Researchers at NTT Photonics Laboratories in Ibaraki, Japan, have invented an arrayed-waveguide device that operates over a wide wavelength range, produces a stable, narrowband output, and can be fabricated on a 4-in. wafer. The current prototpe has 16 channels spaced 1 GHz apart with a frequency deviation of no more than 250 MHz from the design over 150 h using a standard commercial temperature controller. Researchers say the new device will provide a much-needed standard for the S- and L-bands, which are not served by the acetylene- and hydrogen cynacide-cell filters used to calibrate across the C-band.
The new AWG is a folded version of the conventional variety (see figure). It works in a similar way to a one-dimensional diffraction grating or a prism, but the dispersion/wavelength selection is much larger than would normally be the case for either. In a grating, slits of varying refractive index provide the path-length difference, which will usually vary by no more than a few millimeters. In a prism, the maximum path-length difference is related to the size of the optical element. With an arrayed waveguide grating, however, the difference is determined by length of waveguide connected to the input slab and therefore can be arbitrarily long. Because the discrimination of the grating is related to the path difference, only the AWG can satisfy the requirements of DWDM.
Of course, though the waveguides can be arbitrarily long, there is a cost associated with increasing their length to achieve increased discrimination. In particular, the NTT team found that, to make the 16-channel device they were interested in, the device would have to be 22 cm long. By taking the simple step of folding it and carefully designing the geometry to retain the required AWG properties, they were able to reduce the length of the device to less than 7 cm. Initially, the quality of the emission spectrum was poor: because of its large footprint, the refractive index of the waveguides varied significantly across the device. However, they were able to correct for this using a selective ultraviolet phase-compensation technique, with the result that each of the band losses was kept in the 8- to 11-dB range.
Though the initial work has concentrated on a TE-polarization-only AWG, NTT engineers say it should be straightforward to make the device polarization insensitive. They propose to do this, they explain, by putting a half-wave plate in the device and doing the appropriate phase compensation on either side of its line of symmetry.
For more information, contact Kazumasa Takada at email@example.com.
- Kazumasa Takada et al., IEEE J. Lightwave Tech. 20(5) (May 2002).