Photonic simulation has become an integral step in the design cycle of optical networks. Armed with simulation software, a system engineer can alter the physical parameters of optical components and study their impact on network behavior. To ensure validity of the photonic simulation, it is important that the virtual model reflects the actual physical properties of the network.
Recent software interfaces have been developed between design-and-simulation tools and physical measurement equipment to capture essential characteristics of the optical components and import them into virtual models. The accurate modeling of newly developed components helps system integrators understand the components' functional importance in the network and increase efficiency of design.
As manufacturers of DWDM components and subsystems, we are always searching for the most advanced instruments that not only measure optical parameters of our products, but also are able to translate them into the qualitative figures of network performance. We used optical-modeling software (VPItransmissionMaker from VPIsystems) and an all-photonic analyzer (81910A from Agilent Technologies) to combine fast and accurate optical measurement with advanced network simulation.
The recently developed data-exchange interface takes the measurement data from the analyzer and generates a virtual model of the measured component. The model reflects all the parameters of the component, such as insertion loss, reflectivity, polarization-delay loss (PDL), group delay, and differential group delay. The virtual component can then be incorporated into the network model, and the software will simulate the network behavior. The software enables analysis and verification of optical links design in a very cost-effective manner. It includes extensive libraries of numerical models for passive and active optical components such as fiber, lasers, photodiodes, and erbium-doped fiber amplifiers (EDFAs).
The ability to accurately measure components and immediately model their performance in the network results in significant cost savings compared to building a test bed and running network experiments with expensive equipment and real photons. This is especially critical in testing of multichannel devices, such as DWDM multiplexers. A 40-channel optical bit-error-rate (BER) test station with tunable lasers would be a costly alternative to a virtual model.
In addition to significant cost savings, virtual prototyping also shortens the component design cycle, giving system integrators understanding of the network operation changes when altering component parameters. The interface between the analyzer and the software enables live measurements and real-time data export from the component to the system model.
The real-time measurement is also important when testing tunable components. The ability to tweak component parameters, measure them in real time, and run simulations in a sweep mode gives network engineers instantaneous feedback. This real-time feedback is invaluable as it reveals how the physical parameters of components would influence the network performance merits, like transmission distance or BER.
We have been using the software to model the behavior of our DWDM mux (Lightchip) in customer networks. Consequently, we have helped our customers identify the required specifications for the mux/demux to make a knowledgeable purchasing decision. In one instance, our customer—a tier-two telco equipment supplier—asked us to demonstrate the impact of various DWDM filter shapes on the network performance and to choose an optimum filter shape based on the specific laser drift and the targeted quality of transmission. The application was a 40-channel DWDM metropolitan network.
Using the analyzer in conjunction with the software, we generated two models of the muxes/demuxes. The models were based on the two filter shape options that were available. Both filter options had flattened tops, but differed in width and minimum insertion loss (see Fig. 1).
The filter shape tailoring is possible in diffraction-grating muxes/demuxes because of the use of micro-optic elements in the beam path. The models contained the data on optical transmittance and phase response of each of 40 channels within the wavelength range of 1520 to 1565 nm. Because the mux/demux has almost identical frequency/phase response across all channels, we elected to measure only one channel and generate the rest by shifting the measured filter in frequency domain. The alternative approach of measuring every channel would have taken significantly longer time, but the virtual model would have been more accurate.
The goal of the simulation was to study BER performance of an OC-192 (10-Gbit/s) system as a function of laser drift and optical-filter shape. The BER degradation happens when a drifting wavelength passes through a number of optical add/drop nodes without an optical-to-electrical-to-optical (OEO) conversion. As the laser drifts, the signal spectrum no longer falls within the nominal passband of the mux/demux. Consequently, the higher harmonics begin to cut off, causing transmission errors.
The maximum allowed laser drift significantly limits the number of concatenated optical filters that the signal can pass without an OEO conversion, and therefore the maximum transmission distance. To keep the network cost low and avoid OEO conversion, it is important to understand the system implication of the DWDM filter shape and choose the mux/demux parameters accordingly.
The simulation setup includes a point-to-point DWDM link with three optical add/drop (OAD) nodes. Each OAD represents a mux/demux pair and an array of optical switches in between. For simplicity, we did not model any added or dropped signals, but studied the expressed transmission instead, as a worst-case scenario (see Fig. 2).
In this implementation, all channels were passing through a concatenation of eight mux/demux filters and total of 80 km of fiber. The system was designed to transmit 40 channels in the C-band separated by 100 GHz. 10-Gbit/s signals were generated by an array of 40 externally modulated lasers with the output power of 2 mW. The nodes were separated by 20 km of SMF-28 fiber with no chromatic-dispersion compensation. Four EDFAs were used to boost the power before the light is launched into the fiber.
Our multiplexers and demultiplexers were implemented as component sweeps. This allowed switching between the two measured components with different filter functions during the course of emulation. We modeled two cases, with all muxes and demuxes in the network having either one or another filter option. The setup also included a parameter sweep, which varied the wavelength of the laser in Channel 20 (194.3 THz, or 1542.94 nm).
The wavelength was detuned from +150 to –150 pm from the nominal in 50-pm decrements. Each wavelength sweep was performed for every option of mux/demux filter shape. At the end of each simulation cycle, the signal data was stored in a binary file for future analysis and BER plots generation. The system performance was evaluated using a virtual BER estimator (see Fig. 3).
In this network configuration, DWDM mux/demux with the Filter Option 1 can tolerate ±100-pm laser drift and Filter Option 2 tolerates a wider detuning of ±150 pm, without degradation of signal quality. The BER plots show that with the Filter Option 1 and the wavelength drift of ±150 pm, the targeted BER of 10 to 16 cannot be achieved.
Based on the results of simulation, we recommended that our customer use mux/demux Option 2 in a combination with less expensive lasers having ±150-pm maximum wavelength drift. This mux/demux has slightly higher insertion loss, but because of its widened passband, tolerates larger wavelength drift, thus reducing the cost of transmitter.
This design example illustrates how a combination of accurate component measurement and system simulation led to a wise purchasing decision that was based on qualitative data analysis. Alternatively, a lab trial would have taken significantly longer and would have involved a major investment in equipment at the early stage of development. The required specifications of the DWDM mux/demux were determined based on data captured from real component characteristics and an accurate performance analysis of the customer's system that incorporated these components.
The convergence of real and virtual prototyping has shortened the design cycle of our customer's system, while allowing us to avoid costly network experiments. Additionally, it has added confidence in the reliability of the system architecture.
Yuri A. Yudin is a senior fiberoptic systems engineer at Confluent Photonics, 5 Industrial Way, Salem, NH 03079. He can be reached at email@example.com.