pda-challenges for an emerging industry

pda-challenges for an emerging industry

Photonic design automation allows the combination of various simulation paradigms to achieve real photonic capabilities.

HARALD HAMSTER, BNeD GmbH, and JUAN LAM, Hewlett-Packard Co., HP EEsof Div.

Driven by the explosive demand for bandwidth and the rapid innovation of new optical transport technologies such as dense wavelength-division multiplexing (dwdm), photonic design engineers and telecommunications managers face renewed challenges to meet performance, cost, and delivery targets. Recently, the first professional software design tools have come to market to help users cope with the increasing complexity and sophistication of optic designs.

The telecommunications industry in general and the users and providers of optical equipment in particular are experiencing revolutionary changes. The three main drivers for this transformation are the ever-increasing demand for communications services, rapid technological advances, and the changes in the competitive landscape brought about by deregulation. Traffic on the Internet is currently growing at triple-digit percentage rates annually. Web-based traffic is expected to surpass traffic on circuit-switched networks in less than four years.

These developments are already putting severe pressure on many telecommunications operators to upgrade the bandwidth of their existing fiber networks through technologies such as dwdm. The commercial introduction of erbium-doped fiber amplifiers in the mid-1990s and the more recent emergence of dwdm systems have not only effected significant improvements in performance (e.g., bandwidth throughput), but have also significantly lowered unit costs. In Europe and the United States alike, deregulation has considerably heightened the pace of the industry; many young and aggressive companies are springing up to build and supply advanced fiber-optic networks. They challenge incumbent players in their core business.

Design challenges

These trends have led to an increased complexity and sophistication of optical designs. However, the everyday challenges for managers and engineers to meet ever-stiffer performance, cost, and time-to-market targets remain. Telecommunications companies are faced with the task of integrating new transport technologies with their legacy infrastructure and management systems. Complex choices arise between different logical and physical architectures, such as 1:1 versus 1:N protection, as well as evolving new protocols and standards, i.e., Internet protocol (next-generation), Asynchronous Transfer Mode, and dwdm. These decisions will have great impact on current and future investment patterns.

For systems houses that need to develop these architecturally and technologically complicated solutions, the engineering challenges are tremendous. A cornucopia of trade-offs must be analyzed to achieve the optimal transmission characteristics. Power range, fiber non-linearities, temperature stability, gain shaping, dispersion management, and many other effects must all be taken into account simultaneously and balanced against each other.

Component manufacturers are expected to deliver double-digit annual price decreases to their customers while the process of making these components becomes more and more intricate. At the same time, quality and tolerance targets are growing tighter. For example, the explosion in components associated with many-channel dwdm systems will bring new levels of diversity and complexity to the manufacturing process.

Fundamentals of pda

Many problems associated with these challenges can be addressed with the help of software design tools, which have become an essential part of today`s development process in the semiconductor and electronics industry. Similarly, photonic design-automation (pda) tools have been developed for automating design in the photonics industry. In a wider sense, pda is a collection of design methodologies, software tools, and services used to engineer complex photonic networks or products.

pda may be used at various levels (see Fig. 1). Applications in network planning, system engineering, component development, and device design require very different levels of accuracy in the representation of network elements. The complexity may range mathematically from very simplistic models for the design of a whole network with tens of thousands of elements, to the numerically very intensive solution of Maxwell`s equations that may be required for the accurate design of a single device on the wave level.

Fiber-optic communications systems have evolved to a point where the nature of the physical problems associated with photonic modeling poses particular challenges. A hands-on approach has become prohibitive; parallel simulation of 4-wave mixing, amplified spontaneous emission (ase) noise, and polarization-mode dispersion cannot be done with spreadsheet software. Photonic designs require a particularly large simulation bandwidth.

Simulations of a 100-channel system, for example, may easily require a bandwidth in the tens of terahertz and several million data points if a reasonable statistic for bit-error-rate simulations is to be achieved. Consequently, the development of efficient algorithms and a careful choice of hardware platforms and simulation engines are of great importance. The validation of the resulting code is a crucial issue as well. Unfortunately, due to the complexity of the software and the related physical systems, experimental validation is, in general, very time-consuming and tedious.

Benefits and applications

pda will deliver value to users in at least four main applications:

serving as stand-alone simulators solving specific problems in either network, system, component, or device design;

providing an integrated approach to simulation, allowing a seamless combination of different simulation paradigms such as optical and electrical co-simulation;

allowing links to real and measured components and devices;

providing a basis for easy information exchange among industry participants through common standards for data formats.

Stand-alone simulators for networks, systems, or components have become common. Most of these tools have been developed for use by researchers at both commercial and non-profit institutions. The serious technological problems in tool development that were outlined here were often solved at these institutions through the use of outstanding and innovative approaches. Figures 2 and 3 illustrate two examples of typical problems that can be analyzed with state-of-the-art stand-alone simulators. System performance is readily estimated considering dispersion maps, ase, or 4-wave mixing. Simulations that have previously been difficult to perform become quite simple. Examples are the estimation of bit-error rates under changing temperature conditions or the assessment of statistical component tolerances on a whole system through Monte-Carlo simulations. Recently, a number of companies have come to market with suitable products, and it is likely that these tools will serve as work platforms to the user community.

Large increases in productivity can be achieved if design tools can bridge across several application areas. Simulation engines capable of integrated and layered simulations have been available for a few years. One example is Ptolemy, which was developed in a multiyear research effort by the University of California at Berkeley. Hierarchical simulation has found applications in many areas, including its use as a design platform for mobile communications networks. However, simulation capabilities for multiple levels of abstraction have so far found no role in photonics simulation. This is about to change with the introduction of the first professional photonic design tools. To illustrate the importance of this development, three areas are outlined for which integrative features appear most crucial:

The first area concerns integration along the signal path, i.e., the ability of a tool to allow the co-simulation of electrical and optical signals within a single design environment. Such a capability is essential in order to model the performance of a communications system accurately. For example, the noise characteristics of an electrical receiver can be the crucial factor in determining overall system performance. Tools with these important capabilities are already available (see Fig. 4). The first of these is the HP Photonic System Designer, which is based on Ptolemy. It combines the RF and microwave design capabilities of HP EEsof`s Advanced Design System with optical simulation technology and optics libraries from BNeD.

The second area is applications where simulations across different portions of the value chain must be covered. For example, a designer working at a component manufacturer may need to simulate the system performance of his product, or a network planner at a telecommunications company may want to estimate how a new point-to-point system will integrate into the company`s network. The individual technologies to create such tools already exist, and products with such capabilities are expected to emerge quite soon. These products must be tailored toward the different needs of network operators, system houses, and component suppliers. A prerequisite is common simulation platforms to allow integration. Model libraries will be needed that allow the use of different, accurate models as explained in Figure 1. A user can therefore easily adjust the desired level of detail and computing time.

The third area covers the ability to simulate concurrently across different network layers. For a simulation deep in the physical layers, it is often sufficient to use a synchronous approach to the simulation, i.e., the timing and order of signals are given at the start of the simulation. However, in the higher layers (i.e., link and network layers in the osi reference frame), the protocol-dependent "burstiness" of the traffic must be taken into account. A simulation engine like Ptolemy is well-suited to this kind of approach because it allows the incorporation of traffic-dependent "discrete" events (see Fig. 5).

An essential feature of pda tools will be their ability to integrate with actual network elements. This integration can be achieved through libraries that characterize the performance of actual, commercial components, similar to the device libraries found in PSpice. These "virtual data sheets" will allow engineers to use vendor-specific models rather than generic models, which may require a great number of free parameters to be set.

Another way that real devices can be incorporated is by basing the simulation on models derived from measured data. This method gives the best correspondence between simulation and real performance. To enable an easy transfer of measurement data to pda tools, it will also be important to have interfaces to test and measurement equipment at hand. The latest commercial products provide the first links from real devices to users.

There is a great need among industry participants for common data formats and standards so they can exchange product information and simulation results easily. For example, in large (system-integration) projects such as undersea cable construction, it should become possible for a purchaser to perform virtual vendor benchmarks. In such a process, electronic specifications and simulation results would be compared, thus reducing costly and lengthy evaluation periods in test beds and laboratories. However, open and common standards do not yet exist. It will probably be some time before any kind of data-exchange standard provides a common platform to the industry at large.

The pda industry is just beginning to take shape. Many of its future features and products can only be imagined at this time. However, it seems likely that professional tools will deliver great productivity increases to the user community. Many new advantages that could not have been realized previously are certain to emerge through pda for the benefit of end-users. u

Fig. 5. High-level network layers can be simulated by mixed-domain simulation engines such as Ptolemy.