmit side or the receive
side determines if an eye-diagram or a bit-error-ratio (BER) is more
appropriate.
In either case, a reference filter followed by reference optical receiver is
required.

2 - 20

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Optical spectrum analyzer with filtered

SMF output and DCA provide spectrum
and signal measurements
multichannel input

Agilent 86141B H17 OSA

w/SMF filter mode
Agilent 86100 Infinium DCA

An implementation of such a measurement system is an OSA with a filtered

output capability such as the Agilent 86141B H17 followed by an Agilent
86100 Infinium Digital Communications Analyzer (DCA).
The OSA can perform OSNR, channel power and other spectral
measurements. The DSA measures extinction ratio and eye-mask
parameters with a standardized rate-specific reference receiver.

2 - 21

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Key measurements
Channel power

Eye mask compliance

Channel OSNR
Tilt and peak-to-peak deviation

Extinction ratio

Using the built-in channel analysis application in the OSA, a tabular

summary of wavelength, channel power, and OSNR is obtained. With the
Infinium DCA, compliance to an optical eye-mask is tested.

2 - 22

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Optical spectrum analyzer with filtered

SMF output and DCA provide spectrum
and signal measurements
multichannel input

Agilent 86141B H17 OSA

w/SMF filter mode
Agilent 86130A BitAlyzer
Agilent 83446A Lightwave Receiver

2 - 23

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Key measurements
Channel power

Channel BER

Channel OSNR

Channel Q-factor

Tilt and peak-to-peak deviation

2 - 24

Making OSNR Measurements in a

Modulated DWDM Signal Environment

Summary

OSNR is a valuable measure of system optical

performance

At 10-Gb/s and 50 GHz channel spacing, modulation

spectra overlap putting a limit on the OSNR measured

value that is not improved with narrower OSA RBW.

A more-definitive measure of signal integrity is an

optical eye-mask or BER measurement on each
individual channel

In summary, OSNR can be a useful measure of system optical performance

but a combination of high data rate and narrow channel spacing limit OSNR
as a measure of ASE build-up. The limit imposed by modulation sidebands
is not improved by using a narrower OSA resolution bandwidth.
To obtain a more-definitive measure of signal integrity or quality at a
multichannel interface, it is necessary to filter each channel and perform
optical eye-diagram or BER measurements.

2 - 25