Manufacturing Matters with Tunable Lasers

Paula M. Powell

One problem faced by manufacturers of full-band-tunable laser modules is the need for close cost parity with existing fixed or partially tunable devices. A key step in achieving this goal involves standardizing as much of the fabrication and assembly process as possible.

Bookham Technology plc in Abingdon, UK, recently used such an approach when it introduced a high-power broadband tunable laser module based on a monolithically integrated digital-supermode distributed Bragg reflector laser and semiconductor optical amplifier. The surface-ridge device is fabricated using conventional InP processing (see figure). In addition, its optical packaging uses many processes, automated equipment and piece-parts common with some of the firm’s other products.

This full-band-tunable laser module, based on a monolithically integrated four-section distributed Bragg reflector laser and a semiconductor optical amplifier, tunes much like an integrated set of three-section reflector lasers would.

The laser module is fully tunable across the C-band. Besides the semiconductor optical amplifier, it has four key sections: a rear grating comb reflector, a linearly chirped front grating, and gain and phase sections. In essence, the laser module can be narrowband-tuned by adjusting (scanning) the current to the rear grating, while holding the current to the front grating at a fixed value. Using the phase section for longitudinal mode tracking then enables continuous tuning. The modules can be tuned to any of 80 channels at 50-GHz spacing with minimum channel powers of 13.5 dBm and a side mode suppression ratio greater than 40 dB. One benefit of this strategy is the capability for rapid calibration.

Gratings for this module and its predecessor, based on a three-section distributed Bragg reflector laser, are etched into the substrate structure using direct-write electron-beam lithography. The process can repeatably write the regular period and phase change required to enable the overall reflection responses necessary for tuning. Other conventional etching techniques either don’t offer the required precision or are just not repeatable enough to support a high-yield production process.

Although the technique takes longer than other etching processes, the increase in manufacturing time is offset by the fact that chips in stock can be immediately pulled against any customer wavelength requirement. The bin-fill exercises required with fixed or thermally tunable lasers to ensure adequate wavelength coverage also are eliminated.

An equally critical component of the yield equation involves screening and testing the chip-on-carrier assembly to ensure high yields as the stage components enter final assembly. As is fairly common with this laser type, after active-layer and epitaxy growth, etching and other fabrication processes, passivation and metallization, the wafer is cut into bars (or stripes). At this stage, the performance of the bars usually is tested. The testing is of a somewhat rough nature because of difficulty in accessing individual parts, adding heat sinks, etc.

Once the finished laser is bonded to the substrate, though, it is easier to handle. It is at this stage, when one has added only the cost of the bonding process and the substrate material to the equation, that it is first possible to perform detailed, accurate testing of individual laser properties. It therefore makes sense to take on any additional cost of testing at this point instead of during final assembly, when packaging and the thermoelectric cooler are added. Those elements can account for more than three-quarters of the material cost of the final product.

Acknowledgment

The editors wish to thank Dave Woodcock, a product line manager at Bookham Technology plc in Abingdon, UK, for his contribution to this article.