Wavelength locker

A wavelength locker is a small optical frequency reference built into a laser transmitter to keep its emission on a fixed frequency, usually a channel of a dense WDM grid. A few percent of the output is tapped and sent through a filter whose transmission changes steeply with frequency, most often a low-finesse etalon; the ratio of the filtered to the unfiltered photocurrent is an error signal, and a control loop adjusts the laser temperature or a tuning current to hold that ratio at a set point. The aim is to keep the channel within a few gigahertz of its nominal frequency over life and temperature, a small fraction of a 50 or 100 GHz channel spacing.

Why lasers need one

A DFB laser drifts by about 0.1 nm/°C, which is 12.5 GHz/°C at 1550 nm. A thermoelectric cooler holds the chip temperature, but the relation between chip temperature and wavelength changes slowly with aging and with drive current, and the thermistor does not sit exactly at the active region. Tunable lasers are harder still: their wavelength depends on several section currents and on temperature, and the calibration table that sets a channel is only accurate to within some gigahertz. The locker closes the loop on the optical frequency itself. Coarse WDM, with 20 nm channels, tolerates the full drift and needs neither cooler nor locker.

Etalon locker and its error signal

The etalon thickness is chosen so that its free spectral range equals the channel spacing, so that every grid channel sits at the same point on a transmission fringe. For fused silica with group index $n_g$ = 1.462 at 1550 nm, a 50 GHz free spectral range requires

$$d = \frac{c}{2 n_g\,\Delta\nu_\mathrm{FSR}} = 2.05\ \mathrm{mm}.$$

The Airy transmission of a lossless etalon with mirror reflectivity $R$ is

$$T(\nu) = \frac{1}{1 + \mathcal{F}\sin^2(\pi\nu/\Delta\nu_\mathrm{FSR})},$$ $$\mathcal{F} = \frac{4R}{(1-R)^2}.$$

Lockers use low reflectivities to obtain a broad, nearly linear slope. With $R$ = 0.40 the finesse is 3.3 and the transmission swings between 1 and 0.18. The lock point is set midway, at $T$ = 0.59, which lies 6.4 GHz from the transmission peak; there the transmission changes by 0.071 per GHz, or 12 % of its value per GHz. A 0.1 % error in the photocurrent ratio therefore moves the locked frequency by about 8 MHz, and a 1 % error, from unequal aging of the two photodiodes or a change in tap ratio, by about 84 MHz.

Because the etalon response repeats every free spectral range, the locker only holds the laser on the nearest fringe of the correct slope. The coarse setting from temperature and current must place the laser within roughly a quarter of the free spectral range of the intended channel before the loop is closed; otherwise it locks to a neighbor.

Temperature of the reference

The etalon itself drifts. For solid fused silica, with a thermo-optic coefficient of about 1 × 10⁻⁵ K⁻¹ and an expansion coefficient of 0.55 × 10⁻⁶ K⁻¹, the fringes move by about 1.4 GHz/K at 1550 nm. Lockers therefore mount the etalon on the same temperature-controlled platform as the laser, use athermal designs that balance the two coefficients, or use an air-spaced etalon, whose drift with a fused-silica spacer is about 0.11 GHz/K.

Other discriminators

The edge of a thin-film filter, two etalons in quadrature, or a fiber Bragg grating can replace the single etalon. On silicon and indium phosphide chips, the locker is often a ring resonator or an unbalanced Mach-Zehnder interferometer with monitor photodiodes, the same structures used to lock the rings themselves (see ring resonator thermal tuning and locking). In the laboratory, a wavemeter or a stable reference cavity serves the same purpose with higher accuracy and lower speed or greater bulk.

Pitfalls

Back-reflections from the etalon into the laser cause feedback and must be avoided by tilting the etalon or using an isolator, but tilt raises the free spectral range through the $\cos\theta$ factor in its denominator, so the angle is fixed during alignment. Dark current and offset in the photodiode electronics become frequency errors at low tap power. Finally, the locker regulates only the mean frequency; it does not narrow the linewidth.

Common questions

What is the difference between a wavelength locker and a wavemeter?

A wavemeter measures an absolute wavelength with high accuracy over a wide range but is too large and slow to sit inside a transceiver. A locker gives a fast, compact error signal around fixed set points and relies on a prior calibration for its absolute frequency.

Does a wavelength locker narrow the laser linewidth?

No. Its loop bandwidth is usually far below the frequencies that determine linewidth; it removes slow drift.

References: B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, 3rd ed. (Wiley, 2019); L. A. Coldren, S. W. Corzine and M. L. Mašanović, Diode Lasers and Photonic Integrated Circuits, 2nd ed. (Wiley, 2012); E. Hecht, Optics, 5th ed. (Pearson, 2017).