WO1993005554A1

WO1993005554A1 - A method of suppressing relative intensity noise in coherent optical systems, such as communication receivers

Info

Publication number: WO1993005554A1
Application number: PCT/DK1992/000255
Authority: WO
Inventors: Cathal Mahon; Rune J. S. Pedersen; Morten HØGDAL; Gert Schiellerup
Original assignee: Nkt Research Center A/S
Priority date: 1991-09-02
Filing date: 1992-08-28
Publication date: 1993-03-18
Also published as: DK153791A; AU2571292A; DK153791D0

Abstract

A method of suppressing relative intensity noise in coherent optical systems, and especially local oscillator-generated intensity noise, the so-called RIN, in systems including heterodyne receivers by inserting a birefringent fibre (14), the length of which is adjusted such that the local oscillator noise cannot interfere with the desired signal/the signal to be received.

Description

Title: A method of suppressing relative intensity noise in coherent optical systems, such as communication receivers. Technical Field

The present invention relates to a method of suppressing relative intensity noise, the so-called RIN, from a laser or a corresponding coherent light source, in particular a semiconductor laser in a coherent optical communication system and especially local oscillator generated relative intensity noise (RIN) in optical communication systems using heterodyne receivers. The intensity noise of the semiconductor lasers can be very important for optical systems. The noise spectrum of the semiconductor laser is typically rather flat at frequencies far below the so-called relaxation resonance frequency so as subsequently to rise particularly sharply at the above-mentioned relaxation resonance frequency. Fig. 10 illustrates an example of a measured relative intensity noise spectrum.

In connection with an optical communication system ineluding a heterodyne receiver, the local oscillator-generated intensity noise can become the limiting factor for the receiver sensitivity. Local oscillator-generated relative intensity noise, subsequently referred to as RIN, is generated in the photodetector in form of a beat signal between the optical carrier wave and the noise from the local oscillator (described by Yamamoto, Y.: lEEE Journal of Quantum Electron. Vol. QE-19, pp 34-46). In this connection it should be kept in mind that the ideal condition is that the receiver is limited by shot noise.

Accordingly, RIN presents a problem when weak optical signals transmitted from for instance semiconductor la ser diodes are to be detected. RIN appears in digital communication systems inter alia in form of an increased tendency to bit errors. An increase of the bit-errorrate, BER, for a predetermined received optical power corresponds to a reduction of the receiver sensitivity. BER represents therefore a measurement of the noise applying to a communication system.

When RIN is of no importance, i.e. when RIN is negligible, it is possible to achieve the above ideal condition where the receiver is limited by shot noise by an increase of the local oscillator laser power entering the photodetector until the shot noise dominates compared to the thermal noise of the receiver. This is due to the fact that both the shot noise power and the intermediate frequency signal power are proportional to the local oscillator power which implies that the signal to noise ratio at the intermediate frequency level is increased until the shot noise limit has been reached. Subsequently, the signal/noise ratio becomes constant. As the noise power in the receiver originating from RIN is proportional to the local oscillator power squared, the above results can only be achieved provided RIN is far more insignificant than the thermal noise at the local oscillator power by which the shot noise-limited detection is achieved. If RIN instead is comparable with or larger than the thermal noise, nothing but finding a minimum in BER at a predetermined received signal power is achieved. An example of the latter situation is shown in Fig. 8, where BER is measured versus the local oscillator power entering the photodetector by a change of the coupling efficiency, i.e. by means of a coupler presenting a coupling coefficient which can be varied while the injection current of the laser is kept constant so that the operating conditions, the output power, the output frequency, and the noise spectrum of the laser are steady. It appears from Fig. 8 that the optimum LO power is about 0.35 to 0.40 mW with the used arrange ment. The optimum found depends on both the receiver used and the local oscillator used.

Background Art

Noise suppression is an item which has been and still is the object of much concern, said concern also applying to relative intensity noise in semiconductor lasers. Reference is made below to papers of relevance for the present invention.

An optical balanced heterodyne receiver with a local oscillator (LO) and two detectors is known from a paper by Gregory L. Abbas, et al: "A Dual-Detector Optical Heterodyne Receiver for Local Oscillator Noise Suppression" in Journal of Lightwave Technology, LT-3, No. 5, Oct. 1985. The receiver is designed to suppress local oscillator noise. The suppression is achieved by means of two detectors generating two electric signals and by utilizing the fact that RIN in the signals from the two detectors is strongly correlated because the noise originates from the same LO, and a substantial amount of the noise can therefore be electrically cancelled. The cancellation requires, however, a careful matching of the two signal paths from the splitting of the optical signal until the cancellation occurs. The latter procedure requires a careful measurement and adjustment of the path lengths, i.e. the fibre lengths, as well as of the amplitude and the phase-response of the photodiodes as well as of their sensitivity.

The above procedure is particularly difficult at high bit rates where the bandwidth requirement is excessive because the matching must be good also at the highest frequencies involving the shortest wavelength.

An interferrometric technique for reducing the intensity noise in semiconductor lasers is described in a paper by M.A. Newkirk and K.J. Vahala: "Amplitude-Phase Decorrelation: A method for reducing intensity noise in semiconductor lasers", IEEE Journal of Quantum Electronics, Vol. QE-27 No. 1, Jan. 1991, pp 13 to 21. The technique is based on the fact that amplitude and phase-fluctuations are (inherently) correlated so that the amplitude fluctuations can be reduced by allowing the laser light to pass through a component with a suitable frequency versus amplitude response. As the technique utilizes the fact that a correlation exists between the amplitude and phase fluctuations, it is, of course, only efficient where such a correlation does in fact exist. The paper indicates that the optimum suppression is achieved at low frequencies, whereas the suppression decreases proportional to increasing frequencies so as to vanish completely at frequencies exceeding the relaxation resonance frequency. This is due to the fact that the correlation existing due to pertubations of the carrier density cannot be maintained in a period much shorter than the relaxation oscillation damping time. Accordingly, it is impossible to achieve an efficient suppression at high frequencies and indeed at the relaxation resonance frequency of the laser, where RIN is dominating. This involves heavy limitations for the design of a heterodyne optical system with a high intermediate frequency.

DE-OS 26 50 500 discloses a polarization-colour-filter device for separating discrete lines in a laser beam by changing the direction of polarization of the laser beam. A quartz body is used for rotating the direction of polarization of the laser beam, and the thickness of the quartz body in the direction of propagation of the beam is chosen such that the two discrete laser beams to be separated are polarized in two orthogonal directions and subsequently sent through a polarization filter or a polarization splitter.

DE-OS 29 OS 752 discloses an optical transmission filter with a slightly birefringent optical monofibre. The filter is made of a fibre with a solid core of silica glass which per se can present a linear birefringence of 100 rad/m. The fibre is a monomode fibre, in which the light can only propagate in two orthogonal polarization states when 2 π d/λ (n_K ² - n_M ²) < 2,4, where n_K represents the refractive index of the fibre core and n_M represents the refractive index of the fibre cladding. The optical fibre is typically a weakly guiding monomode fibre, where the difference between the two refractive indexes is only about 0.5%. The fibre is fixed on a number of locations where it is non-rotatably fixed. As a result, the fibre is split into a number of delay sections and a number of intermediate coupling sections. Each section presents a predetermined torsion with the result that the fibre has a transmission response comparable with a Sole-filter made of birefringent crystal plates and two linear polarizors arranged in front of and behind the birefringent crystal plates.

EP 0 018 874 discloses a multispectral op to-electronic receiver system with a large dynamic range for the received light intensity and with optical means for reception and focusing of the light beams so as to generate an image of an observed field on an op to-electric transducer. The system includes further a spectral filter for the selection of a spectral band as well as an attenuation filter capable of stepwise attenuating the light intensity and being position controlled by an assigned device. The spectral filter comprises several separate elementary filters arranged on a support and each selecting a specific spectral band. The attenuation filter comprises as many identical elementary attenuation filters as the spectral filter, each of said elementary attenuation filters of the attenuation filter being associated with an elementary spectral filter in such a manner that each elementary filter provides the same attenuation at the same setting.

Description of the Invention The present invention suggests a method of suppressing RIN, i.e. relative intensity noise, from a laser or a corresponding coherent light source, especially a semiconductor laser in a coherent optical communication system, and the method is characterised in that a birefringent optical member is inserted in the optical signal path, and that the length of the signal path in the birefringent optical member is adjusted such that the polarization of the aimed/desired optical laser signal and the relative intensity noise (RIN) of the laser are substantially orthogonal at the exit of the birefringent member. As explained more detailed in the detailed portion of the description, the birefringence ensures that the member can possess the necessary frequency-dependent polarization transfer function.

The features stated in the characterising clause of claim 1 ensure that the RIN signal and the laser signal aimed at cannot co-operate or, in other words, interfere. Accordingly, mixing products cannot be generated by these signals on the surface of a succeeding detector. In order to understand the latter it should be kept in mind that the E-field vectors of the desired and the undesired signal, respectively, are orthogonal when they hit the detector surface and cannot therefore be mixed in the detector.

The birefringent member is preferably a high birefringent fibre, a so-called HI-BI-fibre. The length of the HI-BI-fibre is adjusted such that the desired laser signal and the undesired laser-RIN-signal are orthogonal at the exit of the fibre, and especially on the surface of a succeeding detector. A polarizor may optionally be mounted on the exit of the birefringent fibre and in front of the detector. The polarizor is adjusted such that the undesired noise signal is blocked and the desired signal is allowed to pass substantially freely.

It is, however, preferred to omit the polarizor in order to avoid the requirements presented to the control and stability of the arrangement, said requirements being a proviso for a correct operation of a polarizor.

The method has several potential fields of application, such as systems using direct detection to sensors and optical communication and especially for heterodyne transmission systems for coherent communication.

For the sake of completeness it should be mentioned that it is known from a paper by G. Schiellerup and R.J.S. Pedersen: "ASK transmitter for high bit rate systems", SPIE, Coherent Lightwave Communications, Vol. 1372, pp 27 to 38, to utilize a HI-BI-fibre followed by a polarizor for the conversion of a frequency-modulated signal into an amplitude-modulated signal. The purpose of the known use is quite different from the use according to the present invention, and such a use of the fibre does not allow a splitting up of the signal and the noise. Accordingly, the known use differs highly from the use according to the present invention. According to a particularly advantageous method according to the invention, the birefringent member or fibre may be inserted in the receiving section of a coherent optical communication system of the type operating according to the heterodyne principle and receiving an information-carrying laser signal by means of a local oscillator laser and a detector, which operates as a mixer providing an intermediate frequency signal - the polarization of the local oscillator signal is adjusted so as to assume that partially equal amounts of power are coupled to the two axes of the birefringent fibre - and the length of the birefringent fibre may be adjusted such that the local oscillator-laser signal and the undesired laser-noise signal are orthogonal on the surface of the detector. The polarization for the information signal and for the local oscillator signal should simultaneously be adjusted such that the two desired signals achieve the same polarization at the exit of the birefringent member so as to effectively provide an intermediate frequency signal.

The use of a high-birefringent fibre in an heterodyne receiver is indeed known from a paper by C.J. Mahon, G. Schiellerup, R.J.S. Pedersen entitled: "Simple, all- fibre optical image-rej ection mixer for optical heterodyne receivers" published on the 16th European Conference on Optical Communication, ECOC90 , in Amsterdam, pp 1029 to 1032. The fibre was here dimensioned and placed such that a suppression of the image-signal was obtained. The obtained advantages are that the communication channels of a multi-channel coherent communication system can be more densely stacked, i.e. more channels within the same bandwidth, and thus a more efficient use of the bandwidth is obtained. As it appears from the above explanation, the paper deals with a use of the optical fibre which is completely different from the use according to the present invention. Another embodiment of a heterodyne receiver with a suppression of the image frequency is described in EP-A2 0 245 026. The described suppression is, however, obtained by way of principles differing highly from the principles of the present invention.

In addition to suppressing the image signal, an image rejection receiver suppresses also the ASE, i.e. amplified spontaneous emission, located at the image frequen cy, which is described in greater detail in a paper by B. F. Jørgensen, B. Mikkelsen, C.J. Mahon, S. Saito: "Simple method to improve the dynamic range of optical amplifiers in coherent optical communication systems with heterodyne receivers", Electronics Letters, Vol. 27, No. 7, 1991, pp 611 to 612. The beat signal between the local oscillator and ASE from the optical amplifier was reduced to one half by removing the ASE at the image frequency. Consequently, the dynamic range of the ampli- fier used is enlarged considerably. As it appears from the above explanation, such a use differs also essentially from the use according to the present invention.

The invention relates furthermore to an arrangement for carrying out the method according to the invention and dealt with in claim 9, as well as to a heterodyne receiver, a mult i-channel communication system, and a polarization diversity receiver as dealt with in claims 10, 11, and 12, respectively.

Brief Description of the Drawings

The invention is described in greater detail below with reference to examples and the accompanying drawings, in which

Fig. 1 illustrates an arrangement with a laser, a HI-BI- fibre, and a detector, Fig. 2 illustrates the same arrangement with a polarizor placed in front of the detector,

Fig. 3 illustrates a similar arrangement with several HI-BI-fibres and polarizors alternately arranged,

Fig. 4 illustrates a typical optical heterodyne receiver, in which a HI-BI-fibre is inserted in the detector branch, Fig. 5 illustrates a measurement arrangement for the measurement of BER,

Fig. 6a outlines the varying polarization along a birefringent fibre,

Fig. 6b shows an example of possible polarization of signals at the entrance and the exit of a birefringent fibre,

Fig. 7 shows the transmission response of the intermediate frequency section used in the measurement arrangement of Fig. 5, as well as the calculated suppression of RIN,

Fig. 8 shows a curve illustrating an example of BER versus the local oscillator power,

Fig. 9 shows examples of measurement results, and

Fig. 10 shows an example of a measured RIN spectrum.

Best Modes for Carrying Out the Invention The following examples are all to be considered parts of communication systems, where only units relevant for the present invention have been shown for the sake of clarity. Fig. 1 shows an example of an arrangement including a semiconductor laser LD, a lens 12, an isolator ISO, a lens 12', an HI-BI-fibre HI-BI, and a detector 16. The two lenses 12, 12' are used for coupling from the laser LD and coupling into the HI-BI-fibre HI-BI. The isolator ISO prevents reflected light from being fed back into the laser, said reflected light otherwise optionally having a significant influence upon the performance of the laser. From a very simplified point of view it can be said that the signal from the laser LD in principle comprises two components: a pure (desired) signal at a predetermined frequency and a noise peak at a slightly different frequency. The difference frequency is the so-called relaxation resonance frequency.

If the laser signal was detected directly without the HI-BI-fibre, a "beat noise" component would be generated due to the mixing of the two frequency components on the detector surface.

This noise can be suppressed by the use of the HI-BI- fibre. At the entrance of the fibre 14, the signal and the noise components have identical polarizations. However, due to their mutual frequency difference, the two components experience the polarization rotation in the transmission through the fibre differently. By orienting the HI-BI-fibre so that an equal amount of light is coupled into the two fundamental axes of the HI-BI-fibre (for instance by coupling a linear polarization 45° to the two axes), the polarization of the light changes from the entrance 31 of the fibre to the exit 32 of said fibre, cf. Fig. 6b. When the polarization is observed after a predetermined length of path, the polarization is constant provided the frequency of the light is constant, cf. the text on Fig. 6a. According to the invention the length of the fibre is chosen such that the frequency difference between the carrier frequency and the noise frequency has exactly the effect that the polarization of one of the two frequencies is orthogonal to the polarization of the other frequency, whereby the two components are orthogonal at the exit of the fibre where the two components hit the detector surface. Consequently, the two components cannot co-operate on the detector surface, and no mixing products can originate from these components. The noise arising when no birefringent fibre is used due to the interference between the two components, viz. signal and noise, has thus been eliminated by the use of a suitable length of the fibre.

As signal and noise are separated with respect to polarization, the noise can also be removed purely optically by means of a polarizor after the HI-BI-fibre and before the detector, cf. Fig. 2. In the latter case, the polarization of the desired signal must be a linear polarization at the exit of the HI-BI-fibre, and then the polarizor is rotated so as to allow free passage of the desired signal and to block the passage of the undesired signal, which illustrates the fundamental principle.

Suppression of noise is, however, perfect for only one specific frequency difference. Since real laser signals are spectral distributions and not single frequencies, the real noise suppression is substantial, but not perfect.

The arrangement shown in Fig. 2 is almost identical to the arrangement of Fig. 1 with the only difference that a polarizor 18 is inserted between the HI-BI-fibre 14 and the detector 16. The polarizor 18 can be rotated in such a manner that it stops the undesired portion of the optical signal. As the desired signal is orthogonal to the undesired noise after its passage through the HI-BI- fibre, the polarizor can be adjusted such that it only allows passage of the desired laser signal.

The above arrangement can be additionally extended as indicated in Fig. 3 by adding more HI-BI-fibres 14, 24, ..., each fibre being followed by a polarizor 18, 28 .... Each section has a frequency to polarization transfer function, and when coupled in series the de sired transfer function is generated by adjusting the polarizors and the individual lengths. In this manner the suppression is achieved in a predetermined frequency band.

Fig. 4 illustrates a completely different use of the principle of the invention, said use being of major practical importance. The Figure shows a simple diagram of an arrangement including the signal laser TL, the local oscillator laser LD, the coupling lenses 12, 12', isolators ISO, polarization controllers POL CONT, a fibre coupler FC, a length of HI-BI-fibre, and a detector. The lenses and the isolators are used as described above.

The arrangement simulates a typical heterodyne receiver, where information, if any, is detected at the difference frequency between a possible information-carrying signal SI from a signal transmitter TL and a local oscillator signal LO from the local oscillator LD.

The polarization controllers are used for adjusting the polarization on the entrance of the HI-BI-fibre individually for the signal SI and the local oscillator LO.

The two signals are added in the coupler, whereafter the total signal is coupled into the HI-BI-fibre. The polarization of SI and LO is adjusted by means of the polarization controllers so that their polarizations are identical at the exit of the HI-BI-fibre.

The length of the fibre in the example shown in Fig. 7 is chosen such that noise frequencies at a frequency distance from the local oscillator signal corresponding to the intermediate frequency obtain a polarization at the exit exactly orthogonal to the polarization of the local oscillator at the same exit. The noise frequencies are therefore suppressed as shown in Fig. 7, where the RIN-suppression, the curve A, as well as the response of the intermediate frequency filter in the curve B are shown. The receiver presents typically a band-pass response, the centre of which is the difference frequency.

As a general rule the fibre length must be chosen such that the integral of the product of the noise signal and the intermediate frequency response above the relevant frequency range assume minimum or in other words: The integral of the resulting noise in the IF-receiver is minimized. In the above situation it is practical first to consider two extreme, highly simplified situations:

1. If it is assumed that the receiver has no bandwidth restriction, i.e. in practise it presents a lowpass response with a suitably high cut off frequency, the noise suppression operation can be explained as follows:

The length of the HI-BI-fibre is selected as before, so the signal-noise beat signal is suppressed maximally at the relaxation resonance frequency of LO. The polarization controller, POLCONT. 20, in the signal arm is adjusted to its optimum interference, i.e. the heterodyning efficiency, between the local oscillator signal LO and the received signals SI. This is obtained when the polarization of the local oscillator signal is parallel to the polarization of the received signal on the detector surface, i.e. when the two E-field-vectors are parallel. 2. If it is assumed that the local oscillator noise has a flat frequency-independent power spectrum, i.e. white noise, and the receiver has a true band-pass re sp ons e :

In this case the optimum noise suppression is obtained when the length of the HI-BI-fibre is selected so as to provide an optimum noise suppression at the intermediate frequency in the centre of the band-pass.

A real application includes a local oscillator with a relaxation noise peak and a receiver with a band-pass response, i.e. a combination of the above two situations. Here the optimum fibre length depends on a careful estimate of the overlap between the various responses. In systems involving a high intermediate frequency, the optimum fibre length corresponds probably to a point on the right side of the relaxation noise peak, as the relaxation resonance frequency must be assumed to be slightly below the intermediate frequency. In order to explain the invention in greater detail, some calculation examples are stated below.

EXAMPLE A In the following a field is calculated, said field being generated at the detector 16 by a signal with the angular frequency ω₀ transmitted into the fibre 14 of the arrangement of Fig. 1. The signal from the laser diode

LD can be written as follows: {¹⁾

where e_o and e_e are unit vectors of the ordinary and the extraordinary axis, respectively. Here an equal excitation of the two axes is assumed. If it is desired to consider the possibility of varying the polarization state of the signal on the entrance of the fibre, the signal can be written as follows:

(1a)

where θ_1d defines the polarization condition at the entrance of the fibre. When the fibre length is called L, the velocity of the light c, and the refractive index along the ordinary axis n_o, the propagation time of the signal propagating along the ordinary axis is: T_o = n_o L/c (2a) and of the signal propagating along the extraordinary axis :

T_e = T_o + Δn L/c = T_o + τ (2b) where Δn represents the change in refractive index from the ordinary to the extraordinary axis, and where τ = Δn L/c.

The field at the fibre exit is then :

α_o and α_e being the attenuation coefficient in the fibre

It is assumed that α_o = α_e = α . This does not imply any restrictions as an unbalance can be compensated for by changing the weighting of the excitation on the two axes. The above has the following result for the signal at the exit of the fibre: (5)

In order to understand the physical meaning of the above expression, reference is made to Fig. 6a showing how the polarization varies along the birefringent fibre.

In the Figure, θ_1d = 0°, and the Figure shows the polarization for four different lengths of the fibre (+/- an integer number of wavelengths),

ω₀τ = p • 2 π, ω₀τ = π/2 + p • 2 π ,

ω₀τ = π + p • 2π, ω₀τ = 3π/2 + p • 2π,

It appears, that as far as ω₀τ = p · 2 π is concerned, the field is linearly polarized towards the vector

e_o + e_e. lf the fibre length is increased by π/2, a circularly polarized field is obtained.

In case the fibre is ω₀τ = π + p · 2 π , the field is again linearly polarized, but orthogonal to the vector e_o + e_e , and in case ω₀τ =3 π/2 + p · 2 π, a circular polarization is again obtained.

Thus the polarization of the field or the light on the exit of the fibre is determined by the polarization of the field on the entrance of the fibre as well as of the length of the fibre.

EXAMPLE B This example deals with the arrangement shown in Fig. 2 where a polarizor 18 is placed in front of the detector 16. When it is assumed that the direction of polarization for the polarizor 18 is parallel to e_o + e_e, the following result is obtained:

E_0p = E₀·(e_o+e_e)·(e_o+e_e)=K1·(1+e^-jω0^τ)· (e_o+e_e)·e^jω0(t-T₀)

A detection by means of a photodiode results in generation of a photocurrent I₁ which is proportional to the vector of Poynting: P = E x H, i.e. I_pd α |P| α E_0p · E_0p* which accordingly results in a detector current (disregarding θ_1d): I_pd α |E_1d|² (1+cos(ω₀τ))/2 = '| E_{1 d}'| ² cos² ( (ω₀τ ) /2 ) showing that the optimum detection condition with a polarizor is obtained for ω ₀τ = 2 pπ where ω ₀ is the angular frequency desired to be detected.

EXAMPLE C

As the laser signal, as mentioned, usually also includes an unwanted noise peak at a slightly different angular frequency ω', the fibre length is advantageously adjusted such that ω₀τ = 2p π, and

ω'τ = π + 2p'π at the angular frequency ω' at which the noise is usual ly concentrated and which should therefore be avoided.

It appears from the above, that the frequency difference Δω = ω₀-ω ' between the desired angular frequency ω₀ and the noise frequency ω ' determines the length of the fibre, as

Δω τ = 2pπ - π - 2p'π = π + 2p"π = Δ ω Δn·L/C based on which it is possible to determine the optimum length of the fibre to be:

L = (2p"+1) • π • c/(ΔωΔn) where p" merely need to be an integer or 0.

When the possibility of another polarization condition θ_1d on the entrance of the fibre is considered, the expression of the detector current I_pd is modified to be:

I _{p d} α |E_1d|² (1+cos(ω_0τ-θ_1d))/2 = |E_1d|² cos² ((ω₀τ-θ_1d)/2)

As the polarization state of the two signals ω₀ and ω' probably is the same because both signals originate from the same laser, the same result as above for the optimum fibre length is obtained.

EXAMPLE D Now the situation in a heterodyne receiver with two signals with different angular frequencies ω ₀ and ω ₀ + Δ_ω is considered, the detector current for the two mixed signals being calculated. The expressions and the terminology are analogous with the above . θ_a and θ _b represent the polarization states of the respective signals on the entrance of the fibre. On the exit of the fibre the signals can be written as (equivalent to (5))

(9a) ) ( 9b)

The detection results in a generation of the photocurrent I₂ = K₂ (E_a + E_b) · (E_a* + E_b*) which can be written as

= K3 + K2 • K_a • K_b •2 [cos(Δω(t-T₀)) +

cos(Δω(t-T₀) - (θ_a-θ_b) - Δωτ)] (10) where K2 and K3 are constants.

In the above expression, first of all the constants are disregarded, and only the part presenting the intermediate frequency signal, i.e. the time and frequency-dependent part of (10) is discussed:

I₂'(t,Δω) = cos(Δω(t-T₀)) + cos (Δω( t-T₀) - ( θ_a-θ_b)-Δωτ) (11) which also can be written as

2 cos[Δω(t-T₀)-(θ_a-θ_b+Δωτ)/2] x cos [(θ_a-θ_b+Δωτ)/2]

It is noted:

(11) can assume amplitude values between 0 and 2 depending on the adjustment of the time-dependent phase part in the second term. When θ_a-θ_b = 0, the expression can be reduced to

I₂' (t, Δω) = 2 cos(Δω(t-T₀)+Δωτ/2) x cos(Δωτ/2) (11b) which in case of Δω τ = 2pπ can be further reduced to:

I₂' (t, Δω) = 2cos(Δω(t-T₀)) i.e. the best possible conditions for detecting an intermediate frequency = Δω _MF at the end of the fibre are obtained when Δω _MF τ = 2pπ ,

For Δω τ = π + 2pπ , the expression can be reduced to I₂' ( t,Δω) = 0,

(i.e. Δω cannot be detected at the end of the fibre).

It means that Δω _RIN as a mixing product of RIN from the local oscillator and the pure local oscillator signal can be eliminated by chosing the fibre length such that Δω_RIN · τ = π+2pπ . As a general rule, it is possible to obtain a constructive or destructive interference, i.e. maximum signal and cancellation, respectively, by adjusting θ_a, θ_b, and τ (i.e. the fibre length) in such a manner that the two branches are of opposite-phase at a predetermined Δω .

In the case of noise suppression in a "solitary" laser, θ_a - θ_b Δ ω is defined at the resonance peak, and the fibre length and consequently τ is selected so that cancellation is obtained at this frequency distance (cf. example C).

However, in the case of the heterodyne detection it is slightly more complicated. Here two beat signals are of importance: 1: Beat between LO-signal and LO-noise. This beat is suppressed as described above by a suitable choice of the length of the birefringent fibre, as θ_a = θ _b .

2: Beat between received signal and LO-signal: In this general case, it concerns a beat frequency, viz. the intermediate frequency IF, different from the beat frequency between the LO-signal and noise. Here a polorization control of (θ a - θ _b) is utilized for ensuring an optimum heterodyne signal at the intermediate frequency.

As two degrees of freedom applies to the system, viz. τ and (θa - θ _b), it is possible to simultaneously optimize both the noise suppression and the heterodyne efficiency.

Measurement

In order to additionally explain the invention, the method according to the invention was subjected to a test by way of experiments in form of a transmission test, where the effect of the method according to the invention is measured as an improvement of the receiver sensitivity, which can be determined as a reduction of the number of bit errors, BER, for a constantly received signal power.

A measurement arrangement as shown in Fig. 5 was used. The arrangement corresponds to the arrangement shown in Fig. 4 apart from the fact that the detector has been replaced by a 2.5 Gbit/s CP-FSK receiver 50 with a front end 51, an IF-section 52, and a demodulator 53. An AFC- circuit, i.e. an automatic frequency control, provides a feed back from the IF-section 52 to the local oscillator LO so as to stabilize the intermediate frequency.

The intermediate frequency used is 10.34 GHz and the bandwidth of the intermediate frequency filter is 7.1 GHz. The transmission response of the intermediate frequency filter appears from Fig. 7A together with a calculated curve for the suppression of the local oscillator noise RIN. A PRBS-generator 55 transmits a pseudorandom sequence of bits modulating the transmitter TL. The received string of bits appearing on the output of the demodulator 53 is compared with the string of bits generated by the PRBS-generator 55, and a counter 56 registers the number of errors and consequently BER. Initially, a reference measurement was carried out without an HI-BI-fibre in order to determine the influence of the LO RIN upon the performance of the receiver system expressed by bit-error-rate BER. The BER was measured as function of received LO-power for a fixed, received signal power and a fixed LO-injection current. Only the coupling of the LO-power was changed in this measurement. The local osc illator-generated RIN is thereby constant, and the RIN transferred to the photodectector is therefore proportional to the transferred local oscillator power.

If RIN was very low, BER would have approached a constant BER for high local oscillator power corresponding to shot noise-limited detection, but when RIN is comparable with the thermal noise, a minimum is achieved on the curve for RIN (BER) versus the coupling when the amount of RIN transferred is equal to the thermal noise. This appears clearly from Fig. 8, in which a minimum is located at a transmitted local oscillator power of about 0.4 mW . This measurement shows the optimum local oscillator power for the specific injection current when no reduction of RIN applies.

Here it should be noted that the RIN generated by the local oscillator is inversely proportional to the output power of the laser in such a manner that an increased injection current involves a smaller RIN for the same local oscillator power. Accordingly, the minimum where RIN transferred is equal to the thermal noise is found at a higher local oscillator power.

The above can be continued until the minimum BER is determined by the shot noise. Such a method may very well imply that the local oscillator must be supplied with an extremely heavy injection current, which is disadvantageous for the durability of the laser. Correspondingly, the maximum photocurrent of the photodetector presents an upper limit to the magnitude of the local oscillator power which can be applied.

An external reduction of RIN is obviously to be preferred, and the following measurements demonstrate the influence of RIN reduction where the reduction has been carried out by means of the method according to the invention. Fig. 9 illustrates curves for the bit-error-rate BER for the 2.5 Gbit/s CP-FSK coherent optical communication system of Fig. 5 with a heterodyne receiver. A reference curve A for the same LO injection current as previously used in Fig. 8, and with a coupled or received LO-power of -4 dBm , i.e. the optimum power, was measured.

Subsequently, a precalculated length of HI-Bl-fibre 14 was inserted in the detector branch, and the polarization of the LO-signal and the information-signal SI was adjusted by means of the polarization controllers 20.

The measurements of BER with an inserted HI-BI-fibre, cf. curve B, showed that the sensitivity of the receiver had been improved by about 2.4 dB , which corresponds to an RIN suppression of 11 dB over the intermediate frequency band. This suppression agrees well with the suppression of 10 dB which could be expected in view of the transfer functions of Fig. 7 and the RIN spectrum of the local oscillator used.

The received LO-power was subsequently increased to 1 mW , i.e. 0 dBm, by increasing the injection current for the LO laser, and the curve C was measured. The increased local oscillator power provided an additional improvement of the sensitivity of 2.5 dB with the fibre inserted as shown by the curve C. When the shot noise-limited detection for an optimum coupled LO power is not reached, the sensitivity improvement obtained by suppressed RIN is alway better than when RIN is not suppressed, and when RIN is suppressed it is furthermore unnecessary to examine whether the coupled local oscillator power is optimum, as it should just be as high as possible.

The last curve D is measured with a photodetector-preamplifier combination providing a substantial reduction of the thermal noise in the receiver. The limited sensitivity improvement of 1 dB indicates that shot noise- limited detection is about to be reached. Table 1 shows measured and calculated noise powers for the four BER-curves of Fig. 9. Measured values are marked by an underlining.

Table 1

Based on the measured improvements in the system sensitivity, the total noise power for the curves B-D are calculated and shown in Table 1 (the curve D is measured under the same conditions as the curve C, but a front end with improved noise performance was used.) As both shot noise and thermal noise are measured for all configurations, the reduced value of RIN can be calculated.

The LO RIN power values are calculated for the curves B, C, and D, and correspond to the expected results bearing in mind the experimental uncertainty in determining the measured noise powers.

It appears from the Table, that LO RIN in the receiver is suppressed so that its contribution to the total re ceiver noise is reduced significantly. Consequently, the performance of the receiver can now approach shot noise limited operation. A 1 dB improvement is obtained in the curve D with respect to the curve C. It appears from the calculated figures that the total suppression of LO RIN would only yield an additional improvement of 0.4 dB in sensitivity for D.

This RIN suppression method can also be easily implemented in a polarization-diversity-receiver-configuration by inserting a polarization beam-splitter in the signal path and by duplicating the coupler-arrangement.

When it is ensured that all channels have the same polarization, the receiver configuration suppresses mutual intermodulation-beat-products between the channels in a multi-channel system like a balanced receiver.

The present invention provides thus a simple method of suppressing local oscillator RIN in coherent optical communication systems with optical heterodyne receivers, and the use has been demonstrated by way of experiments on a test arrangement in a 2.5 Gbit/s CP-FSK system, where an RIN suppression of about 11 dB was achieved. The use of the HI-BI-fibre provided the system with a substantially improved performance and reduced the penalty of the system, (i.e. the reduction of the receiver sensitivity due to RIN) to only 0.4 dBm.

Claims

Claims.

1. A method of suppressing relative intensity noise, the so-called RIN, from a laser or a corresponding coherent light source, especially a semiconductor laser in a coherent optical communication system, c h a r a c t e r i s e d in that a birefringent optical member (14) is Inserted in the optical signal path, and that the length of the signal path In the birefringent optical member (14) Is adjusted such that the polarization of the aimed/desired optical laser signal and the relative Intensity noise (RIN) of the laser are substantially orthogonal at the exit of the birefringent member.

2. A method as claimed in claim 1, c h a r a c t e r i s e d In that the birefringent member (14) is a high birefringent fibre, a so-called HI-BI-fibre.

3. A method as claimed in claim 1 or 2, c h a r a c t e r i s e d in that the polarization state of the signal from the laser is adjusted such that substantially equal amounts of power are coupled into the two axes (e_o + e_e) of the birefringent fibre.

4. A method as claimed in claim 1, 2 or 3, c h a r a c t e r i s e d in that the birefringent member (14) is followed by an optical detector (16) and is of such a length that the desired laser signal and the undesired laser-RIN signal are mutually orthogonal on the surface of the detector (16).

5. A method as claimed in claim 1, 2 or 3, c h a r a c t e r i s e d in that a polarizor (18) is mounted at the exit of the birefringent fibre (14), and that the polarizor (18) is followed by a plurality of alternating birefringent fibres (24, ...) and polarizors (28, ...), where the lengths (14, 24, ....) of the fibres are adjusted such that the combined structure presents a de sired transfer function.

6. A method as claimed in claim 1, 2, 3 or 4, c h a ra c t e r i s e d in that the birefringent member or fibre (14) is inserted in the receiving section of a coherent optical communication system of the type operating according to the heterodyne principle and receiving an information-carrying laser signal (SI) by means of a local oscillator laser (LD) and a detector (16), which operates as a mixer providing an intermediate frequency signal, that the length of the birefringent fibre (14) is adjusted such that the local oscillator-laser signal (LO) and the undesired laser-noise signal are orthogonal on the surface of the detector.

7. A method as claimed in claim 6, c h a r a c t e r i s e d in that the polarization condition of the local oscillator signal (LO) is adjusted such that substantially equal amounts of power are coupled into the two axes (e_o + e_e) of the birefringent fibre.

8. A method as claimed in claim 6 or 7, where the information-carrying signal is carried through a first optical fibre to a fibre coupler also receiving the local oscillator signal through a second optical fibre, and where both signals are transmitted towards the detector surface (16) operating as a mixer, c h a r a c t e r i s e d in that polarization control units (20) are connected to the first (21) and/or the second optical fibre (22), said units being adjusted such that the information-carrying signal (SI) and the local oscillator signal (LO) present the same polarization when they hit the detector surface.

9. An arrangement for carrying out the method as claimed in one or more of the preceding claims, and comprising a coherent light source, an optical birefringent fibre, and a detector, c h a r a c t e r i s e d in that the length (L) of the fibre (14) is adjusted such that the polarization of the noise signal from the light source (LD) is orthogonal to the polarization of the pure local oscillator signal on the detector surface or at the detector, Δω τ = π +2pπ, where τ - Δn L/c, and where Δn is the difference In the refractive index between the ordinary and the extraordinary axis, ω ' being the angular frequency of the dominating noise and Δω being the difference frequency between the local oscillator frequency ω ₀ and ω '.

10. Heterodyne receiver for carrying out the method as claimed in claim 6, 7 or 8, c h a r a c t e r i s e d in that a birefringent member or fibre is inserted between the local oscillator and the detector, and that the length of the birefringent fibre (14) is adjusted such that the local oscillator-laser signal (LO) and the undesired laser-noise signal are orthogonal on the surface of the detector.

11. Multicommunication system, c h a r a c t e r i s e d in that it comprises a plurality of receiver configurations as claimed in claim 9.

12. Polarization-diversity-receiver, c h a r a c t e ri s e d in that it comprises a polarization beam splitter splitting the received signal into two branches, each branch being provided with an arrangement as claimed in claim 9 or a receiver as claimed in claim 10.