Standardized parameters for AWGs would ease system design


Vivek Tandon, Mark Volanthen, Marcel van der Vliet, and Jim Bonar

Economical and compact, arrayed waveguide gratings have found their niche in high-channel-count WDM systems. Standardizing characterization measurements and performance definitions helps accurately calculate system effects and performance.

To date, the most common filters used to multiplex and demultiplex wavelengths in a DWDM system have fallen into two categories: those based on thin-film filter technology and those based on fiber Bragg gratings. However, recent reports from industry analysts predict that the market size and demand for AWGs will exceed those of thin-film filters and fiber Bragg gratings.1 The relatively low insertion losses, lower cost per port, and small size of AWGs are advantageous and particularly applicable to high-channel DWDM systems (typically more than 16 channels). While a thin-film filter or fiber Bragg grating must be configured for individual wavelengths, an AWG can multiplex and demultiplex many different wavelengths simultaneously.

Arrayed waveguide gratings also exhibit low dispersion, which is of increasing importance as DWDM systems move toward higher system line rates (such as 10 and 40 Gbit/s), thereby requiring higher dispersion tolerances. Dispersion issues intensify as multiple filters are concatenated through the system architecture.2

In general, manufacturers must standardize specification definitions to speed the successful adoption of new technologies with multiple suppliers. This need for standardization is particularly true for the characterization measurements and optical performance definitions used when specifying AWGs, since they have implications on the end-system performance and the optical network-link budgets.

Product information sheets and web pages of AWG vendors typically display spectral response graphs. While the graphs readily display some crucial information about the AWG, such as the insertion loss and crosstalk, they often do not sufficiently describe certain parameters, such as the adherence of the channels to the International Telecommunications Union (ITU) grid. As a result, the information can be misleading. In addition to the graph, product data sheets often provide detailed performance-specific actions of the device (see table).

Based on discussions with DWDM network and system designers, we believe there is confusion when comparing AWG devices from different vendors, stemming from the fact that performance parameters have not been universally defined. For example, one company may measure insertion loss as being the minimum loss of the lowest-loss channel; another may define the same insertion-loss parameter to be the maximum loss in any passband, of any channel. The definitions of specifications here have been selected to most fully enable network and system designers to determine the effects of AWG devices on the system-link budget and the overall system performance.

Put simply, an AWG behaves like a diffraction grating and, when used as a demultiplexer, splits an incoming light beam into a series of different wavelengths. Conversely it can be used in the reversed form, as a multiplexer, to combine different wavelengths into a single beam of photons. The design of the AWG is critical to ensure that excellent resolution enables the device to offer frequencies on and off the standard ITU grid. The ITU recommends a channel spacing of 100 GHz, but channel spacings wider (200 GHz) and closer (50 GHz) can also be created.

The function of an AWG demultiplexer is to direct wavelengths launched at the input waveguide to different output waveguides. An AWG can also be used in its reciprocal form as a multiplexer, to combine wavelengths launched into several input waveguides into a single output waveguide.3, 4

An AWG consists of one input waveguide, several output waveguides, and two star couplers connected via an array of waveguides (see Fig. 1, left). Light is confined to the waveguides, but diverges in the star couplers. The star couplers are analogous to lenses and the array of waveguides analogous to a prism, causing wavelengths to be deflected differently (see Fig. 1, right). Depending on their wavelength, each is focused at a different position along the output edge of the second star coupler. Each output waveguide transmits a particular wavelength. By changing the design of the array, the spacing of the output waveguides, and the length of the star couplers, AWGs can be made with a variety of channel spacings, passband widths, and crosstalk values.

The insertion of an optical component into an optical system is a tradeoff: while the component may perform a positive function, it often leads to system-performance degradation. In the case of an AWG, this effect manifests itself as bit-error rate degradation and reduced system power as a result of loss, crosstalk, and dispersion inherent in the AWG. When determining the suitability of an AWG for a particular application, a number of parameters relating to both the amplitude and phase response of the device must be considered, along with other parameters that characterize the performance of an AWG. To appreciate the significance of the values, one has to understand how the parameters are measured.

The starting point when defining the performance of an AWG is to look at its amplitude response for both the transverse electric (TE) and the transverse magnetic (TM) polarization states. Alternatively, the maximum and minimum transmission points of each polarization at each wavelength can be used.

There are a number of methods to obtain the amplitude response of an AWG. The most common is to connect a tunable laser and a polarization controller to the input waveguide and one or more power meters to the output waveguides. The laser steps or sweeps across the wavelength range of interest while the power meters take readings at the required wavelengths.

Measurements of TE and TM can be obtained by setting the input polarization state to TE and then to TM using the polarization controller and recording one scan for each state. A number of measurement equipment vendors supply modules or even complete measurement systems similar to the configuration described previously.

Alternatively, the maximum and minimum response curves can be obtained using the Mueller method.5 The Mueller method is the slower of these two techniques because it requires the laser to scan four times, rather than two. However, unlike the first method described, the absolute state of polarization does not need to be controlled.

All but the last four parameters in the table can be obtained from either the TE and TM curves, or the maximum and minimum response curves. Commercial equipment is available to measure these four parameters.6 The measurements of separate TE and TM, or maximum and minimum response curves, are adhered to by most AWG vendors. However, the way vendors specify the performance of a device from these curves differs widely.

As an example of the confusion faced when evaluating the performance of AWGs, three different definitions of polarization-dependent loss (PDL) are all in use (see Fig. 2). One definition specifies PDL at the ITU center wavelength; another specifies worst-case PDL across the entire passband; and another specifies the difference between peak transmission of the TE and the TM polarization states. There can be a large difference between each of these values, but the correct definition should be the PDL across the passband.

Similar confusions can be found in other parameters. Insertion loss can be quoted as the minimum loss of the lowest-loss channel, or the maximum loss in any passband of any channel, often with a difference of 2 dB or more in the specification for the same device. Crosstalk values are sometimes calculated at the ITU wavelengths and other times over the passband. At times, crosstalk measurements are referenced to the peak transmission, and other times to the minimum transmission in the passband. Further confusion arises from the way vendors include the effects of polarization and channel-spacing accuracy.

Defining parameters such as channel-spacing accuracy is not that important to network designers; they need to know the performance of a device over all the wavelength bands used in the system. Furthermore, network designers should know the worst-case values for every parameter necessary to determine the system-link budget.

A set of customer-oriented definitions is presented (see "Some definitions are becoming de facto standard," p. 40) to enable DWDM network designers to easily evaluate the suitability of a device for their applications. These definitions incorporate the concept of a "clear window," as already used by the majority of AWG vendors and network designers, to define the device performance.

The clear window is defined as a band of wavelengths around each ITU center wavelength. When quoting the performance of an AWG, all specifications must be the worst-case values met within the clear window of every channel and for both polarization states.

Since the specifications quoted are the worst-case values within any clear window of the device, all values quoted are guaranteed for all the usable wavelengths defined by the window. Consequently, there is no need to calculate the effects of passband shape, channel-spacing error, or polarization because these variations are already accounted for in the specifications.

The width of the window is chosen to be the range of wavelengths that the signal could occupy, accounting for the bandwidth of the signal and drift in the wavelength of the laser. For a 100-GHz demultiplexer, a window of 25 GHz is typical. Some specifications are sensitive to the width of window chosen; therefore, when comparing two devices with an identical set of definitions, as a final check, one should ensure that the widths of the clear windows are the same.

When evaluating AWG specifications for an application, a system designer needs to know not only the values of the specifications displayed in the table, but also the definitions of all parameters, the size of the clear window used, and that the clear windows are centered on the ITU grid.

Vivek Tandon is vice president of marketing and business development, Mark Volanthen is product marketing manager, Marcel Van der Vliet is a project leader, and Jim Bonar is product manager at Kymata, Starlaw Park, Starlaw Road, Livingston, EH54 8SF, England. Contact them at 44-1506 426000 or


  1. RHK, Startrax 2000 Conf. (Palm Springs, CA).
  2. C.Caspar et al., Proc. ECOC (1997).
  3. M. K. Smit, Elec. Lett. 24, 385 (1988).
  4. M. K.Smit and C. Van Dam, IEEE JSTQE 2, 236 (1996).
  5. Hewlett Packard Product Note (1996).
  6. D. Derickson, Fibre Optic Test and Measurement, Prentice Hall (1998).


Some definitions are becoming de facto standard
Unfortunately for network and system designers, while AWGs are being increasingly deployed in DWDM optical networks, there are few standards being followed when specifying the optical performance of these passive components. The following set of definitions is increasingly being adopted as the de facto standard by AWG vendors and is the most informative and useful for network designers to identify the impact of an AWG on the system-link budget.

For specifications that apply to a single channel, the value quoted should be for the worst-case channel. The definitions require the response of the device to be obtained for the TE and the TM polarization states, or alternatively, for the maximum and the minimum transmissions at each wavelength.

Insertion loss of a device is the minimum transmission within any clear window for all polarization states. It represents the worst possible loss through the device (see Fig. A).

Insertion loss uniformity of a device is the difference between the insertion loss of the best-case and worst-case channels.

Polarization-dependent loss of a device is the maximum that the transmission can vary over all polarizations at a fixed wavelength over the entire clear window. This parameter in particular is most often misrepresented in product specification sheets.

Passband uniformity is the maximum that the transmission can change with both polarization and wavelength within the clear window of the selected channel.

Adjacent crosstalk of a channel is the highest transmission within an adjacent clear window, referenced to the lowest transmission within the selected channel clear window.

Nonadjacent crosstalk of a channel is the highest transmission within a nonadjacent clear window, referenced to the lowest transmission within the selected channel clear window. The highest and lowest transmissions are determined for any (possibly different) polarization states within each clear window (see Fig. B).

Maximum integrated crosstalk of a channel is the sum of the maximum crosstalk values from all other channels. This occurs when all signals independently align with the wavelengths and polarizations of maximum crosstalk. Maximum integrated crosstalk is most applicable to very narrow spectral signals, because all of the optical power must be precisely at the wavelengths of maximum crosstalk.

Average integrated crosstalk of a channel is the sum of the average crosstalks from all other channels. The average crosstalk of a channel is the mean of the maximum transmission in the clear window of that channel (as shown in green in Fig. B). Crosstalk values are referenced to the mean of the minimum transmission in the chosen channel clear window. Average integrated crosstalk represents the worst-case total crosstalk that could occur for a signal with its power distributed uniformly across the clear window. It is a more appropriate measure of crosstalk for high-bit-rate signals because these higher bit rates cause the power in the signal to be spread across a wider spectrum.

Chromatic dispersion is the worst-case dispersion within any clear window for all polarization states of a device.

Differential group delay is the maximum difference in-group delay between all polarization states within any clear window.

Return loss of a channel is the maximum reflected signal within the clear window of the selected channel for any polarization state, referenced to the incident signal level. Note that this parameter includes the input channel in addition to all output channels.

Directivity of a channel is the maximum signal from that channel measured at any other than the selected channel within any clear window for all polarization states, referenced to the incident signal level.

More in Components