
Yield improvement techniques in the manufacturing of AWG (Cascade) PLC

The AWG Cascade chip offers a step improvement over the conventional
AWG. In this article Broadex Technologies discusses how the AWG Cascade
chip works, where it will be deployed and how it can be manufactured
with a good yield.
BY HENK BULTHUIS, BROADEX TECHNOLOGIES UK
AWG Cascade products consist of two synchronized Silica-on-Silicon Array Waveguided Gratings (AWGs), arranged in series on a single chip. Arrayed Waveguide Gratings are working as a prism that disperse the light coming into the device when coupled to an input fiber. For AWG Cascade (CAWG) chips the dispersion of the two AWGs are synchronized which gives a theoretical zero loss over a wide passband. The small footprint allows the chip to be integrated in small form factor transceivers which make up the workforce of datacenters for data transport, internal between the racks, from rack to rack and even from datacenter to datacenter.
AWG Cascade or CAWG chip in transceiver package
In the illustration below, light enters the chip from a fiber to the input waveguide on the left-hand side. The four receiver channels on the right side of the chip connect to high-speed photodetectors and each photodetector captures a different slice of the optical spectrum. This way the chip is used to demultiplex the 4 channels from the input fiber, each channel carrying its own portion of data. The slicing action of the spectrum for each of the 4 receive channels is displayed in the transmission spectrum.
Layout of an approximately 10 mm long AWG cascade chip for CWDM4 transceiver.
The fiber can now carry 4 times more information compared to using only a single frequency of light. AWGs and CAWGS can be designed to Multiplex and or Demultiplex anywhere between 4 and 96 channels, thus multiplying the capacity of a fiber connection by a factor 4x to 96x respectively without need to grow the fiber plant. Compared to regular AWG, the AWG Cascade produces ultra-low insertion loss, flat-top bandpass shape while maintaining a single mode output. The single mode output allows for efficient coupling to ever increasing high-speed photodetectors which have ever decreasing active area sizes as the speed of switches, or baud rate, increases. Suitable for WDM applications, they can be used as both a MUX, to combine say the light of 4 lasers on a single fiber, or DEMUX, to separate the light from a single fiber to 4 individual detectors.
Figure above: Simulated transmission spectrum of Cascade AWG.
The curved AWG Cascade structure can also be laid out very efficiently to provide up to 16 individual AWG Cascade structures on a single chip, see figure, which is of a similar size (10x20mm) as a typical fiber block array that is conventionally used to couple light from chip to fiber. This dense footprint and the flexibility to adjust the mode field diameter, minimizing coupling loss, is ideal for Co-Packaged Optics (CPO) used in next generation multi-Terabit applications.
Yield improvement for AWG Cascade manufacturing: Even though semiconductor processes are used for manufacturing of PLC (Planar Lightwave Circuits) the yield is not just limited by defects. The wave-type nature of the photons that travel the channels on the chip means that the function of the chip is very sensitive to the exact pathlength that photons travel. When the photons are separated and combined they need to arrive exactly in phase. If the light path traveled by photons in one channel is off by a femtosecond this will already cause destructive interference of the photons causing the chip to fail specification.
During etching of the waveguides the width of the waveguides may vary due to local varying etch chemistry. Other nonuniformities may arise from imaging errors due to photo, wafer bow, resist spin, refractive index and thickness variation during deposition of the layers.