Tracking error

A packaged laser diode almost always carries a monitor photodiode behind its back facet, and the drive electronics almost always run in constant-power mode by holding that photodiode's current fixed. The arrangement assumes that the monitor current is proportional to the useful output from the front facet, with a constant of proportionality that does not change. Tracking error is the amount by which that assumption fails. It is defined as

the change in the front-to-monitor ratio relative to its value at a reference temperature, measured with the laser held at constant monitor current across the operating range, and quoted as a ± bound, for instance ±0.5 dB over −5 to 70 °C, or ±1.0 dB over −40 to 85 °C. Since 1% of ratio drift is 0.043 dB, a ±0.5 dB specification permits the launched power of a "constant-power" transmitter to wander by ±12%, and ±1 dB by ±26%, without any fault in the control loop.

The ratio drifts for physical reasons that can be listed. The monitor sees only the fraction of the back-facet light that lands on its active area, and that fraction changes as the laser's far-field divergence changes with temperature and current, as the package expands and shifts the diode relative to the facet, and as the fraction of light reaching the monitor by reflection off the lens, window, or submount changes. The monitor's own responsivity has a temperature coefficient: for a silicon photodiode near 850 nm it is a few tenths of a percent per kelvin, which alone can produce a drift of order 20% over a 120 K range, close to 1 dB; for the InGaAs monitors used at 1310 and 1550 nm it is a few hundredths of a percent per kelvin, of order 0.1 to 0.2 dB over the same span, which is one reason telecom parts reach tighter specifications than short-wavelength ones. The front-to-back power split itself depends on the facet reflectivities, $P_\mathrm{front}/P_\mathrm{back} = (1 - R_f)\sqrt{R_b}\,/\,[(1 - R_b)\sqrt{R_f}]$, which is 1 for an uncoated pair and 83 for a 5% / 95% coated pair, and both coatings have some wavelength dependence that the laser's temperature-driven wavelength shift samples. Finally the monitor collects spontaneous emission as well as laser light, and the spontaneous fraction is largest near threshold, so the ratio is worst at low power and in the first milliamps above threshold; a specification quoted at rated power says little about operation at a tenth of it.

Two consequences follow for the user. A control loop that must hold launched power to better than the tracking error cannot rely on the monitor alone and needs a tap on the output fiber, which is what analog CATV and high-accuracy instrument transmitters do. And a monitor-based loop responds to a tracking-error drift by changing the drive current, so a rising ratio makes the loop overdrive the laser, which shortens its life and can carry it past its kink-free rating; the derating margin should include the tracking error.

Measurement is a constant-monitor-current sweep. Hold $I_\mathrm{mon}$ at its rated value with the automatic-power-control loop closed, step the case temperature across the range with settling at each point, record the front-facet or fiber-coupled power with a detector whose own temperature is controlled, and take the ratio relative to the reference temperature. Repeat at a low monitor current if the application operates there. A tracking error that changes after burn-in or over life indicates a monitor alignment or window contamination problem rather than a laser problem, and a step in the ratio at one temperature is usually a mode hop changing the far field. The row on a data sheet and what its absence means are discussed in How to read a laser diode datasheet.

References: Telcordia GR-468-CORE, Generic Reliability Assurance Requirements for Optoelectronic Devices, for the monitor-tracking test conditions used in qualification; L. A. Coldren, S. W. Corzine and M. L. Mašanović, Diode Lasers and Photonic Integrated Circuits, 2nd ed. (Wiley, 2012), for the facet power-ratio relation; manufacturer data for photodiode responsivity temperature coefficients (Hamamatsu, silicon and InGaAs). The dB conversions and the front-to-back ratio are computed from the stated values.