GB2375391A

GB2375391A - Measuring grid error in an optical device

Info

Publication number: GB2375391A
Application number: GB0111283A
Authority: GB
Inventors: Daniel John Kitcher
Original assignee: Bookham Technology PLC
Current assignee: Lumentum Technology UK Ltd
Priority date: 2001-05-09
Filing date: 2001-05-09
Publication date: 2002-11-13
Also published as: GB0111283D0; WO2002091052A1

Abstract

Grid error is measured in an optical device such as a demultiplexer (2) having a plurality of channels such as output waveguides (21) carrying light of different frequencies. The light is incident on respective photodiodes of an array (26), each photodiode producing an electrical output representative of the incident light level. Real data representing the location of channel peaks in the photodiode array is generated from these electrical outputs. This data is used to generate a line of best fit of the optical channels to the photodiode array according to a predetermined algorithm and to produce a set of estimated data. The estimated data is compared with the real data to provide an indicator of grid error. A multichip electronics module comprises a first chip constituting the optical device (2) and a second chip for carrying out the processing steps. The output errors may be used to tune the sources of the light to compensate for the errors.

Description

GRID FITTING IN AN OPTICAL DEVICE The present invention relates to grid fitting in an optical device, and more particularly to measuring the deviation from a predetermined grid caused for example by temperature changes which can be difficult to control.

In the following, the invention is described in the context of an integrated chip optical demultiplexer, but can in principle be used with a number of different types of optical components including a variety of optical multiplexers, optical channel power and frequency monitors, demultiplexers etc. All of these components include devices which have a plurality of opto/electric transducers such as photodiodes to provide output signals corresponding to optical signals. A row of photodiodes is used to sense the output from a plurality of output channels and in order to make use of this output it is necessary to have a correlation between the physical location in the output optical spectrum of individual channels and the photodiodes which pick up the optical information of those channels.

In optical components which use an array waveguide grating, one of the difficulties is that the waveguide properties of individual waveguides are exquisitely sensitive to temperature, so that there is a shift in the frequencies guided by the waveguides. This has the effect of shifting the location of output channels in the final optical spectrum output from the waveguide, in a manner which is indeterminate and dependent on the temperature variations. The current proposals for tackling this problem rely on stringent temperature control of packages, and at present, for example, packages for optical channel monitors target 0. 10C temperature stability and accuracy. This is extraordinarily difficult to accomplish with optoelectronics packages rated for environments of 0 to 70 C over lifetimes of in excess of twenty years.

Another difficulty is the requirement to precisely attach photodiodes at the correct locations to pick up predetermined optical channels.

Another problem which can arise in optical systems is that the transmitters can vary in their frequency, for a number of different reasons. That is, a particular channel may be detected, but its detected frequency is outside the tolerance required for correct operation of the WDM system.

It is an aim of the present invention to measure frequency variations in the transmitters.

According to one aspect of the present invention there is provided a method of measuring grid error in an optical device in which a plurality of optical channels at different frequencies are incident on a photodiode array comprising a plurality of photodiodes each operable to convert light energy incident thereon to an electrical value representative of the intensity of the incident light, the method comprising: determining from the electrical values the location of channel peaks in the photodiode array to generate a set of real data; generating from said real data and from said channel peak locations a line of best fit according to a predetermined algorithm of said optical channels to said photodiode array whereby a set of estimated data is obtained; and comparing said real data with said estimated data to provide an indicator of grid error.

The grid error can be reported as an actual deviation in frequency for each channel number, or it can be reported as an error status of 1 for no frequency deviation and 0 for a frequency deviation. In either case the reported grid error can be used to control light sources providing said optical channels thereby to alter the frequency of those light sources to bring them back into line with the expected grid.

Another aspect of the invention provides An optical device comprising a dispersive optical guide which guides a plurality of optical channels at different frequencies and a photodiode array comprising a plurality of photodiodes each operable to convert light energy incident thereon to an electrical value representative of the intensity of the incident light, the optical device further

comprising a system for measuring grid error having: means for determining from the electrical values the location of channel peaks in the photodiode arrangement to generate a set of real data; means for generating from said real data and from said channel peak locations a line of best fit according to a predetermined algorithm for said optical channel to said photodiode array whereby a set of estimated data is obtained; and means for comparing said real data with said estimated data to provide an indicator of grid error.

A still further aspect provides a multi-chip optoelectronics module comprising: a first chip constituting an optical device with a dispersive optical guide which guides a plurality of optical channels at different frequencies and a photodiode array comprising a plurality of photodiodes each operable to convert light energy incident thereon to an electrical value representative of the intensity of the incident light ; and a second, processing, chip which is operable to receive said electrical values and to determine therefrom the location of channel peaks in the photodiode arrangement to generate a set of real data, to determine a line of best fit according to a predetermined algorithm for said optical channel to said photodiode array whereby a set of estimated data is obtained and to compare said real data with said estimated data to provide an indicator of grid error.

A system controller can be included which receives the reported grid error and supplies control feedback signals to alter the frequency of the input optical channels.

For a better understanding of the present invention, and to show how the same may be carried into effect, reference will now be made by way of example to the accompanying drawings in which: Figure 1 illustrates a prior art demultiplexer chip with more than two outputs per signal ; Figure 2 is a schematic diagram illustrating in block diagram form the interface components of the invention;

Figure 3 illustrates the output table for a user; Figure 4 illustrates the optical input to the demultiplexer of Figure 1; Figure 5 is a frequency graph illustrating the relative grid errors; Figure 5A is a schematic graph illustrating some of the photodiode array outputs; Figure 6 is a schematic diagram of an alternative optical arrangement which can use the invention; Figure 7 is a plotted graph showing a straight line fit; and Figure 8 is a flow chart.

Figure 1 illustrates a demultiplexer chip 2 wherein the chip boundary is denoted by reference numeral 4. In the chip 2 shown in Figure 1, a dispersive waveguide array 11 consists of a plurality of curved waveguides 12. The demultiplexer is formed as an integrated chip on a planar substrate. The substrate may be formed with silicon on insulator and the waveguides may be ridge waveguides of the type shown in US Patent No. 5,757, 986. The array 11 is a dispersive array of ridge waveguides formed on the chip 2. Each of the waveguides has a straight input section 15 and a straight output section 19. Line 13 indicates the junction between the straight input sections 15 and the curved sections 12. Similarly the line 14 indicates the junction between the curved sections and the straight output sections 19. In this case the input and output ends of the array 11 are symmetrical. The straight input sections 15 incline inwards towards each other so as to point to the focus position 17 at the end of an input waveguide. The input waveguide is selected from the group of input waveguides labelled N inputs in Figure 1. Each waveguide is referenced respectively 16a, 16b... 16N. The input waveguide to measure is selected using the associated input selector switch 18a, 18b... 18m. The reference numeral 16R denotes a reference input the function of which will be described later. The array of input waveguides is labelled 16.

Similarly the straight output sections 19 are inclined towards each other so as to form a focus in region 20 adjacent the entrance to an array of N output waveguides 21. The individual output waveguides are labelled respectively 21a...

21 N. The geometry of the input and output ends of the array each form part of a similar Rowland circle arrangement. The input ends of the straight waveguide sections 15 lie on an arc forming part of a larger circle 22 having its centre coincident with the end 17 of the centre of the input waveguide array 16. The outputs of the input waveguides 16a... 16N lie on the circumference of an inner circle 23 having half the radius of the larger circle 22. Similarly at the output end of the array 11, the ends of the straight waveguide sections 19 terminate on an arc forming part of a larger circle 24 having its centre coincident with region 20 forming a focus for the output of the dispersive array. The output waveguides 21 are also arranged to terminate in an arc lying on the smaller inner circle 25 which has half the radius of the outer circle 24. Due to the dispersion within the array 11 being dependent on wavelength, the demultiplexed output channels are focussed on an arc of the circle 25 adjacent the output waveguides 21. The channels are closely spaced at the focal line and are too closely positioned for effective detection by respective photo diodes in the output detectors 26. For this reason the array of output waveguides 21 detect the output channel images formed on circle 25 and transmit the optical signals to more spaced locations at the edge 27 of the chip where they are detected by an array of diodes 26. In the described embodiment the array of diodes has 128 diodes, but it will be appreciated that any suitable number can be used appropriately matched to the number of optical channels and the required density of photodiodes per channel.

Figure 2 is a schematic block diagram illustrating the information flow. A multichip module 1 comprises an optics chip 2 and a processing chip 6. As has already been described, the chip 2 receives an optical input which is labelled 3 in Figure 2. The array of photodiodes 26 picks up an optical output from the chip 2 and converts it to an electrical output which is labelled 5 in Figure 2, each photodiode converting the optical energy incident thereon to an electrical signal representing its intensity. This is supplied to the processing chip 6 which processes the electrical signals in a manner which will be described later. The processed signals are supplied to a user output which is labelled 8 in Figure 2 to a system controller 3. The system controller provides control feedback signals to

control light sources, e. g. lasers supplying the optical input. The photodiodes are arranged to provide output signals corresponding to light detected in the output channels of the output array waveguide 21. In this context, each channel has a certain carrier frequency, referred to as its main or centre frequency and represented by the intensity peak for that channel. The purpose of the photodiode array 26 is to pick up light output from the output waveguide array 21 and convert it into electrical signals so that the power on each transmitted channel can be established. The processing chip 6 processes the electrical signals 5 received from the photodiode array 26 and generates a table which correlates the output powers in association with the respective channels. Such a table is shown in Figure 3, where C1, C2... Cn represent the channel numbers; P1, P2... Pn represent the output powers and F1, F2... Fn represent the channel frequencies. In addition an error check is reported against each channel which has a status of 1 if the channel peak frequency is what would be expected and 0 if it has shifted from its grid position as discussed more specifically herein.

Figure 4 is a diagram illustrating the optical spectrum of the optical input 3 on a graph of intensity versus frequency (in THz). It is a well known property of optical communications systems using multiple optical carrier frequencies on a common fibre (DWDM) that the carrier frequencies are always close to a designed list of frequencies, arranged as a so-called ITU grid. In the described embodiment, for example, the grid comprises frequency values of 196.2 to 192.1 THz, with one channel existing on each 0.1 THz spacing.

The detector array output for each such three central channels (at frequencies of 194.0, 194.1 and 194.2 THz) is illustrated in Figure 5. It will be appreciated that similar peaks exist for each of the frequencies between 191 and 196 THz in the ITU grid.

As already mentioned, Figure 5 illustrates the optical spectrum to be measured, zoomed in on three central channels. The optical spectrum contains signals with so-called channel peaks. In an ideal environment, it would be possible to align

the photo-detector array with the output array waveguide 21 so that each channel peak (central frequency) would always land on certain photodiodes.

Figure 5 shows the optical spectrum if two of the incoming signals are on the grid at 194. 0THz and 194.2THz. The third signal is incorrectly tuned. This illustrates a frequency grid error-one of the signals is off the 0. 1THz grid by 8. As described below, a measurement of grid error 8 is sufficient to send a request to that transmitter to correct its tuning by-c, A separate effect is tuning of the channel measurement device. Changing the temperature of the device of Figure 1 shifts all outputs by -10GHzlC for silicon.

This results in an offset which is conventionally minimised through precision temperature control. In a temperature stabilised measurement system, each photodiode (dj, dj, dk, d s... in Figure 5) has a unique optical frequency marking the centre of the passband which it captures, and this can be calibrated. However, the software fitting technique described herein removes the need for precision temperature control because the photodiodes are not presumed to capture particular frequencies. As described in the following, if all outputs are shifted due to a temperature change of the measurement device, then this shift is tracked by the fit line. Deviations from the fit line are not affected by the temperature of the measurement system.

The operation of the processing chip 6 to take account of these errors will now be described. Each photodiode in the photodiode array 26 converts the optical power incident on it into an electronic analogue signal. The analogue data is converted to digital data for subsequent processing. Thus the electrical output 5 of the photodiode array is a set of digital signals representative of the optical powers incident on each of the photodiodes in the array-see Figure 5A. The digital signals from the photodiodes are processed according to a peak location algorithm to determine the location of the channel peaks with respect to photodiode numbers. This is done for example using an algorithm described and explained in US Patent No. 6,002, 479. The output of this algorithm is an array of

photodiode numbers versus channel peaks, as illustrated in Figure 7. Photodiode number is transformed to relative frequency using calibration data of the demultiplexer to obtain the fit graph of Figure 7. This information is used to produce a straight line best fit according to an RMS algorithm. Such best fit algorithms are well known in the art and will not be described further here. That straight line represents the line of best fit for the channel peaks with respect to the diode numbers. The graph of relative frequency of Figure 7 is converted to absolute frequency if it is presumed that the fit lies on the nearest 100GHz ITU grid.

By comparing the "real" data received from the photodiode array with the straight line fit data generated through the RMS algorithm, frequency errors in channel peaks can be estimated and reported and/or compensated for.

Thus from a comparison between the straight line fit and the "real" results the difference between the measured frequency of each channel and the expected frequency (according to the straight line fit) can be derived. Thus, a frequency error 8 can be derived, being the difference between the measured frequency and the estimated frequency according to the straight line fit, and this information can be used to tune the light sources on the input side. The actual value of the frequency error 8 can be reported, or merely that there is an error in the frequency using the status values 1 and 0. The frequency errors can be established notwithstanding the shift of the spectrum which might have taken place with respect to the photodiode array due to temperature changes. Figure 8 is a flow chart illustrating the above described technique.

The technique described herein accommodates bad or missing data. The fit line can be calculated from a subset of the measurements which passed certain quality requirements which may comprise: signalN is present? signalN has sufficient power above the measurement system noise floor ?

signalN lands on defect-free photodiodes? Some signals will not meet all of these criteria and will be ignored when the fit line is generated. If a signal is present but of low power, then an estimated s can be available in the output, marked by a warnings list. The channel monitor system should not risk pulling the fit line toward this noisy data point. In Figure 7, points encircles are either too weak to measure accurately or otherwise bad. The remaining points are sufficient to construct the fit line.

A related method is to replace the RMS best fit procedure using weighting of 1 for good points and 0 for noisy points with a weighted RMS fit. For instance the power of signals may be used as the weighting factors, so that the least noisy data have a higher weighting in the fit procedure.

For example, consider the situation where there are forty frequencies on a channel spacing of 100 GHz. Assume that the frequencies are specified to with +10 GHz. Assume also that on taking data from the photodiode array, twenty frequencies are missing or are at a power level too low to be usefully detected.

Taking therefore twenty good measurements of frequency at a tolerance of 10 GHz, this is equivalent to twenty measurements at a standard deviation of 3 GHz. Therefore the accuracy of the fit is 3 GHz/20 (for a root mean square algorithm), which is equal to 0. 5 GHz to one standard deviation, or a maximum deviation of 1.6 GHz.

This level of accuracy compares well with the corresponding accuracy of a semiconductor optical demultiplexer which has inherent temperature tuning to O. I'C, the achievable accuracy in that situation being : t1 GHz.

Figure 6 illustrates a diagram of an alternative optical configuration which can be used with the present invention. In Figure 6, like numerals denote like parts as in Figure 1. Only schematic detail is illustrated however compared to Figure 1. In the chip 2 of Figure 6, there is an input array waveguide 16 of the type illustrated

and explained in relation to Figure 1. Similarly there is a dispersive array waveguide 11 which guides the light from the input waveguide 16 to an output area. Then, in place of the output waveguide array 21, a mirror 30 picks up light output from the array waveguide 11 and reflects it onto a continuous photodiode array 26. In this embodiment, the photodiode array can include 256 photodiodes.

The technique described herein is particularly useful because as the number of channels in use increases, the scheme becomes increasingly accurate since the fit depends on an average of the grid error of carrier frequencies of the link to monitor. Thus, it can keep up with the increasingly stringent tuning requirements as channel spacings decrease without extra packaging costs being incurred.

Moreover, as customer frequency tolerances are tightened by a factor of two, then the method automatically locks to a tolerance improved by the same factor.

As more channels are added, the number n of frequencies to fit to increases, and the locking of the device improves proportionally to square root of n.

To use a software grid fitting technique, it is desirable to confirm that the temperature of the measurement device lies within a certain range. For example, to use a device similar to Figure 1 with two closely spaced photodiodes per signal, the temperature should lie within 0. 1 channels ; about li for 100 GHz channel spacing. The software method enhances the accuracy of such a device. For a device similar to Figure 6 with a continuous output, the temperature should lie within about 3 channels, 30C for 100GHz channel spacing, limited by the temperature change required to shift spectra beyond spare photodiodes at extremes of the photodiode array.

For an OCM which might stray outside of the temperature range, the technique can still be used by determining initially which channels are present so that the fit knows which channel numbers to assign to the incoming signals and report which has deviated from the expected grid.

The output errors can be used by the system controller to tune the light sources providing the optical input to compensate for the detected frequency error.

The described method uses a best-fit line to relative frequency vs. channel number as the reference to obtain the relative frequency errors of transmitters in a WDM optical communication system. Given the large number of transmitters in use, the best fit line is likely to be extremely close to the standard ITU grid, so that deviations from the fit indicate absolute grid errors with acceptable precision.

If the time to re-tune the transmitters towards higher frequencies (responsive to the reported errors) is not the same as to tune transmitters towards lower frequencies, then it is possible that more than half of the transmitters deviate from the standard ITU grid in the same direction, and will pull the fit line away from the ITU grid. It is possible that overheated laser transmitters will gradually pull the tuning of all of the other transmitters since the fit line shifts by the mean error. Two methods are proposed in the following to prevent the fit from wandering off the standard: Method 1 A fraction of the transmitters in the system are specified to contain their own internal frequency locking components, and the best fit line increases the weighting to the measurements from self-locked lasers if the tuning starts to pull away from the standard ITU frequencies. This allows the software fit method to provide useful error correction requests to remote laser controllers. This drags other lasers into line with a few frequency locked lasers. In normal use, all of the lasers contribute to the fit to give the best averaging away of random errors, but if the locked lasers all deviate from the fit in the same direction then there is a high risk that the fit is wrong, so the weighting to the locked laser signals should be increased to 100% and ordinary signals ignored in order to drag the ordinary lasers back to grid. Once the locked lasers are not all to one side of the fit line, then normal use can resume.

If some parts of the network use locked lasers but most do not, then the channel monitor can store a temporary"calibration fit"whenever a locked signal is included in a measurement set and use this memory to check that a set of ordinary lasers are close to the same set of frequencies as the locked lasers.

The channel monitor is temporarily able to recognise if a whole set of ordinary lasers have drifted off grid and report the required re-turning.

One example of a frequency locked laser is a Nortel frequency locked laser containing a two photodiode monitoring the back facet of a DFB. The ratio of signals in the monitor photodiodes measures laser frequency according to a glass microoptic Etalon. The temperature control of the laser is in closed-loop feedback control to maintain the specified frequency, and is advertised as having < +3GHz tolerance over lifetime.

Method 2 One precision reference frequency (such as reference input 16R in Figure 1) is introduced via an additional input waveguide simultaneously to the WDM signals which are being measured. The reference source may be remote from the channel monitor device and may be shared among several channel monitors to ensure that standards are agreed throughout the network. The reference source is permanently set to a frequency at one end of the channel list. If the reference source is outside of the measurement system tolerance (e. g. 3GHz) of the fit line the fit is wrong and the reference offset should be added to the relative grid errors. Re-tuning of the transmitters based on the referenced measurements will bring the fit line back toward the reference point. When this is within the measurement tolerance of the channel monitor then normal use can resume.

Although at least sequence fit has been described for generating the line of best fit for measured results, other fitting methods could be used.

The technique described herein can also be used to measure relative frequency errors in an optical communication system which uses a known set of frequencies of irregular spacing. In this case, the graph of signal number vs. relative frequency must use real numbers rather than integers to replace "channel number". For instance, a"channel 2. 4" signal at 194.24THz could be allowed, provided that the software is expecting a channel there and a demux with a continuous output such as Figure 6 is used.

Claims

CLAIMS : 1. A method of measuring grid error in an optical device in which a plurality of optical channels at different frequencies are incident on a photodiode array comprising a plurality of photodiodes each operable to convert light energy incident thereon to an electrical value representative of the intensity of the incident light, the method comprising: determining from the electrical values the location of channel peaks in the photodiode array to generate a set of real data; generating from said real data and from said channel peak locations a line of best fit according to a predetermined algorithm of said optical channels to said photodiode array whereby a set of estimated data is obtained; and comparing said real data with said estimated data to provide an indicator of grid error.
2. A method according to claim 1, wherein the grid error is reported as a deviation in frequency for each channel number.
3. A method according to claim 2, wherein the frequency deviation is used to tune a light source providing the optical channel on which that frequency deviation is detected.
4. A method according to claim 1, wherein the grid error is reported as an error status for each channel number.
5. An optical device comprising a dispersive optical guide which guides a plurality of optical channels at different frequencies and a photodiode array comprising a plurality of photodiodes each operable to convert light energy incident thereon to an electrical value representative of the intensity of the incident light, the optical device further comprising a system for measuring grid error having:

<Desc/Clms Page number 15>

means for determining from the electrical values the location of channel peaks in the photodiode arrangement to generate a set of real data; means for generating from said real data and from said channel peak locations a line of best fit according to a predetermined algorithm for said optical channel to said photodiode array whereby a set of estimated data is obtained; and means for comparing said real data with said estimated data to provide an indicator of grid error.
6. An optical device according to claim 5, wherein the comparing means is operable to provide an indicator of grid error as a deviation in frequency for each channel number.
7. An optical device according to claim 5 or 6, which comprises a plurality of light sources for inputting said optical channels.
8. An optical device according to claim 7, which comprises a system controller for receiving said indicator of grid error and for selectively altering the frequency of one or more of said light sources in dependence on said grid error.
9. A multi-chip optoelectronics module comprising: a first chip constituting an optical device with a dispersive optical guide which guides a plurality of optical channels at different frequencies and a photodiode array comprising a plurality of photodiodes each operable to convert light energy incident thereon to an electrical value representative of the intensity of the incident light ; and a second, processing, chip which is operable to receive said electrical values and to determine therefrom the location of channel peaks in the photodiode arrangement to generate a set of real data, to determine a line of best fit according to a predetermined algorithm for said optical channel to said photodiode array whereby a set of estimated data is obtained and to compare said real data with said estimated data to provide an indicator of grid error.