How to Calibrate and Cross-Check an Optical Power Meter

Scope

This article describes how to check that an optical power meter is reading correctly between its formal calibrations, and how to estimate the uncertainty of a power reading. Formal calibration, traceable to a national standard, is done by a calibration laboratory; the checks here catch the drift, damage, setting errors and nonlinearity that occur in between. Choosing a meter in the first place is covered in Choosing a Laser Power Meter.

In short: zero the meter, compare it with a reference meter at the same wavelength and power, confirm the wavelength setting, test linearity by adding two beams, check the repeatability of fiber connections, and combine the contributions into an uncertainty. Record the results so that a drift shows up as a trend.

1. Zero

With the input blocked by a cap (not merely the beam turned off), zero the meter under the same conditions as the measurement: the same range, temperature and, for a thermal head, after it has settled. Repeat after a change of range or after the head has warmed. An offset that cannot be zeroed out on a photodiode head suggests light leaking in or damage.

2. Compare with a reference

Keep one meter, recently calibrated and used only for checks, as the reference. Measure the same stable source with both, at the wavelength and power level of interest, swapping the heads into the same position or fiber connection. Use a source whose output is stable over the minutes the comparison takes; a laser with a polarization-independent isolator, or a broadband source through a filter, reduces fluctuations from back-reflection and polarization.

The ratio of the two readings is the meter's correction relative to the reference. A difference within the combined calibration uncertainties of the two meters is consistent; a larger one, or one that grows from check to check, calls for recalibration.

3. Wavelength setting

Measure a source of known wavelength with the meter set correctly and then set to a neighboring wavelength. The change should match the ratio of the responsivities at the two settings; at constant quantum efficiency the responsivity is proportional to wavelength, so 1310 versus 1550 nm differs by 0.73 dB. A meter that shows no change has a fault in its wavelength correction; one that shows a much larger change near its band edge is behaving normally, which is the reason to set the wavelength carefully there.

4. Linearity by superposition

A meter is linear if the reading for two beams together equals the sum of the readings for each alone. With a source split into two paths, each with its own shutter, and recombined onto the detector:

1. Read beam A alone, $P_A$.
2. Read beam B alone, $P_B$.
3. Read both together, $P_{AB}$.

The nonlinearity at the combined level is $P_{AB}/(P_A + P_B) - 1$. With readings of 0.500 mW, 0.500 mW and 0.997 mW, it is −0.3%. Repeating with attenuators at a series of levels steps the check across the meter's range; nonlinearity usually appears near the top of a photodiode head's range, as it approaches saturation, and at the boundaries between gain ranges. The two beams must not interfere at the detector, so they are made mutually incoherent by a path difference longer than the source's coherence length, or by orthogonal polarizations.

A simpler, less complete check is to insert a calibrated attenuator, for example a 10 dB neutral density filter of known transmission, and confirm that the reading falls by that amount at several power levels.

5. Connector repeatability

For fiber measurements, disconnect and reconnect the fiber to the meter's adapter ten times, cleaning the connector each time as in Fiber Connector Inspection and Cleaning, and record the spread. A spread of a few hundredths of a dB is typical of clean connections; a larger one points to a dirty or damaged connector, a worn adapter, or a fiber end that does not sit at the same position on the detector each time.

6. Uncertainty budget

Combine the independent contributions as a root sum of squares. An example for a photodiode meter at 1550 nm, in percent of the reading:

In decibels, 1% is 0.043 dB, so the combined uncertainty is about 0.12 dB. For a loss measurement made with the same meter at both ends (reference and device), the calibration term largely cancels and the repeatability and linearity terms dominate; see Measuring Insertion Loss and Return Loss.

Records

Log the date, the reference meter and its calibration date, the source, the wavelength, the readings and the ratio for every check. A log turns a single surprising reading into a trend that shows whether the meter has drifted or the source has changed.

References: International Organization for Standardization, ISO/IEC Guide 98-3:2008, Uncertainty of measurement: Part 3: Guide to the expression of uncertainty in measurement (GUM); D. Derickson (ed.), Fiber Optic Test and Measurement (Prentice Hall, 1998).