CAD optimizes component design at many levels

Aug. 1, 2000

Zhengyu Huang

Brent K. Whitlock

Robert Scarmozzino

Advances in dense wavelength-division multiplexing (DWDM) are rapidly increasing the amount of available fiberoptic bandwidth. As fiberoptic communications systems improve, the requirements they place on component specifications become more and more stringent. Component design to meet these requirements benefits from the use of computer-aided design (CAD) tools to model and simulate devices.

Levels of abstraction

There are CAD tools available for both electronics and photonics at many levels of abstraction. A component design can be modeled at a variety of different levels depending upon what details are known of the component`s physical design parameters and what details are required from its simulation. Generally speaking, models at higher levels of abstraction have less physical design detail and can be simulated fairly rapidly. Conversely, models at the lowest levels of abstraction have a great deal of physical design detail and require much-longer simulation run times.

System simulators are at a high level of abstraction. Lightwave system simulators can be used to design fiberoptic communications systems using component models with parameters that correspond to their system-level performance. For example, optical filters can be modeled with a specified frequency response. Circuit simulators are at a lower level of abstraction and allow more-detailed modeling of components, such as transmitters and receivers. Device simulators are at the lowest levels of abstraction and allow very detailed modeling of both active and passive components such as laser diodes and optical waveguides.

A hierarchical CAD environment uses tools and models at different levels of abstraction to optimize the trade-offs between computational efficiency and model complexity when optimizing designs. The benefits of this hierarchy include reduced design-cycle time, increased capability to handle design complexity, decreased redesign cycles between prototypes and shipping product, increased manufacturing yield, and increased product reliability and longevity. The design optimization process cuts across the boundaries of the design hierarchy as design parameters in each of the levels affect and are affected by parameters and simulation results at other levels of the hierarchy.

In a mixed-level environment, simulators at different levels of the hierarchy are used together to tie design parameters at one level more closely to performance at another level of abstraction. Such environments are common in the electronics field, and the explosion of photonics in the marketplace has generated increased interest in developing equivalent tools in this area. For example, the Photonics CAD consortium is an Advanced Technology Program sponsored by the National Institute of Standards and Technology (Gaithersburg, MD) that is working on developing a multilevel framework and tools for photonics CAD.1

An example of a WDM component whose design benefits from mixed-level simulation is the wavelength router. An arrayed waveguide (AWG) router is an integrated optical wavelength multiplexer based on an arrayed waveguide grating. As the key device in WDM optical systems, an AWG router can offer many different functions. Its basic demultiplexing and multiplexing capabilities are as follows. For a typical N × N AWG router, a set of N different channel wavelengths (almost equally spaced) can be input into each of N input ports, and each wavelength is output to a different output port. On the other hand, if each of the N wavelengths is input into its corresponding, but different, input port, all the wavelengths will be output together at a particular output port determined by the order of the input wavelength set. In this manner, the AWG multiplexes the four separate wavelength input signals into a single DWDM output signal. In addition, by reconfiguring the AWG router, it can realize add/drop multiplexing (ADM). Moreover, because at any channel wavelength, different input ports always connect to different output ports, a single AWG router can function as a full N × N interconnection.

An N × N AWG router is composed of N input/output waveguides and two identical star couplers connected by an array of waveguides. The operating mechanism of an AWG router is as follows. The beam propagates through the input waveguide and becomes divergent in the input star coupler, which is a free propagation region. Then the beam is coupled into the waveguide array and propagates through the individual array waveguides to the output coupler. The optical length difference between adjacent array waveguides is designed to be a constant, integer multiple of the center wavelength of the router. Thus, the outgoing beam from the array waveguides has a wavelength-dependent wavefront tilt. After propagating through the output star coupler, the beam will be focused at a wavelength-dependent position and coupled into different output waveguides depending on the wavelength.

Simulation at the device level

Arrayed waveguide routers are passive optical devices that are generally designed using a device-level CAD tool (see Fig. 1).2-4 For example, a 4 × 4 AWG router can be simulated using the beam propagation method (BPM).5 Simulation of such a device in its entirety, while possible, is complicated due to the large spatial domain required by the array and because it entails propagation at large angles from the optical axis. A superior method breaks the problem up into sections that can be simulated separately and combined afterwards via a matrix-like approach. Input/output coupler and waveguide sections are simulated via standard BPM, and the waveguide array is represented simply by a laterally varying phase, derived from the array geometry. The BPM simulation stops at the beginning of the array, the phase of the wavefront is modified, and then the BPM resumes at the beginning of the output star coupler. This process is repeated at each wavelength until the full spectrum is produced.

The simulation produces the resulting spectrum of a 4 × 4 AWG router based on a silica waveguide structure. The wavelength channel spacing is 0.4 nm (50 GHz), the waveguide width is 6 µm, and the number of arrayed waveguides is 40. The characteristics of this AWG wavelength response-including channel spacing, passband width and flatness, channel center wavelength and interchannel crosstalk, and so on-can all be adjusted by choosing appropriate device parameters. For example, the passband width can directly be changed by varying the ratio of the effective mode width (We) and the separation of the input/output waveguides in the star couplers (Di). It should be noted that there are other methods of controlling the passband width that rely on changing the mode shape; the simpler method presented here is used only for illustration.

Mixing levels

In order to illustrate a mixed-level simulation, device-level results are used in the system simulation of a WDM network employing AWG wavelength routers as the multiplexer and demultiplexer. One of the important system design considerations is the variation in the laser-transmitter wavelengths of each of the channels, especially when the network contains multiple transmitters and receivers at a given wavelength channel. Depending on the lasers used in the system, the AWG may need to be designed to allow for greater or smaller variation in wavelength from the ideal. This requires that each channel be optimized to have a wide enough passband with minimal loss and crosstalk between the channels. As mentioned before, the passband width can be enlarged by increasing We /Di . Using device-level BPM simulation, the spectral responses for three cases with different We /Di are calculated (see Fig. 2).

Next, system-level simulations are performed in order to optimize the device design based on system-level performance. The four-channel network consists of externally modulated laser transmitters multiplexed by an AWG, sent over 30 km of fiber, demultiplexed by a second identical AWG, and received and analyzed by four receivers and BER tester modules. The topology for such a network can be modeled in a system-level CAD tool (see Fig. 3).

This network is simulated to study the power penalty associated with variations in the transmitter wavelengths. The simulation sweeps the laser wavelength parameter through the range of values that it may take depending on time, temperature, and manufacturing variations. Bit-error rate curves versus input optical power to the AWG mutiplexer are generated for each of the wavelength values in the sweeps (see Fig. 4). This system-level simulation is performed for each of the device designs with varying We /Di . In general, the device simulation can either be called by the system simulation as needed or be performed ahead of time for each of the planned device parameter sweeps. Here, at the system level of abstraction, only the wavelength response is input as the mux/demux parameters.

In sum, these examples illustrate that mixed-level simulation allows the component designer to choose appropriate parameters in the device design to optimize the system performance in the target system. This is important for evaluating the impact of component design on system performance, especially when the system performance is related simultaneously with different aspects of device performance.

REFERENCES

1. For additional information on PCAD, see www.pcad-team.com.

2. C. Dragone, IEEE Photon. Tech. Lett. 3(9), 812 (Sept. 1991).

3. H. Takahashi et al., Electron. Lett. 26(2), 87 (Jan. 1990).

4. M. K. Smit and C. van Dam, IEEE J. Selected Topics in Quantum Electronics 2(2), 236 (June 1996).

5. For a recent review of BPM and other methods, see R. Scarmozzino et al., IEEE J. Selected Topics in Quantum Electronics 6(1), 150 (Jan./Feb. 2000).

Zhengyu Huang is staff scientist, Brent K. Whitlock is senior scientist, and Robert Scarmozzino is chief scientist at RSoft Inc., 200 Executive Blvd., Ossining, NY 10562; www.rsoftinc.com. For more information, contact Brent Whitlock at [email protected].
FIGURE 1. A 4 × 4 arrayed waveguide (AWG) router is modeled in a device-level computer-aided design tool. The software breaks the model down into sections and simulates them separately. (Image from RSoft BeamPROP software)
FIGURE 2. Spectral responses for a 4 × 4 AWG router with different passband widths are obtained as a result of using different values of the ratio of effective mode width to separation of the input/output waveguides (We/Di). The individual wavelength responses of each of the four output ports are overlaid. In the first case (top), the passband is significantly widened; however, the interchannel crosstalk is relatively high (19 dB). While the second and third cases both have very low channel crosstalk levels (less than 43 dB), the passband width in the second case (center) is still twice of that in the third case (bottom). In these simulations, random phase error in the arrayed waveguides-which usually limits crosstalk to higher levels-has been neglected to simplify the presentation.
FIGURE 3. A system-level CAD tool models the topology for a four-channel WDM network. (Image from RSoft LinkSIM software)

FIGURE 4. Link-level simulation shows the required input power to the multiplexer to achieve a certain bit-error-rate (BER) performance. The graph illustrates the power penalty due to a 0.5-Å variation in the transmitter wavelength. The AWG used in the simulation corresponds to the second case in Fig. 2. For a BER of 10-10, when the channel wavelength is varied by +0.5 Å from the filter center, the power penalty is approximately 1.9 dB. Similar simulations were performed for the first and third cases; the corresponding power penalties are 0.9 dB and 5.6 dB, respectively. Because the first case has a high level of crosstalk (which will cause problems in other tests), the second case is determined to have optimized device parameters to realize the minimized power penalty for wavelength variation.

Sponsored Recommendations

March 12, 2025
Join us for an engaging discussion with industry experts on the intersection of AI and optics. Moderated by Sean Buckley, editor-in-chief of Lightwave+BTR, this panel will explore...
March 25, 2025
Explore how government initiatives and industry innovations are transforming rural broadband deployments, overcoming cost and logistical challenges to connect underserved areas...
April 9, 2025
As transceiver speeds increase, so do thermal challenges. Discover key insights into innovative cooling solutions that ensure optimal performance and reliability.
March 7, 2025
In today’s hyperconnected world, rolling out and managing profitable, high-performance networks for access and transport will require innovative architectural approaches. The ...