Share and Explore the Tech Inside You!!!

OSNR: What does this mean;Why do we need and How to take care of it?

Signal to Noise Ration (SNR)  is not an unknown terminology for Engineers and Tech professionals who are dealing with Digital or Analog form of Communication.Here we will  explore the aspect of SNR in Optical Fiber Communication  space known as Optical Signal to Noise Ratio(OSNR).


Warning! Keep patience while scrolling down to read as this article may seem long as this is important topic.Whole content is collected from free available trusted sources and can be downloaded or shared.Copy/Paste mistakes may be observed and can be corrected once notified.Content may seem non-structured but there is content for everyone from beginner to professional so reader can absorb what he wants.


Bingo! Let's start it.



Some handy definition of OSNR to pick :-

  • OSNR [dB] is the measure of the ratio of signal power to noise power in an optical channel .


  • OSNR is the short form of Optical Signal to Noise Ratio. It is key parameter to estimate performance of Optical Networks. It helps in BER calculation of Optical System.
  • OSNR is important because it suggests a degree of impairment when the optical signal is carried by an optical transmission system that includes optical amplifiers.
  • If we know the OSNR and the bandwidths, we can find Q and the BER
  • It can be seen as the QoS at the physical layer of optical networks. OSNR is directly related to bit-error rate, which will lead to packet losses seen by higher layers.
  • OSNR indirectly reflects BER and can provide a warning of potential BER deterioration.
  • OSNR has long been recognised as a key performance indicator for amplified high-speed transmission networks to ensure network performance and reliability and it is related to many design parameter such as number or repeater/amplifiers ,reach ,available modulation formats etc.

Now let's explore it in more detail:-


Optical signal-to-noise ratio (OSNR) is used to quantify the degree of optical noise interference on optical signals. It is the ratio of service signal power to noise power within a valid bandwidth.When the signal is amplified by the optical amplifier (OA), like EDFA, its optical signal to noise ratio (OSNR) is reduced, and this is the primary reason to have limited number of OAs in a network.

The OSNR values that matter the most are at the receiver, because a low OSNR value means that the receiver will probably not detect  or recover the signal. The OSNR limit is one of the key parameters that determine how far a wavelength can travel prior to regeneration.

OSNR serves as a benchmark indicator for the assessment of performance of optical transmission systems. DWDM networks need to operate above their OSNR limit to ensure error – free operation. There exists a direct relationship between OSNR and bit error rate (BER), where BER is the ultimate value to measure the quality of a transmission.


The value of OSNRout that is needed to meet the required system BER depends on many factors such as the bit rate, whether and what type of FEC is employed, the magnitude of any crosstalk or non-linear penalties in the DWDM line segments, etc.Below picture highlights OSNR as one of the important parameter  in a DWDM link.


Optical amplifiers such as erbium-doped fiber amplifiers (EDFAs) are normally employed in optical networks to compensate for the transmission losses over long distances. However, besides providing optical gain, EDFAs also add unwanted amplified spontaneous emission (ASE) noise into the optical signal. Furthermore, the cascading of EDFAs results in accumulation of ASE noise. ASE noise is typically quantified by OSNR and is one of the most important parameters to be monitored in optical networks since the BER is directly related to the signal OSNR Furthermore, it also plays a pivotal role in fault diagnosis and as a measure of general health of links in an optical network.










OSNR may change due to signal power changes or higher repeater noise levels due to aging.


Now lets read the Legacy method to measure it:-


The traditional method to measure OSNR is defined in the IEC 61280-2-9 standard and is known as the interpolation or out-of-band method, as shown in Figure-1 below


OSNR for a Point-to-Point Link


 NFstage is the noise figure of the stage, h is Plank's constant (6.6260 × 10-34), ν is the optical frequency 193 THz, and Δf is the bandwidth that measures the NF (it is usually 0.1 nm).


OSNR =158.9+ Pin.dBm NF10log(Br )

where OSNRdB = optical signal to noise ratio of the optical amplifier, dB

Pin.dBm = average amplifier input signal power (DWDM systems use single-channel power), dBm

NF = amplifier noise figure, dB

h = Planck's constant 6.626069 1034, Js

f = signal center frequency, Hz

Br = optical measurement bandwidth RBW, Hz

If the measurement optical bandwidth can be assumed to be 0.1 nm (12.48 GHz),

OSNR = 58 + Pin.dBm NF

OSNRF.dB =158.9+Psource.dBm −Γ−NF10log(Br)10logN

where OSNRF.dB = final OSNR seen at the receiver, dB
source.dBm = average source signal power into the first span

(DWDM systems use single-channel power), dBm NF = amplifier noise figure, the same for all EDFAs, dB Br = optical measurement bandwidth, Hz
= number of amplifiers in the fiber link excluding the booster
Γ = span loss, the same for all spans, dB


Above equation provides the actual mathematical calculation of OSNR. This calculation method has quite a few approximations in which we can still find the system OSNR to a great degree of accuracy. In a multichannel WDM system, the design should consider OSNR for the worst channel (the one that has the worst impairment). The worst channel is generally the first or last channel in the spectrum.

we can see that the EDFA gain factor G is not considered. That is because OSNR is a ratio, and the gain acts equally on signal and noise, canceling the gain factor in the numerator and denominator. In other words, although EDFAs alleviate the upper bound on transmission length due to attenuation, by cascading EDFAs in a series, the OSNR is continuously degraded with transmission length and ASE (from EDFAs). This degradation can be lessened somewhat by distributed Raman amplifiers (DRAs).


Now see what bliss Raman Amplifier can do here:-


Addition of Raman and OSNR change:-




As we can see from above equation the factor GRA in the numerator actually enhances the OSNR of the system.(stages) could be considered as EDFA hops here.


OSNR-based design essentially means whether the OSNR at the final stage (at the receiver) is in conformity with the OSNR that is desired to achieve the required BER. This also guarantees the BER requirement that is essential for generating revenue.




This method works well for networks up to 10G, without any Reconfigurable Optical Add-Drop Multiplexers (ROADM).


But traditional way of measurement don’t work anymore in High Speed Communications:-


However, IEC 61280-2-9 isn’t feasible for 100G+ signals as well as ROADM networks.

Figure 2 illustrates 100G channels spaced 50GHz apart, which is a common spacing in modern submarine (and terrestrial) networks. Polarization-Multiplexed (Pol-Mux) 100G+ signals are typically wider (require more optical spectrum) than legacy On-Off-Keyed (OOK) 10G signals, meaning they could overlap with neighboring channels. Accordingly, the midpoint between channels no longer consists only of noise, but rather of signal plus noise. Thus, the IEC method applied to 100G+ Pol-Mux signals will therefore lead to an overestimation of noise and inaccurate measurement date leading to incorrect decisions.


Figure 2IEC 61280-2-9 Method Fails with Dense Pol-Mux 100G+ signals

Figure 3 illustrates a 100G signal that has gone through a ROADM, with the green area showing the channel bandwidth. Given filters inside a ROADM, the noise at the midpoint between channels will be carved (or filtered), leading to an underestimation of the noise level, if the IEC 61280-2-9 method is used, meaning this method is not feasible in ROADM-enabled coherent submarine networks.

Figure 3IEC 61280-2-9 Method Fails in ROADM 100G+ Pol-Mux Networks


To address the issues described in Figures 2 and 3, in-band OSNR was introduced around 2009 to support OSNR measurements of 10G signals in ROADM networks and 40G OOK signals. However, this method can’t be applied to coherent, Pol-Mux 100G+ signals, because of technical reasons beyond the scope of this blog. Consequently, Pol-Mux OSNR techniques have been introduced to support 100G+ signals, which is the topic of the latest standards.


Appropriate standards for Pol-Mux OSNR measurements?


There are two standards providing relevant guidelines for OSNR measurements of Pol-Mux signals. They are the China Communications Standards Association (CCSA) YD/T 2147-2010 standard and the IEC 61282-12 standards, which was recently introduced in February 2016. Both standards provide a future-proof definition of OSNR, which can be applied to any type of signal, at any data rate, including super-channels and flexible-grid signals. Specifically, the IEC-61282-12 standard specifies that:



  • s(λ) is the time-averaged signal spectral power density, not including ASE, expressed in W/nm
  • ρ(λ) is the ASE spectral power density, independent of polarization, expressed in W/nm
  • Br is the reference bandwidth expressed in nm (usually 0.1nm if not otherwise stated)
  • and the integration range in nm from λ1to λ2is chosen to include the total signal spectrum

The only drawback of these two standards is that a careful application of their formulae requires turning off channels, to access the Amplified Spontaneous Emission (ASE) noise floor, which isn’t possible on an in-service lest we upset end-users! Fortunately, in-service Pol-Mux OSNR methods have been introduced.

below table summarizes the correct OSNR method for each type of signal.


Data Rate

ROADM Present?

OSNR Method

Works on in-service network?

≤10G signals


IEC 61280-2-9


≤10G signals


In-band OSNR


Non-coherent 40G signals

Yes or No

In-band OSNR


Coherent 100G+ signals

Yes or No

Pol-Mux OSNR (IEC and CCSA standards)


Coherent 100G+ Signals

Yes or No

In-Service Pol-Mux OSNR



Table 1: OSNR Measurement Methods for Various Signal Types


Using the wrong OSNR measurement method for a given signal can have a significant impact on results, as it can lead to errors ranging from a few dBs up to 10dB, potentially leading to future outages. Using the proper method guarantees the right OSNR measurements are achieved leading to accurate network modeling, link simulation, and maintenance of ongoing submarine cable network performance.



Summarizing the importance of OSNR and its proper measurement method?

Some of the benefits of OSNR testing, including avoiding network outages, optimizing troubleshooting times, and ensuring optimal terrestrial and submarine cable performance. OSNR will become even more critical at data rates beyond 100G, because of the more stringent OSNR thresholds that will be required. Several OSNR methods have been introduced over the years, so the key takeaway is that the right OSNR measurement method must be used on a specific signal type to get accurate results.




Introduction to OSNR for high speed communication


The OSNR is the signal-to-noise ratio (SNR) measured in a reference optical bandwidth, where frequently a bandwidth Bref of 12.5 GHz is used corresponding to 0.1 nm wavelength. The OSNR relates to the Es N0 and Eb N0 as


where Bref is the previously introduced reference bandwidth, RS corresponds to the symbol rate of the transmission, r is the aforementioned rate of the code with r = kn, and qcorresponds to the number of bits mapped to each modulation symbol.

The use of high-order modulation formats requires a higher optical signal-to-noise ratio (OSNR), which may result in a significantly reduced achievable transmission distance .Following diagram shows the OSNR estimation stage for High Speed Optical Communication.


Lets talk about OSNR Penalty now:-

OSNR penalty is obtained from the BER curves and determined at a particular BER. A value of the OSNR penalty is obtained by comparing the values of OSNR before and after  the change of the parameters, which are under test, as given by



Gripple = penalty due to DWDM amplifier gain ripples

OSNRpenalties = various transmission penalties due to CD, PMD, PDL, etc. (note these penalties maybe different for 100G vs. 400G)



Calculation of Q-Factor from OSNR

The OSNR is the most important parameter that is associated with a given optical signal. It is a measurable (practical) quantity for a given network, and it can be calculated from the given system parameters. The following sections show you how to calculate OSNR. This section discusses the relationship of OSNR to the Q-factor.

The logarithmic value of Q (in dB) is related to the OSNR 

In the equation, B0 is the optical bandwidth of the end device (photodetector) and Bc is the electrical bandwidth of the receiver filter.

Q is somewhat proportional to the OSNR




Gray coding is assumed for all formats. For PM-BPSK the exact formula is:

For PM-QPSK the exact formula is:

For PM-16QAM and PM-64QAM, respectively, the following formulas are approx- imate, but their accuracy is better than ±0.05 dB of OSNRNL over the range

10−1– 10−4:


EDFA Noise – Why Input Power Matters






Optical signal suffers more than only attenuation. In amplitude, spectrally, temporally signal interaction with light- matter, light- light, light-matter-light leads to other signal disturbances

such as :-

  •  Power reduction
  •  Dispersion
  •  Polarization
  • Unbalanced amplification

Thus leading to random noise, which causes misalignments, jitter and other disturbances resulting in erroneous bits, the rate of which is known as bit-error-rate



Because of all possible influences outlined bits transmitted by source and bits arriving at the receiver may not have the same value. In actuality a threshold value is set at the receiver, above the threshold refers to a logic “one” and below threshold refers to a logic “zero”.


In order to measure BER in photonic regime, the optical signal is converted to electrical signal.

Example: Assuming a confidence level of 99%, BER threshold set at 10-10 and a bit rate of 2.5 Gb/s the required number n is 6.64 x 1010


Given the OSNR, the empirical formula to calculate BER for single fiber is

Log10 (BER) = 10.7-1.45 (OSNR)




In an experimental environment where factors such as loss, dispersion, and non-linear effects are excluded, if the OSNR is less than the specified threshold, the pre-FEC BER will be excessively large and uncorrectable bit errors will be generated. The OSNR  threshold in this case is called B2B OSNR tolerance.




Some mathematical aspects of OSNR:-


Calculating OSNR:

As can be seen from the definitions above, two quantities must be known to compute OSNR: the Total Signal Power, and the amount of ASE Noise Power present in a 0.1nm bandwidth.


Measuring the ASE power:

When the ASE noise floor is clearly visible left and right of the optical signal, the ASE Noise Power at the signal wavelength can be interpolated from two measurements made left and right of the signal.


Alternately, when the noise floor is not visible left and right, the optical signal needs to be removed temporarily in order to allow the measurement of the ASE Noise Power at the signal wavelength.

This is the case when some filtering devices implemented somewhere in line are removing some of the noise between channels (Example = WSS).

This could also be the case if the modulated signal bandwidth is so large so that the tail of adjacent signals overlaps the open space between them – masking the noise floor.



Note that power values are frequently provided in dBm by the OSA - whether the measurements is made using integrated power function or a user-specified resolution bandwidth.  To convert the values in dBm to mW, the following relation must be used:



Measuring the Total Signal Power:

When making this measurement it is important to use a bandwidth that is large enough to capture the entire signal:

If using an OSA with variable resolution bandwidth, this means that the resolution bandwidth has to be set larger than the width of the signal.

If using a integrated- power function between vertical markers, the markers have to be set to include the entire signal bandwidth.  


Note that when measuring a DWDM spectrum, the power of each DWDM signal cannot be measured independent of ASE noise present in the measurement bandwidth.

 Ie., the value that is actually measured is: Total Signal Power + ASE Noise Power.

To get the Total Signal Power only, the ASE noise content (measured separately in the previous step) must be subtracted from the measurement. Since the measurement bandwidth used to measure the noise on its own (ASE BW) may be different from the bandwidth used to measure the signal (Signal BW), a factor is added to the equation.  This removes the correct amount of noise from the measurement.



Signal BW is the bandwidth used to measured the signal,

ASE BW is the bandwidth used to measure the ASE Noise Power only (e.g., 0.1nm).


Calculating OSNR from measurements:

Since OSNR must be reported as signal power with respect to 0.1nm worth of ASE noise, the denominator of the OSNR equation also includes a factor.  This adjusts the amount of ASE noise measured to an amount expected inside a 0.1nm bandwidth:




In a transmission chain, the relative evolution of the optical signal and noise levels is usually characterized by the optical signal-to-noise ratio (OSNR). The OSNR, in a given optical bandwidth, is defined as



The optical signal may be polarized, but the noise is usually not and, depending on the receiver polarization sensitivity, the noise must be considered with a single or two polarizations. Any active (or passive) linear optical device amplifies (or attenuates) simultaneously the incoming signal and the incoming noise



Accordingly with the fluctuation dissipation theorem, it also adds noise, making its output OSNR lower than its input one. This OSNR degradation is expressed in terms of noise factor (NF) defined as:




where G is the optical power gain (or the attenuation coefficient) of the device which is larger (or smaller) than 1. NF is denoted noise figure when it is expressed in dB. Because the relative importance of the added noise strongly depends on the input noise level, NF is only an intrinsic parameter of the optical device (i.e. independent of the input signal and noise) when an input reference noise is defined.

For polarization insensitive devices, input noise and output noise usually do depend on polarization, making noise figure independent of polarization considerations.


Single amplifier noise factor

OSNR at the output of an optical amplifier with an output power Pout and for an optical bandwidth B0 is expressed as:


where m=1 or 2 is the number of polarization modes contributing to noise. It is usual to consider the two polarization modes (m=2) of the noise and to make reference to an optical bandwidth equal to 0.1 nm corresponding to Bo = 12.5GHz at a wavelength of 1550 nm. In this particular situation, the OSNR is expressed in dB as:




Noise factor of a cascade of fibers and amplifiers

Let us consider now the cascade of spans displayed in Figure 3.8 and including NAMP11 optical fiber spans with an attenuation coefficient A5expð2αALÞ and NAMP lumped linear phase insensitive amplifiers with a gain net G and a noise factor NF. For each span a near compensation of the signal attenuation by the gain is assumed, making its net gain GSPAN 5 GA close to 1. Each fiber reduces the sig- nal level and reduces the noise level nearly in the same way. Since the input noise level is far above the appropriated reference level, the attenuation noise of the fiber is negligible and the OSNR is kept nearly unchanged. Each amplifier output POUT restores the input noise of the link but adds the amplifier noise.



The accumulated noise is Namp times larger than single amplifier and we have:





To bridge transoceanic distances while keeping a high OSNR (optical SNR), it is crucial to limit the noise contribution added by the successive amplifiers. The impact of the added noise on the output OSNR can be calculated with:



This equation can also be expressed in a more physical manner:


where N is the amplifier count, and Δλ the width of the filter where the OSNR is expressed. The number “30” at the end of Equation corresponds to the conversion of signal input power from units in watts into units scaled in milliwatts.





Note: Now try to utilise the above concepts and equations whatever way you want .





Calculating Channel Capacity for DWDM links

The maximum data rate (maximum channel capacity) that can be transmitted error-free over a communications channel with a specified bandwidth and noise can be determined by the Shannon theorem. This is a theoretical maximum data transmission rate for all possible multilevel and multiphase encoding techniques.

As can be seen below that the maximum rate depends only on channel bandwidth and the ratio between signal power to noise power. There is no dependence on modulation method.

Rmax =Bolog2(OSNR+1)


    Rmax = maximum data rate for the channel (also known as channel capacity), Gbps

    Bo = optical channel passband, GHz

    OSNR = channel optical signal to noise ratio


For a 62 GHz channel passband (for standard 200 GHz DWDM channel spacing) and an OSNR of 126 (21 dB) the maximum possible channel capacity is 433 Gbps.

As channel bandwidth decreases so does maximum transmission rate. For a 30 GHz channel passband (100 GHz DWDM channel spacing) and OSNR of 126 (21 dB) the maximum possible channel capacity is 216 Gbps.




Calculating Transponder bandwidth

Transponder bandwidth is the product of Modulation , Baud Rate and the Polarisation.

BW=Modulation x Baud x Polarisation 

Following table will give an idea for various bit rates:-



Modulation = 2 (bits/s/Hz)

Baud Rate = 32G

Polarisation= 2


BW= 2 x 32 x 2 =128Gbps




BER and Q relation

The Bit Error Rate (BER) of a digital optical receiver indicates the probability of an incorrect bit identification. In other words, the BER is the ratio of bits received in error to the total number of bits received. Below lists different values for BER and their corresponding errors per bits and over time.

As we know that, the photocurrent is converted to a voltage then measured. The measurement procedure involves a decision as to whether the bit received is a 1 or a 0. The BER is a not only a function of the noise in the receiver and distortion in the system, but also on the decision level voltage,VD that is the threshold level above which the signal is classified as a 1 and below which the signal is classified as a 0. Even an ideal signal with no noise nor distortions has a non-zero BER if the decision level is set too high or too low. For example, if VD is set above the voltage of the 1 bit, the BER is 0.5, assuming equal probability of receiving a one and a zero.


Error per 10E-15 bits

@ 10Gbps, One error in



0.1 msec



0.1 sec



1.7 min



1.2 days


Mathematically, the Bit Error Rate is expressed as

BER = p(1)P(0 1) + p(0)P(1 0)

where p(1) and p(0) are the probabilities of receiving a 1 and a 0, respectively. P(0/1) is the probability of deciding a 0 when the bit is actually a 1, and P(1/0) is the probability of deciding a 1 when the bit is a 0.

Minimum BER as a function of Q 



A short discussion on 980nm and 1480nm pump based Erbium Doped Fiber Amplifiers (EDFA)


The 980nm pump needs three energy level for radiation while 1480nm pumps can excite the ions directly to the metastable level .


(a) Energy level scheme of ground and first two excited states of Er ions in a silica matrix. The sublevel splitting and the lengths of arrows representing absorption and emission transitions are not drawn to scale. In the case of the 4 I11/2 state, s is the lifetime for nonradiative decay to the I13/2 first excited state and ssp is the spontaneous lifetime of the 4 I13/2 first excited state. (b) Absorption coefficient, a, and emission coefficient, g*, spectra for a typical aluminum co-doped EDF.

.The most important feature of the level scheme is that the transition energy between the I15/2 ground state and the I13/2 first excited state corresponds to photon wavelengths (approximately 1530 to 1560 nm) for which the attenuation in silica fibers is lowest. Amplification is achieved by creating an inversion by pumping atoms into the first excited state, typically using either 980 nm or 1480 nm diode lasers. Because of the superior noise figure they provide and their superior wall plug efficiency, most EDFAs are built using 980 nm pump diodes. 1480 nm pump diodes are still often used in L-band EDFAs although here, too, 980 nm pumps are becoming more widely used.


Though pumping with 1480 nm is used and has an optical power conversion efficiency which is higher than that for 980 nm pumping, the latter is preferred because of the following advantages it has over 1480 nm pumping.

  • It provides a wider separation between the laser wavelength and pump wavelength.
  • 980 nm pumping gives less noise than 1480nm.
  • Unlike 1480 nm pumping, 980 nm pumping cannot stimulate back transition to the ground state.
  • 980 nm pumping also gives a higher signal gain, the maximum gain coefficient being 11 dB/mW against 6.3 dB/mW for the 1.48 
  • The reason for better performance of 980 nm pumping over the 1.48 m pumping is related to the fact that the former has a narrower absorption spectrum.
  • The inversion factor almost becomes 1 in case of 980 nm pumping whereas for 1480 nm pumping the best one gets is about 1.6.
  • Quantum mechanics puts a lower limit of 3 dB to the optical noise figure at high optical gain. 980 nm pimping provides a value of 3.1 dB, close to the quantum limit whereas 1.48  pumping gives a value of 4.2 dB.
  • 1480nm pump needs more electrical power compare to 980nm.



The 980 nm pumps EDFA’s are widely used in terrestrial systems while 1480nm pumps are used as Remote Optically Pumped Amplifiers (ROPA) in subsea links where it is difficult to put amplifiers.For submarine systems, remote pumping can be used in order not to have to electrically feed the amplifiers and remove electronic parts.Nowadays ,this is used in pumping up to 200km.

The erbium-doped fiber can be activated by a pump wavelength of 980 or 1480 nm but only the second one is used in repeaterless systems due to the lower fiber loss at 1.48 mm with respect to the loss at 0.98 mm. This allows the distance between the terminal and the remote amplifier to be increased.

In a typical configuration, the ROPA is comprised of a simple short length of erbium doped fiber in the transmission line placed a few tens of kilometers before a shore terminal or a conventional in-line EDFA. The remote EDF is backward pumped by a 1480 nm laser, from the terminal or in-line EDFA, thus providing signal gain



Following are the vendors that manufactures 980nm and 1480nm EDFAs


Techniques/technologies to mitigate compensation in long haul submarine optical networks

Map of transmission impairments in long-haul submarine systems and some techniques/technologies for their mitigation or compensation.





Spectral Hole Burning (SHB) phenomenon in Optical Networks

Spectral Hole Burning (SHB)

Spectral hole burning (SHB) is a major limitation of amplified WDM systems with high channel count. The main reason lies in the fact that there is no possibility of compensating for this effect.

  • Due to the inhomogeneous portion of the linewidth broadening of the dopant ions, the gain spectrum has an inhomogeneous component and gain saturation occurs, to a small extent, in an inhomogeneous manner. This effect is known as spectral hole burning because a high power signal at one wavelength can 'burn' a hole in the gain for wavelengths close to that signal by saturation of the inhomogeneously broadened ions. Spectral holes vary in width depending on the characteristics of the optical fiber in question and the power of the burning signal, but are typically less than 1 nm at the short wavelength end of the C-band, and a few nm at the long wavelength end of the C-band. The depth of the holes are very small, though, making it difficult to observe in practice.


  • In addition, accurate predictions are very difficult to carry out. SHB acts as a selective oversaturation of specific erbium ion classes due to a precise matching of the signal wavelength with their corresponding Stark energy sublevels. Gain contributions of a given ion class to the overall amplifier gain spectrum will be dependent on the specific values of energy of the related Stark sublevel (determined by inhomogeneities in the local electric field in the glass as opposed to on the crystal) and of their population density (i.e. of the related induced saturation). Clearly, the overall gain spectrum of the amplifier may be distorted due to this SHB effect. The best-known induced distortion is the hole induced in the gain spectrum in the spectral vicinity of a saturated channel.
  • This gives rise to a hole in the gain profile around the saturating channel wavelength, whose width is determined by temperature. Increasing temperature will increase this homogeneous broadening (and thus the hole width at the expense of its depth) while lower temperatures will reduce and make this hole deeper in the gain profile. . Since it is not possible to operate the amplifier at a lower temperature where the effect of homogeneous broadening vanishes, the system designer should account for the holes induced by each signal channel in the amplifier gain profile at room temperature


  • SHB does not distort the overall gain profile because the sum of the different contributions has a flat transfer function. Problems may be encountered when some channel powers increase compared to other channels
  • SHB could be seen (wrongly) at first glance as a regulating effect because the most favored channels will see a slightly lower gain due to the SHB they induce. This will indeed slightly reduces the power excursion between channels (the correcting effect being, however, much lower than the effect creating this SHB). The detrimental effect actually comes from the distortions induced in the amplifier gain spectrum due to thermal broadening. Other channels, located a few nanometers aside from the most favored ones will also see an induced reduced gain level, while such channels may not be gain favored like the channels that create the SHB effect. This will result in a decrease in the OSNR of such neighboring channels
  • The SHB effect not only stresses the dynamic range of the system by increasing the burst power but it also degrades the OSNR performance. In particular, whilst the power change is limited to the beginning of the burst (during the formation of the hole), the OSNR impairment is observed at the end of burst,where the spectral holeis already completely formed by the burst.

SHB also has a limiting effect in the implementation of preemphasis of the less-favored channels. This technique consists of increasing the power of the worst channels at the transmitter side, at the expense of the best channels, leading to the same OSNR for all channels at the link output. This can be performed while keeping the EDFA output powers constant and decreasing the transmitted power of the best channels. However, the highest predistortion that can be performed at the link input in order to compensate for a given excursion in output OSNR is limited by SHB.

Difference between EDFAs requirements for Terrestrial(Land) Systems and Submarine Systems.

Compared with requirements for EDFAs for terrestrial applications and for Submarine applications, there are major important differences making the two types of amplifiers definitely two different components.


Terrestrial(Land) systemSubmarine System

•Reliability of land-based equipment is somewhat relaxed, corresponding to a 15-year required lifetime.

• Submarine systems are designed for a 25-year lifetime and a minimum of ship repair that imply reliability and redundancy of all the critical components.

• Terrestrial equipment should enable operation over a wide temperature range of −5, +70°C (and −40, +85°C in storage conditions). 





  This wide temperature range makes it necessary to implement cooling means for the           highest temperatures and compensation means for temperature-sensitive devices.

• In submarine amplifiers, heat is dissipated from the outer side of the repeater container into the sea. Such a container is designed in order to make the heat go through the box from the pump device to the outer side, ensuring moderate temperature in all points. Temperature of the deep sea is indeed around +5°C. Specific care is taken for repeaters located at the coast or in shallow water, in order to guarantee no pump failure while avoiding Peltier cooling. 

For reliability reasons, no glue is used on the optical path. The constant temperature of the devices and the doped fiber incorporated in the amplifier makes it possible to perfectly tailor the gain spectrum of the submerged EDFAs, owing to very accurate equalizing filters and to concatenating hundreds of amplifiers. 

This would not be possible for land-based amplifiers whose gain cannot be guaranteed below 1 dB for a 30-nm bandwidth partly due to such temperature changes (while a few tenths of dB of gain excursion is reached for submarine amplifiers).

• The infrastructure itself of terrestrial systems determines the actual characteristics of the amplifier that needs to cope with important variations of the span loss between two amplifier sites. In addition, for economical reasons, the amplifiers cannot be tailored to cope with this nonuniform link.

• In submarine systems, the link is manufactured at the same time as the amplifiers and much attention is paid to guarantee constant attenuation loss between amplifier values, while the amplifier has been designed to perfectly adapt to the link characteristics.


• There are high gain range (20 to 35 dB) of the amplifiers incorporated in land-based systems and allowed by the margins given on the OSNR due to the reduced total link length. 

Gain equalizers therefore compensate for much larger gain excursion values than in submarine amplifiers and should therefore be located at amplifier midstage in order not to impact their equalizing loss on the amplifier output power.

• On the contrary, such filters can be placed after the single section of doped fiber that composes the amplifier in the case of submarine applications.





Temperature dependency of EDFA Gain for various channels

Here the results are after evaluating the effect of a thermal variation on the output tilt. In this particular set-up the amplifier is kept at room temperature (25 °C) and only the active fiber spool undergo a thermal cycle. Temperature of the EDFA is varied from 0 °C to 65 °C and the amplifier gain is measured at four point : 0, 25, 40 and 65 °C.



As can be seen a 65 °C temperature variation implies a 1.8 dB tilt variation. Considering a reduced temperature range (5-45 °C) the output tilt variation is about 1.1 dB.

It was  tried to investigate the origin of the temperature dependency. First  used a different EDFA with a lower erbium concentration (14 dB/m erbium peak absorption); then tried to reduce the saturation of the EDFA using lower power levels, but in both cases the output tilt variation was very similar to that of Figure .

Temperature variation has also effect on the EDFA efficiency: with high temperature the active fiber is less efficient than at low temperature.

With constant pumps power, a 65°C variation implies a 0.25 dB difference on the output power. To compensate this extra tilt we can act in two way: using the VOA ; heating the EDFA to a constant 65 °C.

First solution requires a thermal sensor to measure the EDFA temperature and a compensation table (stored in the firmware) to act on VOA attenuation.

Second solution requires a heater and special mechanics & software to store the EDFA spool and to keep their temperature constant.


Simplifying what and why of Raman Amplifier.


It's always a wondering situation when we discuss Raman Amplifier;its benefits , requirement and application.I have tried to make it simpler to understand here.

Hope it will help the readers.



  • The Raman amplifier is typically much more costly and has less gain than an Erbium Doped Fiber Amplifier (EDFA) amplifier. Therefore it is used only for speciality applications.
  • The main advantage that this amplifier has over the EDFA is that it generates very less noise and hence does not degrade span Optical to Signal Noise Ratio (OSNR) as much as the EDFA.
  • Its typical application is in EDFA spans where additional gain is required but the OSNR limit has been reached.
  • Adding a Raman amplifier might not significantly affect OSNR, but can provide up to a 20dB signal gain.
  • Another key attribute is the potential to amplify any fiber band, not just the C band as is the case for the EDFA. This allows for Raman amplifiers to boost signals in O, E, and S bands (for Coarse Wavelength Division Multiplexing (CWDM) amplification application).
  • The amplifier works on the principle of Stimulated Raman Scattering (SRS), which is a nonlinear effect.
  • It consists of a high-power pump laser and fiber coupler (optical circulator).
  • The amplification medium is the span fiber in a Distributed Type Raman Amplifier (DRA).
  • Dstributed Feedback (DFB) laser is a narrow spectral bandwith which is used as a safety mechanism for Raman Card. DFB sends pulse to check any back reflection that exists in the length of fiber. If no High Back Reflection (HBR) is found, Raman starts to transmit.
  • Generally HBR is checked in initial few kilometers of fibers to first 20 Km. If HBR is detected, Raman will not work. Some fiber activity is needed after you find the problem area via OTDR.

Common type of Raman amplifier 

  • The lumped or discrete type Raman amplifier internally contains a sufficiently long spool of fiber where the signal amplification occurs.
  • The DRA pump laser is connected to the fiber span in either a counter pump (reverse pump) or a co-pump (forward pump) or configuration.
  • The counter pump configuration is typically preferred since it does not result in excessively high signal powers at the beginning of the fiber span, which can result in nonlinear distortions,

  •  The advantage of the co-pump configurations is that it produces less noise.


As the pump laser photons propagate in the fiber, they collide and are absorbed by fiber molecules or atoms. This excites the molecules or atoms to higher energy levels. The higher energy levels are not stable states so they quickly decay to lower intermediate energy levels releasing energy as photons in any direction at lower frequencies. This is known as spontaneous Raman scattering or Stokes scattering and contributes to noise in the fiber. 

Since the molecules decay to an intermediate energy vibration level, the change in energy is less than the initial received energy during molecule excitation. This change in energy from excited level to intermediate level determines the photon frequency since Δ f = Δ E / h . This is referred to as the Stokes frequency shift and determines the Raman gain versus frequency curve shape and location. The remaining energy from the intermediate level to ground level is dissipated as molecular vibrations (phonons) in the fiber. Since there exists a wide range of higher energy levels, the gain curve has a broad spectral width of approximately 30 THz. 

During stimulated Raman scattering, signal photons co-propagate frequency gain curve spectrum, and acquire energy from the Stokes wave, resulting in signal amplification.

Theory of Raman Gain

The Raman gain curve’s FWHM width is about 6 THz (48 nm) with a peak at about 13.2 THz below the pump frequency. This is the useful signal amplification spectrum. Therefore, to amplify a signal in the 1550 nm range the pump laser frequency is required to be 13.2 THz below the signal frequency at about 1452 nm.

Multiple pump lasers with side-by-side gain curves are used to widen the total Raman gain curve. 

where fp = pump frequency, THz  fs = signal frequency, THz Δ f v = Raman Stokes frequency shift, THz 

Raman gain is the net signal gain distributed over the fiber’s effective length.It is a function of pump laser power, fiber effective length, and fiber area.



For fibers with a small effective area, such as in dispersion compensation fiber, Raman gain is higher. Gain is also dependent on the signal separation from the laser pump wavelength,Raman signal gain is also specified and field measured as on/off gain. This is defined as the ratio of the output signal power with the pump laser on and off.In most cases the Raman ASE noise has little effect on the measured signal value with the pump laser on. However, if there is considerable noise, which can be experienced when the measurement spectral width is large, then the noise power measured with the signal off  is subtracted from the pump on signal power to obtain an accurate on/off gain value.The Raman on/off gain is often referred to as the Raman gain.

In addition, to obtain significant gain, pump power used in distributed Raman amplification is much higher than signal power. Therefore, the pump energy transferred in the process of stimulated emission remains low compared to the involved pump power level in the case of practical distributed Raman amplification. This makes the Raman gain weakly dependent on the total signal power, or on the channel count. This is an advantage in terms of practical implementation, but also requires a perfect control of the pump power. Backward pumping is therefore usually used to average the effects of pump instabilities and its relative intensity noise (RIN).

It is worth pointing out that Raman gain (expressed in dB) that is produced is a linear function of the pump power. This is because there is almost no gain saturation induced by signal power in distributed preamplification, making the amplification process operate as in the small-signal input power regime. This is quite different compared to EDFAs, which are operated in saturation for having high output power. Their output power is then a linear function of the pump power, making their gain, expressed in dB, a logarithmic function of the pump power.

Noise sources

Noise created in a DRA span consists :-

  • Amplified spontaneous emissions (ASE)
  • Double Rayleigh scattering (DRS)
  • Pump laser noise.

ASE noise is due to photon generation by spontaneous Raman scattering.

DRS noise occurs when twice reflected signal power due to Rayleigh scattering is amplified and interferes with the original signal as crosstalk noise.

The strongest reflections occur from connectors and bad splices.

Typically DRS noise is less than ASE noise, but for multiple Raman spans it can add up. To reduce this interference, ultra polish connectors (UPC) or angle polish (APC) connectors can be used. Optical isolators can be installed after the laser diodes to reduce reflections into the laser. Also span OTDR traces can help locate high-reflective events for repair.

Counter pump DRA configuration results in better OSNR performance for signal gains of 15 dB and greater. Pump laser noise is less of a concern because it usually is quite low with RIN of better than 160 dB/Hz.

Nonlinear Kerr effects can also contribute to noise due to the high laser pump power. For fibers with low DRS noise, the Raman noise figure due to ASE is much better than the EDFA noise figure. Typically the Raman noise figure is –2 to 0 dB, which is about 6 dB better than the EDFA noise figure.

Raman amplifier noise factor is defined as the OSNR at the input of the amplifier to the OSNR at the output of the amplifier.

Noise figure is the dB version of noise factor

The DRA noise and signal gain is distributed over the span fiber’s effective length.


Counter pump distributed Raman amplifiers are often combined with EDFA pre-amps to extend span distances. This hybrid configuration can provide 6 dB improvement in the OSNR, which can significantly extend span lengths or increase span loss budget. Counter pump DRA can also help reduce nonlinear effects by allowing for channel launch power reduction.


  Functional Block Diagram for CoPropagating and Counter Propagating Raman Amplifier


                       Co-Propagating                                                                                                                                                Counter-Propagating


Field Deployment architecture of EDFA and RAMAN Amplifiers


Interesting to know




  1. Planning Fiber Optic Networks by Bob Chomycz


View older posts »

BER-to-Q utility



Points for more explanation:-
(980 pump)-This pump wavelength gives the best EDFA noise performance but requires a higher degree of wavelength accuracy because of its narrow absorption band
(1480 pumb) -Good power efficiency due to the small energy difference between 1480 nm and 1550 nm, a broad absorption spectrum that is less demanding on the wavelength accuracy of the pump laser, and lower attenuation in optical fibers for the remote pumping of EDFAs.

thanks , very nice explanation .

One question - Are the FS bytes are used while mapping 10GE client signals into OPU2 and then further ODU2/OTU2 ?

Thanks Sanjay bhai.

Its number of bits per symbol Akshay!Will edit it.Thanks

I am waiting for your posting..Because I am sure that your post would be help me. Thanks

Very Helpful Sanjay. Thanks

Spectral efficiency is defined by bit rate/channel size
So in case of 200g Dp-16qam 37.5ghz then the spectral efficiency should 200/37.5=5.33 bit/sec/hz but in table it's mentioned 4?
Please correct me if I am wrong..

I really liked your article , your article is very
petrified me in the learning process and provide
additional knowledge to me , maybe I can learn
more from you, Thank you for sharing


Flag Counter