dwdm requires accurate measurements
The right equipment is only part of the battle in light of dense wavelength-division multiplexing?s tight tolerances.
Dennis Derickson Hewlett-Packard Lightwave Div.
Fiber-optic links that use dense wavelength-division multiplexing (dwdm) have proceeded from laboratory experiments to commercial use in a short period of time due to the introduction of erbium-doped fiber amplifiers (edfas). International standards bodies are converging on a standard dwdm channel grid with wavelengths spaced at 100 GHz (about 0.8 nm) above and below 193.1 THz (1552.5 nm).1 These channels are chosen to fall within the gain bandwidth of the edfa between 1530 and 1562 nm.
The popularity of dwdm has caused optical component and system manufacturers to specify wavelength accuracy measured to as low as 0.001 nm. These tight accuracies are necessary to ensure system robustness when the worst-case conditions of the optical component wavelength tolerances converge.
High-accuracy wavelength measurements are also required for the proper operation of fiber-optic networks. Wavelength monitors embedded in telecommunications networks are used to monitor channel wavelengths and power to make sure all channels are functioning properly.
These new measurement requirements have caused fiber-optic test and measurement equipment designers to make significant improvements in high-accuracy wavelength measurement techniques. This article will discuss several important techniques that have been developed to allow wavelength measurements with wavelength accuracies on the order of 0.001 nm at 1550 nm.
The instrument most often used for measuring wavelength is the optical spectrum analyzer (osa). The most common type of osa uses a diffraction grating (see Fig. 1). The input fiber is collimated by a mirror and applied to a diffraction grating, which produces a beam whose output angle is dependent on the optical wavelength. The wavelength tuning of the osa is controlled by the rotation of the diffraction grating. The angular position of the grating determines the wavelength that is focused directly at the center of the output slit. The size of the focused spot and the slit determine the resolution of the osa. To sweep across a given span of wavelengths, the diffraction grating is rotated, with the initial and final wavelengths of the sweep determined by the initial and final angles.
To provide accurate wavelength tuning, the diffraction-grating angle must be precisely controlled with extremely small angular resolution (for a 0.001-nm wavelength step, a typical diffraction grating moves by 0.000045°). The angle must also be repeatable over time and temperature. As long as the grating position is repeatable, the absolute wavelength accuracy of the grating-based osa can be accurately calibrated.
Optical encoder technology with interpolation techniques is used in grating-based osas to allow repeatable grating positions that have very fine angular resolution. Even the best grating-based osas are not capable of making sufficiently accurate wavelength measurements unless they are periodically calibrated with a highly accurate wavelength reference source.
Wavelength calibration techniques
An osa can be calibrated by comparing its measurements to a source with a well-known wavelength. Figure 2(a) lists some available laser reference sources that are useful for osa calibration.2-5
The helium-neon laser is a convenient reference source that is available at several wavelengths. The 1523-nm helium-neon laser line is the nearest to the dwdm communications band. Semiconductor lasers can also be used if they are calibrated against or locked to a high-accuracy wavelength reference.
Single-point calibrations allow very accurate measurements near the calibration wavelength, but errors start to accumulate as osa measurements are taken further from this point. Alternatively, a wavelength-tunable laser can be calibrated with a wavelength meter and swept over a range of frequencies. Wavelength meters are the optical equivalent of frequency counters and will be discussed later in this article. Wavelength meters are capable of determining the wavelength of a source to less than 1 part per million. Figure 2(b) illustrates the measurement procedure used to transfer the accuracy of a wavelength meter to an osa. The power from the tunable laser is coupled to both the wavelength meter and the osa. The osa is then forced to read the same wavelength value as was read by the wavelength meter. This calibration method requires a significant investment in equipment. As long as the osa has a consistently repeatable grating positioner, very accurate measurements can be made.
Figure 3a illustrates a method for calibrating osas using a gas absorption cell. Calibration with gas absorption lines has the advantage that such lines have natural constants.6-10
The light from a broadband source such as an edge-emitting light-emitting diode is passed through a glass tube containing a molecular gas. Gas cells absorb radiation near the vibrational and rotational resonances of molecules. The resulting light is collected and passed on to an osa. The strongest absorption occurs at the fundamental resonance frequency for gas molecules, which most often occurs at wavelengths longer than 2 microns.
The available absorption lines for the important 1550-nm dwdm fiber-optic band are limited. The two most promising candidates are acetylene and hydrogen cyanide. The resonances for both of the molecules are overtones of the fundamental vibrational frequencies. Figure 3(b) shows the absorption spectrum for acetylene.6 One set of absorption spikes is nearly uniformly spaced over the 1510- to 1545-nm range. The length of the evacuated tube for this measurement is 5 cm and the gas pressure is 400 Torr. The magnitude of the absorption lines is less than 3 dB for these conditions. The vacuum wavelengths of these absorption lines have been measured to an accuracy of 0.001 nm.
Saki, Sudo, and Ikegami have studied the effect of temperature and pressure for the acetylene absorption band.8 They concluded that temperature sensitivity is less than 100 kHz/K and pressure sensitivity is less than 1.5 kHz/Pascal. An extreme temperature variation would be 100K resulting in a 1-MHz shift. This corresponds to a 0.000008-nm wavelength shift at 1550 nm. The environmental pressure dependence is also unimportant for dwdm applications.
An alternate absorption cell gas is hydrogen cyanide. The absorption characteristics for a 15-cm-long tube at 200 Torr is shown in Figure 3(c). Hydrogen cyanide absorption is centered on the important edfa gain sub-band at approximately 1545 to 1560 nm. For this reason it is a desirable absorption cell for dwdm applications. The only drawback to hydrogen cyanide is its toxicity. Hydrogen cyanide attaches to hemoglobin mole cules in the blood, rendering them unable to carry oxygen. However, the amount of gas present in an absorption cell can be made small enough so that calibrations can be accomplished without hazard.
Air versus vacuum wavelengths
All fiber-optic input spectrum analyzers directly measure the wavelength of light in an air environment. Few manufacturers have undertaken the challenge of building a vacuum chamber around an osa. dwdm wavelength measurements are quoted in terms of optical frequency or vacuum wavelengths. This means that the osa must "make an assumption" about the index of refraction in order to provide an accurate wavelength display. The relationship between these variables is:
c = f lvac
where
lvac = lmnm is the wavelength of light in a vacuum
c = speed of light in vacuum (2.99792458E+8 m/sec)
f =optical frequency
lm =wavelength in medium (e.g., air)
nm =refractive index of the medium
osas directly measure lm. The accuracy of osa measurements directly correlates to the index of refraction, which depends on temperature, pressure, and humidity. To do a perfect conversion between the vacuum wavelength and air wavelength would require an accurate knowledge of the refractive index. The refractive index of air at sea level for a 15°C temperature and no humidity is approximately 1.000273.
Assume that an osa is calibrated with an acetylene absorption cell so that a 1550-nm laser reads exactly 1550 nm at sea level. What type of wavelength error would be introduced by a climb into the mountains with this osa? Figure 4 shows how an osa calibrated at sea level would change its displayed reading as a function of elevation. For a very high mountain, 5000 m, the instrument would be in error by 0.2 nm. This would not be sufficiently accurate for dwdm applications. This means that, for accurate measurements, osas must be calibrated for the environmental conditions that the measurements will be made in.
It is useful to see how the index of refraction varies as a function of the three most common variables: pressure, temperature, and humidity. Figure 5(a) shows how the index of refraction changes as a function of wavelength for three values of temperature.11,12
Figure 5(b), 5(c), and 5(d) show the variability of the index of refraction at the 1550-nm wavelength. Figure 5(b) shows a plot of the index versus temperature with 760 Torr pressure and 0% relative humidity. Figure 5(c) shows a plot of index versus pressure with 25°C temperature and 0% relative humidity. Figure 5(d) shows a plot of index versus relative humidity with 25°C temperature and pressure of 760 Torr. From these three graphs, it is seen that pressure variations have the largest effect on the index of refraction for reasonable values of the environmental variables. The average atmospheric pressure can vary from 760 Torr at sea level down to 400 Torr at a height of 5000 m.
Michelson interferometer wavelength meter
As previously stated, wavelength meters are the equivalent of electronic frequency counters. They offer the most convenient method for highly accurate measurement of dwdm wavelengths. Even after a wavelength calibration, grating-based osas do not match the wavelength accuracy of a wavelength meter.13,14
Figure 6(a) shows the block diagram of a Michelson interferometer wavelength meter with an input for the unknown wavelength signal and an input for a wavelength reference. Light from the unknown input is collimated and directed to the interferometer. The signal is split into two paths with a beam splitter. Both beams are then incident on 100% reflecting mirrors, which bounce the light back toward the beam splitter. These mirrors are most often constructed as retroreflectors so that the beams are reflected back at nearly the same angle at which they are sent into the mirror. Part of the light reflected from the two arms of the interferometer goes back toward the input beam. The other portion of the light is incident on a photodetector.
In wavelength meter measurement operation, the position of the variable-length arm is scanned. Figure 6(b) shows the scanned interferometer output measurements for a 1550-nm distributed feedback (DFB) laser input. This measurement of photocurrent versus mirror position is referred to as an interferogram. The interferogram shows the detector signal alternating from dark to light as the variable-position mirror is scanned. The alternations of dark and light are due to constructive and destructive interference of the two beam paths. The plot of Figure 6(b) shows only a small segment of the interferogram over a 25-micron window. For a narrow spectral width signal (10-MHz linewidth) such as the unmodulated DFB shown in Figure 6(b), the interference signal will remain strong for interferometer delays of many meters.
The interferogram of the unknown signal is compared to that of the interferogram produced by a known wavelength standard such as a helium-neon laser. If the mirror is moved over a specified scan length, L, the interference fringes in both the reference and the unknown arms are counted. The wavelength of the unknown signal can then be calculated by comparing the fringe counts in the unknown and the reference signal paths and taking the ratio of counts:
lu = (NrNu)(nunr)lr
where
lu =the unknown wavelength
lr =the reference wavelength
nr = the index of refraction at the reference wavelength
nu = the index of refraction at the unknown wavelength
Nr = the number of reference counts over a distance L
Nu = the number of unknown wavelength detector counts over distance L
The equation requires that an accurate ratio of the index of refraction at the reference wavelength to the index at the unknown wavelength be stored in instrument memory. A well-designed wavelength meter will allow the unknown wavelength to be measured with an accuracy approaching that of the reference source. Wavelength meters with accuracies of less than 0.001 nm centered at 1550 nm are available.
Figure 6(b) shows the interferogram that results when a single wavelength signal is applied to a Michelson interferometer. What happens if two or more separate laser sources are combined in the wavelength meter input? This situation is very common in dwdm systems. Figure 7(a) shows the interferogram result when lasers of equal power at 1300, 1550, and 1650 nm are simultaneously applied to a Michelson interferometer wavelength meter. The interferogram has considerable complexity compared to that of Fig. 6(b).
Consider what would result if a Fourier transform operation were done on the photocurrent versus distance function that makes up the interferogram of Fig. 7(a). The Fourier transform will produce a plot of the photocurrent magnitude versus spatial frequency, s (measured in cycles per meter), of the interferogram.
Figure 7(b) illustrates the results of performing a Fourier transform operation on the data of Figure 7(a). The temporal frequency shown in the plot (cycles per second) is obtained by multiplying by the speed of light in the medium. The Fourier transform operation on the interferogram allows the wavelength of the signals to be separated and measured individually. Here each of the input signals is individually resolved both in frequency and in power.
The Fourier transform operation does not compromise the wavelength accuracy of the measurement when compared to the fringe counting methods discussed earlier. Fourier transform wavelength meters provide the best solution for simultaneous measurement of dwdm signal wavelengths, with high accuracy on each channel.
Summary
osas can make wavelength measurements to about 0.01-nm accuracy (at 1550 nm) if periodically calibrated. Wavelength repeatability after calibration is an important specification. The most common methods of osa calibration are comparison to a high-accuracy laser standard or to a gas absorption cell.
Wavelength meters provide the most convenient source for very-high-accuracy wavelength measurements. They are capable of wavelength accuracies of less than 0.001 nm at a center wavelength of 1550 nm. Fourier transform wavelength meters are well suited for dwdm measurements because they can simultaneously measure the wavelength of many channels. u
Dennis Derickson is research and development project manager at Hewlett-Packard Lightwave Div., Santa Rosa, CA. He can be reached at (707) 577-2532 or [email protected].
References
1. Bellcore gr-2918-core Document, "Generic Criteria for Sonet Point-to-Point Wavelength Division Multiplexed Systems in the 1550 nm Region," Bellcore, Piscataway, NJ, 1996.
2. Melles-Griot Product Catalog, "Fundamentals of Helium-Neon Lasers," 1770 Ketterling Street, Irvine, CA, 92714.
3. D.A. Jennings, F.R. Peterson, and K.M. Evenson, "Frequency measurement of the 260-THz (1.15 µm) HeNe laser," Optics Letters, Vol. 4, No. 5, pp. 129-130, May 1979.
4. C.gif. Moore, Atomic Energy Levels as Derived from the Analysis of Optical Spectra: Vol. 1, nsrds-nbs 35, Vol. 1 (com-72-51282), p. 77, Dec. 1971.
5. M. Tetu, "Absolute Wavelength Stability Ions," 1997 Optical Fiber Communications Conference, Optical Society of America, Tutorial FE1, pp. 167-220, 1997
6. S.L. Gilbert, T.J. Drapela, and D.L. Franzen, "Moderate-Accuracy Wavelength Standards for Optical Communications," in Technical Digest - Symposium on Optical Fiber Measurements, nist Spec. Publ. 839, pp. 191-194 (1992).
7. P. Varanasi and B.R.P. Bangaru, "Intensity and Half-Width Measurements in the 1.525 µm Band of Acetylene," J. Quant. Spectrosc. Radiat. Transfer 15, 267, 1975.
8. Y. Saki, S. Sudo, and T. Ikegami, "Frequency Stabilization of Laser Diodes Using 1.51-1.55 µm lines of 12C2H2 and 13C2H2," IEEE J. Quantum Electron., Vol. 28, pp. 75, 1992.
9. A. Baldacci, S. Ghersetti, and K. N. Rao, " Interpretation of the Acetylene Spectrum at 1.5 µm," J. Mol. Spectrosc., Vol. 68, pp. 183, 1977.
10. K. Chanm, H. Ito, and H. Inaba, "Optical Remote Monitoring of CH4 Gas Using Low-Loss Optical Fiber Link and InGaAsP Light-Emitting Diode in the 1.33 µm region," Appl. Phys. Lett., Vol. 43, pp. 634, 1983.
11. B. Edlen, "The refractive index of air," Metrologia, Vol. 2, No. 2, pp. 71-80, 1966.
12. E.R. Peck and K. Reeder, "Dispersion of Air," J. Optical Society of America, Vol. 62, No. 8, pp. 958-962, Aug. 1972.
13. G.gif. Obarski, "Wavelength measurement system for optical fiber communications," nist technical report nist/TN-1336, Feb. 1990.
14. J.J. Snyder, "Laser wavelength meters," Laser Focus, pp. 55-61, May 1982.