Measure environmental stress to improve manufacturing yield

April 1, 2000

Joachim Vobis

Dense wavelength-division multiplexing (DWDM) technology imposes tough performance requirements on system components. Possibly the most demanding component in a DWDM system is the optical demultiplexer at the end of the transmission because it has to separate the individual channels.

The "black box" parameters describing its performance are independent of the technology of the device, whether it is realized in integrated optics (phased array), in fiber (fiber grating), or in bulk optics. Add/drop elements and many other filter devices have characteristics that are similar to or are a subset of a demultiplexer.

Most filter measurements start with the bandwidth peak or highest point on the measurement curve, which represents the minimum insertion loss. The x-dB bandwidth is determined as the wavelength difference between the points on the left and right of the filter slope that are x dB down from the peak. The bandwidth at that point (typically 0.2 to 3 dB) defines the passband and the band edges (see Fig. 1).

An ideal filter slope would be symmetrical and flat at the top. Real devices and tiny measurement uncertainties usually cause a peak or pit search to be somewhere in the passband but not necessarily at its center. Therefore, the center wavelength is often calculated as the mean value of the band edges.

Flatness is the peak-to-peak variation of the insertion loss in the wavelength range over which a channel is allowed to operate. This range typically is approximately 20% of the channel spacing (note that the passband of a device must be at least this wide). The steepness describes how fast a filter slope drops from 3 dB down to 20, 30 or even 40 dB. In fiber gratings, it is often required to characterize the falling slope with very high wavelength resolution (in less than 10-pm steps) to detect residual ripples on the filter shape. These ripples can be a measure for the quality of the grating.

In integrated optics with its nonsymmetrical waveguides, components often exhibit strong polarization dependence. In a demultiplexer, this polarization sensitivity shows up as a shift of the center wavelength and/or a variation of the filter bandwidth. As it is not possible to keep the state of polarization constant over the entire optical network, the polarization sensitivity must be made small enough that no critical signal degradation is possible.

Another important parameter for a demultiplexer is the crosstalk between channels, sometimes called channel isolation. It indicates the amount of power that is coupled onto the signal channel from all neighboring channels. As the receiver is a broadband element (PIN diode or avalanche photodiode), it cannot distinguish between bits at the wavelength of interest and bits coming from other channels. If the crosstalk is too high, the bit error rate will increase. Today most demultiplexers have crosstalk figures of at least 25 dB. For more-complex networks, >45 dB may be necessary.


Optical components (0/0 devices) are characterized using a light source and a receiver. Usually the source is characterized versus wavelength using a "zero loss" device such as patchcord, and then the device under test (DUT) is measured. The two results (traces or data arrays) are subtracted to calculate the insertion loss as a function of wavelength.

In recent years, several stimulus/response setups have evolved, each one with its own advantages and drawbacks:

Broadband source and swept receiver. The amplified spontaneous emission (ASE) from an erbium-doped optical amplifier without input signal provides high power density over the wavelength range for which most DWDM components have been specified (for other wavelength ranges, broadband light sources incorporating edge-emitting LEDs can be used). Optical spectrum analyzers (OSAs) can acquire traces faster than most sources can sweep over a given wavelength range. In conjunction with their built-in application to characterize passive components, they are well suited for manufacturing testing (see Fig. 2).

The measurement range depends on the source density and the OSA resolution bandwidth (RBW) and sensitivity. A typical scenario might be an ASE source operating at 0.1 mW/nm and an OSA with 0.1 nm RBW and -80 dBm (10 pW) sensitivity. In this case, a device with up to 60-dB loss can be measured and still have acceptable throughput.

Swept source and broadband receiver. The narrower the channel spacing of the DWDM system, the more precisely the fine structure of a component must be measured. A tunable laser source in conjunction with a power meter is best solution for this problem. Because the tunable laser is extremely narrowband, very fine details can be resolved. Particularly if a multiwavelength meter is added, this combination allows you to scan steep slopes or edges with approximately 1-pm resolution and absolute accuracy of only a few picometers.

Sweeping a tunable laser can be slower than an OSA sweep. However, multiple devices or multiple ports of the same device can often be measured by splitting the powerful laser signal with a star coupler and then using several power meters in parallel. In fact, the latest generation of tunable laser sources accepts several single or dual power-meter sensors in their mainframes. With additional mainframes, this concept can be extended to more than 32 parallel power sensors.

Swept source and swept receiver. A combination of a tunable laser and an OSA provides the best selectivity. They are commonly used for characterizing optical amplifiers but, in limited cases, they can be used for select passive components that have extreme loss ranges. For example, a notch filter cuts out only a small portion of the spectrum and passes on most of the remainder. Residual spontaneous emission from the laser source therefore can limit how deep the notch can be examined. The OSA filters this "offset light." As a result, components can be measured with less than 1-dB insertion loss but more than 80-dB loss in the notch.

Setups including polarization controllers. To characterize the polarization-dependent loss of a device, a polarized source (such as a laser) or a polarizer inserted immediately after a broadband source is required. Then a polarization controller can randomize the state of polarization. The response system (OSA or power meter) then measures the device under test (DUT) multiple times.

Because the DUT is now exposed to many polarization states, it eventually will show its best-case and worst-case loss versus wavelength. Of course, the receiver must have less polarization-dependent loss than the DUT. High-end power sensors in conjunction with a tunable laser source (which is always polarized) are possibly the most accurate alternative.

testing STRESSES

Most devices are designed to be stable even if the ambient temperature, humidity, or pressure changes over time. An add/drop filter, for example, may vary any of its key parameters: center wavelength, bandwidth, insertion loss, crosstalk, or polarization dependence. Furthermore, even permitted small changes should reverse themselves when the environmental conditions return to normal.

If not, then the device may drift out of its specifications within its expected lifetime. This condition sometimes is more difficult to meet over the storage-temperature range (where the components must survive but not necessarily meet specifications) than over the operating-temperature range.

In this instance, the ASE/OSA setup is quick and easy to use. Particularly with a built-in passive component analysis, a technician can manually characterize and document a small series of components during product design or manufacturing without much instrument or programming knowledge. For large quantities, however, it may pay to invest in optical switches and remote software that can control other equipment as well, such as an environmental chamber.

Alternatively, a single tunable laser source signal may be split many times and then measure all components in parallel. Although it may take longer to scan a wavelength range, this approach provides the results for an entire series of DUTs. Because the measurement must be repeated often to cover wide temperature, humidity, and pressure ranges, it pays to minimize the number of measurements as well as the time to do them.

A comprehensive characterization at room temperature may be the chosen approach, for example, but only a few wavelength points selected to which a laser must be tuned (on an OSA, the trace length must be reduced to the smallest number of points that still provide good results within the required accuracy). If data at a particular environmental condition indicate a noticeable shift in the parameters of the DUT, then it might make sense to characterize the device in full.

Polarization dependencies are already time-consuming to measure because of the time required to randomize the light over a sufficient time period. If a filter is very stable-it has practically the same center wavelength, bandwidth, and crosstalk at many different temperatures-then it may be sufficient to test its polarization dependence only at room temperature and at its minimum and maximum temperatures.

OPTImizing choices

As the variety and volume of passive optical components rapidly expand, careful characterization of their properties is essential in building large optical networks. Today`s test solutions have been optimized to meet both accuracy and throughput requirements.

Newer OSAs allow automatic comparison of actual DUT performance with specifications, thereby providing convenience of measuring a series of devices without any programming. Alternatively, a tunable laser with multichannel power meters enables parallel testing of either many components or many outputs of a component. Technical differences in the setups will decide which advantages count most in a given application.

Two documents explaining measurement techniques and ways to solve real-life application issues are available from Agilent Technologies: "The DWDM Component Test Guide" (Literature #5965-3124E) and "Eight methods to characterize passive optical components" (Literature #5968-6304E); in the USA, call 1-800-452-4844.

Joachim Vobis manages the applications engineering team at Agilent Technologies` Lightwave Division, 1400 Fountaingrove Parkway, Santa Rosa, CA 95403. He can be reached at [email protected].

Correction: In the February issue of WDM Solutions, on p. 44, we inadvertently printed the wrong company address and telephone number for Fermionics Lasertech, which has recently relocated. The correct address is 1153 Lawrence Drive, Newberry Park, CA 91320; the telephone number is 805-375-0999. The e-mail address for company president Yetzen Liu ([email protected]) remains the same. Our apologies for any inconvenience caused by the error.
FIGURE 1. Common filter characteristics of a demultiplexer that must be measured during manufacture include bandwidth, band edges, passband, center wavelength, insertion loss, crosstalk, and polarization-dependent loss.

FIGURE 2. During manufacture of passive components, an OSA can be used with a broadband source or swept source to characterize devices in terms of environmental stresses.

Sponsored Recommendations

Coherent Routing and Optical Transport – Getting Under the Covers

April 11, 2024
Join us as we delve into the symbiotic relationship between IPoDWDM and cutting-edge optical transport innovations, revolutionizing the landscape of data transmission.

Scaling Moore’s Law and The Role of Integrated Photonics

April 8, 2024
Intel presents its perspective on how photonic integration can enable similar performance scaling as Moore’s Law for package I/O with higher data throughput and lower energy consumption...

Connecting the world, one fiber at a time

Dec. 12, 2023
The world runs on ethernet: the future of higher speeds

Supporting 5G with Fiber

April 12, 2023
Network operators continue their 5G coverage expansion – which means they also continue to roll out fiber to support such initiatives. The articles in this Lightwave On ...