Measuring Insertion Loss and Return Loss of a Fiber Component

Scope

This article gives the procedures for measuring the insertion loss (IL) and return loss (RL) of a two-port fiber component: a connector pair, a patch cord, a splice, a filter, an isolator, a coupler arm, or a fiber-pigtailed chip. It covers the reference methods and what each includes in the result, launch conditions for multimode fiber, the continuous-wave reflectometer method for RL, and the error sources that matter when the quantities are a tenth of a decibel of loss or fifty decibels of return loss. Loss measured along a link by backscatter is covered in the OTDR entry, the polarization dependence of loss in Measuring PER and PDL, and the separation of coupler loss from waveguide loss on a chip in Coupler Loss De-Embedding.

Definitions

Insertion loss is the reduction in transmitted power caused by inserting the component, and return loss is the ratio of incident to reflected power at a port, both in decibels:

$$\text{IL} \;=\; 10\log_{10}\!\left(\frac{P_\text{ref}}{P_\text{DUT}}\right),$$ $$\text{RL} \;=\; 10\log_{10}\!\left(\frac{P_\text{in}}{P_\text{refl}}\right).$$

Both are positive numbers for a passive component, and a larger RL means less reflection. A reported IL is meaningful only together with the reference method, because the method decides which connectors are counted.

Equipment

Insertion loss

Reference methods

The source is connected to a launch cord and the power meter to a receive cord. What is measured as the reference, before the component is inserted, decides which connections the result includes:

The one-cord method gives the largest number and the three-cord method the smallest, for the same component. When results from two labs disagree by about the loss of one or two connections, the usual cause is that they used different methods.

Procedure

1. Warm up the source and meter for the time their makers specify, and zero the meter with its input capped.
2. Inspect and clean the cord end faces, then take the reference reading $P_\text{ref}$ by the chosen method.
3. Insert the component without disturbing the launch cord, since moving it changes the power delivered and, in multimode fiber, the mode distribution.
4. Read $P_\text{DUT}$ and compute IL.
5. Disconnect and reconnect the component three or more times and record the spread. The repeatability of the connections is usually the largest single term in the uncertainty of a connectorized IL measurement.
6. Re-check the reference at the end. A drift in the reference larger than the repeatability means the source or the setup moved during the measurement, and the run should be repeated.
7. Repeat at each wavelength of interest; IL of couplers, filters and bent fiber changes with wavelength (see macrobend loss).

Launch conditions in multimode fiber

In multimode fiber the loss of a connector depends on how the light is distributed among the fiber's modes. An overfilled launch, with power in the high-order modes near the core edge, loses more at a connector with a small offset than a launch concentrated near the core center. The same connector can therefore give different results with different sources, and the difference is not an error in either measurement. Standardized multimode measurements specify the launch by its encircled flux, the fraction of power within given radii of the core center, and achieve it with a mode conditioner or a qualified launch cord; a mode scrambler alone does not guarantee compliance. Single-mode fiber has one guided mode and no equivalent issue, provided the launch cord is long enough, or looped, to strip light guided in the cladding.

Return loss

Continuous-wave reflectometer method

Light from the source passes through a coupler or circulator to the component; light reflected back from the component returns through the coupler to a second detector. The reading must be calibrated for the coupler's ratios and corrected for the setup's own reflections.

1. Calibrate the reflection path. Terminate the test port with a reflector of known RL. A flat, clean, unconnected PC end face in air reflects 3.4% of the light, a return loss of 14.7 dB (from the Fresnel formula with $n$ = 1.45); many reflectometers use this or an internal reference as their calibration point.
2. Measure the background. Terminate the test port so it returns nothing: wrap the fiber several turns around a small mandrel just past the connector, so the light is lost from the core before it reaches the far end, or immerse the end in index-matching gel. The reading $P_\text{bg}$ is the setup's own reflection and the coupler's directivity leakage.
3. Measure the component. Connect the component and terminate its output the same way, so that only reflections from the component and its input connection are measured.
4. Subtract the background in linear units, not in dB:

$$P_\text{refl} \;=\; P_\text{meas} - P_\text{bg}.$$

A component with a true RL of 50 dB measured on a setup whose background corresponds to 55 dB reads 48.8 dB before subtraction, which is 1.2 dB pessimistic. When the background is within a few dB of the component's reflection, the result becomes a lower limit on RL rather than a measurement.

OTDR method

An OTDR separates reflections by distance and gives the reflectance of each event along a link, which is useful when a component sits among several connectors. Its accuracy for reflectance depends on the fiber's backscatter coefficient and on the instrument's dead zone, and the continuous-wave method is preferred for a single component on the bench.

What limits return loss

For the polish grades, a mated UPC pair is expected to exceed 50 dB and an APC pair 60 dB (see PC, UPC and APC connectors). A component that measures much worse at its input usually has a contaminated or damaged connector or an air gap, both of which show at inspection. Reflections from inside a component, such as a chip facet or a lens surface, are seen in the same reading and can be separated from connector reflections only by distance, with an OTDR or an optical frequency-domain reflectometer.

Common failure modes

Reference method unstated. Two correct measurements differ by one or two connector losses.

Launch cord moved after the reference. The reference no longer describes the launch, and the error is as large as the change in coupling.

Cladding light. A short single-mode cord carries light in the cladding, which a large-area detector collects in the reference but which is lost in the component, so IL reads high. Loop the launch cord or use a longer one.

Coherent interference. A narrow-linewidth laser forms Fabry-Pérot fringes between reflective connectors, and the IL reading moves with temperature and wavelength. Use a source with a broad linewidth, or an isolator, and APC connections where the method allows.

Background not subtracted. RL is underestimated, as in the worked example, and the error is largest for the best components.

Far end not terminated. The reflection from the component's output connector or fiber end returns through the component and is added to the reading.

References: IEC 61300-3-4, Fibre optic interconnecting devices and passive components: Basic test and measurement procedures, Part 3-4: Examinations and measurements, Attenuation; IEC 61300-3-6, Part 3-6: Examinations and measurements, Return loss; IEC 61280-4-1, Fibre optic communication subsystem test procedures, Part 4-1: Installed cabling plant, Multimode attenuation measurement; D. Derickson (ed.), Fiber Optic Test and Measurement (Prentice Hall, 1998), chapters on power measurement and reflectometry.