WO2012072131A1

WO2012072131A1 - Error estimation in optical communication networks

Info

Publication number: WO2012072131A1
Application number: PCT/EP2010/068651
Authority: WO
Inventors: Ernesto Ciaramella; Andrea Peracchi; Giancarlo Prati
Original assignee: Telefonaktiebolaget L M Ericsson (Publ)
Priority date: 2010-12-01
Filing date: 2010-12-01
Publication date: 2012-06-07

Abstract

A method of monitoring error performance of an optical communications signal at a node (20) of an optical communications network (10) comprises converting (151) the optical communications signal to an electrical signal and sampling (152) the electrical signal to obtain digital signal samples. The method processes (153, 155) the signal samples to determine a measure of bit error performance, by determining a statistical distribution of the signal sample values and deriving the measure of bit error performance from the statistical distribution of signal sample values. The method can find a minima value of the statistical distribution and scale the minima value by at least one scaling parameter. The step of determining a statistical distribution of the signal sample values can comprise obtaining a histogram of amplitude values of the signal samples. The statistical distribution of the signal sample values can be a probability density function of the signal samples.

Description

ERROR ESTIMATION IN OPTICAL COMMUNICATION NETWORKS

TECHNICAL FIELD

This invention relates to estimation of bit errors in an optical communication network.

BACKGROUND

An optical transport network (OTN) typically has some form of Performance monitoring (PM) to determine the quality of optical channels within the network. Performance monitoring provides a way of checking that a quality of service (QoS) meets a quality of service that has been agreed with a client.

In a kind of optical network known as an "opaque" optical network, a signal is converted between the optical domain and electrical domain at a node. These opto- electrical (OE) conversions can occur in the signal path at wavelength division multiplexing (WDM) systems or at switches. Current opaque architectures use transceivers at intermediate nodes. These transceivers perform optical regeneration. An optical signal is received, converted to the electrical domain, and converted back to the optical domain. Typically, Forward Error Correction (FEC) is performed at the node to correct any errors that have occurred during transmission. The Forward Error Correction function at a node provides a count of the number of bit errors. Bit error rate (BER) information is a useful measure of the status of the link. The BER parameter is sensitive to the same impairments that affect the QoS and the number of errors is a good measure of the quality of the received signal. The BER information can be used to provide support for fault detection and isolation, performance monitoring and management functions. However, this monitoring solution requires each regenerating node to have a set of N transceivers, each capable of FEC processing, for N WDM channels. This incurs a significant cost and power consumption at the nodes.

Synchronous Digital Hierarchy (SDH) and Synchronous Optical Networking (SONET) requires that errors in transmission are detected and an alarm condition is raised when the BER in a given data stream exceeds a specified level. Typically, the BER alarm is an indication that a channel over which the transmission is being conveyed has degraded, and the transmission is then switched to a different path by a recovery operation. In order to identify the particular network link where the BER degradation arises, it is necessary to implement performance monitoring whenever possible. Particularly, this would be expected at each regeneration site, .i.e. at each segment of the path, rather than simply monitoring end-to-end. A bank of transponders, implementing a FEC scheme, is required at every regeneration node. This increases power consumption and cost of a node.

Some techniques have been proposed to reduce the hardware and software resources required for BER detection without using FEC, by either using electrical or optical measurements. An overview of various techniques is given in the paper "Optical Performance Monitoring", D . C. Kilper et al, Journal of Lightwave Technology Vol.22, No. l, January 2004. Some techniques are based on g-factor estimation. This type of technique is described in a paper "Margin Measurements in Optical Amplifier Systems", Bergano et al, IEEE Photonics Technology Letters, Vol.5, No.3, March 1993 and in ITU-T Recommendation O.201 (07/2003) "Q-factor test equipment to estimate the transmission performance of optical channels". These techniques are based on the variation of the decision threshold voltage far away from the optimum level and the simultaneous measurement of the corresponding errors. Out-of-service measurements interrupt the data stream, which is not acceptable for a working network. Therefore, a node which implements one of these techniques requires an additional clock and data recovery (CDR) unit with variable threshold and the digital output signal of this unit has to be electronically correlated on-the-fly to the output of the CDR in use, which requires high-speed complex electronics. A further disadvantage of these techniques is that they were originally devised to estimate the BER at extremely low values (<10^~12) and are less suited to the BER range for typical terrestrial networks, which is typically in the range between 10^"8 and 10^"3.

SUMMARY

An aspect of the invention provides a method of monitoring error performance of an optical communications signal at a node of an optical communications network. The method comprises converting the optical communications signal to an electrical signal. The method further comprises sampling the electrical signal to obtain digital signal samples. The method further comprises processing the signal samples to determine a measure of bit error performance. The step of processing comprises determining a statistical distribution of the signal sample values and deriving the measure of bit error performance from the statistical distribution of signal sample values.

The measure of bit error performance, such as Bit Error Rate (BER), of an optical signal is a useful parameter for performance monitoring. Obtaining BER in this manner has an advantage of avoiding the need for Forward Error Correction processing at a node, thus reducing the cost, complexity and the power consumption of each network node. The method preserves power efficiency as it makes it possible to remove FEC processing at regeneration sites in the intermediate nodes of an optical mesh network, with FEC processing only being performed at end terminals. The method also allows QoS measurements without interrupting the data stream. It can only require collecting signal samples in the region near the optimum threshold value. Hence, this monitoring could be performed in-service, by the same receiver/regenerator, without any additional equipment.

Advantageously, the step of processing the signal samples comprises finding a minima value of the statistical distribution.

Advantageously, the method further comprises scaling the minima value by at least one scaling parameter. The step of scaling can comprise scaling the minima value by a first scaling parameter to form a first scaled minima value. Alternatively, the step of scaling can comprise scaling the minima value by a second scaling parameter to form a second scaled minima value. The scaled first scaled minima value and second scaled minima value define upper and lower bounds of the measure of bit error performance. The at least one scaling parameter can be determined by performing a calculation to determine the at least one scaling parameter using the statistical distribution function. Alternatively, the at least one scaling parameter can be determined by performing a look-up of a stored scaling parameter value.

Advantageously, the step of determining a statistical distribution of the signal sample values comprises obtaining a histogram of amplitude values of the signal samples.

The statistical distribution of the signal sample values can be a probability density function of the signal samples.

The measure of bit error performance can be used in various ways. A transfer of the optical signal to a different network path can be initiated, based on the measure of bit error performance. Additionally, or alternatively, the measure of bit error performance can be compared with a threshold value and an alarm issued if the measure of bit error performance is below the threshold value.

Advantageously, the method further comprises converting the electrical signal to an output optical communications signal for transmission to another node without subjecting the electrical signal to forward error correction at the node.

The functionality described here can be implemented in hardware, software executed by a processing apparatus, or by a combination of hardware and software. The processing apparatus can comprise a computer, a processor, a state machine, a logic array or any other suitable processing apparatus. The processing apparatus can be a general-purpose processor which executes software to cause the general-purpose processor to perform the required tasks, or the processing apparatus can be dedicated to the perform the required functions. Another aspect of the invention provides machine- readable instructions (software) which, when executed by a processor, perform any of the described methods. The machine-readable instructions may be stored on an electronic memory device, hard disk, optical disk or other machine-readable storage medium. The machine-readable instructions can be downloaded to the storage medium via a network connection.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the invention will be described, by way of example only, with reference to the accompanying drawings in which:

Figure 1 shows a communication network in which an embodiment of the invention can be implemented;

Figure 2 shows a node of an optical communication network;

Figure 3 shows a node of an optical communication network in which an embodiment of the invention can be implemented;

Figure 4 shows apparatus for monitoring error performance of an optical communications signal at the node of Figure 3;

Figure 5 shows sampling points in a received signal;

Figure 6 shows steps of a method of monitoring error performance of an optical communications signal at a node;

Figure 7 shows a graphical representation of a method according to an embodiment of the invention;

Figure 8 shows a simulation model for validating the method; Figure 9 shows simulation results at a first value of energy per bit to noise power spectral density ratio;

Figure 10 shows simulation results at a second value of energy per bit to noise power spectral density ratio;

Figure 1 1 shows processing of a traffic along a signal path through an optical communications network.

DETAILED DESCRIPTION

Figure 1 shows part of an optical communication network 10 in which an embodiment of the invention can be implemented. Nodes 20 are connected by optical links 15. A ring topology is shown as an example, but other topologies (e.g. mesh, star) can be used. The network 10 can be a core network, metro network, or access network. Each of the nodes 20 can be a Reconfigurable Optical Add Drop Multiplexer (ROADM). A node 20 in the network 10 can route communications traffic received at an ingress optical link 15 to an egress optical link 15. In the case of an ROADM, or similar device, nodes 20 are also capable of adding traffic 25 or dropping traffic 26. Nodes 20 can connect to a local network, such as a passive optical network (PON). The network 10 can comprise a network management system 30 which connects 32 to each of the nodes 20. The network management system 30 may be centralised or at least partly distributed. The network management system 30 can send management information to the nodes 20 to configure operation of the nodes. Nodes 20 can send information to the network management system 30, such as alarms and performance monitoring information, such as Bit Error Rate (BER) information. The network 10 can have a control plane, where nodes 20 signal to one another to perform tasks such as setting up paths, tearing down paths, and switching traffic to alternative paths when a fault occurs.

Figure 2 and Figure 3 show two possible forms of node 20 which can be used in the optical communication network 10. Figure 2 shows a node where all traffic is subject to frame processing 56, 66 as the traffic enters and leaves the node 20. Frame processing modules 56, 66 shown in Figure 2 perform forward error correction (FEC) processing of a received signal to correct bit errors. By contrast, Figure 3 shows a node where traffic can pass through the node without being subject to the frame processing 56, 66 of Figure 2. The path 70 through node 20 which avoids framer processing will be called a bypass path. The selection of whether traffic takes the ,

6 bypass path 70 can be determined according to the destination of the traffic. For example, traffic received at input 50 which is destined for an output 26 to a local network is switched 55 to framer 72 and traffic which is destined for another node 20 is switched 55 to electric-optical processing 64. The selection of whether traffic takes the bypass path can be determined according to other factors, such as how far, or via how many nodes, the traffic has travelled since the last time it was processed by a framer.

In Figure 3 an ingress optical link 50 connects to a wavelength demultiplexer 51. An optical signal received on link 50 comprises a set of individually modulated wavelength signals, called lambdas λ. The wavelength demultiplexer 51 demultiplexes the lambdas onto separate fibres 52. An optical-electrical (O-E) converter 54 converts each of the optical signals received on fibre inputs 52 to electrical domain signals. In the case of an intensity modulated (IM) optical signal, intensity (amplitude) of the optical signal represents a data value. The optical-electrical converter 54 is a photodiode having a suitable bandwidth for the expected bandwidth of the optical signal. The photodiode outputs an electrical signal which varies in amplitude over time. An electrical amplifier can follow the photodiode.

Monitoring units 100 process the electrical-domain signals to determine a measure of bit error rate. This will be described in more detail below. A switch 55 switches the electrical signals to a required output. Traffic may be switched directly to an output of the switch, for onward transmission to another node 20, or may be switched to framer 72 for processing. Framer 72 performs frame processing of the electrical signal. Frame processing includes Forward Error Correction (FEC) processing to detect and correct for errors. FEC processing derives a measure of Bit Error Rate (BER) of the received signal, based on the number of detected errors. An egress side of the node 20 has complementary functions to those described on the ingress side, including an electrical-optical converter 64, a set of fibres 62 carrying individual lambdas, a multiplexer 61 and an egress optical link 60. Typically, each direction of communication is carried on separate links 50, 60, although bi-directional communication along a common link is also possible.

Figure 4 shows apparatus 100 for monitoring error performance of an optical communications signal at the node of Figure 3. An input 101 receives an electrical domain signal. The electrical signal is applied to a clock and data recovery (CDR) unit 102, performing regeneration of the signal, and to the BER estimation unit 1 10. CDR unit 102 works at, or near, the optimum threshold. In a transmission system, the data signal is typically detected at optimum, or near-optimum, threshold for the specific received signal. The optimal threshold is determined for each WDM system channel during the pre-setting of the system. The threshold is the amplitude value that determines whether the signal sample is considered to represent a logical "0" or a logical "1". For example, if the sample value is greater than the threshold, then it is taken as representing a logical "1", otherwise it is taken as representing a logical "0".

BER estimation unit 110 comprises another clock recovery unit 1 12. This unit 1 12 samples the data signal and builds the corresponding amplitude histogram over a period of time. A sampler 113 extracts the analog samples from the electrical signal, sampling at the centre of the bit (maximum eye opening). Figure 5 shows the electrical-domain signal and a set of sampling points 120 separated by the bit period Tbit. This process can be described as synchronous sampling. An analog-to-digital converter (ADC) 114 converts each analog sample value to a digital value. Typically, 6 bits is a sufficient resolution, although other sampling resolutions can be used. Digital samples are applied to a processing apparatus 1 15. The processing apparatus 1 15 can be implemented as a field programmable gate array (FPGA), an Application Specific Integrated Circuit (ASIC), a general purpose processor executing code, or any other suitable manner, and will simply be called a processor in the following description. Processor 115 performs mathematical operations on the digital samples. Processor 1 15 uses a store 1 16 to store data. Store 166 can also store software for execution by the processor 115. Processor 115 provides an output 1 17 of the BER and can output an alarm signal 118. The alarm signal can be transmitted with the data to indicate poor data integrity to downstream devices. Additionally, or alternatively, alarm signal 118 can be output to a network management system 30.

Figure 6 shows steps of the overall method of estimating bit error performance. The method begins at step 151 by converting the optical communications signal to an electrical signal (e.g. performed by unit 54 of Figure 3). Step 152 samples the electrical signal to obtain digital signal samples (e.g. performed by sampler 113 and ADC 114 of Figure 4). Step 153 determines a statistical distribution of sample values. This can include collecting the digital samples over a period of time and building an amplitude histogram of the sample values. This can also include obtaining a probability density function (PDF) from the amplitude histogram. An optional, but advantageous, step 154 determines at least one scaling parameter value. The scaling parameter(s) can be obtained by a look-up operation in a store of previously obtained values, or by computation, using the probability density function (PDF). In the following description the scaling parameters are βθ and βΐ . Step 155 determines a measure of Bit Error Rate (BER) from the statistical distribution of sample values. Step 155 advantageously includes a further step 155 A of finding a minima value of the statistical distribution. Step 155 advantageously includes a further step 155B of scaling the minima value found at step 155 A by at least one of the scaling parameters determined at step 154. Advantageously, the method includes a step 156 of comparing the BER with a threshold value, and issuing an alarm, at step 157, if the measured BER is below the threshold value. The alarm can be sent to downstream nodes to warn them that the data is unreliable. The alarm can be sent to network management system 30, or can be used to initiate transfer traffic to an alternative network path, at step 158. The transfer of traffic can be made by the network management system 30 or by control plane signalling between nodes.

The method of processing the signal samples, corresponding to steps 153-155 of the method of Figure 6, will now be described in detail. This method is performed by processor 1 15. As described above, an electrical-domain signal is synchronously sampled. The corresponding amplitude histogram is obtained by using a suitable amplitude size-step (e.g. 50 or 100 bins). The histogram is proportional to the probability density function (PDF) Φ(χ) that describes the statistical distribution of the amplitude x before the decision. Therefore, the histogram can be used to estimate Φ(χ) with good accuracy. In order to formulate the PDF of the sampled values of the received signal X(t), we define the random variable x, representing X(t_k) at an arbitrary sampling time t_h The probability density function Φ(χ) involves the noise PDF, but it also depends upon the presence or absence of the signal pulse. We therefore have to work with conditional probabilities. In particular, if Ho denotes the assumption that the transmitted symbol is 0 and similarly, if Hi denotes the hypothesis that the transmitted symbol is 1 , we can write the mathematical expression of Φ(χ) :

Φ(χ) = P(H₀ )p(x \ H₀) + P{H_X )p{x I H_x ) (i) where p(x\H₀) is the PDF of logical 0s

is the PDF of logical Is. In order to simplify the notation, we define p₀ (x) =P(H₀)p (x \ H₀) and p₁ (x) = P(H₁)p(x \ H₁) .

Therefore we can rewrite (1) as:

Φ(χ) = ρ₀(χ) + _Ρι(χ) .

We will see that this PDF reflects the performance degradation, and from that it is possible to obtain directly a quite good estimation of the real BER of the optical channel. The bit error probability at optimum threshold is:

+ (2)

The regeneration error probabilities P_eo and P_ei corresponding to the decision rule implemented by the comparator are then given by:

and

\ p{x I H, )dx (3) where V is the threshold level. If that threshold is optimum, V=V_opt : this corresponds to the point of intersection of the two curves contributing to the PDF and this choice minimizes the BER. We note that if logical Is and 0s are equally likely, the source digit probabilities P(H₀) and P(Hj) are such that P(H₀)= P(Hi)=½ thus we get:

P_e=½ (P_e0+ P_el).

Now, if we limit our analysis to the region around V_opt, we can reasonably approximate the two contributes to the final PDF as a descending exponential for po(x), and an ascending exponential for pi(x):

ρ₀(χ) *€₀β^~χ/βο (4)

^( ) « e^{x// 1} (5) where Co, Cj, βο, βι are constants and - o^. _{1 Q}

A visual example of this assumption is given in Figure 7. The probability density functions are obtained by means of a standard Monte Carlo simulation for Eb/N₀=l l .25d , where Eb is the received signal mean energy per bit and N₀ is the two- sided noise power spectral density. In the logarithmic scale, the two exponential approximations indicated in (4) and (5) are shown as a linear trend. In order to graphically validate this assumption, we fit a set of PDF data points (in the optimum threshold amplitude region) to a straight-line and we obtain the curves in green (for the PDF of 0s) and magenta (for the PDF of l's). This linear approximation of po(x) and pi(x) is valid for values of x near the optimum threshold V_opt. As a consequence, we can substitute (4) and (5) in (3), and carry out the integration by taking advantage of the well known properties of the integrals of the exponential functions. Therefore we finally obtain:

P_e = oPo(V) + _lPl(V) (6)

We recall this works properly only for V close enough to V_opt. This expression gives an upper and lower bound for the value of BER. Now, we are in the position to estimate these upper and lower bounds for BER, which will be derived in the following. In order to do this, first we note that if β > β₀ , then we can easily see that

P_e≥ β₀Φ(Κ) and P_e≤ Α (*0 , hence:

Alternatively, if β₀ > ft , _e > β, (^ν) and P_e < Α,^φ(7) , hence:

This equation shows that the error probability is up- and low-bounded by two scaled version of the amplitude histogram function Φ(χ); in a logarithmic scale of amplitudes, βοΦ and β]Φ appear simply as two vertical translations of the original function Φ(χ).

Various approaches can be used to determine the coefficients βο and βι .

As an example, they can be derived from (4) and (5), thus giving: βο = ^~ (9) dx

The quantities po(x) and pi(x) are not known, but it is known that Φ(χ) is very close to po(x) when x is much lower than V_opt and it is known that Φ(χ) is very close to pi(x) when x is much higher than V_opt. Thus the above coefficients βο and βι can be approximated using (1) for values of x far from V_opt A distance of one order of magnitude is enough for accurate approximation. This is now stated mathematically:

Φ(χ)

A (11)

άΦ(χ)

dx

Other similar methods can be used, such as pre-estimating the coefficients off-line using an initial estimation of the above at link commissioning.

Numerical validation

In order to validate the above approach to estimating the BER, we use standard Monte Carlo simulations. Figure 8 shows the simulation model of the optical system. The transmitted signal is a standard non-return-to-zero (NRZ) ON-OFF keying (OOK) Gaussian pulse with a 3-dB bandwidth equal to 0.8/T, where T is the bit time. Nowadays, optical systems envisage the presence of optical amplifiers (OAs), and, as a consequence, the signal in the fibre is impaired by a noise that, in the linear regime, can be modelled as additive white Gaussian noise (AWGN). The signal is then optically filtered (ideal optical bandpass filter), photodetected, sampled, and finally processed at the receiver end.

An amplitude histogram Φ(χ) of 50 million samples is computed using a 0.01 amplitude integration step and transmitting a 32-bit de Bruijn sequence, so that all 5-bit interfering patterns are considered. The BER curves are obtained fixing a reference ^ value of Eb/N₀; then, Monte Carlo simulations are carried out for different threshold values V. The BER curves are convex functions with a minimum value at the optimum threshold value V_opt. The aim is to estimate the minimum value of the BER curve, corresponding to (2). This minimum can be calculated following the approach previously illustrated, i.e. according to the following sequence: collect samples; build the amplitude histogram from signal samples; obtain the PDF Φ(χ) after normalisation; estimate the coefficients βο and βι using (1 1)-(12); scale Φ(χ), as explained in (7)-(8). Following this approach, we obtained the numerical results shown in Figures 9 and 10. Figures 9 and 10 show two examples taken for different target BER values (and different Eb/N₀). These figures show curves representing: actual BER, Φ(χ), two scaled curves of Φ(χ). Figures 9 and 10 represent two different values of target Eb/N₀. It can be seen that there is excellent agreement between actual BER and scaled curves of Φ(χ). In both cases, according to (7), we have that β₀≥ β and hence β₀Φ(Υ) < P_e < βι (ν) . In each of Figures 9 and 10, the BER curve are lower- and upper-bounded by the two scaled versions of the PDF Φ(χ).

Advantageously, the two scaling parameters are derived, which should be properly far from the optimum threshold, but not so far away that the exponential approximation is no longer valid. A simple rule-of-thumb is that the approximation of (11) and (12) could be used also for quite different values than the optimum threshold. As an example, we can use them when Φ(χ) is one decade higher than its absolute minimum (on both left and right side of that minimum). That is far enough away from the optimum value that one of the two contributions is strongly dominating, whilst close enough to the optimum value that the approximation still holds.

It has been noted that typical real values of the scaling parameters βο and βι do not fluctuate strongly, and typically within one order of magnitude. Since the technique is aimed at providing the BER within the potential interest range (from 10^"3 to 10^"9) with a limited accuracy (half decade or higher), we can expect that a proper direct estimation of these coefficients may be avoided by relying on approximations, such as using a look-up table that reports typical coefficients for the most typical link configurations.

Figure 11 shows processing of a traffic along a signal path through an optical communications network 10 comprising nodes 21-24. FEC processing is performed at ^ an input to node 21 and an egress from node 24, i.e. at end points of the signal path through network 10. At intermediate nodes 22, 23 along the signal path BER of the signal is monitored using the method described above. It is not necessary to perform FEC processing at intermediate nodes.

Modifications and other embodiments of the disclosed invention will come to mind to one skilled in the art having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. Therefore, it is to be understood that the invention is not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of this disclosure. Although specific terms may be employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation.

Claims

1. A method of monitoring error performance of an optical communications signal at a node of an optical communications network, comprising:

converting the optical communications signal to an electrical signal;

sampling the electrical signal to obtain digital signal samples;

processing the signal samples to determine a measure of bit error performance, wherein the step of processing comprises determining a statistical distribution of the signal sample values and deriving the measure of bit error performance from the statistical distribution of signal sample values.

2. A method according to claim 1 wherein the step of processing the signal samples comprises finding a minima value of the statistical distribution.

3. A method according to claim 2 wherein the step of processing the signal samples comprises scaling the minima value by at least one scaling parameter.

4. A method according to claim 3 wherein the step of scaling comprises:

scaling the minima value by a first scaling parameter to form a first scaled minima value;

scaling the minima value by a second scaling parameter to form a second scaled minima value,

wherein the scaled first scaled minima value and second scaled minima value define upper and lower bounds of the measure of bit error performance.

5. A method according to claim 3 or 4 further comprising determining the at least one scaling parameter by one of:

performing a calculation to determine the at least one scaling parameter using the statistical distribution function;

performing a look-up of a stored scaling parameter value.

6. A method according to any one of claims 3 to 5 wherein the at least one scaling parameter is determined by at least one of:

and

where:

βο, βι are scaling parameters;

x is a sample value;

Φ(χ) is the probability density function of sample values;

V₀pt is an optimum decision threshold for logical values.

7. A method according to any one of the preceding claims wherein the step of determining a statistical distribution of the signal sample values comprises obtaining a histogram of amplitude values of the signal samples.

8. A method according to any one of the preceding claims wherein the step of determining a statistical distribution of the signal sample values comprises obtaining a probability density function of the signal samples.

9. A method according to any one of the preceding claims further comprising initiating a transfer of the optical signal to a different network path based on the measure of bit error performance.

10. A method according to according to any one of the preceding claims further comprising comparing the measure of bit error performance with a threshold value and issuing an alarm if the measure of bit error performance is below the threshold value.

11. A method according to any one of the preceding claims further comprising: converting the electrical signal to an output optical communications signal for transmission to another node without subjecting the electrical signal to forward error correction at the node.

12. Apparatus for use at a node of an optical communications network to monitor error performance of an input optical communications signal comprising:

a converter arranged to convert an input optical communications signal to an electrical signal;

a sampler arranged to sample the electrical signal to obtain digital signal samples;

a processing apparatus arranged to process the signal samples to determine a measure of bit error performance, wherein the processing apparatus is arranged to determine a statistical distribution of the signal sample values and derive the measure of bit error performance from the statistical distribution of signal sample values.

13. Apparatus according to claim 12 wherein the processing apparatus is arranged to find a minima value of the statistical distribution.

14. Apparatus according to claim 12 wherein the processing apparatus is arranged to scale the minima value by at least one scaling parameter.

15. Machine-readable instructions which, when executed by a processor, cause the processor to perform the method according to any one of claims 1 to 11.