Bob: The speech was fascinating.
John: Yes, the description of the Empire State Building was amazing, especially the huge number of rivets that was used in its construction.
Bob: 360,000, wasn’t it?
John: I thought it was more like 380,000.
Bob: Actually, I wrote it down on the back of an envelope, and the speaker said 365,400.
John: Well, I wrote it down too, and I was very careful to record the digits clearly on my table name card—they’re not smeared with mayonnaise like yours are. I have 384,500.
Even if the outcome of this conversation-soon-to-become-argument were important, the only ways to determine the actual number are to see the speaker’s notes or listen to a recording of what really was said. Bob and John need a reliable reference if they are to reduce the 5% uncertainty that separates them.
DWDM
What about achieving a few parts-per-million (ppm) uncertainty? This is the level of accuracy routinely required when dealing with dense wavelength division multiplex (DWDM) measurements. Because of the high cost of adding cable to a communications network, DWDM has become a very popular means of providing more capacity. And this is the wide-bandwidth capacity that is eminently suited to servicing the increasing demand for faster Internet access. A single fiber-optic cable that previously carried only one signal can handle 16, 32, or more by assigning a separate frequency band and a separate laser transmitter to each signal.
Single-mode fibers carry wavelengths between 1,530 and 1,605 nm, split into the C band from 1,530 to 1,565 nm and the L band from 1,570 to 1,605 nm. Two factors account for the extent and location of these bands. First, the ultrapure glass from which optical fibers are made traditionally has had a very small residual OH+ radical content. This impurity causes the fiber to exhibit so-called water bands or OH+ absorption bands around 1,000-, 1,400-, and 1,600-nm wavelengths. Data transmission is confined to the windows between these bands at 850, 1,300, and 1,550 nm.
Secondly, most of today’s erbium-doped fiber-optical amplifiers (EDFAs) operate in the C band. They are excited by pump lasers operating at 980 nm and can provide up to 40-dB gain. The large difference between signal and pump wavelengths allows the two to be easily separated at the output of the amplifier. EDFAs can have gain flatness within 0.5 dB or better across the entire C-band range.
A series of frequencies spanning both the C and L bands has been defined by the International Telecommunications Union (ITU). The frequencies are all separated by 100 GHz, corresponding to about a 0.8-nm wavelength differential, and represent the nominal centers of separate channels. The channels must not interfere with each other. If the tolerance is too great on an individual laser’s frequency, then maybe only half the ITU frequencies can be used with a 200-GHz spacing between each.
The simplest operation of a DWDM system assigns one signal to one wavelength, creating separate channels within a single optical fiber. However, this approach wastes bandwidth unless all signals have a constant, maximum data rate.
A recent study considered the efficiency of a packet over wavelength (POW) approach. Data rates were maximized for each DWDM wavelength by aggregating a large number of separate input signals. The purpose of the study was to determine just how many wavelengths really are required in a DWDM system. Given the assumptions made in the study, the conclusion was reached that as few as four different wavelengths would be sufficient.1
Because of inefficient wavelength assignment and the traditional restriction to operation within the C band, other researchers have proposed larger and larger numbers of more closely spaced wavelengths. Many input signals can be mixed, but present optical demultiplexing technology limits most practical systems to 64 or fewer wavelengths. Crowding many wavelengths into the 75-nm wide combined C and L bands puts more emphasis on accurate frequency control.
In a recent article, William Gornall, vice president of technology at Burleigh Instruments, said, “To add channels, some systems have reduced the channel spacing to 50 GHz, and 25-GHz spacing is not far off. They will require long-term stability of each laser wavelength to better than 10 pm.”2
Taking a different approach, Ghislain Lévesque, the product manager for optical spectrum analyzers at EXFO, commented, “It generally is accepted that system vendors will increase the number of channels by using a wider spectral range instead of lower channel spacing. For example, using the 1,300- to 1,650-nm spectral window with 0.4-nm channel spacing (50 GHz) gives 875 channels. This is why we are concentrating on the spectral window rather than the resolution.”
Also postulating the future availability of a much wider spectral window, a report on research conducted at Virginia Tech stated that the range of 1,260 to 1,610 nm is available for use now that fiber manufacturers have all but removed the water bands from the spectrum. The report included the assumption that fiber amplifiers and diode lasers probably will be developed within this band to completely fill it with usable bandwidth.3
A continuous band from 1,260 to 1,610 nm still is in the future, but the operating range of EDFAs already has been extended to the L band, opening up these wavelengths for DWDM use. However developments progress, the positioning of multiple wavelengths will remain critical to multiplexed system performance even if very high multiplex ratios can be avoided.
Seeing the Light
So, how do you accurately determine wavelength? To achieve the required low level of uncertainty, two basic types of measurement systems are used as well as laser and absorption-cell references.
OSAs
According to Stephen Blazo, a design engineer at Tektronix, “Diffraction grating-based systems are used in the majority of optical spectrum analyzers (OSAs). These systems, since they generally rely on mechanical movement of a grating, do not have the absolute accuracy that can be provided by an interferometer. But they do have a superior dynamic range and can be tailored for a specific application such as narrow or wide wavelength scan and fast or slow sweep.”
Figure 1 shows the basic operation of a single-pass OSA. To improve performance, light that has been diffracted once can be further split by reflecting it again from the same or another grating, prompting the OSA classifications of single-, two-, or four-pass. The penalty paid for achieving greater dynamic range in this way generally is higher mechanical/optical complexity and fragility. As a result, rugged OSAs intended for field use usually will have lower dynamic ranges than ones designed for laboratory use.
Mr. Blazo described some of the properties of a Tektronix OSA: “The instrument uses a proprietary four-pass optical system that can achieve a dynamic range of greater than 67 dB for lines separated by 100 GHz, greater than 60 dB for 50-GHz separation, and more than 50 dB for 25 GHz. The built-in acetylene absorption cell allows resolution of 10 pm and absolute accuracy of 20 pm.” He also cautioned that the instrument should be recalibrated if the room temperature changes by more than a few degrees or if the instrument is moved or bumped.
Michelson Interferometer-Based Wavemeters
Multiwavelength meters (MWMs) are based on an interferometer. An interference pattern is formed between two unequally delayed light beams split from the unknown output of a fiber-optic cable. Varying the delay between the two paths produces accurate measurements of power and wavelength peaks relative to each other.
Absolute accuracy is achieved by mixing the precisely known output of a HeNe laser with the unknown signal, as shown in Figure 2. Separate wavelengths are not clearly separated as in a grating-based OSA, but rather must be determined by examining the interference pattern. This task is performed by a digital signal processor.
“The interferometric MWM can achieve an accuracy of typically 3 pm over its entire spectral range from 1,450 nm to 1,650 nm,” said EXFO’s Mr. Lévesque. “Power uncertainty and resolution are 0.5 dB and 2 pm, respectively, in a current EXFO meter. This great wavelength accuracy has a price: the dynamic range of the MWM is limited to +10 to –30 dBm.”
Tektronix’s Mr. Blazo agreed. “The absolute accuracy of the MWM depends on the stability of the HeNe laser. This laser system has a long history of use and stability easily sufficient to allow the stated accuracy of 2 pm in a Tektronix MWM. If the need for even higher absolute accuracy arises, a stabilized HeNe laser can have a long-term stability of 0.1 pm or better.”
MWMs are inherently accurate because they compare the unknown input light with a reference laser output. OSAs must be calibrated immediately before making a measurement to achieve the highest accuracy, but they have a wide dynamic range. Table 1 compares the strengths and weaknesses of the two complementary types of instruments.
Reference Cells
Molecular absorption is the physical phenomenon underpinning the calibration sources provided by Wavelength References. Marc Davis, the engineering manager, said, “12Carbon acetylene and 13carbon hydrogen cyanide have been widely researched and identified by the National Institute of Standards and Technology (NIST) as primary standards for wavelength calibration in the DWDM band. The hydrogen-cyanide absorption spectrum covers the entire C band with calibration points spaced about 0.8 nm apart (Figure 3). The acetylene spectrum spans 1,510 nm to 1,540 nm, but the absorption is very strong and the cell is more compact.”
In addition to strong absorption lines, both types of cells exhibit less than 100 kHz/°C temperature dependence—that’s about 0.5 parts per billion/°C stability. Absolute accuracy of the reference is 0.3 pm, and the cells are intended for calibration of optical spectrum analyzers.
The width of the absorption lines can be varied by changing the pressure of the gas to match the equipment being calibrated. The minimum linewidth in a 100-torr (1 torr = 1 mm of mercury) cell is about 6 pm, although the absolute accuracy is much better than this. Consequently, these calibration cells are suitable for use in systems operating with 25-GHz or even 10-GHz channel spacing.
According to a Wavelength References application note, NIST has measured hydrogen-cyanide absorption bands to an uncertainty of 0.6 pm using 100-torr cell gas pressure [100 torr = 13,300 Pa (Pascals) = 0.13 atm (atmospheres) = 133.3 bar]. However, using a 10-torr pressure resulted in only 0.12-pm uncertainty because of the corresponding reduction in line width.
Although working with hydrogen-cyanide cells sounds dangerous, a standard Wavelength References cell contains only about 1.0 mg of the substance. This quantity is comparable to the 0.4 mg of hydrogen cyanide per liter of blood normally occurring in a healthy individual as a result of metabolism and does not represent a hazard in the unlikely event that a cell should break.
Lasers
A single line emission standard locks a distributed feedback (DFB) laser to a molecular absorption line. “This product provides a signal in the DWDM band for verification of wavemeters,” Mr. Davis explained. “An absolute accuracy of 0.2 pm can be achieved using a microprocessor-based locking technique that is quite robust as compared to earlier analog methods.”
The wavelength accuracy achievable without reference locking depends largely on controlling the temperature of the DFB laser module. For example, although DFB lasers are available with wavelengths corresponding to the ITU series frequencies, an individual laser may have a 0.1-nm/°C temperature coefficient.
According to a Lucent data sheet, “The wavelength of the [D2525Pxxx] laser can be temperature-tuned for more precise wavelength selection by adjusting the temperature of the internal thermoelectric cooler.” Typically, manufacturers are attempting to achieve 0.01°C control or about 1-pm wavelength stability.
DFB semiconductor lasers and traditional HeNe lasers are totally different types of devices. The characteristics of the HeNe lasers are inherently stable and very insensitive to environmental changes, resulting in their use as primary references.
References
- Bannister, J., et al, “How Many Wavelengths Do We Really Need? A Study of the Performance Limits of Packet Over Wavelength,” Optical Networks, Vol. 1, No. 2, April 2000, pp. 17-28.
- Gornall, W., “Extreme Photonics: Wavelength Measurement,” Photonics Spectra, February 2000, pp. 94-98.
- Cox, C., et al, Erbium Doped Fiber Amplifiers, a thesis project at Virginia Tech, April 27, 1998. dvorak.mse.vt.edu/faculty/hendricks/mse4206/
projects98/group08/index.html. - Vigot, S., “How to Choose the Right DWDM Instrument,” Application Note ANOTE016.1AN, EXFO, 1998.
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2000 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
August 2000