Polarization dependent loss measurement

Information

  • Patent Grant
  • 11644381
  • Patent Number
    11,644,381
  • Date Filed
    Friday, May 7, 2021
    3 years ago
  • Date Issued
    Tuesday, May 9, 2023
    a year ago
  • Inventors
  • Original Assignees
  • Examiners
    • Lee; Hwa Andrew
    Agents
    • Baratta Law PLLC
    • Baratta, Jr.; Lawrence A.
Abstract
There is provided a method for measuring the PDL of a DUT as a function of the optical frequency ν within a spectral range, which uses a single wavelength scan over which the input-SOP varies in a continuous manner. The power transmission through the DUT, curve T(ν), is measured during the scan and the PDL is derived from the sideband components of the power transmission curve T(ν) that results from the continuously varying input-SOP. More specifically, the Discrete Fourier Transform (DFT) of the power transmission curve T(ν) is calculated, wherein the DFT shows at least two sidebands. At least two sidebands are extracted and their inverse DFT calculated individually to obtain complex transmissions (ν), =−J . . . J, where J is the number of sidebands on one side. The response vector |m(ν) of the DUT is derived from the complex transmissions (ν) and a matrix determined by the continuous trajectory of the SOP of the input test lightwave; and the PDL of the DUT as a function of ν (PDL curve) is derived therefrom.
Description
TECHNICAL FIELD

The present description generally relates to polarization dependent loss (PDL) measurement, and more particularly to measuring polarization dependent loss of a device under test using a single spectral scan.


BACKGROUND

Conventional methods for Polarization Dependent Loss (PDL) measurement include the Polarization Scanning technique and the Mueller Matrix Method.


In the Polarization Scanning technique, the Device Under Test (DUT) is exposed to virtually all States Of Polarization (SOP) and the PDL is determined from the maximum and minimum power transmission through the DUT. Because the Polarization Scanning technique needs to expose the DUT to many states of polarization, the PDL is measured at a single wavelength at a time. Measuring the PDL over a broad wavelength range can quickly become impractical.


The classic Mueller Matrix Method (see U.S. Pat. No. 5,371,597 to Favin et al.) aims at determining the first row of the Mueller matrix (4 elements) that characterizes the DUT, from which the PDL of the DUT is deduced. To this end, a lightwave having successively a small number (e.g. 4) of distinct, known input states of polarizations (SOP) is launched into the DUT, and the corresponding small number of power transmissions are measured (power of the lightwave at the output of the DUT divided by the power at the input of the DUT). To obtain the PDL characterization over wavelength, distinct wavelength scans need to be performed sequentially for each of said small number of distinct input-SOPs (minimum of 4), wherein the input-SOP is fixed (constant as a function of time) during each individual scan.


The classic Mueller Matrix Method is prone to substantial errors should the DUT be unstable over the interval between said sequential wavelength scans, e.g., due to moving patch-cords, in which case the PDL measurement can be utterly wrong. A single wavelength scan approach would resolve this problem by making the SOP of the input test lightwave vary rapidly during the wavelength scan, so the DUT response to a plurality of input-SOPs would be measured at each wavelength within a much shorter time.


A method was proposed in the art (see U.S. Pat. No. 6,856,386 to Anderson) to measure the PDL based on the Mueller Matrix Method but using a single wavelength scan, by changing the input-SOP during the wavelength scan in a cyclic manner. In this method, the input-SOP can be changed using either an active or a passive device. However, one drawback of this method is that it requires that the SOP of the input test lightwave be either carefully controlled or monitored using a polarimeter and, in the case of a passive SOP generator, the combination of a polarimeter and additional polarization controller in an feedback loop, which significantly increases the difficulty of implementation, complexity and cost of the test hardware.


There therefore remains a need for an improved single wavelength scan PDL measurement method.


SUMMARY

There is herein provided a PDL measurement method which uses a new set of equations, i.e. different from that of the classic Mueller Matrix Method, as well as new data acquisition scheme and data processing to derive the PDL from a single wavelength scan (or spectral scan) over which the input-SOP varies. As explained in the detailed description, the proposed new set of equations can be viewed as a generalization of the mathematics of the Mueller Matrix Method. When compared to the classic Mueller Matrix Method, the new set of equations is applied to derive the same result, i.e., the 4-element response vector of the DUT (elsewhere referred to as the first row of the Mueller matrix) as a function of optical frequency (or wavelength), from a different data set obtained by a different data acquisition procedure, i.e., a single wavelength scan over which the input-SOP varies continuously over the scan instead of a few sequential scans with fixed SOPs.


Therefore, in accordance with one aspect, there is provided a method for measuring the PDL of a DUT as a function of optical frequency ν within a spectral range, which uses a single wavelength scan over which the input-SOP varies in a continuous manner, i.e. not a finite number of discrete points, but a continuous trajectory on the surface of the Poincaré sphere as a function of ν. The power transmission through the DUT, curve T(ν)=Pout(ν)/Pin(ν) is measured during the scan, where Pin(ν) and Pout(ν) represent the power of the ligthwave at the input and output of the DUT respectively. The PDL is derived from the resolved sideband components of the power transmission curve T(ν) that result from the continuously varying input-SOP. More specifically, the Discrete Fourier Transform (DFT) of the power transmission curve T(ν) is calculated, wherein the DFT shows at least two resolved sidebands in the delay (τ) domain (four in the hereindescribed embodiment), where delay τ is the conjugate variable of ν in the Fourier transform (the same way as signal-frequency f is the conjugate variable of time t). At least two sidebands are extracted individually, i.e. selected by applying a bandpass filter in the delay domain, and their inverse DFT calculated to obtain individual complex transmissions custom character(ν), custom character=−J . . . J, where J is the number of sidebands defined as the number of resolved peaks on a single side of the DFT, e.g. side τ>0. The transmissions custom character(ν) are complex in the sense that they are complex numbers (real and imaginary parts, or equivalently, amplitude and phase). The 4-element response vector |m(ν)custom character of the DUT, elsewhere referred to as the first line of the Mueller matrix, is derived from the complex transmissions custom character(ν) and a matrix custom character that represents a system of equations to be solved. As well known, a system of equations can be advantageously solved by arranging the known coefficients into a matrix, and inverting said matrix. In the context of RSB, said matrix custom character only depends on the continuous trajectory of the input-SOP (continuous curve on the surface of the Poincaré sphere). Matrix custom character thus constitutes a generalization of the corresponding matrix A of the classic Mueller Matrix Method which similarly represents a system of equations and only depends on the finite number of known discrete input-SOPs. Of note is that neither matrix A nor custom character is the Mueller matrix; the Mueller matrix characterizes a DUT, thus is independent of any particular set of input-SOPs. Finally, the PDL of the DUT as a function of ν (PDL curve) is derived from said response vector in the same way as in the classic Mueller Matrix Method.


The provided method allows for asynchronous sampling of the power transmission curve T(ν) with respect to the variation of the input-SOP, i.e. the sampled points do not need to fall on specific input-SOPs. This property allows to circumvent the need for monitoring the input-SOP with a polarimeter in the case of an active SOP generator (see active SOP generator approach of U.S. Pat. No. 6,856,386), or to additionally control the SOP through a feedback loop in the case of a passive SOP generator (see passive SOP generator approach of U.S. Pat. No. 6,856,386).


Accordingly, measuring the power transmission curve T(ν) comprises data sampling that is allowed to be asynchronous with respect to the SOP variation of the input test lightwave, i.e. the sampled points do not need to fall on a pre-determined set of input-SOPs, and consequently, the most usual and simple kind of sampling can be performed, i.e. sampling at equal intervals dν between successive samples, regardless of the input-SOP at the values of ν corresponding to the samples.


The phase of the SOP of the input lightwave along the continuous trajectory, i.e. where exactly along the trajectory is the point that represents the SOP at a given value of ν, may be either known or measured at each sampled point.


In the provided method, said phase can be measured live through appropriate data processing of the output of a single polarizer (referred to as the phase reference signal Pφ in the detailed description), no need for a full polarimeter nor additional polarization controller and feedback loop.


In accordance with another aspect, there is provided a method for measuring the PDL of a DUT within a spectral range. The method comprises:

  • generating an input test lightwave which State Of Polarization (SOP), as represented by a point on the surface of the Poincaré sphere, varies according to a continuous trajectory while being spectrally scanned over the spectral range and launching the input test lightwave into the DUT;
  • measuring the power transmission through the DUT, curve T(ν), during a spectral scan of the input test lightwave;
  • calculating a Discrete Fourier Transform (DFT) of the power transmission curve T(ν), wherein the DFT shows at least two sidebands;
  • extracting said sidebands and computing the inverse DFT of each said sidebands to obtain complex transmissions custom character(ν);
  • computing the response vector |m(ν)custom character of the DUT from said complex transmissions custom character(ν) and a matrix custom character determined by the continuous trajectory of the SOP of the input test ligthwave;
  • deriving the spectrally-varying PDL of the DUT from said response vector |m(ν)custom character.


In accordance with another aspect, there is provided a system for implementing the provided PDL measurement method. The apparatus comprises:

  • a tunable light source to generate a test lightwave and spectrally scanning the test lightwave over the spectral range;
  • a State Of Polarization (SOP) generator for varying the SOP of the test lightwave according to a continuous trajectory on the surface of the Poincaré Sphere while the test lightwave is spectrally scanned over the spectral range, the test lightwave to be launched into the DUT;
  • a power transmission measurement apparatus for measuring the power transmission through the DUT;
  • an acquisition device for sampling values of the power transmission through the DUT during a spectral scan of the test lightwave and produce the measured power transmission curve T(ν);
  • a processing unit configured for:
    • calculating a Discrete Fourier Transform (DFT) of the power transmission curve T(ν), wherein the DFT shows at least two sidebands;
    • extracting said sidebands and computing the inverse DFT of each said sidebands, to obtain complex transmissions Tcustom character(ν);
    • computing the response vector |m(ν)custom character of the DUT from said complex transmissions custom character(ν) and a matrix custom character determined by the continuous trajectory of the SOP of the input test lightwave;
    • deriving the spectrally-varying PDL of the DUT from said response vector |m(ν)custom character.


Throughout this specification, spectrally-varying values, functions and equations are expressed as a function of the optical frequency ν. It is however noted that such variation could equivalently be termed as a variation as a function of wavelength λ, where λ=c/ν or the wave number σ=ν/c. Similarly, the expressions “wavelength scan” and “spectral scan” are used interchangeably herein to denote a scan that varies the optical frequency ν and the wavelength λ of a lightwave over a spectral range.


In this specification, unless otherwise mentioned, word modifiers such as “substantially” and “about” which modify a value, condition, relationship or characteristic of a feature or features of an embodiment, should be understood to mean that the value, condition, relationship or characteristic is defined to within tolerances that are acceptable for proper operation of this embodiment in the context of its intended application.


In the present description, and unless stated otherwise, the terms “connected”, “coupled” and variants and derivatives thereof refer to any connection or coupling, either direct or indirect, between two or more elements. The connection or coupling between the elements may be mechanical, physical, operational, electrical or a combination thereof.


Further features and advantages of the present invention will become apparent to those of ordinary skill in the art upon reading of the following description, taken in conjunction with the appended drawings.





BRIEF DESCRIPTION OF THE DRAWINGS


FIG. 1A is a graph showing a continuous trajectory, on the surface of the Poincaré sphere, of the SOP of an input test lightwave as a function of optical frequency ν, in accordance with one embodiment for which J=4.



FIG. 1B is a graph showing the discrete Fourier transform of the example trajectory of FIG. 1A.



FIG. 1C is a graph showing the power transmission curve T(ν) through an example DUT, in response to the input-SOP trajectory of FIG. 1A.



FIG. 1D is a graph showing the discrete Fourier transform of the example power transmission curve T(ν) of FIG. 1C.



FIG. 1E is a graph showing the discrete Fourier transform of the example trajectory of FIG. 1A as shown in FIG. 1B only with additional annotations.



FIG. 1F is a graph showing the discrete Fourier transform of the example power transmission curve T(ν) of FIG. 1C as shown in FIG. 1D only with additional annotations.



FIG. 2 is a schematic illustrating the physical principle of a passive SOP generator, in accordance with one embodiment.



FIG. 3 is a schematic showing a passive SOP generator in accordance with one practical implementation of the principles illustrated in FIG. 2, which is based on segments of polarization-maintaining fiber.



FIG. 4 is a schematic showing a passive SOP generator in accordance with another practical implementation of the principles illustrated in FIG. 2, which is based on polarization-splitting Mach-Zehnder interferometers.



FIG. 5 is a schematic showing a passive SOP generator in accordance with yet another practical implementation of the principles illustrated in FIG. 2, which is based on all-bulk polarization-splitting Mach-Zehnder interferometers.



FIG. 6 is a block diagram showing a PDL measurement system in accordance with one embodiment which employs a single spectral scan and a passive SOP generator.



FIG. 7A is a schematic showing a power transmission measurement apparatus in accordance with an embodiment which is all-fiber based.



FIG. 7B is a schematic showing a power transmission measurement apparatus in accordance with another embodiment which is based on bulk components.



FIG. 8 is a schematic showing an acquisition device in accordance with one embodiment.



FIG. 9 is a schematic illustrating the relationship between the governing Stokes vectors, ŝo, p0 and p1, and the coupling angles (φš0, φp characterizing the 2-segment passive SOP generator.



FIG. 10 is a flowchart showing the steps of the resolved-sidebands (RSB) processing method.



FIG. 11 shows a pseudocode expressing in more details Procedule_3 of the method of FIG. 10.



FIG. 12 shows a pseudocode expressing in more details Procedule_2 of the method of FIG. 10.



FIG. 13 shows a pseudocode expressing in more details Procedule_1 of the method of FIG. 10.



FIG. 14 shows a pseudocode expressing in more details Procedule_0 of the method of FIG. 10.





It will be noted that throughout the drawings, like features are identified by like reference numerals.


It should also be understood that when the appended drawings are denoted as schematics, elements of the drawings are not necessarily drawn to scale. Some mechanical or other physical components may also be omitted in order to not encumber the figures.


The following description is provided to gain a comprehensive understanding of the methods, apparatus and/or systems described herein. Various changes, modifications, and equivalents of the methods, apparatuses and/or systems described herein will suggest themselves to those of ordinary skill in the art. Description of well-known functions and structures may be omitted to enhance clarity and conciseness.


Although some features may be described with respect to individual exemplary embodiments, aspects need not be limited thereto such that features from one or more exemplary embodiments may be combinable with other features from one or more exemplary embodiments.


DETAILED DESCRIPTION
General Description of the Method

The present paragraph exposes a case only as an explanatory step intended to help the reader to understand the proposed method, but there is no foreseeable advantage in doing this step as such. In the classic Mueller Matrix Method, a number J of wavelength scans are performed sequentially with fixed input-SOPs, meaning that the input-SOP does not change during the scan. Now, imagine that a number J of wavelength scans are similarly performed sequentially but with input-SOPs (i.e., SOPs of the test lightwave) that, instead of being fixed SOPs, oscillate sinusoidally as a function of the optical frequency ν, e.g., describing distinct circles on the surface of the Poincaré sphere as a function of ν. Then for each scan j, j=0 . . . (J−1), we obtain a transmission curve Tj(ν) representing the response of the DUT to said oscillating SOP (instead of the response to a fixed one). Since, in the classic Mueller Matrix Method, the set of J transmission curves obtained with J fixed SOPs is enough information to deduce the desired result, it follows that the set obtained with J oscillating SOPs, thus more individual values of SOPs in a way, is also enough. Of course, deducing the result in the case of oscillating SOPs requires a new set of equations, different from the one of the classic Mueller Matrix Method.


Part of the PDL measurement method described herein can be seen as a generalization of the mathematics of the classic Mueller Matrix Method in the context of J responses to J oscillating input-SOPs instead of fixed, constant input-SOPs. In the classic Mueller Matrix Method, the 4-element response vector of the DUT, |m(ν)custom character, is obtained by solving a linear system of equations, where the four unknowns to be determined are the four elements of the response vector. As for any linear system of equations, the coefficients that multiply the unknowns in the different equations can be grouped to form a single matrix, say matrix A (not to be confused with the Mueller matrix), and the solution of the system of equations is obtained by first computing the inverse of A. In the classic Mueller Matrix Method, said matrix A only depends on, and is deduced from the known set of J fixed SOPs. In the case of oscillating SOPs as described above, it can be shown that there still exists such a matrix representing the system of equations to be solved, which is the matrix custom character. Notably, it has a different number of rows, and its elements are complex numbers instead of real numbers. But the important point to note is that this equivalent matrix indeed exists and only depends on, and is deduced from the known set of, in this case, oscillating SOPs. This constitutes the main aspect of said generalization of the mathematics of the classic Mueller Matrix Method.


However, as mentioned hereinabove, there is no advantage in making J sequential scans with oscillating SOPs instead of fixed one. Indeed, the goal is to get the result using a single scan. Therefore, there is herein proposed to step up from the above explanatory case to one where the responses Tj(ν) to J oscillating input-SOPs are obtained both simultaneously and separately from a single scan. For this to be, the responses Tj(ν) should be separated as, by analogy, multiple communication channels are separated in the frequency domain. Using an input-SOP that follows a suitable continuous trajectory on the surface of the Poincaré sphere as a function of ν allows to generate responses Tj(ν) that each occupies its own band in the delay domain r around a given carrier-delay ±τ1, ±τ2, etc. How to generate such trajectory will become clear in the following description of embodiments. It is noted that since the power transmission curve T(ν) is a function of optical frequency ν, the abscissa of the Fourier transform of T(ν) is delay τ (not signal-frequency f) and the oscillating components of T(ν) appear as resolved sidebands in the delay domain (τ) (not the signal-frequency domain (f)).


The above explains the name given to the method, i.e. the Resolved-SideBands (RSB) method. Resolved because said sidebands are separated enough in the delay domain (τ) for the different corresponding signals, the responses Tj(ν), not to overlap. Yet more clearly, the Fourier transforms of the responses Tj(ν), which are functions of delay τ, do not overlap in the delay domain. An enlightening analogy can be made with a much more familiar case; the carriers of TV channels which should be separated enough in the frequency domain for adjacent signals spectra (in the Fourier transform of the transmitted/received electromagnetic waves) not to overlap, enough to avoid significant “crosstalk”, this being determined by the required bandwidth of the signals that modulate the sinusoidal carriers. Similarly, in the RSB method, the equivalent of the time-varying signal bandwidth is the “DUT extent” in the delay domain (τ), i.e. how large is the Fourier transform of the DUT transmission curve TDUT(ν). For example, if the DUT is a narrow optical filter such as a WDM filter, the narrower TDUT(ν), the larger its extent in the delay domain. Of course, in order to avoid “crosstalk” the spacing between the sideband-carriers are set larger than the maximum extent of the DUT that is expected to be encountered. This spacing is a characteristic of the SOP generator that one can set in accordance with the expected application of the instrument, i.e. large enough for the sidebands not to overlap for DUTs having the largest expected extent (typically DUTs that have the narrowest and/or sharpest transmission TDUT(ν)).


As in the classic Mueller Matrix Method, the result of the measurement of the RSB method is the 4-element response vector |m(ν)custom character (elsewhere named first line of the Mueller matrix) as a function of optical frequency ν. From the response vector |m(ν)custom character, the PDL can be derived as known in the art, in the same way as in the classic Mueller Matrix Method. The polarization-dependent center wavelength (PDCW) and the polarization-dependent bandwidth (PDBW) of the DUT can also be derived as known in the art from |m(ν)custom character in cases where the DUT is such that these parameters have a defined meaning, e.g., optical bandpass filters.


It is worth mentioning that there is a formal equivalence between the mathematics of the classic Mueller Matrix Method and the mathematics of the RSB method, mainly through the existence of said equivalent matrix custom character (the matrix that represents the system of equations to be solved): the most notable difference is that, contrary to the corresponding matrix (again, not the Mueller matrix) of the classic Mueller Matrix Method, matrix custom character of RSB has complex elements (real and imaginary parts), and the measured transmission terms custom character(ν), custom character=−J . . . J, where J is the number of sidebands, are also complex-valued (amplitude/phase of the oscillating responses) instead of real-valued. The notation custom character is used to clearly distinguish the complex phasor custom character(ν) from the real response Tj(ν) (real part of custom character(ν)). Representing such oscillating signals by complex phasors makes the mathematics of RSB much simpler and concise than would otherwise be.


The proposed PDL measurement method is now further described with reference to FIGS. 1A to 1D which are used to illustrate the oscillating components of T(ν) appearing as sidebands in the delay domain (τ) (see FIG. 1D-1F).



FIG. 1A shows an exemplary continuous trajectory of the SOP of an input test lightwave as a function of optical frequency ν for which J=4. The trajectory shown in FIG. 1 is one among an infinity of possible trajectories. The SOP at a given value of ν corresponds to one point on the surface of the Poincaré sphere (a representation widely used in the art to depict the SOP). The position of said point is represented by the 3-element unit Stokes vector ŝ(ν)=(s1(ν) s2(ν) s3(ν))τ, where the three coordinates s1, s2, and s3 are the Stokes parameters. The trajectory is the overall path followed by the input-SOP on the surface of the Poincaré sphere.



FIG. 1B shows the DFT (discrete Fourier transform) of the example trajectory of FIG. 1A; since ŝ(ν) is a 3-element vector, there are in fact three such DFTs, one DFT for each of the Stokes parameters (coordinates); but to simplify, FIG. 1B regroups the three curves into a single one given by the root-mean-square value of the three DFTs at each point ν, which suffices to illustrate the DFT of the example trajectory ŝ(ν) of FIG. 1A. In this exemplary case, the DFT consists of J=4 discrete sidebands, where J is defined as the number of discrete tones on one side of the Fourier transform (e.g. the τ>0 side). Of note is that a suitable trajectory of the input-SOP for the proposed RSB method should be such that its Fourier transform shows a small number J of discrete sidebands; J=2 is the minimum because J=1 implies that the trajectory is a circle and thus all SOPs are coplanar, in which case there is no solution to the system of equations (matrix custom character is singular), J=4 is our preferred choice, and J>4 (no theoretical upper limit) is useless, and may even have indirect drawbacks in practice. As an example, if using a lot more sidebands than basically needed, then T(ν) has a much larger extent than needed in the delay domain. This notably implies that, when scanning ν at a given constant scan-speed in a practical embodiment, the photodetector electronics should have a larger bandwidth, which generally results in both a larger noise spectral-density and smaller signal-to-noise ratio for a given input power.


Because the example trajectory of FIG. 1A is periodical, the discrete sidebands in its Fourier transform are equally spaced. But the RSB method does not require a periodical trajectory, nor equal spacing of the sidebands. However, it may be more judicious in practice to target a substantially equal spacing in order to best separate the sidebands from one another within some given total delay range.



FIG. 1C shows the power transmission curve T(ν) through an example DUT, in response to the input-SOP trajectory of FIG. 1A. The example DUT is an optical filter having a passband of 30 GHz (FWHM) and a PDL of 3 dB.



FIG. 1D shows the DFT of the power transmission curve T(ν) shown in FIG. 1C. As seen by comparing FIG. 1B and FIG. 1D, there is a one-to-one correspondence between component custom character of the SOP trajectory (custom character=−J . . . J, custom character≠0) (FIG. 1B, 1E) and sideband Γ in the DFT of T(ν) (FIG. 1D, 1F), located at delay values custom character(custom character=−J . . . J, custom character≠0). In other words, each sinusoidally oscillating input-SOP component custom character (complex phasor) (FIG. 1B, 1E) generates a sideband custom character around carrier-delay custom character in the DFT of T(ν) (FIG. 1D, 1F). In the special case where the custom character=0 part of FIG. 1B is null, as for this example trajectory, the custom character=0 part of the DFT of T(ν) (FIG. 1D) is the DFT of the average transmission that would be measured with unpolarized input-light (smoother curve DOP=0 in FIG. 1C, where DOP means Degree of Polarization).


The Mathematics of RSB

The mathematics of RSB is now described in detail, with reference to FIGS. 1E and 1F, which are the same as FIGS. 1B and 1D but with additional annotations. Throughout this specification we conveniently use the bra-ket notation defined as follows: a ket |vcustom character is a column vector that may have any number of either real or complex-valued elements, and the corresponding bra custom characterv| is the transposed complex-conjugate of |vcustom character, thus a row vector. Finally, the bracket notation custom characterv/v′custom character represents the scalar product defined as,

bra: custom characterv|=(v*0 v*1 v*2 . . . v*n−1), where * means complex-conjuguate.
ket: |vcustom character=(v0 v1 v2 . . . vn−1)τ, where τ means transposed.
scalar product: custom characterv|v′custom character=v*0v′0+v*1v′1+v*2v′2+ . . . +v*n−1v′n−1


Although seldom mentioned explicitly, all PDL measurement methods, including the classic Mueller Matrix Method, assume that the DUT response to an input lightwave is linear. Indeed, the Mueller matrix of the DUT is by definition a linear transformation of the input Stokes vector into the output Stokes vector. Otherwise, i.e. a non-linear DUT, the very parameter to be measured, PDL, does not even exist as commonly defined. For instance, PDL at a given value of ν would be multi-valued, notably depending on the power of the input lightwave. The RSB method makes this same basic assumption as a matter of course.


The general form of such linear response is, as far as only the transmitted power is concerned,

T(ν)=custom characters(ν)|m(ν)custom character  (1)

with T(ν)=Pout(ν)/Pin(ν), where Pin(ν) and Pout(ν) are the power of the lightwave at input and output of the DUT, |m(ν)custom character=(m0(ν) m1(ν) m2(ν) m3(ν))τ is said response vector of the DUT measured by the method, and custom character(s(ν)|=(1 s1(ν) s2(ν) s3(ν))τ is the 4-element bra representing the 100% polarized input-SOP, where the last three elements of custom characters(ν)| are those of the 3-element unit Stokes vector ŝ(ν)=(s1(ν) s2(ν) s3(ν))τ, |ŝ(ν)|=1.


Mathematical note: The most general linear transformation mapping a vector (input-SOP) to a scalar (transmission T) is by definition a scalar product, as (1) is. The response vector embodies this mapping even if Mueller-matrix calculus would not exist; this is to say that reference to the Mueller matrix is incidental, including in the classic method said of the Mueller matrix; it happens that the response vector is indeed identical to the first line of the 4×4 Mueller matrix, but this matrix is nowhere used in the calculations.


As exposed hereinabove with reference to FIGS. 1A to 1D and illustrated in FIG. 1E, a continuous trajectory ŝ(ν) of the input-SOP that is suitable for RSB is the sum of (2J+1) components, i.e.,

ŝ=custom character(custom character=−J . . . J), custom character=(custom character)τ  (2)

where ω=2πν, and custom character is the complex vectorial amplitude of component custom character in the DFT of the known trajectory ŝ(ν) of the input-SOP. Observe that custom character is a 3-element complex-valued vector; in (2), the three complex amplitudes corresponding to the three individual Stokes parameters (s1, s2, s3) are simply wrapped up in this single, complex-valued vector. It allows to write it all in a single equation instead of three, as in (2), but also everywhere else below. Analytical expressions of the custom character for specific trajectories will be given in the description of the preferred embodiment. Of note is that the custom character can of course be measured; having analytical expressions for the preferred embodiment is convenient, but not required. Now, since response (1) is linear, one component custom character of ŝ(ν) in (2) generates one component custom character of T(ν) by mere definition of linearity, i.e.,

T(ν)=custom character(custom character=−J . . . J)  (3)


Therefore, each component custom character of ŝ(ν) generates one and only one term custom character in sum (3), which translates as sideband custom character around carrier-delay custom character in the DFT of T(ν) as shown in FIG. 1F. Thus (1) implies that complex transmissions custom character(ν) are given by,

custom character(ν)=custom character|m(ν)custom character  (4)

where custom character| is the 4-element bra corresponding to custom character, i.e.,

custom character|=(custom character) where custom character=1 if custom character=0, 0 otherwise.  (5)


Observe the presence of Kronecker symbol custom character in (5). This reflects that the average over ω of an oscillating SOP component custom character with custom character≠0 is null.


Equation (4) is a compact expression of a system of (2J+1) equations (custom character=−J . . . 1), one equation for each value of custom character, in which the 4 unknowns are the 4 elements of the response vector |m(ν)custom character. Also, equation (4) shows that there is one such system of equations for each value of ν. This is made more apparent by writing the scalar product (4) at length in terms of the four elements of both custom character| and |m(ν)custom character; one equation, i.e. for one value of custom character at one value of ν, then reads as,

custom character·m0(ν)+custom character·m1(ν)+custom character·m2(ν)+custom character·m3(ν)=custom character(ν)  (6)

which is one among a system of (2J+1) equations (custom character=−J . . . J) where (m0(ν) m1(ν) m2(ν) m3(ν)) are the 4 unknowns to be determined, (custom character) are the known coefficients depending only on the known continuous trajectory ŝ(ν) of the input-SOP, and custom character(ν) is the measured complex transmission curve custom character extracted from sideband custom character in the DFT of T(ν) as exposed previously. As well known, a system of equations can be solved by arranging the known coefficients into a matrix, and inverting this matrix. In RSB, said matrix is noted custom character, and following from equations (4) and (5) the system of equations at a given ν is wrapped up as,

|custom character(ν)custom character=custom character|m(ν)custom character  (7)

the solution of which is the measured response vector at optical frequency ν, given by,

|m(ν)custom character=custom character−1|custom character(ν)custom character  (8)

where custom character−1 is the generalized inverse of custom character. For the sake of illustration, we show below the explicit expressions of custom character and |custom character(ν)custom character in our preferred case with J=4, and then list their relevant general characteristics:









=



(



0



z
4.1
*




z
4.2
*




z
4.3
*





0



z
3.1
*




z
3.2
*




z
3.3
*





0



z
2.1
*




z
2.2
*




z
2.3
*





0



z
1.1
*




z
1.2
*




z
1.3
*





1



z
0.1




z
0.2




z
0.3





0



z
1.1




z
1.2




z
1.3





0



z
2.1




z
2.2




z
2.3





0



z
3.1




z
3.2




z
3.3





0



z
4.1




z
4.2




z
4.3




)







(
v
)





=

(





4
*



(
v
)








3
*



(
v
)








2
*



(
v
)








1
*



(
v
)








0



(
v
)








1



(
v
)








2



(
v
)








3



(
v
)








4



(
v
)





)






(
9
)








The general characteristics of ket |custom character(ν)custom character are:

    • |custom character(ν)custom character is an ordered list of all (2J+1) measured complex transmissions custom character(ν) at ν, disposed in a column vector. Thus it has (2J+1) rows=number of equations.
    • Row i of |custom character(ν)custom character (i=0 . . . 2J) is the measured custom character(ν) with custom character=(i−J).
    • custom character(ν)=custom character(ν) (property used in writing equation (9), only to save data processing time)


      The relevant general characteristics of matrix custom character are:
    • custom character has 4 columns (number of unknowns) and (2J+1) rows (one row for each equation).
    • Each row i of custom character (i=0 . . . 2J) is simply the bra custom character| given by equation (5) with custom character=(i−J). In words, each row of custom character is one of the oscillating component of the input-SOP represented by the corresponding bra (row vector).
    • custom character


But then custom character is not a square matrix and thus has no inverse in a strict sense, reflecting the fact that there are more equations than unknowns (more rows than columns). It is to be understood that having more equations than minimally required is by no way a drawback of any sort. Only that mathematically speaking we have to use a generalized inverse of custom character, here defined as follows,

custom character−1=B−1custom character with B=custom charactercustom character ( means transposed complex-conjugate)  (10)


The response vector |m(ν)custom character that RSB seeks to measure is given by equations (8) and (10).


This completes the description of the part of RSB that can be said to generalize the mathematics of the classic Mueller Matrix Method. Indeed, although the compact form of equation (11) below is seldom seen in the literature, in the normed classic Mueller Matrix Method the solution can be similarly written as,

|m(ν)custom character=A−1|T(ν)custom character  (11)

where A is a real (no imaginary part) 4×4 matrix which rows are the 4 real bras representing the 4 fixed input-SOPs, and |T(ν)custom character is a 4-element vector consisting of the 4 real transmissions, measured at the sameν but obtained sequentially in time (i.e. one distinct scan for each fixed input-SOP, one after the other). So written, equations (8) and (11) are found to be formal mathematical equivalents.


Finally, to make sure that a possible misunderstanding is avoided: A and custom character are indeed matrices, involved in the calculations of their respective methods, but neither of them is the Mueller matrix. The name Mueller Matrix Method may be misleading for that matter, since the full Mueller matrix is nowhere used in the calculations, only its “first line”, thus a much simpler vector which we name response vector, whereas indeed a matrix is involved, matrix A, be it explicitly or implicitly (solving a system of equations).


In some embodiments, a suitable trajectory of the input-SOP as a function of ν as described hereinabove is generated using a passive SOP generator. Such a passive SOP generator advantageously generates by laws of Physics a trajectory showing the desired small number of sidebands (J=2, 3 or 4). However, in other embodiments, suitable trajectories of the input-SOP may be generated by using an actively controlled SOP generator. Such actively controlled SOP generators are generally known in the art and commercially available and will therefore not be herein described in detail. Actively controlled SOP generators are also much more complex, bulky and costly.


Passive SOP Generator


FIG. 2 shows a passive SOP generator in accordance with one embodiment that is used to illustrate the physical principle. The passive SOP generator is based on a two-segment PMD emulator, abbreviated as the 2-segs from hereon. A single segment is not enough for generating a suitable trajectory (because then J=1), whereas more than two is unnecessary. As illustrated in FIG. 2, an input polarized ligthwave (i.e. the output of a spectrally scanned polarized light source) successively passes through a first segment Seg0 and a second segment Seg1. Each segment Seg0, Seg1 is characterized by a Principal State of Polarization (PSP) that is constant as a function of ν and a Differential Group Delay (DGD) that may or may not be constant over the relevant spectral range but is herein assumed constant for simplicity. Each segment Seg0, Seg1 acts as a 3D rotation on the 3-element Stokes vector that represents the SOP. The rotation is characterized by a fixed rotation axis p (the PSP) and a rotation angle φ (phase), that varies linearly as a function of ν, which slope is the DGD (more precisely, DGD=dφ/dω, where ω=2πν). The output-SOP ŝ(ν) is thus given by the product of these two rotations applied to the input-SOP ŝo. In one embodiment, the first and second segments are embodied by pieces of Polarisation Maintaining Fibers (PMFs). Referring to FIG. 2, the input-SOP ŝo and the PSPs p0, p1 of respective segments Seg0, Seg1 are 3-element unit Stokes vectors. The DGDs of segments Seg0, Seg1 are respectively noted δτ0 to and δτ1. The coupling angles φŝo and φp are respectively the angle between ŝo and p0 and the angle between p0 and p1. Parameter r is the ratio of the DGD δτ0 of segment Seg0 over the DGD δτ1 of segment Seg1. The 2-segs may be designed to obtain a ratio r that is substantially an integer value such as 1, 2 or 3, so as to advantageously obtain substantially equally-spaced sidebands. There is however no requirement that the ratio r be a strictly integer value.


A 2-segs always generate 4 sidebands, located at delay values δτ0, δτ1, (δτ0+δτ1) and (δτ0−δτ1) in the delay domain. This implies that for ratios like r=1 and 2 some of the four are superimposed (located at the same delay value). Thus, the number of distinct (not superimposed) sidebands in the Fourier transform of the transmission T(ν) depends on the ratio r. For example, an integer ratio r≥3 will generate J=4 distinct sidebands, a ratio r=2 will generate J=3 distinct sidebands and a ratio r=1 will generate J=2 distinct sidebands only. In other words, both the number J and relative locations of the distinct sidebands, at delay values referred to as custom character(custom character=−J . . . J, custom character≠0) in this specification, depend on the ratio r of the 2-segs.


In one embodiment, the passive SOP generator is designed with a ratio r˜3, which advantageously generates J=4 substantially equally spaced sidebands, and the targeted coupling angles are φŝo90° and φp=90°. Almost any other coupling angles may be used, except 0° or 180°, but the above-mentioned choice minimizes the sensitivity of the measurement to both additive noise and uncertainty on the knowledge of the actual coupling angles.



FIG. 3 shows a passive SOP generator 300 in accordance with one practical implementation of the 2-segs of FIG. 2 which is based on PMF fibers only. The passive SOP generator 300 comprises two concatenated pieces of PMF 302, 304 and an input polarizer 306. It also comprises an optional phase reference arm 310.


The input polarizer 306 sets the SOP at the input of the first piece of PMF 302, and thus the coupling angle φŝo. The two pieces of PMF 302, 304 are fused at an angle φp between their respective PSPs (p0 and p1). Of note is that the specific values of the coupling angles φŝo and φp are not critical in the sense that no high precision alignment is required, as long as the actual angles are known or determined afterwards. The actual coupling angles φŝo and φp can be determined afterwards through a calibration process, e.g. using a polarimeter.


Respective lengths of PMF pieces 302, 304 determine the value of their DGDs δτ0, δτ1, thus also the ratio r. Here again, there is no high precision requirement in cutting the PMF pieces to a given length. The two DGDs may be effectively measured live through appropriate data processing of the phase reference signal Pφ(ν). More to the point, the two effective DGDs determine the phase of the SOP of the input lightwave along the trajectory, i.e. where exactly along the trajectory is the point that represents the SOP located at a given value of ν. This phase may be either known or measured at each sampled point. One may target design values of angles φŝo and φp, DGD δτ0 and DGD δτ1 with reasonable care, but the true angles and DGDs after assembly do not need to be precisely equal to the design values.


The phase reference arm 310 comprises a coupler 308 used to extract a small portion of the light at the output of the second PMF piece 304 toward a polarizer 312 (analyzer â) followed by a photodetector 314, which output constitutes a phase reference signal Pφ(ν). This phase reference signal would not be needed in a perfect world where the DGDs and corresponding phases φ0(ν), φ1(ν) (see FIG. 2) would be exactly known. In a real non-ideal world, the phase reference serves the following purpose: from the sampled signal Pφ(ν), said phases φ0(ν), φ1(ν) (angles of rotation) are deduced at each sampled point of the spectral scan through appropriate data processing. There is therefore no need for the DGDs to be accurately known in advance, infinitely stable and constant as a function of ν. It is noted that the input polarizer 306 and phase reference arm 310 may be made of bulk components or optical fiber components (i.e. polarizers, coupler or beamsplitter).


The angle φâ between â and p1, or in words between the analyzer axis and the PSP of the second PMF piece 304, is set by rotating polarizer 312 (or by an angular fusion splice between the second PMF piece 304 and the PMF-fiber pigtail of the polarizer 312). In practical embodiments φâ is set to roughly 45° (not critical, but it should not be 0°, 90° or 180°, because then either one of the oscillating components of the signal Pφ(ν) that are analyzed in the data processing to measure said phases live would have zero amplitude).


It is noted that, although simple, the embodiment of FIG. 3 may suffer from a non-negligible drawback is some cases. It will be understood that the narrower the DUT transmission curve in the optical frequency ν domain, the broader the extent of its Fourier transform in the delay τ domain. Therefore, if the intended application should allow DUTs having a large extent in the delay τ domain, such as narrowband WDM filters for example, the DGDs of the PMF pieces 302, 304 may need to be at least δτ1˜1 ns and δτ0˜3 ns (assuming r=3). Since the DGD per unit length of a typical Panda fiber is about 1.5 ps/m, the lengths of the PMF segments would be respectively 0.67 and 2.0 km, for a total of 2.67 km. Such length of PMF can be exceedingly expensive.


But larger DGD values, e.g. in the nanoseconds order of magnitude or even greater if needed, may be obtained by replacing the PMF pieces 302, 304 by devices 402, 404 which may be named Polarization-Splitting Mach-Zehnder interferometers PS_MZ.



FIG. 4 shows a passive SOP generator 400 in accordance with another practical implementation of the 2-segs of FIG. 2, which is based on such PS_MZs. This bulk-and-fiber PS_MZ embodiment overcomes the above-mentioned drawback of the all-fiber embodiment of FIG. 3. The optical path is split in two arms and then recombined using two Polarization Beam Splitters (PBS) 416. For each PS_MZ devices 402, 404, a path length difference (delay) between the two arms is set by a piece of fiber 418 inserted in one of the arms, preferably a piece of PMF fiber in order to minimize and stabilize the effective loss of the corresponding arm. The DGD of each PS_MZ device 402, 404 is equal to the delay between the two arms, which is roughly equal to the path length of said piece of PMF fiber. For example, a 3-ns DGD may be obtained using a PMF length of about 60 cm for a typical fiber with a refractive index of 1.5, which is indeed much less than the 2 km needed in the embodiment of FIG. 3. The PSP is determined by the polarization axis of the PBSs. Multiple collimating lenses used along the optical paths are denoted c1 to c8.


The passive SOP generator 400 also comprises an optional phase reference arm 410. As in the embodiment of FIG. 3, a small proportion of the light at the output of the second PS_MZ device 404 is extracted by a coupler 408 toward polarizer 412 followed by photodetector 414 which output constitutes the phase reference signal Pφ(ν).


The coupling angle φŝo is set by rotating the assembly comprising the input polarizer 406 and collimator c0. The coupling angle φp is set by the angled fusion between the two pieces of PMF 420, 422 attached respectively to collimators c3 and c4. The PSP of each of these two pieces of PMF 420, 422 is aligned with the polarization axis of the PBS that faces the collimator to which they are attached.


The angle φâ between â and p1, or in words between the analyzer axis and the PSP of the second PS-MZ, is set by rotating polarizer 412.


The lengths of the PMF pieces 418 (denoted PMF0 and PMF1) mainly determine the DGDs of the two PS_MZs 402, 404: δτq˜nLq/c, wherein Lq is the PMF length and n is the mean group index of the fiber (˜1.5). For an accurate pre-calculation of the lengths before cutting the fibers, the small light path lengths in air, into the PBSs and into the collimator lenses should be taken into account. It is noted that the use of PMFs in the PS_MZ 402, 404 is optional although advantageous to properly align the polarization that comes out of the first PBS with the polarization axis of the second PBS, in order to minimize and stabilize the loss of this path (a misalignment has no other adverse effect).


Collimators c3 and c4 and their attached pieces of PMF are inserted for the purpose of simplifying the assembly. The two assemblies constituting the two PS_MZ 402, 404 may then be mounted on a single plane. Otherwise there should be two planes making an angle θ=φp/2 (see FIG. 5).



FIG. 5 shows a passive SOP generator 500 in accordance with another practical implementation of the 2-segs of FIG. 2, which is based on all-bulk PS_MZs.


The passive SOP generator 500 comprises two all-bulk PS_MZs 502, 504. The optical path is split in two arms and then recombined using two Polarization Beam Splitters (PBS) 516. For each PS_MZ 502, 504, a path length difference (delay) between the two arms is set by a free-space optical path that is made longer in one of the arms 518, using a mirror arrangement. The DGD of each PS_MZ 502, 504 is equal to the delay between the two arms. The passive SOP generator 500 also comprises an input polarizer 506, and again, as in the embodiment of FIG. 3, a small proportion of the light at the output of the second PS_MZ 504 is extracted by a beam splitter 508 toward a phase reference arm 510 which output is the phase reference signal Pφ(ν).


The coupling angle φŝo is set by rotating the input polarizer 506. The coupling angle φp is set by rotating the mounting plane of the second PS_MZ 504 (normal û1) with respect to the mounting plane of the first PS_MZ 502 (normal û0). The angle θ between normal û1 and normal û0 is θ=½φp (θ represents the physical angle between the normal vectors of two planes, whereas the coupling angles φ refer to angles between Stokes vectors such as ŝo, p0, p1, â; Stokes vectors are not limited to representing linear polarizations; when linear SOPs are involved, a corresponding physical angle do exists, but equal to half the angle between the Stokes vector).


It is noted that the two PS_MZs 502, 504 may alternatively be mounted on the same plane by inserting a retardation waveplate (λ/2) in-between. The coupling angle φp is then set by rotating the waveplate. Although apparently convenient, there is a drawback; waveplates that would be substantially achromatic (constant retardation as a function of ν) over a very large scanning range (e.g. from 1250 to 1650 nm) are rare and exceedingly expensive.


It will be understood that the above-described implementations of the passive SOP generator may be varied by mixing portions thereof. For example, the phase reference arm 310 or 410 may be replaced by the bulk phase reference arm 510 and vice versa. The bulk delay lines 518 of the all-bulk embodiment 500 may also be replaced by segments of PMF and associated collimator lenses c1-c2, c5-c6 as in the bulk and fiber embodiment 400.


Measurement System


FIG. 6 shows a PDL measurement system 600 in accordance with one embodiment which employs a single spectral scan and a passive SOP generator.


It should be noted that the RBS method, notably its new mathematics (see equations (1) to (10)) constituting a generalization of the mathematics of the classic Mueller Matrix Method, is more general than this embodiment, is not limited to this embodiment and does not require data to be acquired using a single wavelength scan. It rather allows dealing optimally with this much more challenging case. Data may as well be acquired using multiple scans sequentially in time with different oscillating SOPs, and whether the SOP generator is active or passive. The PDL measurement system 600 is still mostly advantageous compared to the classic Mueller Matrix Method in that the PDL can be derived from a single wavelength scan over which the input-SOP continuously varies instead of a few sequential scans with fixed SOPs.


The PDL measurement system 600 comprises a tunable light source 602, an SOP generator 604, a power transmission measurement apparatus 606, an acquisition device 608 (also referred to as data sampling), a processing unit 610 implementing the RSB data processing, a data output 612 for displaying, saving in memory or otherwise outputting the measurement results.


The tunable light source 602 generates a test lightwave and spectrally scans the test lightwave over the spectral range over which the PDL is to be measured.


The SOP generator 604 is used to vary the SOP of the test lightwave according to a continuous trajectory on the surface of the Poincaré sphere while the test lightwave is spectrally scanned, before the test lightwave is launched into the DUT. Although, as explained hereinabove, passive SOP generators offer many advantages, in some implementations it may still be chosen to use an active SOP generator.


The power transmission measurement apparatus 606 measures the power of the ligthwave at both the input and output of the DUT, i.e. Pin(ν) and Pout(ν) respectively, so that the transmission through the DUT, T(ν)=Pout(ν)/Pin(ν), can be computed.


The acquisition device 608 simultaneously samples values of Pin(ν), Pout(ν) and optionally Pφ(ν) during a spectral scan of the test lightwave.


The processing unit 610 computes the power transmission curve T(ν) from the sampled power Pin(ν) and Pout(ν) as T(ν)=Pout(ν)/Pin(ν), and processes said computed curve T(ν) and optional phase reference output Pφ(ν) in accordance with the RSB method to derive therefrom the spectrally-varying PDL of the DUT.


The data output 612 outputs the measured characteristics which, in addition to the spectrally-varying PDL may include the polarization-dependent center wavelength (PDCW) and the polarization-dependent bandwidth (PDBW) of the DUT.


Tunable Light Source

The tunable light source 602 generates a test lightwave and spectrally scans the test lightwave over the spectral range over which the PDL is measured. The tunable light source 602 may be based on a tunable narrowband single-mode laser or, alternatively, a broadband light source (e.g. a super-LED) followed by a tunable optical filter (also named “monochromator”).


It will be understood that the term “single-mode” intends to comprise a laser source which output spectrum substantially consists of a single peak, which peak always has some finite width in practice, the laser linewidth.


The term “narrowband” refers to said laser linewidth. In the case of the broadband source and optical filter embodiment, in refers to the width of the passband of the optical filter. The maximum acceptable linewidth depends on the intended application as it determines the minimal spectral resolution of the measurement. For example, the characterization of a broadband DUT, having a constant SOP-averaged transmission and constant PDL over a large range of ν, is less stringent on the linewidth than other DUTs including WDM optical filters having a 3-dB bandwidth as small as 25 to 30 GHz. To give some figures: for example, in the case of the passive SOP generator described hereinabove for which the DGDs of the 2 segments are δτ0˜3 ns and δτ1˜1 ns (case r=3), which are large enough for said DUTs to be measured with negligible crosstalk between the sidebands, the resulting extent of T(ν) in the delay domain requires that the laser linewidth be smaller than about 10 MHz (but up to 5 times this value may be acceptable if a compensation of the lowpass filtering effect of the finite laser linewidth is applied by the data processing based on the knowledge of the laser lineshape).


The term “tunable” intends to mean that the optical frequency ν of the laser or the center optical frequency ν of the tunable filter passband in the broadband source alternative, can be set to a plurality of values over the spectral range of the measurement. The spectral range may be determined by the application or be selected by the user depending on the spectral range over which the PDL of the DUT is to be characterized. In any case, the RSB method requires (to obtain a single point of PDL(ν)) that the spectral range at least covers about one cycle of the SOP trajectory (e.g. ˜1 GHz for δτ1=1 ns).


The spectral scan typically involves that, starting from a given initial optical frequency νo at t=0, the optical frequency ν is tuned to increase or decrease in a substantially continuous manner as a function of time t and with a substantially linear variation as a function oft, i.e. ν(t)=νo+vs t, where vs is the scanning speed. Such spectral scan may be obtained using what is widely referred to as a “swept-wavelength system”, which allows to sample data over a very large number of values of ν in a very swift way. It is however noted that such continuous scan is not necessarily required by RSB. Indeed, data could as well be acquired by varying the optical frequency ν in a random order, then re-ordered in ascending order afterward, as long as the power transmission T(ν) of the DUT is obtained as a function of ν. But in practice, it is obviously more convenient to proceed with a continuous scan.


That being said, in practice, sampling data (Pin(ν), Pout(ν), Pφ(ν)) with a constant step dν in optical frequency or re-sampling in the data processing (using interpolation) to obtain such constant step dν, allows computing the Discrete Fourier Transforms (DFT) using the Fast Fourier Transform algorithm (FFT), although a very much slower general DFT computation does not require the samples to be equally spaced.


Power Transmission Measurement

The power transmission measurement apparatus 606 splits the incoming ligthwave in two paths for the purpose of measuring the power of the ligthwave at both the input and output of the DUT, i.e. Pin(ν) and Pin(ν) respectively, so that the transmission through the DUT, T(ν)=Pout(ν)/Pin(ν), can be computed afterward. Power transmission measurement apparatuses commonly used in the classic Mueller Method, or in simpler insertion loss measurements, may be used for this purpose.



FIG. 7A shows a power transmission measurement apparatus 700A in accordance with an all-fiber embodiment. FIG. 7B shows a power transmission measurement apparatus 700B in accordance with a bulk embodiment. In either cases, the apparatus 700A/700B is configured to measure both the optical power Pin of the test lightwave at the input of the DUT and the optical power Pout of the test lightwave at the output of the DUT to allow calculation of the power transmission T=Pout/Pin.


The power transmission measurement apparatus 700A comprises an optical fiber power coupler 702, which may consist of a singlemode fiber coupler, to extract a portion of the input test lightwave toward a first photodetector 704 in order to measure values of input optical power Pin. A second photodetector 706 received the test signal lightwave at the output of the DUT to measure values of output optical power Pout.


The power transmission meaurement apparatus 700B is similar to apparatus 700A but for the optical fiber coupler 702 that is replaced by a bulk beam splitter 712 and a pair of collimating lenses 708, 710.


Acquisition Device

The power transmission curve T(ν) is obtained by sampling both Pout and Pin while scanning the optical frequency ν. Data Sampling can be applied using conventional data sampling. FIG. 8 shows an acquisition device 800 in accordance with one embodiment using such conventional sampling. Each signal Pφ, Pin, Pout to be sampled, is filtered using an Anti-Aliasing Filter (AAF) 802 and sampled using an Analog-to-Digital Converter (ADC) 804. Data samples are cumulated in memory 806.


Delay Scale (τ), signal-frequency scale (f) and sampling frequency (fe)

Referring to the FIG. 1D, the signals to be sampled correspond to a given extent of the DFT of the transmission curve T(ν) in the delay domain, which is independent of the scan speed vs. When performing a spectral scan with a constant scan speed, the extent of the sampled signal in the signal-frequency (f) domain (single-sided bandwidth) depends on both the extent of T(ν) in the delay domain and the scan speed vs (the mapping is f=vs τ). Defining the following:

    • τ+: total extent of T(ν) in the delay domain, containing all sidebands (e.g. 4 in the example where δτ0=3 ns and δτ1=1 ns);
    • τg: spacing between sidebands (e.g. 1 ns in the example where δτ0=3 ns and δτ1=1 ns);
    • f+: extent of the signal-frequency scale containing all sidebands;
    • fe: sampling frequency


Assuming for example δτ0=3 ns and δτ1=1 ns, then with r=3, the total extent in the delay domain, τ+, is about 5 times the spacing τg between sidebands (τ+˜5 τg). Since f+=vsτ+, then f+˜5 vsτg. To avoid aliasing artefacts, the sampling frequency should be fe≥2 f+. Therefore, if vs=10 THz/s for example, i.e. vs.λ˜80 nm/s@λ=1550 nm, then f+=50 kHz and fe should be greater than 100 kHz.


The following care should be taken regarding the anti-aliasing filter 802. It is a well-known and good practice to limit the bandwidth of the signal (including wideband noise and/or high-frequency spurious) before sampling. But here, care should be taken to make the filter response substantially flat over the range −f+ to f+ because any attenuation of the sidebands by the anti-aliasing filter 802 will result in a bias on the measured PDL. Indeed, the PDL is directly represented by the amplitude of the sidebands (relative to the mean DC part). So only 1% attenuation translates directly to roughly −1% bias. A good choice in that case may be a second-order filter which gives a flat response over a large range relative to its 3-dB cut frequency.


Care should also be taken to sample all signals (Pφ, Pin, Pout) synchronously (negligible delay between them). Of course, the requirement becomes more stringent as the scan speed increases.


Analytical Expression of Matrix

Before describing the RBS processing in detailed steps, it is worth exposing an analytical expression of matrix custom character that may be implemented in the algorithm in the case of embodiments using the passive SOP generator. Remind that matrix custom character represents the system of equations to be solved by the RSB processing in accordance with equation (8), and only depends on the known trajectory of the input-SOP as stipulated by (9). Yet more precisely, what is given below in equations (13) are analytical expressions of the complex vectorial amplitudes custom character that appear in the general expression (9) of matrix custom character.


The trajectory ŝ(ν) of the SOP at the output of the 2-segs of the passive SOP generator, which is characterized by said complex vectorial amplitudes custom character as specified in (2), only depends on two parameters, the two coupling angles φŝo and φp, or equivalently the three governing Stokes vectors ŝo, p0 and p1. Remind that ŝo is the SOP at the input of the 2-segs, whereas p0 and p1 are the PSPs of first and second segment respectively. Said analytical expressions of the custom character given below derive from the laws of Physics. The specific set of equations (13) assumes that the following conditions are met:

  • 1) r≥2.5, where r is the ratio of the DGDs of the first and second segment of the 2-segs (r=δτ0/δτ1). Then the number of sidebands is J=4. Remind that the above embodiment targets r=3;
  • 2) The three governing Stokes vectors, ŝo, p0 and p1, are coplanar; this condition is easily met in practice with virtual perfection if the corresponding SOP and PSPs are each simply set by the polarization axis of a linear polarizer or PBS, or the PSP of a PMF fiber, as in all embodiments of the passive SOP generator shown previously in FIGS. 3, 4 and 5. For the setup of FIG. 4 however, care must be taken to ensure that the PSPs of the two accessory pieces of PMF 420, 422 inserted between the two PS_MZ for production convenience (both PS_MZ on a single plate) are as specified in hereinabove, i.e. properly aligned with the polarization axes of the corresponding PBSs (say within 1° for the above statement of virtual perfection to still hold).


Notes



  • a) If 1≤r<2.5, the number of distinctly-resolved sidebands and their order/disposition in the delay domain τ differ from what is shown in FIG. 1. However, it does not mean that ratios such as r=1 or r=2 cannot be targeted if one finds some advantage to it. The embodiments herein described in details target r=3, but other cases such as r=2 or r=1 are possible, but more complicated. It is considered within the reach of one skilled in the art to derive analytical expressions of the custom character, and thus matrix custom character, that are applicable in these cases if necessary. The custom character can also be measured directly, for example by performing a spectral scan and measuring the varying SOP ŝ(ν) at the output of the 2-segs with a polarimeter, and performing a DFT of the measured SOP.

  • b) The three governing Stokes vectors, ŝo, p0 and p1, are defined within a common rotation applied to the three. This means that the definition of custom character and the measured PDL are the same if a global rotation is applied to the true Stokes vectors, only the angles between them, the coupling angles, matter. That is the reason why we can express the custom character as a function of only the two known coupling angles φš0 and φp. Accordingly, we arbitrarily set p11=(0 0 0)T in the expressions below. The physical p1 does not need to be equal to û1. It is simply equivalent to a global rotation of the three vectors such that p1û1 after the rotation is applied.




FIG. 9 illustrates the relationship between the governing Stokes vectors, ŝo, p0 and p1, and the coupling angles φš0 and φp, wherein (û1 û2 û3) are the cartesian base vectors of the standard 3D Stokes vectors space.


Setting p11 as mentioned above, ŝo, po and p1, are computed as,

p1=Stokes(0,0)
p0=Stokes(−φp,0)
ŝo=Stokes[(φšo−φp),0]  (12)

where Stokes(φ, ψ) represents the specification of a 3-element unit Stokes vector in terms of spherical coordinates. Of note is that ψ=0 (“equator” of the Poincaré sphere) is the circle encompassing all possible linear SOPs.


Then the announced analytical expressions of the complex vectorial amplitudes custom character, with custom character=−4.4 since here J=4 for the specified r≥2.5, reads as,

z0=(cos φšo cos φp)p1
z1=½[cos φšo(p0−cos φpp1)]+i½[cos φšo(p1×p0)]
z2=¼[ŝo−cos φ′šp1+cos φšo(cos φpp1−p0+vxx]+i¼[(p0−p1ŝo+cos φšo(p1×p0)]
z3=½(cos φ′š−cos φp)p1
z2=¼[ŝo−cos φ′šp1+cos φšo(cos φpp1−p0+vxx]+i¼[(p0−p1ŝo+cos φšo(p1×p0)]
custom character=custom character  (13)

where we defined cos φ′š=(p1·ŝo) and vxx=p1×(p0×ŝo). Whenever the above specified conditions 1) and 2) are met, the values (13) of the custom character can be put in (9) to construct matrix custom character and then compute the response vector of the DUT according to (8) and (10).


Processing of the Resolved-Sidebands (RSB) Method

After some usual data conditioning of the sampled data, e.g. cropping ends having transients, resampling and/or decimation if necessary or useful, etc., the inputs to the processing of RSB are:


(1) a vector T of N values Tn=T(νn) of the power transmission curve, n=0 . . . (N−1), at optical frequencies νn with a substantially constant step dν between successive samples. T is computed beforehand from the sampled outputs Pinn) and Poutn) as Tn=Pout.n/Pin.n.


(2) a vector Pφ of N values Pφ.n=Pφn) of the phase reference output Pφ(ν) at the same optical frequencies vνn as for T.


The processing of these inputs by RSB is described in detail below in the form of comments, definitions and other pertinent information put around the actual procedures applied to real data.



FIG. 10 shows the main steps of the RBS processing, comprising an optional step 1002 of retrieving the phases custom charactern) of the sideband-carriers with custom character=1 . . . J, n=0 . . . (N−1), using said vector Pφ as input; a step 1004 of extracting individually each sideband from the DFT of said vector T to compute complex transmissions T custom charactern), custom character=−J . . . J, n=0 . . . (N−1); a step 1006 of computing the response vector of the DUT at each νn, |m(νn)custom character, from the complex transmissions custom charactern) and said matrix custom character according to (8) and (10); and a step 1008 of computing PDL(νn), the PDL curve of the DUT, from the response vector |m(νn)custom character, n=0 . . . (N−1).


Step 1002

In step 1002, a processing unit receives the sampled phase reference output Pφn) or, more specifically, a vector Pφ of N values Pφ.n=Pφ)(νn) of the phase reference output at optical frequencies νn, n=0 . . . (N−1). From these values, the procedure of step 1002 derives the phases custom charactern) of the sideband-carriers with custom character=1 to J, n=0 . . . (N−1), which are used in step 1004 (Procedure_2).



FIG. 11 expresses the procedure (Procedure_3) of step 1002 in more details.


Step 1004

In step 1004, the processing unit receives the sampled power transmission curve T(νn) or, more specifically, vector T of N values Tn=T(νn) of the power transmission curve at optical frequencies νn, n=0 . . . (N−1). From these values, the procedure of step 1004 extracts individually each sideband from the DFT of said vector T to compute complex transmissions custom charactern) with custom character=−J . . . J, n=0 . . . (N−1).


The procedure involves:

  • a) calculating a Discrete Fourier Transform (DFT) of the power transmission curve T(ν), wherein the DFT shows at least two sidebands; and
  • b) extracting individually each sideband from the DFT of transmission curve T(ν) to compute complex transmissions custom character(ν).


More specifically, at the end of Procedure_2 the processing unit reshuffles the resulting custom charactern) data into a single vector |custom character of N values |custom charactern at sampled points n, wherein each |custom charactern is itself a vector whose elements are the 9 values custom charactern) where −4≤custom character≤4. This reshuffling is not fundamentaly required, but it eases the next calculations in Procedure_1 by making them faster than otherwise, as well as making this procedure very compact from a programmatic point of view.



FIG. 12 expresses the procedure (Procedure_2) of step 1004 in more details.


Step 1006

In step 1006, the processing unit receives the complex transmissions custom character(ν) derived in step 1004, or more specifically vector |custom character, and computes the response vector |m(ν)custom character of the DUT.


More specifically, the processing unit computes |m(νn)) from vector |custom charactern obtained in step 1004 and known matrix custom character defined hereinabove, for each sampled point n=0 . . . (N−1)).



FIG. 13 expresses the procedure (Procedure_1) of step 1006 in more details.


Step 1008

In step 1008, the processing unit computes the spectrally-varying PDL of the DUT from the response vector |m(ν)custom character.



FIG. 14 expresses the procedure (Procedure_0) of step 1008 in more details.


The input to the procedure is a vector |mcustom character of N values |mcustom charactern=|m(νn)custom character of the response vector, n=0 . . . (N−1), wherein each |mcustom charactern is itself a 4-component vector. The output may comprise a vector PDL or PDLdB of N values PDLn or PDLdB.n of the PDL of the DUT at optical frequencies νn, where n=0 . . . (N−1). Of note is that, in accordance with the most widespread definition and tradition, the PDL value returned by Procedure_0 is expressed in dB.


The embodiments described above are intended to be exemplary only. The scope of the invention is therefore intended to be limited solely by the appended claims.


Glossary of Main Parameters and Variables



  • ν: optical frequency

  • τ: delay (Fourier-transform conjugate variable of ν)


  • custom character: number of wavelength scans with fixed input-SOPs in the classic Mueller Matrix Method

  • J: number of sidebands on one side of the Fourier transform in the resolved-sidebands method (RSB)

  • T(ν): power transmission curve as a function of ν


  • custom character(ν): complex transmission associated to sideband-carrier custom character as a function of ν, wherein custom character=−J . . . J


  • custom character: carrier-delay of sideband-carrier custom character (similar to carrier frequency but here, the carrier is characterized by its delay instead of its frequency)

  • |m(ν)custom character: response vector of the DUT as a function of ν


  • custom character: equivalent matrix of RSB representing the system of equations to be solved


  • custom character: complex vectorial amplitude of the sideband-carrier custom character

  • Pφ(ν): phase-reference output

  • Pin(ν): optical power of the test lightwave at the input of the DUT

  • Pout(ν): optical power of the test lightwave at the output of the DUT

  • δτ0, δτ1: differential group delay (DGD) of first and second segment of the 2-segment passive SOP generator, respectively

  • p0, p1: principal state of polarization (PSP) of first and second segment of the 2-segment passive SOP generator, respectively

  • r: ratio of the DGDs (r=δτ0/δτ1) of first and second segments of the 2-segment passive SOP generator

  • φšo, φp: coupling angles characterizing the 2-segment passive SOP generator

  • ŝ: 3-component unit Stokes vector representing the SOP of the 100% polarized input test lightwave.

  • τ+: total extent in the delay domain containing all sidebands (e.g. 4 in the exemple where J=4);

  • τg: spacing between sidebands (e.g. 1 ns in the exemple where δτ0=3 ns and δτ1=1 ns);

  • f+: extent of the signal-frequency scale containing all sidebands (˜½ signal bandwidth or single-sided bandwidth);

  • fe: sampling frequency


Claims
  • 1. A method for measuring the Polarization Dependent Loss (PDL) of a Device Under Test (DUT) within a spectral range, comprising: generating an input test lightwave which State Of Polarization (SOP), as represented by a point on the surface of the Poincaré sphere, varies according to a continuous trajectory while being spectrally scanned over the spectral range and launching the input test lightwave into the DUT;measuring the power transmission through the DUT, curve T(ν), during a spectral scan of the input test lightwave;calculating a Discrete Fourier Transform (DFT) of the power transmission curve T(ν), wherein the DFT shows at least two sidebands;extracting said sidebands and computing an inverse DFT of each said sidebands to obtain complex transmissions (ν);computing the response vector |m(ν) of the DUT from said complex transmissions (ν) and a matrix determined by the continuous trajectory of the SOP of the input test lightwave by solving the linear system of equations defined by |(ν)=|m(ν) wherein |(ν) is constructed from said complex transmissions (ν);deriving the spectrally-varying PDL of the DUT from said response vector |m(ν).
  • 2. The method as claimed in claim 1, wherein the continuous trajectory of the SOP of the input test lightwave over the spectral range is known or predetermined and the matrix derived therefrom.
  • 3. The method as claimed in claim 1, wherein the phase of the SOP of the input lightwave along the continuous trajectory is determined at each sampled point.
  • 4. The method as claimed in claim 3, wherein said phase is determined from a phase reference signal measured at the output of a polarizer receiving a portion of the input test lightwave.
  • 5. The method as claimed in claim 1, wherein measuring the power transmission curve T(ν) comprises sampling data asynchronously with respect to the SOP variation of the input test lightwave.
  • 6. The method as claimed in claim 1, wherein the SOP of the input test lightwave is varied using a passive SOP generator.
  • 7. The method as claimed in claim 6, wherein the passive SOP generator comprises a first and a second polarization-splitting Mach-Zehnder interferometer.
  • 8. The method as claimed in claim 7, wherein the ratio r of the differential group delay (DGD) of the first segment over the differential group delay (DGD) of the second segment is substantially equal to an integer value.
  • 9. The method as claimed in claim 8, wherein the ratio r is substantially equal to 3.
  • 10. The method as claimed in claim 1, further comprising deriving at least one of the polarization-dependent center wavelength (PDCW) and the polarization-dependent bandwidth (PDBW) of the DUT.
  • 11. A system for measuring the Polarization Dependent Loss (PDL) of a Device Under Test (DUT) within a spectral range, comprising: a tunable light source to generate a test lightwave and spectrally scanning the test lightwave over the spectral range;a State Of Polarization (SOP) generator connected to the tunable light source and configured for varying the SOP of the test lightwave according to a continuous trajectory on the surface of the Poincaré Sphere while the test lightwave is spectrally scanned over the spectral range, the test lightwave to be launched into the DUT;a power transmission measurement apparatus connected to the SOP generator and comprising a first photodetector and a second photodetector for measuring power values of the test lightwave at the input and the output of the DUT, respectively;an acquisition device connected to the power transmission measurement apparatus and comprising an analog-to-digital converter sampling the measured power values during a spectral scan of the test lightwave to produce a measured power transmission curve T(ν);a processing unit connected to the acquisition device to receive said measured power transmission curve T(ν) and configured for: calculating a Discrete Fourier Transform (DFT) of said power transmission curve T(ν), wherein the DFT shows at least two sidebands;extracting said sidebands and computing an inverse DFT of each said sidebands to obtain complex transmissions (νv);computing the response vector |m(ν) of the DUT from said complex transmissions (ν) and a matrix determined by the continuous trajectory of the SOP of the input test lightwave by solving the linear system of equations defined by |(ν)=|m(ν) wherein |(ν) is constructed from said complex transmission (ν); andderiving the spectrally-varying PDL of the DUT from said response vector |m(ν).
  • 12. The system as claimed in claim 11, wherein the continuous trajectory of the SOP of the input test lightwave over the spectral range is known or predetermined and the matrix derived therefrom.
  • 13. The system as claimed in claim 11, further comprising a phase reference arm to determine a phase of the SOP of the input lightwave along the continuous trajectory at each sampled point, said phase reference arm comprising a polarizer followed by a photodetector.
  • 14. The system as claimed in claim 11, wherein values of the power transmission through the DUT are sampled by the acquisition device asynchronously with respect to the SOP variation of the input test lightwave.
  • 15. The system as claimed in claim 11, wherein the SOP generator is a passive SOP generator.
  • 16. The system as claimed in claim 15, wherein the passive SOP generator comprises a first and a second polarization-splitting Mach-Zehnder interferometer.
  • 17. The system as claimed in claim 16, wherein the ratio r of the differential group delay (DGD) of the first segment over the differential group delay (DGD) of the second segment is substantially equal to an integer value.
  • 18. The system as claimed in claim 17, wherein the ratio r is substantially equal to 3.
  • 19. The system as claimed in claim 11, wherein processing unit is configured for further deriving at least one of the polarization-dependent center wavelength (PDCW) and the polarization-dependent bandwidth (PDBW) of the DUT.
US Referenced Citations (10)
Number Name Date Kind
5227623 Heffner Jul 1993 A
5298972 Heffner Mar 1994 A
5371597 Favin et al. Dec 1994 A
6650406 Allard et al. Nov 2003 B1
6856386 Anderson et al. Feb 2005 B2
6888624 Caplette et al. May 2005 B2
6888625 Anderson May 2005 B2
20040070759 Andeson Apr 2004 A1
20040184029 Anderson Sep 2004 A1
20090141274 Szafraniee et al. Jun 2009 A1
Foreign Referenced Citations (1)
Number Date Country
102539119 Jul 2012 CN
Non-Patent Literature Citations (3)
Entry
Agilent Technologies, “Polarization Dependent Loss Measurement of Passive Optical Components,” Application Note, Mar. 14, 2002, pp. 1-12.
Keysight Technologies, “PDL Measurements using Polarization Controller,” Application Note, Dec. 1, 2017, pp. 1-16.
Telecommunications Industry Association, “Measurement of Polarization Dependence of Insertion Loss of Single-Mode Fiberoptic Components,” by a Mueller Matrix Method, Tia Document, FOTP-198, TIA-455-198, Dec. 2002, pp. 1-24.
Related Publications (1)
Number Date Country
20210356359 A1 Nov 2021 US
Provisional Applications (1)
Number Date Country
63024741 May 2020 US