OPTICAL NETWORK, AND METHOD, FOR THE TRANSMISSION OF WDM
CHANNELS.
This invention relates to optical networks for the transmission of wave division multiplexed (WDM) channels, and to a method for the transmission of WDM channels.
WDM optical networks require power levelling of the optical channels periodically throughout their transmission to ensure that no channel exceeds the dynamic range of the detector when it is finally received, or creates non-linearities within transmission elements (such as four wave mixing in transmission fibre and spectral hole burning in EDFAs) These power fluctuations between different wavelength channels are unavoidable and stem from various reasons, the most significant being optical amplifier residual gain variations, and wavelength dependent loss in components and in the transmission fibre itself. If uncorrected the gain variation, or lack of flatness, will build up throughout the system and may limit the maximum system performance.
Fixed gain variations in a line transmission system can be compensated for as the system is commissioned. This can be achieved by a number of techniques, for example, by pre-emphasis of the channels, i.e. changing the transmitted power of each channel in anticipation of the gain it will receive, or by installation of fixed gain flattening filter elements. Fibre Bragg gratings or dielectric filters are commonly used as fixed gain flattening filter elements. However, such a passive approach to gain flattening usually requires assumptions to be made about the source and destination of each channel, which may not be valid in more advanced communications systems based on reconfigurable optical add/drop multiplexers or optical cross-connects. A passive gain
flattening filter will also not compensate for effects such as span ageing and polarisation dependent loss which change the wavelength dependent loss of the system over both long and short timescales. To compensate for these temporal gain variations dynamic levelling is required.
Existing methods of dynamic levelling involve the implementation of variable optical attenuators into each wavelength path. There could be one variable optical attenuator placed immediately before each receiver, or more efficiently a dynamic gain equaliser may be placed at intermediate points in the transmission span: a combination of these may be employed. Such a dynamic gain equaliser typically consists of a demultiplexer- variable optical attenuator array-multiplexer combination. All the WDM channels come into the dynamic gain equaliser on a single fibre, and are split by the demultiplexer so that each channel takes a spatially separate path. Variable optical attenuators in these paths attenuate each channel to the desired level, determined by separate additional monitoring means, and the multiplexer then combines the separate paths back into one output fibre.
There are a number of drawbacks with dynamic gain flattening using the current approach. The most significant is that the channels must be attenuated to produce a flat output. This means than even if only one channel is too low, all the others must be attenuated, significantly reducing the total output power from the dynamic gain equaliser. This power loss must be recovered with further amplification, which degrades the optical signal-to-noise ratio and limits the number of spans over which the system can operate. Further, active monitoring is required to allow the output power from the
levelling element to be fed back to the attenuator to ensure that the desired output spectrum is maintained. This adds significant cost and complexity to the design.
US-A- 6 292 288 discloses a form of amplified levelling, using a multiple wavelength pumped Raman amplifier, in which a feedback signal reduces the wavelength dependency of gain to such an extent that the use of a gain flattening filter is not required. However, the Raman gain profile from a single pump wavelength is quite broad (~8THz), so complex, or highly sloped, gain profiles are impossible i.e. the differential gain between adjacent channels is limited. Also the single pumped Raman gain shape is reasonably complicated and the pump power required changes as more pump wavelengths are added. Therefore, calculating the pump power ratios required to realise a particular gain shape is a complex problem and the amplifier does not lend itself to being used as a rapidly reconfigurable dynamic gain equaliser.
US-A-5 225 922 discloses an optical transmission system in which the different gain of each channel of a WDM input is compensated for by separate optical amplification of each of the individual WDM channels under the control of a feedback signal generated by a detector for the respective amplified channel.
WO-A-01/45304 discloses an optical transmission system in which each channel of a WDM input is provided with a separate optical amplifier operated in a condition of saturation and pumped by a shared optical pump. While this reduces the initial cost of the system because of the sharing of the expensive optical pumps, the cost does include many other expensive components, such as optical amplifiers.
The invention provides an optical network for the transmission of WDM channels, comprising a demultiplexer for splitting the channels into a plurality of optical paths, the demultiplexer having a coarse stage for splitting the WDM channels into a plurality of sub-bands, and a fine stage for splitting the sub-bands into the plurality of optical paths, each optical path passing through an optical amplifier arranged to be in a condition of saturation to reduce variation in the signal powers of the incoming channels, a separate source being provided for pumping the optical amplifiers associated with each sub- band, wherein each optical amplifier is arranged to amplify more than one channel.
The invention also provides a method for the transmission of WDM channels, comprising the steps of demultiplexing the channels into a plurality of optical paths using a coarse stage for splitting the WDM channels into a plurality of sub-bands and a fine stage for splitting the sub-bands into the plurality of optical paths, each optical path passing through an optical amplifier being in a condition of saturation to reduce variation in the signal powers of the incoming channels, and pumping the optical amplifiers associated with each sub-band using a separate source, including the step of each optical amplifier amplifying more than one channel.
The passing of multiple channels through each optical amplifier provides substantial savings in the cost of the optical network.
The variation between highest and lowest output power may be less than 3dB, for example, for an input signal power variation of up to lOdB.
The common pump source may be diode or fibre lasers, and the optical amplifiers may be EDFAs.
An optical network, and a method, for the transmission of WDM channels in accordance with the invention will now be described in detail, by way of example, with reference to the accompanying drawings, in which:
Figure 1 is a diagram of a part of the optical network;
Figure 2 is a graph showing the characteristic of gain with input power of the optical amplifiers of the optical network of Figure 1;
Figure 3 is a graph showing the characteristic of output power with input power of the optical amplifiers of the optical network of Figure 1; and
Figure 4 is a graph showing the spectrum of input power of a typical WDM input signal.
Referring to Figure 1, the optical network consists of three main parts, namely, a demultiplexer and a multiplexer, indicated generally by the reference numerals 1 and 2, and a plurality of amplifier arrays, one of which is indicated generally by the reference numeral 3. The demultiplexer 1 receives incoming DWDM (Dense Wave Division Multiplexed) channels on fibre 4, and the multiplexer 2 outputs the amplified DWDM channels on fibre 5.
A banded approach is employed. The incoming DWDM channels to be flattened are first partially demultiplexed into M sub-bands by coarse demultiplexer 6. The sub-bands are then each subsequently further demultiplexed by fine demultiplexers 7, only one of which is shown, onto N fibres, one of which has the reference numeral 8.
The further demultiplexed channels then pass through respective optical amplifiers, one of which is indicated by the reference numeral 9. The optical amplifiers 9 are erbium doped fibre amplifiers (EDFAs).
The EDFAs 9 of each sub-band are all pumped from a common high power pump laser 10. The pump laser 10 is split with a 1:N coupler 11, for example, a combination of a number (N-l) of fused fibre 3dB couplers, or a passive planar lightwave circuit, so that all optical amplifiers 9 are supplied with constant pump power regardless of the input power present.
Figure 2 shows the gain of each EDFA with input signal power, and Figure 3 shows the resulting variation of output power with input signal power. For very low input signal powers, the EDFA amplifies the input power linearly, but the available gain reduces as input power increases. Point S represents a 3dB reduction in amphfication compared to that linear region. This means that the gain accorded to progressively larger signals becomes progressively less, reflecting a reduced population of erbium ions in an excited state. The pump laser could be arranged so that the expected i.e. nominal input signal level corresponded to the level S at which the EDFA enters saturation. Equally, the
EDFA could be run deep in saturation to provide a better levelling function, but this would have the result of a lower gain. To have higher gain, the EDFA could be run just below the saturation point, where some levelling function will still take place. A 6-lOdB range of input powers could correspond to a l-2dB range of output powers.
The gain of a saturated amplifier is dependent on the input power as well as the pump power. If the input power drops below the nominal level, the saturation of the amplifier is reduced and the gain will increase. Conversely if the input power to an amplifier is above its nominal level then the amplifier will be driven deeper into saturation (or gain compression), reducing the signal gain. The reduced gain means (Figure 3) that the output power of each EDFA 9 remains largely constant for a large range of input signal powers.
Further, because all the EDFAs of the sub-band are supplied with equal pump power, due to the 1:N coupler, the output powers of the EDFAs remain approximately equal irrespective of the input signal powers. The pump power to each amplifier cannot be changed independently of the others. By making the pumps 10 and couplers 11 for the other sub-bands the same as for the sub-band described, the output powers are brought to the same level regardless of the input signal power level without the requirement for complex monitoring and feedback (or feedforward in the case of pre-leveller monitoring) algorithms.
The levelled signals are then recombined together, in a similar manner to the demultiplexing. Thus, the individual channels are first multiplexed into sub-bands with
fine multiplexers, one of which has the reference numeral 12. The M sub-bands are then multiplexed using coarse multiplexer 13, to produce the output, levelled, WDM. With this implementation, the multiplexer and demultiplexer costs are spread over the life of the system, whereas the pump costs are spread over the number of equipped channels. Each sub-band has its own pump laser, so that if the number of WDM channels is increased at some stage in the future, the bulk of the additional cost would be in the provision of one or more additional pumps and combiners.
The EDFAs could be formed by a mini-Erbium doped fibre amplifier array. For example, this could be an eight-way pump used to drive eight individual EDFAs, resulting in a much smaller package.
Referring to Figure 4, in accordance with the invention, two adjacent channels use each EDFA 9. This reduces the expense of the levelling apparatus considerably. An incoming spectrum to be flattened may have the variation of power with wavelength shown in Figure 4, for example, varying between power levels A and B. The maximum variation between adjacent channels may be between power levels C and D. The laser pumps are arranged so that power levels C and D correspond to the region surrounding the saturation point S of the EDFAs 9.
The levelling function of the invention will still occur, but only to the degree of the flatness of channels that are co-amplified, that is, the output flatness will largely vary only between power levels C and D. This is significantly less than the input flatness (power levels A to B), and is suitable for some applications.
This embodiment can be realized by arranging that a pair of individual fibres 8 supply each amplifier, or by arranging that the fine de-multiplexer 7 splits the sub-band into pairs of channels, not necessarily adjacent in wavelength.
It is possible for three or four channels to use each EDFA. The final flatness will be limited to what it was within a group before the levelling apparatus. Thus too many channels using the same EDFA would reduce the achievable flatness in the output. All channels in a group must be present in the levelling apparatus, for otherwise the partially populated sub-band would appear to be a low power full sub-band and receive an incorrect gain.
As an example, the levelling apparatus could level 80 DWDM channels (with 50GHz channel spacing) on the transmission fibre 4 split into eight sub-bands of ten channels. The first stage demultiplexer 6 splits the input to 8 x 10 channels (M=8). The levelling apparatus thus requires a maximum of eight pump lasers, but initially only one need be provided (able to power the first ten channels), giving a low initial cost.
The second stage demultiplexer 7 de-multiplexes each ten channel group into five pairs of adjacent channels (N=5). Thus there are five sub-band amplifiers 9 each amplifying a pair of wavelengths. This is also a cost reducing measure, as the amplifiers are one of the most expensive components of the levelling apparatus. Other split ratios are possible, depending on trade-offs between cost, pump powers, flatness, levelling granularity, and the amount of active controls required. For example, 160 DWM
channels could be split into eight sub-bands of twenty channels feeding five sub-band amplifiers. The four channels could be adjacent to each other in frequency.
The pumps 10 could be high power diode lasers or high power fibre lasers. For an initially small number of channels, the diode pumps would be the less expensive option.
The technique of implementing power flattening described above has significant advantages over alternative methods. It requires no knowledge of the actual individual incoming channel powers, so that expensive monitoring and feedback mechanisms are avoided. This in turn removes the time constant associated with control electronics and provides a very fast flattening function.
The invention takes advantage of differential amplification to achieve dynamic gain flattening. Each group of channels receives a different amount of gain dependent on its input power, so that the output power of all the channels from the device are substantially equal. In this way each channel receives the minimum amount of amplification and attenuation necessary to achieve a flat output. Minimising the total amplification required also minimises the noise power added to the channel in the form of in-band amplified spontaneous emission.
Various technologies could be used for the optical amplifier array in place of EDFAs, for example, semiconductor optical amplifier arrays, erbium doped waveguide amplifiers (EDWAs). The important aspect is that they are driven with constant output power, regardless of input channel power. This is most simply achieved with an
amplifier operated with constant pump power. Their performance is not critical, gain flatness can be much worse than would normally be required for a WDM in-line amplifier. Similarly the per-channel output power from individual amplifiers can be much higher than is realistically achievable from broadband amplifiers, e.g. 15dBm from a single channel amplifier. While modest in itself, this amounts to over IW of total output power from a forty channel amplifier.
The invention is most suitable for implementation in reconfϊgurable optical add/drop multiplexer or optical cross-connect nodes where the multiplex and demultiplex functions must be present, although it could be implemented in point-to-point transmission systems. Further, metropolitan applications may benefit most, as these systems require a high level of add/drop availability and are traditionally most highly cost-sensitive.
The multiplexers 12, 13 and demultiplexers 6,7 may be realised with various technologies (e.g. arrayed waveguide gratings, bulk diffraction gratings, cascaded interleavers, cascaded dielectric filters, thin film filters, and Echelle gratings) and there is no requirement that they be identical.