The broadband revolution is here.If you aren’t using streaming video, voice over IP (VoIP), and the Internet 24/7, you really have missed out.
Providing the very large bandwidth requirements of these types of services is the forte of fiber-optic networks. The actual optical fiber has almost unlimited capacity, and new ways of carrying increased amounts of data over existing fibers continue to be developed. For example, wavelength division multiplexing (WDM) or dense WDM (DWDM) effectively multiplies the number of communications channels by sending many wavelengths of light down each optical fiber.
The G.692 Recommendation on Optical Interfaces for Multichannel Systems with Optical Amplifiers, issued by The International Telecommunication Union (ITU), standardizes the frequencies of light to be used in DWDM systems. Spaced 200, 100, 50, or 25 GHz apart, they range from 197,100 GHz to 186,700 GHz, corresponding to wavelengths from 1,521.020 nm to 1,605.737 nm. This range includes both C and L bands, 1,528 nm to 1,565 nm and 1,566 nm to 1,606 nm, respectively. Even tighter spacing, down to 12.5 GHz and less, is being developed.
A related trend is increasing data rates. A few years ago, SONET’s OC-48 2.5-Gb/s data rate was the state of the art. Today, the use of 10-Gb/s OC-192 is growing, and OC-768 40-Gb/s systems are in the early stages of deployment.
But, higher data throughput is only part of the technical-economic equation. Network operators also would like to increase the length of fiber between amplifiers in new installations to reduce cost. Improved fiber and recently available high-power laser transmitters make this possible.
DWDM systems represent an elegant and apparently straightforward solution to the need for greater bandwidth. However, when the number of channels, the data rate, the fiber length, and the transmitter power are increased, what previously were second-order optical effects suddenly become very important.
Component Testing
None of the components used to build DWDM systems is completely linear. For example, fibers exhibit dispersion, which is defined as a change in group delay caused by a change in some other parameter. The two primary effects in fiber are chromatic dispersion, sensitivity of group delay to light wavelength, and polarization mode dispersion (PMD), the change in group delay caused by division of light between fast and slow polarization modes.
In addition to the fiber, transmitters, amplifiers, filters, multiplexers, gratings, receivers, and couplers suffer from various impairments. The demand for higher-performance networks requires more rigorous testing because component specifications have become critical. Systems simply won’t work if the dispersion is too high or the power imbalance among erbium-doped fiber amplifier (EDFA) output channels is too great.
Test procedures and test equipment and calibration also have been highlighted. If the test setup is incorrect, it’s very easy to measure the polarization loss of a connector rather than that of the device under test (DUT). This problem is even more acute when small effects and cross-dependencies are being examined. The input that drives most DWDM component testing is the light from a laser source.
Tunable Laser Sources
When DWDM channels were widely spaced, broadband sources could be used to characterize components. An optical spectrum analyzer (OSA) provided the required wavelength discrimination.
The progression to tunable laser sources (TLSs) was described by Karl Merkel, an application engineer at Agilent Technologies’ Optical Communication Measurements Division. “As DWDM channel spacing decreased and the width of the filters in passive components narrowed, OSAs no longer could provide the necessary wavelength resolution. In addition, narrowing the resolution bandwidth (RBW) effectively reduced the source power, so the dynamic range was greatly limited in high-resolution applications,” he said.
“Today, testing passive DWDM components is the largest use of high-performance TLSs,” Mr. Merkel continued. “The use of bit rates up to 40 Gb/s means that the dispersion characteristics of the fiber, amplifiers, and even the passive components must be tested as well as the loss vs. wavelength.”
A TLS provides a precisely positioned and very narrow spectral line width. Unfortunately, TLSs have a background noise level caused by spontaneous emission. Not all of the electrons excited to higher energy levels are stimulated to emit photons. Instead, some portion of the electrons will spontaneously emit photons as they return to the ground energy state.
The integral of this random, broadband noise power across the tuning range is the total source spontaneous emission (SSE), a basic limitation to the measurement dynamic range (Figure 1). This is the same phenomenon responsible for the background noise in EDFAs, termed amplified spontaneous emission (ASE).
By following the DUT with an OSA, for example, the measurement bandwidth can be reduced to the OSA’s resolution bandwidth (RBW). This means that less spontaneous emission is included in the measurement, but because the TLS line width is very narrow, typically all the signal power is included.
The signal-to-SSE ratio (S/SSE or STSE) usually is footnoted to highlight the OSA RBW, for instance, S/SSE (0.1 nm). The TLS manufacturer’s goal is to sufficiently reduce SSE so that an OSA is not required in the test setup. Generally, S/SSE will degrade as the tuning range widens, so a narrow range often will be quoted in a specification footnote.
As a laser’s line width becomes smaller, the coherence of the light becomes greater. One consequence of this effect can be increased sensitivity to backscattered light or reflections from a connector at the laser output, leading to instabilities. An isolator may be required after the laser to correct the problem, and many TLSs have one built in. A broad-spectrum source exhibits low coherence and typically is insensitive to reflections but cannot characterize narrowband devices.
Generally, coherent sources will exacerbate whatever polarization dependencies may exist within a DUT. So, if significant polarization mode effects are suspected, for example in an insertion-loss measurement, it may be better to use a broadband uncorrelated LED source followed by a narrowband detector such as an OSA. Alternatively, many TLSs provide a coherence control that reduces coherence by broadening the output line width.
“It is interesting to note that using a source with a very narrow line width does have some important implications if you want to be certain of test-system accuracy,” commented EXFO product manager Marie-Hélène Côté. “The test system has to satisfy the Nyquist criterion to avoid aliasing the data. This translates into some precise requirements for the system’s optical bandwidth, which includes the laser line width and the effects of the detector’s electrical bandwidth.
“The sampling rate of the system must permit at least two samples inside a band corresponding to the system’s optical bandwidth,” she continued. “As an example, EXFO’s IQS-12004B integrated test system uses a fiber laser with a typical 12-pm line width and a 5-pm sampling step.”
Lasers also are characterized by the amount of optical power they produce. Higher output power in a test instrument means that you can run more than one test in parallel by appropriately splitting the light output. Or, you may need a high-power level just to simulate a long-haul transmission system.
Achieving low S/SSE usually involves reducing output power, although this dependency has been improved in new models. A TLS may provide a separate low-SSE output port optimized for SSE rather than power. Another port on the same instrument may supply 10 dB greater power, for example, but with higher SSE.
Minimum output power also depends on the tuning range. For example, consider the Agilent Technologies Model 81640B TLS with a 145-nm tuning range. Minimum power from the low-SSE port is guaranteed to be ³-7 dBm from 1,520 to 1,610 nm but only ³-13 dBm over the complete 1,495- to 1,640-nm range.
Distributed Feedback (DFB)
The DFB semiconductor laser is based on a distributed Bragg grating etched onto the active layer. A tuning range of ±1 nm can be achieved by varying the operating temperature and, to a lesser degree, the drive current. A large number of DFB lasers are required to simulate a DWDM system.
You also would need a multiplexer wide enough to accommodate all the channels and an overall means of power control to ensure that all frequencies are at the same power level, for example, for EDFA testing. Several manufacturers provide modular optical test instruments that allow you to build such a test system.
Laser diodes often are modulated by controlling their drive current. But at high data rates, this creates an undesired chirp in the output light frequency. Consequently, most high-frequency modulation testing, 10 GHz and above, is done with external cavity lasers and separate modulators.
External Cavity Lasers (ECLs)
As the name implies, ECLs are mechanically tuned lasers. They are designed to have the wavelength range required to test DWDM communications components and to be stable, low noise, and compact. Two other characteristics also tend to be at the top of ECL datasheets: freedom from mode hopping and fast sweeping.
Many closely spaced modes exist simultaneously below the threshold of sustained stimulated emission, and none is dominant. Above the threshold, one mode becomes dominant and is characterized by near monochromaticity and high coherence. A very large number of wavelengths of this frequency exactly correspond to the cavity length, other modes having been suppressed. A small change in operating conditions can cause mode swapping: a sudden shift from one output wavelength to a slightly different one.
The TLS traditionally has been a bottleneck in automated optical component testing, according to Graham Sperrin, marketing manager at Anritsu. “With the introduction of faster scanning, support circuitry has been developed to prevent abrupt mode hopping. An active mode-hop suppression-control system detects any wavelength drift while in the fast scan mode and uses this feedback to control the external cavity length.”
The importance of mode hopping and fast tuning was further explained by Michael Minneman, president of dBm Optics: “In passives test, the driving factors are speed of wavelength sweep, the S/SSE ratio that limits the depth of filter [stop band] that can be characterized, and the capability of the laser to be truly mode-hop free.
“Virtually no laser sold over the last 10 years is always mode-hop free over its entire wavelength range,” he continued. “The 15-pm to 40-pm jump in wavelength that occurs during a mode-hop event creates major problems. Today, there are measurement systems that detect and correct wavelength mode hops.”
Fiber Laser
A fiber laser is similar to a Fabry-Perot laser since hundreds of closely spaced modes exist simultaneously. There is no attempt to establish single-mode operation.
EDFAs and fiber lasers are closely related. In both cases, electrons are excited to higher energy levels by pumping an erbium-doped fiber with light from a separate laser diode. In the case of the EXFO FLS-2600B Fiber Laser, this is done at 980 nm in the reverse direction. One design of fiber laser uses a ring cavity defined by a thin-film interference bandpass filter and a fiber-optic splitter that couples power to the output (Figure 2).
Like an EDFA, the fiber laser has an inherent amount of ASE (SSE) across a relatively wide bandwidth. However, because fiber lasers contain a tunable filter, the amount of broadband emission at the output is reduced, giving rise to a high S/SSE ratio. This means that wide dynamic range measurements such as the ratio of power in occupied to unoccupied DWDM channels can be made.
Benefits of a fiber laser include insensitivity to parasitic etalon effects. The large line width corresponds to a medium coherence length of 15 cm to 30 cm, considerably less than the typical 1-m to 2-m fiber pigtail or patchcord length used during testing. This means that an etalon possibly formed by a small residual reflection at a connector and another several meters away at the surface of a photodetector would have no effect.
On the downside, the line width of the EXFO FLS-2600B Fiber Laser Source, for example, is about 1.3 GHz, 1.3% of the 100-GHz ITU channel spacing or 2.6% of the 50-GHz spacing. Reducing the line width so the laser is suitable for 25-GHz and 12.5-GHz systems requires more development. Also, currently a fiber laser cannot provide greater than about +5 dBm output power, limiting its test applications.
“Fiber lasers are optimized for most passive component testing,” added EXFO’s Ms. Côté. “For applications where high-frequency modulation is necessary, such as in the simulation of a transmitter in system testing, a fiber laser is not suitable. Being a multimode laser, modulation induces beats along with the signal, which renders measurements impractical.”
Further Considerations
As shown in Table 1, each type of laser presents a different set of relationships among the main parameters. For example, the wide tuning range and very narrow line width of an ECL are offset by its cost and relatively low power. A fiber laser has a much greater line width, but the cost is lower than that of an ECL.
Highlighted cells denote positive attributes—a fiber laser’s high S/SSE or an ECL’s large tuning range. However, the values listed typically are worst-case numbers and may improve under different operating conditions. For example, the EXFO FLS-2600B output power increases to 0 dBm if the tuning range is changed to 1,515 nm to 1,610 nm instead of 1,510 nm to 1,612 nm because of the rapid drop-off in gain of the EDFA bandpass characteristic. Similar qualifications exist for the other listed parameters.
As Mark Wippich, product marketing manager for network tunables at New Focus, commented, “The DFB laser should be considered for network testing because it is cost-efficient. However, it is slow and doesn’t tune over a wide range. The fiber laser is useful for component testing. It tunes over a wider range and has lower noise than the DFB, but its line width is too large to characterize narrowband passive components. The ECL has fast tuning for high test throughput and provides flexibility within a network application, but it costs more than the other types.”
Tuning speed obviously relates to test throughput, but faster is not always better. In open-loop test systems, the linearity of the TLS’s wavelength vs. time tuning characteristic affects accuracy. According to Patrick McCormick, engineering manager at ILX Lightwave, “ECL products typically use a continuous sweeping motor that generates triggers at specified time intervals. The motion is not stepped. This presents timing-related challenges on the sense side of the measurement.
“The delays for both the trigger processing and the associated delay of the measurement must be accounted for to ensure that the power measurements are accurately associated with their respective wavelengths,” he explained. “Using a stepped approach, where the tunable laser waits for the sensor measurement to be completed, will minimize the error. Micro electromechanical systems (MEMS) technology used in two ILX lasers provides stepping capability and high accuracy but over only a 40-nm range.”
Some open-loop TLSs specify high absolute accuracy but with a footnote calling for user calibration and constant-temperature operating conditions. More recent closed-loop test systems rapidly measure the TLS output power and wavelength as each test progresses. They have an onboard reference so absolute accuracy can be guaranteed.
Duwayne Anderson, an applications engineer at Textronix, referred to this type of system in his comments. “Although wavelength accuracy and power stability are important to measurement results, they are not really that important for test systems,” he said. “Swept wavelength meters working alongside the lasers calibrate the wavelength, and power monitors correct for power stability.”
Reference
Guide to WDM Technology Testing: A Unique Reference for the Fiber-Optic Industry, EXFO, Quebec City, 2000.
Acknowledgements
These companies provided material for this article:
Agilent Technologies
www.rsleads.com/208ee-191
Anritsu
www.rsleads.com/208ee-193
dBm Optics
www.rsleads.com/208ee-195
ILX Lightwave
www.rsleads.com/208ee-194
Tektronix
www.rsleads.com/208ee-192
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2002 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
August 2002