4 views

Uploaded by api-19786391

- IRJET- Parkinson’s Disease : A Case Study
- ME6504-Metrology and Measurements
- Speech Enhancement Using Spectral Subtraction Based on SFM1
- ML090160226.pdf
- 5-HDIL
- 14_02-0068
- Key Paper __5
- 775-782
- GH_FFT
- Optical Swetching ppt
- Pasco Microwave Optics WA-9314B
- ICRA96
- Peak Value
- Micro Ondas
- icecs06
- Sound Sence
- Comparators (1)
- anote093-ang_LOSS
- anote099-ang
- anote094-ang_LASER

You are on page 1of 5

INTERFEROMETRIC METHOD

Normand Cyr, Ph.D., Senior Optical Researcher

The interferometric method has always been the only accepted and proper method to test PMD in the field. Historically, all PMD instruments

based on the interferometric method have been driven by the TIA/IEC standard document FOTP-124, entitled Polarization-Mode Dispersion

Measurement for Single-Mode Optical Fibers by Interferometry. Although this standardized technique remains the correct way to test PMD,

it is important to remember that it also has its limitations. Other methods do exist, but none are reliable or adequate for testing PMD in the

field, where conditions can vary quickly due to vibrations, patchcord movement, gusts of wind, etc. But what if the standard method just got

a lot better and a lot more powerful?

This application note introduces an exact analysis, based on the fundamentals of the standard method, but using a slightly modified setup,

largely extending the application domain of the interferometric approach.

The interferometric method presented in the FOTP-124 describes the actual test setup and defines a possible mathematical method to extract

the PMD information from the raw data. Since this approach became generalized, as time went by, the industry no longer remembered that

this mathematical method was just that: one of many possible methods. Therefore, it has been used and accepted somewhat blindly.

The method described in the standard and, first and foremost, the mathematical analysis behind it, relies on a number of assumptions that

render the standard method fully applicable only in a limited number of cases. Among these underlying assumptions are the following:

Ideal random-coupling DUT with an infinite coupling ratio

A perfectly Gaussian interferogram

A smooth, ripple-free, gaussian-like spectrum at the receiver end

PMD x spectral width much larger than 1

When these conditions are met, or nearly met, the standard method gives accurate results. But what happens when either one of the

assumptions is not respected? Of course, the instrument still displays a result, but, in this case, how reliable is it really? Unfortunately, industry

professionals have forgotten about most of these underlying assumptions and have, therefore, fully trusted the displayed results. And since

no other mathematical analysis that could bypass these assumptions was developed, who can really blame them?

However, due to these stringent assumptions, the standard interferometric PMD-measurement method, as described in FOTP-124, has severe

limitations which stem from the basic theory that supports the method and its embedded analysis. This severely restrains the measurement

conditions and the application domain in which the standard analysis remains fully valid.

The first, and most important, assumption on which the standard analysis relies is the presumed Gaussian-shaped interferogram (which, in

itself, is based on another assumption; that is, an ideal random-coupling DUT). But actually, most interferograms are not Gaussian, even when

the DUT is a long fiber link. They often tend to be somewhat flattened; sometimes they look like a rectangular window, or even have a concave

shape with a depression at the center, in which case the Gaussian assumption would be totally false. This is clearly stated in paragraph 6.2

of the FOTP-124:

“6.2 Accuracy: Accuracy is related to the capability to precisely fit the interferogram with the Gaussian function”…

www.exfo.com

Telecommunications Test and Measurement

Application Note 096

Consequently, a non-Gaussian interferogram produces false results. Moreover, if the assumptions are not strictly met, it is not specified by

how much the result may be off target; for instance, in cases of mixed coupling, or when interferograms deviate significantly from the assumed

Gaussian shape, which is a frequent occurrence.

A second important limit comes from the assumption that the spectrum at the receiver end is perfectly smooth, ripple-free and Gaussian-like.

The interferometer in the instrument creates interference between the two principal states of polarization (PSPs). Around a position of zero

delay (center of the interferogram), the auto-correlation of the source peak corresponds to the spectrum, which is basically the Fourier

transform of the spectrum itself. The resulting interferogram is the addition of both this auto-correlation (containing no PMD information) and

the cross-correlation.

As a first, obvious consequence, it was thus important that this auto-correlation have minimal impact on the total interferogram, to hide or

falsify results as little as possible. A broad Gaussian source creates a single narrow auto-correlation peak, easily removable from the total

interferogram. A spectrum with abrupt changes or ripples creates a complex interferogram, which is added to the cross-correlation and adds

unknown error to the analysis (using the standard method). PMD analyzer vendors were aware of this and did offer smooth, polarized

broadband sources, but testing through devices, which altered the spectral shape, was not recommended.

However, even by optically removing the auto-correlation peak, the problem was not entirely solved. Although less obvious, the most basic

issue regarding spectrum smoothness is that the width of the cross-correlation itself depends on the width of the spectrum's auto-correlation

(Fourier transform). And this type of dependence is not known or specified within the mathematical framework of the standard method.

Therefore, testing through optical add/drop multiplexers (OADMs) or erbium-doped fiber amplifiers (EDFAs) was impossible, even if a specially

configured interferometer was used to remove the auto-correlation peak.

In the new proposed method, all severe limitations are eliminated, thanks to a theoretical framework that simply removes the stringent

assumptions from the basic theory. Now, the formula that links PMD to the interferogram is exact in all cases, with any interferogram shape

and any type of coupling, and without any of the previous stringent assumptions. As an added bonus consequence, the exact analysis allows

us to specify by how much the standard analysis is off target when interferograms are not Gaussian. Nonetheless, this new approach still fully

respects the FOTP-124 recommendation. Remember that FOTP-124 states that what is described therein is "a" method for deducing PMD

from the observed interferograms, not "the" only imaginable method. In cases of a tested fiber that respects the conditions stated previously

in the FOTP-124, both mathematical approaches will render the same results.

For the first time, an interferometric PMD analyzer will offer truly reliable PMD measurements (not assumed ones) that compare exactly, and

that are rigorously traceable to concrete wavelength-averaged DGDs (rms value), measured with a reference test method (RTM) such as the

Poincaré Sphere Analysis (PSA) or the Jones Matrix Eigenanalysis (JME). This is valid for any coupling ratio, as well as any interferogram and

spectrum shape. Moreover, measurement of PMD = 0 is also offered; and as a last, very powerful benefit, it allows for testing of entire links

that include EDFAs and OADMs.

While the purpose of this application note does not extend to explaining the mathematics behind the machine, the sections below will provide

you with a general idea of how all the above is possible.

www.exfo.com

Telecommunications Test and Measurement

Application Note 096

Setup

Our new patent-pending setup for measuring PMD is a slightly modified version of the setup defined in the FOTP-124 (basic interferometer).

Starting from the standard setup, a polarization beam splitter is added at the interferometer output. Two interferograms along the two

orthogonal polarizations are sampled by separate detectors.

To detectors

By adding the signal from the two detectors, the auto-correlation interferogram remains, while the cross-correlation interferogram cancels out.

Inversely, by subtracting the signals, the auto-correlation is canceled out while the cross-correlation interferogram remains. The point is to get

both interferograms separately, without one interfering with the other (no matter what the spectral shape of the signal entering the

interferometer). The new exact analysis that removes all previously mentioned limitations needs the two interferograms, separately.

Exact Analysis

A new mathematical approach has been developed, which is not based on the assumption of a Gaussian-shaped cross-correlation.

This approach also allows for the elimination of the offset generated by ill-shaped, unsmooth or excessively narrow spectra, as is the case

when measuring entire links with amplifiers for example. How so? First, a mathematical equation is applied separately on both the squared

cross-correlation envelope and the squared auto-correlation envelope, which, thanks to the modified setup, has been measured separately

without one interfering with the other. The rms width of both s-envelopes (s- stands for "squared") are obtained. But, according to the new

formula given below, the squared rms width of the auto-correlation s-envelope acts as an offset, which is subtracted from the result, completely

eliminating the effect of ill-shaped spectra. Thus, with this new modified setup and new mathematical approach, any source shape can be

used, including sources that were altered by EDFAs, allowing you to test the whole network at once, instead of section per section, which, in

turn, saves considerable amounts of time and money.

Non-Gaussian Interferograms

Non-Gaussian interferograms are very common (see image below) and typically produce flat-top shapes, far from a strictly Gaussian shape.

In some case, for example, the true PMD (rms–DGD) is 10 ps. The standard analysis gives 7.5 ps in such a bad scenario; i.e., 25 % off target

3

PMD = 2 (σ 2

x σ 20 )

www.exfo.com

Telecommunications Test and Measurement

Application Note 096

(but still, there can be worse cases). This can mean the difference between making the right decision or the wrong one; for example, a network

can be determined suitable for 10 Gb/s transmission when in fact it is not. An error like this can have a huge impact on performance and,

most importantly, can be very costly. But, from now on, such errors can be avoided.

As mentioned previously, we can now precisely determine by how much the standard analysis is off target. This is expressed as a Gaussian

Compliance Factor and is simply defined as the standard analysis result divided by the exact analysis result.

As an illustration of how this works with EDFAs, consider the following real example of a network tested using this method. The network was

looped and accounted for a total of 844 km. It contained 11 EDFAs altogether. The spectrum at the output of the entire looped link was as

follows:

Because PMD varies statistically as a function of the input state of polarization (SOP), results mentioned hereafter come from averaging 30

measurements at different SOPs, using a polarization scrambler.

www.exfo.com

Telecommunications Test and Measurement

Application Note 096

All the sections of the link were measured separately, and the sum of the sections (square root of the sum of the square of each section) gave

a PMD value of 9.06 ps. The end-to-end measurement (each section plus contribution of each EDFA) resulted in a PMD of 9.08 ps. To truly

validate the quality of the result, known PMD emulators were added at the end of the link (5 ps or 10 ps).

The fact that the obtained result was always within 1 % of the expected result (when doing the quadratic sum) proves that testing accurately

through EDFAs is now possible.

Corporate Headquarters > 400 Godin Avenue, Vanier (Quebec) G1M 2K2 CANADA Tel.: 1 418 683-0211 Fax: 1 418 683-2170 info@exfo.com

Toll-free: 1 800 663-3936 (USA and Canada) www.exfo.com

EXFO America 4275 Kellway Circle, Suite 122 Addison, TX 75001 USA Tel.: 1 800 663-3936 Fax: 1 972 836-0164

EXFO Europe Le Dynasteur, 10/12 rue Andras Beck 92366 Meudon la Forêt Cedex FRANCE Tel.: +33.1.40.83.85.85 Fax: +33.1.40.83.04.42

EXFO Asia-Pacific 151 Chin Swee Road, #03-29 Manhattan House SINGAPORE 169876 Tel.: +65 6333 8241 Fax: +65 6333 8242

EXFO China Beijing New Century Hotel Office Tower Room 1754-1755 Beijing 100044 P. R. CHINA Tel.: +86 (10) 6849 2738 Fax: +86 (10) 6849 2662

No. 6 Southem Capital Gym Road

Appnote096.2AN © 2005 EXFO Electro-Optical Engineering Inc. All rights reserved. Printed in Canada 05/05

- IRJET- Parkinson’s Disease : A Case StudyUploaded byIRJET Journal
- ME6504-Metrology and MeasurementsUploaded byVijay Shankar
- Speech Enhancement Using Spectral Subtraction Based on SFM1Uploaded byvinay
- ML090160226.pdfUploaded byAmin Kotb
- 5-HDILUploaded byAwad Almalki
- 14_02-0068Uploaded byIgor Gjorgjiev
- Key Paper __5Uploaded byDharsan Sathya
- 775-782Uploaded byCalvin Szeto
- GH_FFTUploaded byjbusowicz
- Optical Swetching pptUploaded byShivenduTomar
- Pasco Microwave Optics WA-9314BUploaded byJoaquin Casanova
- ICRA96Uploaded bybhagyashreemali
- Peak ValueUploaded bygeorgie_borgie
- Micro OndasUploaded byVictor Carrera
- icecs06Uploaded byanis_ben_aicha5483
- Sound SenceUploaded byMatthew Tucker
- Comparators (1)Uploaded byflirtymint

- anote093-ang_LOSSUploaded byapi-19786391
- anote099-angUploaded byapi-19786391
- anote094-ang_LASERUploaded byapi-19786391
- CS184_V04[1]Uploaded byapi-19786391
- anote098-ang_OSNRUploaded byapi-19786391
- Victoria+Sdh+Sonet+Pdh+DsUploaded byshahramkarimi76
- anote073-ang_OTDRUploaded byapi-19786391
- anote087-ang_PMDUploaded byapi-19786391
- anote092-ang_FIBERUploaded byapi-19786391
- anote090-ang_LOSSUploaded byapi-19786391
- anote084-ang_FIBERUploaded byapi-19786391

- Ahmed Khan CV.docxUploaded bysuhaib asim
- !0 Japanese What-s to Learn BEGIN HEREUploaded byarykamori
- Greasing PlanUploaded byJerry Higdon
- Phonetics & PhonologyUploaded byRogerio Azeredo
- CV - EnglishUploaded byperholm
- MSL DrivematrixUploaded byBéla Kovács
- Acceptance of Criteria of Concrete 100Uploaded byLucky Chopra
- HRMUploaded bySuman Poudel
- Hotels4u PillarPost NewYorkUploaded byJo_HC_BPD
- Generalized-Dynamic-VSC.pdfUploaded byAhmed Zidan
- Chapter 3 Philosophies & Framework Uploaded PptUploaded bybarona_paul
- AAS) Atomic Absorption Spectroscopy-2nd Edition..Uploaded bySarah Permata Sari
- SKF Instruments for Condition MonitoringUploaded byLiliTiraxtur
- Bhendi Bazaar Cluster RedevelopmentUploaded byAjay Walia
- Vintage Airplane - Apr 2009Uploaded byAviation/Space History Library
- ACM-3311_QIGUploaded byAlexander Pl
- Livret 6 AnglaisUploaded byJenniDelgado
- DasUploaded byHari Baskar
- Phoolan Devi Murder Case JudgementUploaded bySampath Bulusu
- Iso HPLC SystemUploaded byAnil Chawla
- Flavonoids From Buddleja ParvifloraUploaded byalinuml
- angryminis_minirulebook_v2Uploaded byIndaBor
- mc-ty-scalarprod-2009-1.pdfUploaded byMohamed Yahia
- Excel CheatsheetUploaded bypeeyushthakur
- Canada Fsw-june 2014 Coe PDFUploaded byPrateek Dubey
- NZ Property Investor December 2017Uploaded bymaze
- UK China Collaborative Literature Review on Toxic Metals in Industrial Symbiosis in the Cement Industry (DRAFT)Uploaded byShashwat Omar
- Batalla de Dau Tieng 1969.pdfUploaded byClarin.com
- 892103 the 2001 EuroVan With the 2.8-Liter 24-Valve VR6 Engine and Electronic Stabilization ProgramUploaded byeduardorojas007
- Strategic_Management.pdfUploaded byjeevarvinder