Reduction of Interferometer Noise in an Optical Network

1. Field of the Invention

This invention relates to the reduction of crosstalk in an optical network and in particular concerns the reduction of interferometric noise such as incoherent beat noise.

2. Related Art

Interferometric noise may occur when an optical wave sheds a fraction of its power, which later interferes with its parent after having experienced a temporal delay. When the temporal deiay greatly exceeds the source coherence time, the interferometric noise is termed incoherent beat noise. If this condition is not met, the noise is partially coherent or coherent in nature. Several architectures in which such interference takes place have been identified, including delay elements in all-fibre optical bit rate limiters, reflections and Rayleigh backscatter in multi-stage amplified links, and crosstalk in wavelength division multiplexed (WDM) and optical time division multiplexed (OTDM) switching networks.

OTDM switching networks have been demonstrated to support high bandwidth traffic, and are additionally transparent to bit rate, coding format and wavelength. They may be configured to switch continuous high bit rate services such as high definition television (HDTV) or bursty traffic in asynchronous transfer mode (ATM) cells. OTDM switching networks employing amplitude shift keying (ASK) digital transmission/direct detection may be constructed from 2.times.2 switching elements interlinked by optical fibre delay lines (see the specification of PCT/GB92/00400). Inter-channel crosstalk arises in such networks because, in practice, the crosspoints do not possess perfectly isolated outputs. Every TDM channel emerging from the network is corrupted by the unwanted crosstalk waveforms.

In dealing with such unwanted crosstalk, known systems have concentrated on minimising incoherent noise-free crosstalk. The degradation resulting from this crosstalk may be modelled by a "sum of intensities" approach.

In accordance with the invention, it has been appreciated that interferometric noise, in particular incoherent beat noise crosstalk, can be more significant than incoherent noise-free crosstalk. Incoherent beat noise crosstalk is intensity noise arising from the interference of incoherent waveforms which originate either from very closely-matched wavelength sources such that their beat frequency falls within the receiver bandwidth, or from a single source with a differential delay in the optical network. Incoherent beat noise has been found to induce larger power penalties and bit error rate (BER) floors. The magnitude of the performance degradation induced is dependent upon the crosspoint isolation and the size and configuration of the network. It has been found, however that an isolation of -15 dB, limits the network size to three or four crosspoints, and this is a severe limitation when considering the design of switching networks.

In accordance with the present invention, it has been appreciated that interferometric noise, and in particular incoherent beat noise power arising when crosstalk and data originate from the same source, can be reduced in the optical network by varying the centre frequency of the source.

More particularly, in accordance with the invention there is provided a method of reducing interferometric noise power in an optical network through which data is transmitted as a stream of bits during successive bit periods by an optical source with a given centre frequency, to a receiver for receiving a frequency base band, comprising causing a variation in the centre frequency of the source such as to cause a redistribution of the interferometric noise power from the base band to higher frequencies, thereby reducing noise in the base band.

The method of the invention has particular application to the reduction of incoherent beat noise power.

The source conveniently comprises a laser diode and it has been found that according to the invention the centre frequency can be controlled by controlling the duration of periods when the laser is switched on. For a laser diode, it has been discovered that the centre frequency decreases during periods when the laser is switched on and increases during periods when the laser is switched off. This is likely to be due to heating of the laser chip, although other possible phenomena cannot be ruled out.

The network presents both a main signal path and, because of unwanted crosstalk, a subsidiary attenuated delayed signal path for the data transmitted to the receiver so that laser radiation from the main path that occurs during a given bit period suffers interference from laser radiation delayed from a preceding period. By producing the aforesaid rise and fall in frequency, the frequency difference between the interfering signals can be maximised, which causes a redistribution of the incoherent beat noise power from the base band to higher frequencies, which reduces the in-band noise.

Furthermore, the data may be coded so as to maximise the frequency difference between the centre frequencies for the interfering bit periods for which the laser is switched on.

The invention also extends to a network with reduced incoherent beat noise power, including an optical source with a given centre frequency operative to transmit data through the network as a stream of bits during successive bit periods; a receiver for detecting a frequency base band; and means for causing a variation in the centre frequency of the source such as to cause a redistribution of the incoherent beat noise power from the base band to higher frequencies, thereby reducing noise in the base band.

In order that the invention may be more fully understood an embodiment thereof will now be described by way of example with reference to the accompanying drawings in which:

FIG. 1 is a schematic illustration of an experimental apparatus, which lead to the invention;

FIG. 2A is a graph of a 32 bit test pattern used in the apparatus of FIG. 1, together with an interfering copy delayed by 16 bits and the x's show positions where interferometric noise is expected;

FIG. 2B illustrates the signal at the output, exhibiting interferometric noise, with a receiver bandwidth of 2 GHz;

FIG. 2C illustrates the output when the receiver has a bandwidth of 615 MHz matched to that of the data transmission;

FIG. 3A is a graph illustrating the optical output of the laser for a relatively low bit rate (9.72 Mbit/s);

FIG. 3B is a graph illustrating the variation in laser centre wavelength that occurs for the bit pattern shown in FIG. 3A;

FIG. 4A is a graph showing a relatively high bit rate (622 Mbit/s) optical output from the laser;

FIG. 4B illustrates the variation of the centre frequency of the laser for the output signal shown in FIG. 4A;

FIG. 5 illustrates schematically an optical network that operates in accordance with the invention; and

FIG. 6A illustrates as a solid line the modelled output centre frequency of the laser shown in FIG. 5 as a function of position within the bit sequence, and a corresponding 16 bit delayed signal, which produces interference, is shown in hatched outline;

FIG. 6B illustrates the transmitted bit sequence; and

FIG. 6C illustrates the difference between the laser frequency and the frequency of the bit of the same sequence delayed by 16 bits (i.e. the frequency difference between the solid and hatched lines of FIG. 6A).

Referring firstly to FIG. 1, this shows an experimental set-up which was used to model the characteristics of an optical switching network (or any other optical network in which incoherent beat noise crosstalk occurs). Light from a laser diode 1 is launched into an optical fibre 2 to a receiver 3 such as a photodiode that produces an electrical output 4. The laser diode, in a conventional manner, is supplied with a dc bias from a source 5. In addition, the laser diode is directly modulated by a modulator 6, resulting in the optical output as shown in the lower graph of FIG. 2A. In this example, a 32 bit test pattern was generated at 622 Mbit/s. The wavelength of the laser is typically of the order of 1538 nm.

In a switching network, a plurality of switching elements are usually connected together in a matrix. In theory, a single optical path is provided for a particular signal but in practice, crosstalk occurs between the various channels, resulting in the simultaneous existence of signal and delayed crosstalk waveforms that result from multiple paths to a particular receiver. In FIG. 1 this is simulated by a second optical fibre 7 that is coupled to the fibre 2 at positions 8, 9 in a manner known per se. An attenuator 10 together with a delay 11 are connected in the path of fibre 7 and typically, a 16 bit delay occurs for signals passing through the fibre 7. Thus, at the junction 9, the main signal travelling on fibre 2 suffers interference from an attenuated signal delayed by 16 bit periods, which simulates crosstalk in a switching matrix. This interfering signal is shown in the upper part of FIG. 2A. The output of the receiver 3 is shown FIG. 2B and it will be seen that the interfering signals produce incoherent beat noise crosstalk 12 in the output signals. The output shown in FIG. 2B illustrates the full bandwidth response of the receiver(2.0 GHz). If however the receiver 3 is tuned so as to have a narrower bandwidth, corresponding to the baseband of the data, it is found that the incoherent beat noise crosstalk is reduced, as shown in FIG. 2C. Arrows 13 indicate the positions at which the incoherent beat noise crosstalk is reduced as compared with the graph of FIG. 2B. Thus, it can be seen that at the position of arrows 13, there is an effect which causes the incoherent beat noise to redistribute its frequency spectrum outside of the base band of the receiver 3. The present invention makes use of this effect in order to minimise incoherent beat noise crosstalk. It has been found that the redistribution mechanism is related to the modulation sequence applied to the laser 1. This was investigated by measuring the centre wavelength of the modulated optical waveform produced by the laser 1, for different data bit rates. The centre wavelength was monitored using a scanning Fabry-Perot etalon (not shown).

Referring to FIG. 3A, when the modulator 6 was operated to produce a relatively low optical bit rate of 9.72 MBit/s it was found that the centre wavelength of the laser 1 varied with time as shown in FIG. 3B. It is postulated that the centre wavelength of the laser 1 is a function of its operating temperature and the device is subject to heating during the "on" periods of modulation. Thus, during the "one" bit period shown in FIG. 3A, the modulating signal from modulator 6 heats the laser and causes a progressive decrease of its operating frequency (with a corresponding increase in wavelength). Conversely, during a "zero" period of the bit waveform, in the absence of the modulating signal, the laser 1 cools with a corresponding increase in its operating centre frequency. Thus, the operating frequency cyclicly increases and decreases for successive "zero" and "one" bits, following a RC circuit like dependence with time. For the device under test, it was found that the 1/e time constant was equal to 22.5 ns.

Referring now to FIG. 4A. This shows the optical output when the relatively high bit rate 622 Mbit/s modulation is applied to the laser 1. The variation in wavelength of the optical output is shown in FIG. 4B. The relatively high modulation frequency is significantly greater than the reciprocal of the thermal time constant of the laser and as a result, there is a relatively small change (c.f. FIG. 3B) in the operating wavelength of the laser with time, from bit period to bit period. There is a sharp frequency transition between next adjacent zero and one bits, commonly called transient chirp, but this is not significant for the present analysis.

In accordance with the present invention, it has been appreciated that the dependence of the optical centre frequency of the laser on the bit rate can be used to maximise the redistribution of the incoherent beat noise power to frequencies outside of the base band of the receiver 3, thereby to reduce the in-band content. For a given data transmission bit rate, the optimum thermal time constant for the laser can be determined. The laser source should be chosen so that its time constant aligns as closely as possible with the optimum value. In this way, it is possible to maximise the frequency difference between the centre optical frequency that occurs during a particular "one" bit period, and that which occurs for the interfering delayed signal (in this example, a 16 bit period delayed signal). The frequency difference results in a redistribution of the incoherent beat noise power to higher frequencies outside of the base band of the receiver 3.

A simple example of a switching network which optimises filtering of incoherent beat noise power in this way, will now be described with reference to FIG. 5 which is based on the network shown in FIG. 13 of PCT/GB92/00400. The switching matrix consists of a plurality of 2.times.2 crossbar switching elements S1,1 . . . S1,7;S2,1 . . . S2,7 arranged in first and second rows R1,R2.The network allows optical signals that comprise bit trains in successive time slots applied to inputs I1,I2 to be selectively switched, reordered in terms of time slot and directed to outputs O1,O2 selectively. Thus, for example, considering switch S1,1 the input to I1 can be switched to an input of either switch S1,2 or S2,2 of row R1 or R2. Conversely, switches S1,6 and S2,6 can switch signals back between the rows R1 and R2. Successive switches in each row are connected by a direct optical fibre connection with a relatively short delay time e.g. connection C1, and also by a connection which incorporates a longer delay e.g. connection C2, in such a manner to enable signals from different time slots to be reordered in time. Reference is directed to PCT/GB92/00400 for a fuller explanation.

In FIG. 5, a DFB laser 100 feeds an optical bit stream to an optical fibre 200 that is connected to input I1 of the switching matrix S1. Output O1 is connected to a receiver 300, typically a photodiode and associated low pass filter tuned to the baseband of the data bit stream.

Considering now the bit stream produced by laser 100, when it is desired to switch the bit stream to the receiver 300, the switching elements S1,1 . . . S1,7 are set so as to provide a main, straight-through path for the signals, as illustrated by arrows 20. Additionally, due to the fact that the crossovers do not provide complete isolation between the various channels, crosstalk occurs, which is illustrated schematically by a second, longer path indicated by hatched line 21 between the laser 100 and the receiver 300. Thus, a delayed, attenuated version of the bit stream additionally reaches the receiver 300, which interferes with the signals in the main signal path 20.

In order to produce the modulated bit stream, the laser 100 is biased by a source 500, and is direct modulated by means of modulator 600 in the same way as described with reference to FIG. 100. In accordance with the invention, the laser 1 is selected to have a thermal time constant which will optimise suppression of incoherent beat noise power for the bit rate that is specified for transmission through the system, with the result that incoherent beat noise power in the baseband signal received by receiver 300 is minimised. For a given data transmission bit rate, the optimum thermal time constant for the laser can be determined. The laser source should be chosen so that its time constant aligns as closely as possible with the optimum value. In this way, it is possible to maximise the frequency difference between the centre optical frequency that occurs during a particular "one" bit period, and that which occurs for the interfering delayed signal (in this example, a 16 bit period delayed signal). The frequency difference results in a redistribution of the incoherent beat noise power to higher frequencies outside of the base band of the receiver 300.

The receiver 300 is tuned to have a bandwidth closely corresponding to the data baseband so that the incoherent beat noise power that is shifted to higher frequencies, is not detected and hence a reduction in noise occurs in the signals detected by the receiver.

In an experiment, six DFB laser chips constructed both from bulk and multiple quantum well (MQW) material of both ridge and buried heterostructure configuration, and mounted both p-side up and p-side down, were investigated in respect of the dependence of the laser thermal parameters on the laser construction. Although all the lasers demonstrated the RC like frequency change as a function of bit period, the thermal parameters were most favourable in a laser constructed as a p-side up buried heterostructure DFB. The time constant was approximately 35 ns in all cases. Generally, the p-side up mounted lasers showed a greater frequency change than their p-side down counterparts.

The laser source should be driven with a large modulation depth--this maximises the temperature difference between zero and one bits, and, additionally, maximises the temperature difference between the interfering one bits. In order words the laser should be biased at or below threshold. In networks which are limited by dispersion in addition to interferometric noise, a compromise between biasing above threshold, to damp relaxation oscillations of transient chirp, and maintaining a large modulation depth must be found.

Referring again to FIG. 5, an alternative method of inducing a time variation in the optical frequencies of the bits is to use a phase modulator. This is shown by way of example connected to the second input I2 of the switching network. A bit modulated optical signal stream from a laser source 22 illustrated schematically, is fed through a phase modulator 23, external to the laser source 22 to impart frequency variations between the interfering "one" bits.

An example of the waveforms produced in FIG. 2 will now be described with reference to FIG. 6. Referring to FIG. 6A, this shows the calculated frequency variation for the centre frequency of the laser, in response to a modulated bit stream as shown in FIG. 6B. Mathematically, the sequence dependence of the noise may be understood by considering the modulation history of the laser chip and its affect on the laser frequency. As far as the laser frequency is concerned, the laser was modelled by a linear first order system whose input and output are respectively the laser drive current and the laser optical frequency. The linear system itself is assumed to be made of a `capacitance` shunted by a `resistance` and is described by the following equation: ##EQU1## where f is the optical frequency C is the `capacitance`

R is the `resistance`

t is the time

The solid line in FIG. 6A corresponds to the signal stream 20 for the straight-through signal path, whereas the dashed line corresponds to the crosstalk signal 21. In this example, the crosstalk signal is delayed by 16 bit periods. FIG. 6C shows the wavelength difference between the two signals shown in FIG. 6A. The stars in FIG. 6B show positions at which incoherent beat noise power is to be expected within the baseband defined by the bandwidth of the receiver; however, it can be seen that, as compared with FIG. 2A, the incoherent beat noise power has been distributed out of band for the majority of bit periods in the bit stream by means of the invention, thereby producing a significant reduction in noise.

From the foregoing it will be understood that if for a given bit period, the coherence time of the optical source is reduced, the incoherent beat noise power is redistributed from the baseband to higher frequencies, thereby reducing the in-band content.

Another approach to exploit the phenomena, is to suitably code the transmitted data. For example, the number of consecutive ones or zeros in the transmitted bit stream can be limited to a predetermined number below which the resulting heating effect produces a saturation temperature. Alternatively, the data bits may be encoded into longer sequences.