GB2172765A - Modal noise in optical systems - Google Patents

Abstract

A technique for reducing the modal noise in a digital multimode transmission system produced e.g. by a mode selective connector, comprises optimising the launch into the laser pigtail, so that the modal power distribution is uniform, by offsetting a butt pigtail end (in a plane parallel to the laser junction) in a direction parallel to or perpendicular to the line of the laser junction. <IMAGE>

Description

SPECIFICATION Modal noise in optical systems This invention relates to modal noise in digital multimode transmission systems and particularly to reducing modal noise.

Techniques such as r.f. biassing and controlled reflections are commonly postulated as schemes for improved modal noise performance of semiconductor lasers in digital transmission systems.

However these schemes are not easy to implement and it is an object of the present invention to reduce modal noise in a simple yet effective way.

According to the present invention there is provided an optical source device for e.g. a digital multimode transmission system, the source having a junction and an optical waveguide coupled to it for receiving the launched optical energy and laterally offset from the source junction whereby to minimize modal noise level in the system due to a mode-selection-loss component in the system.

According to another aspect of the invention there is provided an optical transmission system having a mode-selective-loss component, an optical source and an optical waveguide coupled to the source for receiving the launched optical energy and laterally offset such as to minimise or at least reduce modal noise level in the system due to the component.

In order that the invention can be clearly understood reference will not be made to the accompanying drawings, in which: Figure 1 is a block diagram of a laser characterisation system for assessing modal noise; Figure 2 shows the modal-noise-limited predicted bit error rate for samples of 1 .3m lasers, using the system of Fig. 1, Fig. 2A showing representations of the probability density functions typically measured; Figure 3 shows the predicted bit error rate versus axial offset of the pigtail to fibre joint; Figure 4A represents the far field spot size for a laser butt jointed to a fibre pigtail for maximum coupling efficiency; Figure 4B corresponds to Fig. 4A but shows the pigtail fibre offset, according to an embodiment of the invention, in a direction parallel to the laser junction.

Figure 4C is similar to Fig. 4B but shows the offset in a direction perpendicular to the laser junction in accordance with another embodiment of the invention.

Figures 5A, 5B and 5C near field crosssectional power profiles for the butt launch, respectively, of the lasers arrangements of Figs. 4A, 4B, 4C.

Figure 6 is a graph of connector loss versus axial offset for the arrangement of Fig. 4, Figure 7 is a plot of the predicted average BER versus connector offset for the arrangement of Fig. 4, and Figure 8 is a plot of the slope of the laplacian (h.f. PDF) versus connector offset.

Recent work (M.J. Lum et al, Proc 10th ECOC, Stuttgart, September 1984.) on modal noise in multimode systems has shown that the amplitude probability density function (PDF) of a received modal noise is composed of a combination of the statistics of mechanically and spectrally induced modal noise. The statistics of the noise are non-stationary and the level of noise depends strongly on the laser drive and temperature, fibre configuration, and reflections back into the source (P.R.

Couch et al, Proc 9th, ECOC, Geneva, October 1983.). The system may explore these parameters during its lifetime so that its performance is unsatisfactory for periods which may occur only infrequently. The need to predict long term system performance makes it necessary to time compress the modal noise characterisation of lasers.

The following describes a modal noise characterisation system (CS) which explores the parameters which influence modal noise and takes a laser measurement form which the modal noise limited error probability in a transmission system can, for the first time, be directly predicted with confidence.

Fig. 1 shows the characterisation system in which the laser is modulated by a low frequency square wave. The laser temperature and pigtail bend conditions are explored (temperature control and flexer) before a mode selective loss join to a 1 km length of "standard" fibre. The input to receiver presents insignificant mode selective loss (MSL) and the receiver bandwidth is selected to be comparable with the required transmission system bandwidth. A sampling oscilloscope enables gating of the "1" or "0" level prior to their PDF measurement by a computer-controlled analyser which controls the display of the sampling oscilloscope.

Referring to Fig. 2A, this shows the probability distribution functions (PDF) obtained for a modulated pigtailed source experiencing a 3dB mean mode selective loss at the first connector. The receiver in the characterisation system of Fig. 1, is a.c. coupled and typically the probability count spans 5 or 6 decades. This is obtained during a few minutes of sampling.

The mechanically induced low frequency fluctuations are shaped equally between the mean "0" and the "1" levels. High frequency fluctuations, due to spectral variations, appear around the mean "1" level. The resulting PDF (probability distribution functions) on the "1" level is a combination of the low and high frequency fluctuations.

The average bit error rate (BER) prediction is then possible by extrapolation of the straight line skirts of the PDF on the "1" level down to the 50% decision threshold. Comparison of the slopes of the high frequency PDF's pro vides a useful measure of the manifestation of spectrally-induced modal noise (such as partition noise, frequency chirping. This technique enables analysis of modal noise independently of the artifacts of the transmission system.

The relevant transmission system parameters can then be optimised to achieve best modal noise performance. The characterisation system also enables determination of error rate margin for "good" modal noise lasers and/or for small MSL. From a transmission system measurement, information of error probabilities better than 1 in 10'0, which would reveal the system modal noise margin, cannot be made in a reasonably short period.

Fig. 2 also shows the predicated modalnoise-limited error probability from measurement, for samples of pigtailed devices from batches of nominally identical devices. The measurements were made with a 3dB modeselective loss at the pigtail to fibre join.

Fog. 3 shows the predicted error probability versus axial offset at the mode selective loss point and is typical for many lasers.

The virtue of aiming to keep the mode selective loss below 1dB is emphasised from these measurements.

The pigtail to a semiconductor laser is usually aligned for maximum power coupling efficiency. The accepted procedure is to align a fibre with a lensed end so that the centre of axes of the semiconductor laser and fibre are coincident. For alignment of 50,um core fibre the fibre lens tip to laser chip distance is 25-50,um. Under circumstances where the pigtail being aligned is butt ended the same coincidence of centre of axes is required for maximum coupling efficiency.

Both the above launch configuration result in the launched power being weighted significantly in the low order guide modes of the fibre. Where connector mode selective loss cannot be kept below 1dB this results in significant high frequency modal noise in digital multimode systems.

Fig. 4A represents the far field spot size for a laser butt-joined to a fibre pigtail for maximum coupling efficiency. The shaded areas represent the spot size for a channel substrate planar type laser with laser to fibre butt-end spacing 15m and the fibre core diameter is 50mm. Fig. 5A shows the near-field laser pigtail cross-sectioned power profiles at subthreshold, threshold, lasing and an attentive lasing profile with different bend conditions (flexer).

The modal noise level at a mode selective component e.g. a connector, is a function of the modal power distribution in the laser pigtail. By optimising the launch into the pigtail so that the modal power distribution is uniform, we have found that the modal noise level is reduced to make digital multimode fibre systems impairment-free from this noise, for a wide range of mode selective component losses.

The desired modal power distribution is obtained by offsetting a butt pigtail end in a plane parallel to the laser junction. This technique enables the use of a laser which does not have many lasing modes in its spectrum, and one whose modes could exhibit partitioning noise, in a digital multimode transmission system.

Fig. 4B shows an embodiment where the axis of the pigtail fibre is offset in a direction parallel to the laser junction. This results in reduced coupling efficiency but enables the excitation of higher order, in particular skew, modes. The resulting fibre cross sectional power profile is shown in Fig. 5B and is uniform at the threshold level, the sub-threshold level also being shown above. High order (azimuthal) skew modes have radiation patterns which can exhibit many circumferential spots and this is advantageous. The launch penalty due to misalignment is about 2.0dB.

Fig. 4C shows another embodiment where the pigtail fibre axis is offset in a direction perpendicular to the laser junction with attendant loss in coupling efficiency. This offset enables the launch of some higher order radial modes, obtaining a near-parabolic cross-sectional "modal" power distribution. Fig. 5C shows the evolution of the resulting fibre cross-sectional power profile from sub-threshold through threshold to lasing, being approximately Gaussian at threshold.

Thus modal noise level can be reduced by increasing the number of fibre modes (degrees of freedom) which carry the particular level of power (increased uniform excitation).

The approach outlined in Figs. 4B and 4C has shown that for moderate to high (greater than 1dB and up to 3dB) connector losses, an offset butt launch at the laser produces a significant modal noise performance improvement. This approach has the advantage over other potential "solutions" in that it could be readily implemented during packaging.

Fig. 6 shows the relative mean first connector loss versus a calibrated axial offset of the pigtail and adjoining fibre for the arrangement of Figs. 4A, 4B and 4C, respectively, and designated accordingly curves 6a, 6b and 6c.

Fig. 7 is a plot of the predicted average bit error rate (BER) versus the first connector offset for the three launch conditions of Figs.

4A, 4B and 4C, measured at 1dB loss intervals (BER is predicted from the extrapolated slope of the PDF measured on the "1" level), the curves being designated 7a, 7b and 7c, respectively.

Fig. 8 is a plot of the slope of the laplacian (high frequency PDF) versus first connector offset. The slope here is defined as the number of decades between the mean "1" level to the intercept of the slope with the mean "0" level.

The experiments were carried out with pla nar package IRW laser devices.

The modal noise evaluation used a first connector 3dB loss criterion for several reasons: (i) This loss is felt to be highly probable or typical of a maximum, system connector loss.

(ii) Measurement around 3dB loss offers greater repeatability and therefore requires smaller number of repeat measurements. Fig.

6 shows that although the 3dB loss is not the optimum fibre offset to give an equal loss for a wide selection of modal power distributions, it is fairly close to this optimum region of 18-19.5 ,um offset. The 3dB connector loss produced symmetric PDF and measurement for this loss has enabled a close tie-up with theoretical modelling.

In order to assess fully the merits and implications of varying the modal power distribution we have to compare the connector loss versus offset as well as the subsequent modal noise characterisation.

The cross-sectional power profiles of laser pigtails of Fig. 4A, 4B and 4C were measured by suitable imagining of the fibre near field onto an infra-red Hammamutsu camera system. Fig. 4A shows clearly how evolution of the laser near field with bias current, alters the power profile in the fibre. The final profile during lasing operation shows the dominance of power in low order fibre modes. Due to the symmetry of launch these are believed to be confined mainly to radial modes. Examination of the near field speckle shows most of the power to be in 1-5 sports depending on fibre bend conditions.

Goodman's work (Goodman, J.W. et al.

"Analysis and measurement of modal-noise probability distribution for Step index optical fiber", Optics Letter, Vol.5/6, Aug 1980.) and (Goodman, J.W. et al. "Frequency Dependance of Modal Noise in multimode optical fibres". J. Oct Soc. Am. Vol 7/8, Aug 1980.) on step index fibres has shown that the signal to noise (S/N) is directly proportional to degrees of freedom. The number of degrees of freedom can be approximated by the number of speckle spots. For graded index fibres connector loss will be strongly dependent on modal power distribution among high order (high radius and angle) modes. Therefore a launch optimisatiop which, by distribution of modal power, increases the number of speckle spots would have to be carefully weighted so as not to incur excess connector loss penalty.

Asymmetry in the laser near field enables, by offsetting a butt launch in a plane parallel to laser junction (lateral offset), to launch higher order azimuthal (or skew) modes. Fig.

4B shows the uniform power distribution of one such launch with a launch penalty from the fully aligned butt of 1.5 to 3dB. A further lateral offset beyond this produces unfavourable launch of high angle, high radius modes only. Examination of the near field power distribution during fibre laser alignment, however, ensures a repeatable desired alignment condition. With sufficient, independent control of tri-axial movement of the butt pigtail a procedure of obtaining the desired alignment can be achieved by measuring the decrease in coupling efficiency (as measured at the pigtail output) from the fully aligned butt launch.

A perpendicular offset of 1.5dB gave the intermediate (near parabolic) power distribution of Fig. 4C.

Modal noise statistics are non stationary: the noise may be low for much of the time but can be very high for certain combinations of fibre position and laser temperature. The average BER prediction obtained by time compression of the two main variables of fibre configuration and temperature are believed to be close to long-term system convergence. To date the most useful indication of short-term behaviour has been from an actual transmission system experiment where the laser temperature is loosely controlled, but changes sufficiently to explore bad temperature regions.

The measurement gives a system BER versus time (10 sec intervals).

The IRW devices examined so far have shown predicted BER and transmission system results of modal noise limited performance in the region of 1 in 105 to 1 in 109, for a 3dB first connector loss. Owing to the non-stationary nature of modal noise a figure in the range of approximately 1 in 1013 to 10'6 would be required for satisfactory system operation.

Examination of the results of BER versus offset (Fig. 7) shows a BER of 1 in 105 to 1 in 108 for connector losses of 3dB or greater and a well aligned laser/fibre. All three launches predict a "modal noise free" zone for connector losses up to 1.5dB. These results clearly emphasise the value of keeping connector losses to at least around the 1dB region. Launches 4A and 4C, however, show a rapid change in BER in the region leading up to 2dB connector loss. Launch 4B showed a small "excess" loss under good fibre-fibre alignment, possibly due to some cladding mode power. This showed a slightly worse BER for the 1dB loss; however the BER does not deteriorate beyond 3X10" for a loss as high as 5dB. The upturn of the curves for very high loss can be accounted from modal power distribution and offset considerations.

The predicted BER curves for the two offset conditions 4B, 4C give little indication for preference between the two. Examinations of the complementary plots of the slopes of the high frequency noise of Fig. 8 show clearly the lower noise of 4B.

The above measurements have been repeated for one other IRW device, and two other IRW devices have been measured for 3dB connector loss with and without launch offset yielding similar results. The improved performance of the offset launch has been checked to an output level of OdBm.

In conclusion therefore it has been shown that an offset butt launch, for a 3-4dB power penalty with respect to an ideal lens launch, offers a scheme for the improved modal noise performance of IRW devices which can easily be implemented on the production line. The improvement gives a predicted average BER of better than 1 in 10'0 for a 3dB connector loss.

The offset may not minimise modal noise level for very small mode selective losses i.e.

losses 1 dB optical, but the level of modal noise with offset implemented is still acceptable at these low loss levels (system requirement bit error ration 10-9).

The amount of offset in comparison with the optical fibre diameter is of the order of 10%.

The near-field cross-sectional profiles of Figs. 5A, 5B and 5C have been obtained during alignment of pigtail and laser chip with the laser operated below threshold, to give a power profile which does not vary with fibre movement (i.e. laser "incoherent" therefore speckles are not generated at the fibre output). Such a measurement in-situ during packaging, using an infra-red camera had an imaging system is quite straight-forward. This, together with monitoring of coupled power as measured at the pigtail output, can be used to obtain the desired offset for a particular MPD profile.

Claims (12)

1. An optical source device for e.g. a digital multimode transmission system, the source having a junction and an optical waveguide coupled to it for receiving the launched optical energy and laterally offset from the source junction whereby to minimise or at least reduce modal noise level in the system due to a mode-selective-loss component in the system.

2. A device as claimed in claim 1 wherein the coupled waveguide is offset in a direction parallel to the source junction.

3. A device as claimed in claim 1, wherein the coupled waveguide is offset in a direction perpendicular to the source junction.

4. A device as claimed in claim 1, claim 2 or claim 3, wherein the source comprises an inverted rib waveguide laser.

5. A device as claimed in claim 4, wherein the axial spacing between the junction and at adjacent end of the waveguide lines in the range 10-50 im.

6. A device as claimed in any preceding claim in any preceding claim, wherein the lateral offset from the centre of the junction lies in the range 10-50Am.

7. A device substantially as hereinbefore described with reference to Figs. 4B and 4C of the accompanying drawings.

8. A system incorporating a device accord ing to any preceding claim.

9. An optical transmission system having a mode-selective-loss component, an optical source and an optical waveguide coupled to the source for receiving the launched optical energy and laterally offset such as to minimise or at least reduce modal noise level in the system due to the component.

10. A system as claimed in claim 9, wherein the component is a connector.

11. A system substantially as hereinbefore described with reference to the accompanying drawings.

12. A method of aligning a pigtail with an optical source to provide a desired offset, comprising operating the optical source and adjusting the pigtail alignment until an output parameter value corresponds to a predetermined offset, and fixing the pigtail relative to the source.