This invention relates to an optical transmissio system.

A known optical transmission system includes a interferometer and a source of optical signals. Th interferometer comprises a four port optical coupler havin first and second input ports and first and second outpu ports, an optical coupling means coupling the first and secon output ports and including an optical non-linearity, and a optical amplifier. The source of optical signals is couple to the first input port of the interferometer.

An optical input signal coupled to an input port o such an interferometer is split into two portions by th optical coupler, which portions counter-propagate round th coupling means, for example an optical fibre loop, to retur to, and recombine at, the coupler. For a symmetric coupler, the optical path along the coupling means is the same for the two portions. So, for a 50: 50 coupler and a symmetrically positioned amplifier, the portions recombine such that the input signal emerges from the port to which it was originally input. The input signal is said to be "reflected" by the interferometer. For this reason, this configuration is often described as a loop mirror, the "loop" being the optical coupling means. The specification of our co-pending International patent application, publication number WO 88/02875, describes an interferometer having a non-linear optical coupling means, namely a silica optical fibre loop, in which the symmetry of the two counter-propagating directions along the coupling means is broken to provide a differential non-linear effect (and so is called a non-linear optical loop mirror or NOLM). This can be achieved in various ways. For example, a non-50: 50 coupler can be used. In this case, the intensities of the signal portions coupled into the ends of the waveguide loop are not equal. When the input signal is of sufficient intensity, the signal portions propagating in opposite directions around the waveguide experience different

refractive indices. This results in the two counter propagating signal portions experiencing different phas shifts, so that, when, the signals return to the couplin means, they have an. intensity-dependent relative phase shift. The intensity dependence of the relative phase shift result in a device whose output at an input port is, as is wel known, an oscillatory function of the intensity of the inpu signal. Any signal exiting the second input port Cthat is t say the port to which the input signal is not coupled) is sai to be "transmitted" by the interferometer.

A further way of breaking the symmetry of a NOLM is discussed in an article entitled "Nonlinear Amplifying Loop Mirror", by N E Fermann, F Haberl, M Hoffer, and H Hochreiter, Opt. Lett., 15, p.752, (1990), in which an amplifier is placed asymmetrically within the non-linear loop close to one of the output ports of the optical coupler, which in this case is a 50: 50 coupler. Such an arrangement improves the performance of the conventional NOLM, in particular by better exploitation of the waveguide loop non-linearity, as it can be accessed by a smaller input signal. The experiments described in the Fermann et aϊ article were carried out at low signal powers, and at repetition rates which did not saturate the gain of the amplifier. It was there noted, however, that amplifier saturation leads to a reduction in the overall gain of the device although, owing to the low pulse fluenσes, amplifier saturation in each individual pulse could still be neglected. Such a device is called a non-linear amplifying loop mirror (NALM).

Such NOLMs and NALMs can provide pulse shaping in optical transmission systems, and in particular provide pedestal suppression. Thus, these devices have the potential for the suppression of inter-pulse radiation, and for filtering bits in long-distance, all-optical communications systems. Such applications are discussed in an article entitled "Pulse Shaping, Compression, and Pedestal Suppression employing a Non-Linear Optical Loop Mirror" by Smith, N J Doran, and P G J Wigley, Opt. Lett., 15, p.1294 (1990).

A NALM could provide amplification in addition to suc pulse shaping in an all-optical communications system However, if the NALM has an oscillatory output, the intensit of the input signal must be relatively constant in order t avoid reflection by the loop mirror.

One way of removing the oscillatory output of a NALM i disclosed in an article titled "All-Optical Pulse Compressio Using Amplifying Sagnac Loop" by R A Betts, S J Frisken, C Telford and P S Atherton in Electronics Letters Vol 27 No.1 (9th May 1991). In their apparatus, the non-linear element i the loop is a semiconductor laser amplifier (SLA). Thi provides a saturating non-linearity which suppresses th oscillatory behaviour to provide a linear but risin response. A NALM which provides an approximately constan output would be more attractive for use in optica communications systems.

The present invention provides an optical transmissio system comprising an interferometer and a source of optica signals, the interferometer comprising a four-port optica coupler having first and second input ports and first an second output ports, an optical coupling means coupling th first and second output ports and including an optical non linearity, and an optical amplifier, the source of optical signals being coupled to the first input port of th interferometer, wherein the system is such that the optical signals saturate the amplifier thereby suppressing any oscillatory output, and such that the power of the optical signals is sufficient to switch an input signal coupled to the first input port to the second input port. This optical transmission system achieves an approximately constant output over a range of intensities of input signal, so that a range of intensities of input optical pulse will all be switched to the output of the interferometer. Moreover, the signals will be amplified to an approximately constant intensity. The optical transmission system, therefore, provides amplification of the signal, as well as pulse shaping and noise filtering as described in the

article by Smith et al. This is of particular application t optical communications systems.

The optical source may be a pulsed laser, in which cas the optical transmission system of the present inventio provides, at the second output port, noise-filtered optica pulses of substantially constant peak power, even for what ma be variable peak power input pulses. The system of th invention could, therefore, be used as a repeater in a lon distance optical communications link, for example a submarin link.

The optical amplifier may comprise part of the couplin means, as described with reference to the NALMs referred to above, or may be coupled to the first input port to amplify the input signals prior to their being switched. In this latter case, the interferometer will require the symmetry to be broken by, for example, a non-50: 50 coupler as the amplifier no longer forms part of the coupling means.

The interferometer may include an optical fibre loop, although other forms of waveguide may be used, for example, a waveguide formed in a planar substrate such as lithium niobate.

In the case of an optical fibre interferometer, the optical amplifier is conveniently an optical fibre amplifier spliced to the fibre forming the loop. Alternatively, a semiconductor laser amplifier may be employed.

The optical fibre of the loop may be made of material exhibiting the desired non-linearity, or a separate non-linear element may be included in the loop. For example, a highly non-linear element may be incorporated to shorten the loop length, for example a semiconductor laser amplifier.

The invention also provides a method of using an interferometer which comprises a four-port optical coupler having first and second input ports and first and second output ports, an optical coupling means coupling the first and second output ports and including an optical non-linearity, and an optical amplifier, the method comprising coupling a source of optical signals to the first input port of the interferometer in such a manner that the optical signals

saturate the amplifier thereby suppressing any oscillator output, and such that the power of the optical signals i sufficient to switch an input signal coupled to the firs input port to the second input port. Embodiments of the present invention will now b described, by way of example, with reference to th accompanying drawings, of which:

Figure 1 is a schematic representation of a non-linea optical loop mirror having an amplifier coupled to an input port;

Figure 2 is a schematic representation of a non-linear optical loop mirror including an optical amplifier asymmetrically positioned within the waveguide loop;

Figure 3 is a graph showing the power circulating in the loop for the non-linear optical mirror shown in Figure 1, and the resultant non-linear phase difference produced by various input peak powers;

Figure 4 is a graph showing the peak output power from the embodiment of Figure 1 as a function of peak input power for three pulse repetition rates;

Figure 5 is a graph of the non-linear phase of the embodiment of Figure 1 as a function of input peak power;

Figure 6 is a graph showing the gain of the embodiment of Figure 1 as a function of input peak power compared to the gain provided by the amplifier of the embodiment of Figure 1 alone;

Figure 7 is a graph showing the compression ratio provided by the embodiment of Figure 2;

Figures 8a and 8b are reproductions of an oscillogram showing the auto-correlation traces of input pulses with substantial inter-pulse radiation and pedestal free, compressed pulses amplified by the embodiment of Figure 1, respectively;

Figure 9 is a graph showing the auto-correlation width compression ratio as a function of input power; and

Figure 10 is a graph showing the gain verses average input power of the embodiment of Figure 1.

Referring to the drawings. Figure 1 shows an optica transmission system formed from a Sagnaσ loop interferomete 2 which comprises a four-port, fused-fibre optical coupler having first and second input ports 6 and 8, and first an second output ports 10 and 12. The output ports 10 and 12 ar optically coupled by an optical fibre loop 14. Th interferometer 2 is conveniently formed from a single optica fibre 14, two portions of which are fused to form the couple 4. In this embodiment, the loop 14 comprises an 8.8 km lengt of dispersion-shifted fibre with a dispersion zero around 1.55 μm obtained from Corning Corporation. The nature of this fibre ensures that pulse shaping due to propagation effects is negligible. Fibre polarisation controllers 16 are also included in the loop 14 to adjust the device to reflection mode at low powers.

A 30m long erbium doped fibre amplifier (EDFA) 18 is spliced to the output port 10 of the fibre coupler 4. An optical fibre coupler 20 is used to couple pump radiation for the EDFA 18 from a high-power MQW semiconductor laser 22 with a maximum pump power of the order of 50 mW at 1.48 μm. Under these conditions, the EDFA 18 has a small signal gain of 28 dB, and a time-average saturation power of 24 μW. For the above loop parameters, and an effective loop length of 7 km, the saturation power of the amplifier is of the order of 0.6mW.

An optical source 24 (an actively mode-locked semiconductor laser providing pulses at 1.545 μm of about 12 ps duration at a repetition rate of 2.5 GHz and a mean power of about 50 μW) is connected to the input port 6. The measured time-bandwidth products of the pulses produced by the laser 24 are at best 0.4.

It can be easily shown that, for the configuration of Figure 1, the square pulse trans issivity, T, is given by

T = P _t /P _j = G{1 - 2α(l - α) [1 + cos [(1 - α)G - α]φ]} .(1)

where φ(= 2πn ₂ P _r L/λA _tff ) is the non-linear phase shift, P _t and P _j are the transmitted and input powers respectively, α is the

power coupling coefficient of the coupler, L is the loo length, λ is the wavelength, n ₂ is the non-linear (Kerr coefficient (= 3.2 x 10 ^*20 m ² /W), A _ef , is the effective fibre cor area, and G is the power gain (P _out /P _in ) of the amplifier. Th switching power of the device, P _Sa , (= λA _eff /2n ₂ [(l - α)G - α] is derived by setting the argument of the cosine function t π. The use of the amplifier 18 to break the loop symmetr provides low switching powers, together with absolute pedesta suppression for α = 0.5. As an example, for G _ss = 30 dB, α 0.5 and L = 10 km, P _Sa is of the order of 0.25 mW (A _eff = 5 μm ² , λ = 1.55 μm).

Considering now the effect of gain saturation of th configuration shown in Figure 1, and assuming a gain of th form 1 + G _ss /(1 + P/P _sat ) where G _ss is the small signal gain an P _sat is the input power at which the gain is compressed by 3 dB. This simple equation describes well all the measured EDF characteristics for low to medium powers (<1 mW average), an also remains physically accurate in the highly-saturated regime. The influence of the gain saturation is best described with reference to Figure 5, where the argument of the cosine function in equation (1), that is to say the non¬ linear phase difference between the counter-propagating waves, is plotted against P _| , for G _ss = 30 dB, p _sat = 0.001 (= P _Sa ), α =0.5 and n ₂ L/λA _ef( = 1. At high input powers, the non-linear phase difference becomes clamped to πG _ss P _sat , which can also be expressed as πP _sat /P _Sa since P _Sa , is approximately equal to 1/G _SS for large G _ss . Therefore, by choosing P _sat = P _Sa we limit the maximum non¬ linear phase difference to π. The evolution of the non-linear phase is apparent in Figure 6, which shows the computed gain characteristics for sech ² intensity profile pulses (dashed curve). For comparison, the fibre amplifier gain characteristics are also shown for the same values of G _ss and P _sat (full curve). At low input powers, the device is in reflecting mode, and hence the small signal gain is well suppressed. As the input power is increased, however, the device approaches a transmitting state, and the efficiency closely follows that of the EDFA 18

for P, > P _Sa . It is expected that the varying response of th loop throughout the pulse gives rise to incomplete switchin and pulse shaping. Although this is largely responsible fo the small (2-3 dB) reduction in efficiency relative to th EDFA 18 at high powers, the loop amplifier benefits from puls compression and low-level light suppression.

In addition, since amplifier saturation gives rise t a non-linear phase difference which, over the power range o interest, is relatively constant, the pulse shapin characteristics are fairly insensitive to the input level. This is one of the key aspects of the present invention, an is illustrated in Figure 7, where the compression ratio (τ _out /τ _fn ) is shown to vary only from 0.55 to 0.75 over five decades of input power. This is in stark contrast to the complex pulse shaping previously observed for loop mirror configurations, where the input power can cycle through the sinusoidal output of such prior art NOLMs. Although not obvious for the range of input power in Figure 7, τ _0Ut A _in tends to unity for low power (linear) operation. Pulse durations are inferred from the second harmonic auto-correlation measurements. The auto-correlation shape of the transmitted pulses does not change significantly as a function of the input power, this being clearly illustrated in Figure 9, where the ratio of the input and output correlation widths is plotted against the input power for average powers up to 3.5 mW (120 W peak). It should be noted that the ratio of about 0.55 varies by less than 20% over a range of power of the order of 200 x P _Sa . The device gain follows the trend described in Figure 6, with a maximum of 17 dB occurring at an average input power of 50 μW (1.6 mW peak). The performance is well in keeping with that indicated by Figure 6, bearing in mind a 3 dB loss associated with the loop fibre 14 and a lower (28 dB) EDFA gain. It should also be noted that the measured tim -bandwidth products of the filtered pulses are essentially the same as the input.

A further clear demonstration of the intensity filtering properties is shown in Figures 8a and 8b. Here, the amplified, shortened (to 6 ps) and pedestal-free output

(Figure 8b) is shown for input pulses with substantial inte pulse radiation (Figure 8a). This behaviour is observed ov the total range of input power.

A further embodiment of the present invention is sho in Figure 2, in which the erbium amplifier 18 of Figure 1 now coupled to the input port 6 of the interferometer 2. Li elements are given the same reference numerals as in Figure 1 In this case, the symmetry of the NOLM is broken by us of a non-symmetric coupler 24 in place of the symmetric 50: 5 coupler 4 of Figure 1, and the switching power P _sb is that o the standard loop mirror (= P _Sa with G = 1) divided by the gai of the amplifier 18. The ratio of the switching powers of th devices of Figure 1 and Figure 2 is, therefore, given by

^p _Sb / ^p s _a = ft ¹ " ^α > ^G " «]/[(! - 2α)G] . (2

One can see that, for large G (which is generall true), equation (2) simplifies to Ps / ^p sa ⁼ ^ ~ ^α )/( ¹ ~ ^2α ) For a value of α = 0.4, the switching power advantage of th device of Figure 1 is at most a factor of 3. However, th real benefit of the device of Figure 1 is realised as approaches 0.5. In this case, since the fibre amplifier 1 breaks the loop symmetry, low switching powers are maintained together with absolute pedestal suppression for α = 0.5. A an example, for G _ss = 30 dB, α = 0.5 and L = 10 km, P _Sa is o the order of 0.25 mW (A _eff = 50 μm ² , λ = 1.55 μm). For th device of Figure 2, however, as α approaches 0.5 the switchin power rapidly goes to infinity.

Referring now to Figure 3, there is shown the powe circulating in the two counter-propagating directions as function of input peak power for the embodiment of Figure 1, where (a) is the power circulating anti-clockwise between th port 12 and the input of the erbium amplifier 18, and (b) i the power circulating in the loop in a clockwise directio from the erbium amplifier to the output port 12. The soli curve (c) of the graph of Figure 3 shows the non-linear phas shift of the pulses circulating in the two directions roun the loop as a function of input peak power; and, as can b

clearly seen, the non-linear phase shift becomes substantial constant at higher peak powers.

Referring now to Figure 4, there is shown a graph the peak output power from the port 8 as a function of pe input power (in W) of an optical signal input at the inp port 6 at three different pulse repetition rates f equal to 1 2 and 3 kHz. In this case, for f equal to 1 kHz, P _sat equal five times the switching power, P _Sa , of the interferometer o Figure 1. It can be seen that the output power is a oscillatory function of the input power. As the saturatio power moves closer to the switching power with increasin frequency, the peak output power becomes more nearly constant for peak input powers corresponding to P _Sa . It can b seen then that, if the interferometer 2 is operated in a optical transmission system such that amplifier saturatio occurs at approximately the power necessary to switch th input power to the second input port 8 at the first switchin peak, then approximately constant output power is achieve above the switching power. This provides pulse shaping an amplification characteristics which are relatively insensitive to the input power of the optical signals from the optical source.

Referring now to Figure 10 there is shown a graph of the gain of the embodiment of Figure 1 as a function of the average input power of the optical signals from the optical source 24.