Calibrate FO power meters to the limit

When someone mentions calibration, engineers probably think of voltage, frequency, and other fundamental electrical measurements. Light measurements may not come to mind unless these engineers work with fiber-optic (FO) test equipment. Like most other measuring instruments, FO instruments require periodic calibration.

Figure 1. A fiber-optic power meter, such as the Anritsu MT9810A, provides a sensor head and separate instrumentation. This instrument uses a thermoelectric sensor to convert the light it receives from a fiber-optic cable or a laser beam to an electrical quantity. Courtesy of Anritsu.

The standard used to define optical power exists as a piece of laboratory equipment within a national technical laboratory (see, "A cool measurement technique," p. 12). These labs include the National Institute of Standards and Technology (NIST) in the US, the National Physical Laboratory (NPL) in the UK, and the Physikalisch-Technische Bundesanstalt (PTB) in Germany. NIST, for example, can perform calibrations at 670, 780, 850, 980, 1300, and 1550 nm across power ranges of 10 to 200 µW. Typically, the uncertainties of these measurements range from 0.5% to 1.25% (Ref. 1).

Figure 2. A calibration chain leads from a standard maintained at a national lab down to commercial cal labs and, finally, to a user’s instrument.

Remember, no matter how well NIST calibrates a transfer standard, its measurements always include uncertainty. As any standard measurement gets "transferred" down the calibration chain, these uncertainties increase, because each of the instruments in the chain contributes its own uncertainty. The calculation of overall uncertainty involves computing the root of the sum of the squares of individual uncertainties (Ref. 2). You won't need to calculate these uncertainties, but you must know they exist and how they affect the measurements produced by FO power meters.

So, an end user who receives a calibrated FO power meter may find the instrument has an uncertainty of anywhere from 1% (0.04 dB) to 2% (0.09 dB), or even more. As a general rule, you can make good lab and production-line measurements with FO power meters that have an uncertainty of 2% or less. In most cases, power meters used for field testing don't need such a low uncertainty. (Uncertainties noted in this article refer to those with a confidence limit of 95%.)

Information provided by Fotec (Medford, MA) shows how uncertainties can increase to ±5% along the calibration chain from the physical NIST standard to a final product (Table 1). Comparisons of readings from two such end products could vary by as much as 10% because one product could read 5% high, and the other could read 5% low. In practice, though, readings don't show such large disparities, because differences tend to average out to smaller values when you make several measurements (Ref. 3).

Keep in mind that power-meter manufacturers and cal labs cannot control how you use specific equipment. Thus, the uncertainties they specify only relate to the instrument and not to its use in a specific test system. Accessories such as cables and connectors also can contribute to uncertainties. Single-mode fiber (preferred for telecom applications) has a diameter of about 9 µm, so to make good, reproducible measurements, you must ensure that the tiny fiber properly aligns with the detector in the power meter. Simply removing the connector from a power meter, repositioning the fiber, and reconnecting the fiber can change the power reading by 5% (0.2 dB).

Dirty connectors and fiber ends also contribute to uncertainties. If you plan to establish an FO measurement system, use well-characterized "golden" or measurement-quality jumper cables. Also, reducing the movement of fibers in a test setup will reduce any mechanical contributions to uncertainty. Maintaining a constant temperature and relative humidity during testing will help maintain uncertainties at reasonable levels as well.

An absolute-power measurement provides information about the power of one optical signal at one power level. But almost all users want to make accurate power measurements over the range an instrument provides, and they want to compare power levels from different sources. Those types of measurements require a linear response to power. Of course, the linearity measurements come with their own uncertainties (Table 2). The uncertainties increase at lower power levels due to the difficulties inherent in accurately measuring small quantities of light. The electrical characteristics of a power meter may contribute to the uncertainties, but for the most part, instrument suppliers have minimized these effects and provide instruments with good linearity characteristics.

A common power-linearity calibration technique involves using two light sources, usually lasers, and as strange as it may seem, neither of these sources must produce a calibrated power output. The laser sources must provide a stable output, though. In a typical calibration setup, as shown in Figure 3a, a technician first sets one laser to provide an arbitrary power level to the meter. Then, he or she removes the optical signal and substitutes the signal from a second source. By adjusting the output of the second source until it equals the power output from the first source, as indicated by the meter, the technician ends up with two sources of equal power. 

Figure 3. a) A linearity calibration uses two adjustable light sources that produce equal outputs. One signal produces a reference power, and combining the signals produces 3-dB (2X) steps that provide the next reference points for the calibration. b) You also can use one light source and a splitter.

Combining the two equal-power light sources in an FO coupler should produce a new reading 3 dB (2x) higher than the reading from each individual source. If the reading from the combined light sources does not equal twice the value of an individual source, the technician can adjust the power meter accordingly to achieve the two-fold power reading. The technician can continue to increase the power in 3-dB increments to calibrate the entire power range—within the uncertainty of the meter. A similar technique differs only in that one light source and a splitter or coupler produce the two light signals (Figure 3b).

So far, the absolute and relative calibrations have taken place at a single wavelength. Although semiconductor detectors exhibit a nonlinear response across FO wavelength bands, manufacturers have characterized their performance. Thus, power-meter manufacturers can apply a correction factor to adjust power readings at wavelengths not checked during a routine calibration.

If need be, you can have a lab perform a calibration at the 850-, 1300-, and 1550-nm wavelengths commonly used for FO communications. The cal lab also can calibrate a power meter at other wavelengths. Such a calibration uses a white-light source and a monochromator to produce light at the required span of wavelengths. This type of monochromator setup presents its own set of problems, such as low optical power levels (the narrower the spectral range, the lower the power), power instability over long calibration times, and the need to use external optics (Ref. 4).

By knowing and properly applying the uncertainty values determined during calibration, you can ensure FO power meters provide meaningful measurements, and you can properly interpret the results. Although FO power meters may provide many digits on their displays, you have to take into account uncertainties to arrive at a span or range of values in which the actual value should exist. In most cases, you can't take all the meter's digits at face value.