Introduction - Optical Signal Processing by Silicon Photonics - page 5

interconnected in an uninterrupted manner and therefore side-emitting, edge-
emitting, or conducting optical devices can be eagerly integrated on the same sub-
strate. In the case of parallel integration, the chip is constructed by developing the
columns of devices, doing so the surface- or bottom-emitting devices can effec-
tively be used whereas in the case of hybrid integration IP technology the devices
are fabricated using both serial and parallel integration on the same substrate.
Additional elements can be developed separately or directly attached to the IP cir-
cuit to access the control of the optical signals. In addition, both active and passive
devices may be required to be located on the same substrate and therefore hybrid
IP integration demands multilayered IP circuits and components to be produced on
a single substrate such that they must be compatible with three-dimensional struc-
tures of other IO or IP devices.
The enabling technologies for IP mainly depend on silica-on-silicon (SOS), in
which the structure of waveguide comprises three layers, named as; the buffer, the
core and the cladding. The real benefit of SOS is the ability to apply wafer-scale,
planar lithography and processing techniques to integrate substantial numbers of
functions either as arrays of identical devices or in the form of customized cir-
cuit configurations on single or multiple chips. This integration capability offers
an efficient platform for the implementation of typical fiber-based functionalities
such as optical power splitters or combiners, couplers, wavelength-selective cou-
plers, multiplexers/demultiplexers and optical gain elements. Furthermore, optical
switches and controllable attenuators based on the thermo-optic effect can also be
fabricated [ 2 ].
The Four Wave Mixing (FWM) in optical fibers can be both useful and harm-
ful, this depends on the application. It can be harmful as it is capable of inducing
crosstalk in WDM communication systems and limiting the efficiency of such sys-
tems. Nevertheless, FWM can be made avoidable by using asymmetrical channel
spacing's or using fibers having bulky enough GVD that the phase of FWM pro-
cess is not matched over long lengths of fiber.
The FWM is the process that becomes fairly efficient if the phase-matching con-
dition is fulfilled in the sense that the efficiency can be made to go beyond one.
From practical point of view, more power is appeared at the new wavelength in
comparison with the power of the signal happened to fall at the input end. This is
un-surprisingly true if we note that the pump beam provides energy to both the
idler wave and signal wave at the same time. The usage of FWM for conversion
of wavelength engrossed significant consideration during the 1990s because of its
potential application in light-wave wavelength division multiplexing (WDM) sys-
tems. If a pump beam along with a pulse train of signals is injected together, that
contains a sequence of “1” and “0” bits which is pseudo-random inside a parametric
amplifier, the wave is generated as a result through FWM only when the pump and
signal are presented after one another. This results the idler wave that appears in the
sequence form of a pulse train containing “1” and “0” bits as the signal. In conse-
quence, FWM is capable of transferring the signal data to the idler wave at a wave-
length that is new with perfect reliability. It can make a signal even more improved
in term of quality by reducing the noise intensity.