edfa measurements ensure wdm performance
edfa measurements ensure wdm performance
Both direct and indirect test methods can measure edfa performance. But cost and accuracy tradeoffs can determine which method is right for your application.
C. Hentschel Hewlett-Packard GmbH
Erbium-doped fiber amplifiers (edfas) are increasingly used for amplifying wavelength-division multiplexed (wdm) signals. Accordingly, the lightwave communications industry is carrying out an intensive discussion on suitable and economic test methods. Fundamentally, a distinction can be made between direct and indirect testing. Direct testing methods measure the gain and noise figure with essentially the same multiwavelength source as in the real communications system. This solution is accurate, but expensive, considering the large number of channels projected for the new wdm communications systems. On the other hand, indirect testing can reduce the number of lasers necessary. This type of testing involves a reduced set of sources to saturate the edfa and a probe source to test the spectrum.
Assume that the edfa being tested is designed to amplify a dense wdm signal with a channel spacing of 0.8 nm, as defined by the International Telecommunication Union (itu). edfa measurement usually means gain and noise figure measurement. The measurement of the output powers (thereby the gains) can be accomplished by measuring the powers of the amplified signals. More difficult are the channel noise figures, for which a precise measurement of the amplified spontaneous emissions (ase) is needed.
For an ideal optical amplifier with so-called homogeneous broadening, the saturation state (thereby the ase spectrum) is fixed when the total output power is fixed.1 In this case, the ensemble of wdm input powers could be replaced by just one source, which produces the same output power, dramatically reducing the cost of edfa testing. In practice, however, the ase spectrum changes when only one input signal is applied. The reason is spectral hole-burning in the edfa due to inhomogeneous broadening.2 This effect reduces the ase in the vicinity of the signal (see Fig. 1).
The width of the spectral hole caused by a single input signal is 3 to 5 nm, with depth of 0.5 dB maximum depending on the amount of gain compression caused by that signal. In contrast, a wdm signal creates a number of overlapping holes, resulting in a broader hole of less depth, so a single signal cannot generate the same saturation state as the real wdm signal.
Due to the concerns about spectral hole-burning, many scientists believe that the full wdm signal must be applied to the edfa to generate the same saturation state as in the real system. In this situation, one way of determining the ase levels is interpolation, for example, by taking one ase sample to the right, one ase sample to the left of the channel under test, and interpolating between them. But in a dense wdm situation with 0.8-nm channel spacing, interpolation becomes difficult because of the limited spectral selectivity of typical optical spectrum analyzers (OSAs). Two different techniques for ase measurement can be used to overcome this problem: the well-known time-domain extinction technique and the recently introduced signal substitution technique (patent pending).
In this technique, all sources are modulated on and off with a relatively high frequency: 1 MHz, for example. The principle arrangement is shown in Figure 2. Acousto-optic modulators (aoms) are used for the modulation. During the signal-off period (aom1 "off"), aom2 is "on" to feed the ase to the optical spectrum analyzer. However, during the signal-on period, aom2 is "off" to prevent the large amplified signals from saturating the OSA-internal photodetector. This way, undisturbed access to the ase powers, directly at the signal wavelengths, is possible. Another advantage of this technique is that the spontaneous emission of the sources (SSE), an undesired part of the measured output spectrum, is virtually eliminated because the ase is measured during the sources` "off" phase.
Although an intelligent technique, time-domain extinction suffers from high loss in the input paths. Three decibels are lost because of the modulation of the input signal, and another 6 to 7 dB are lost through the insertion loss of aom1. The overall loss of around 10 dB seriously limits the available edfa input power. A booster edfa, as a means to increase the input power level, makes it difficult to control the sources individually.
The new signal substitution technique also measures the ase directly at the signal wavelength. To accomplish this, the channel under test is switched off to open spectral space for the measurement of the ase sample (see Fig. 3). To substitute the missing input signal, the powers of the neighboring channels are increased. Using the neighboring channels means that the "spectral hole" is unchanged, because the missing power is added near the original wavelength. The test equipment is the same as in Figure 2, except the two aoms are not necessary.
Increasing the powers of the two neighboring channels is done on the basis of constant output power (also called constant photon flux). If the power increase for both neighboring channels is to be the same, then this quantity can be calculated when all three channel gains, G, are known from prior gain measurement:
After taking the ase sample, the SSE must be subtracted, which is accomplished by prior SSE measurement. The advantage of this technique is its simplicity, combined with high input power. Compared to time-domain extinction, a power penalty of only 2 dB is needed as headroom for the additional powers of the neighboring channels. Around 8-dB higher input power can be generated.
Both "direct" techniques--time-domain and signal substitution--are based on using one source per wdm channel. A reduction of two, so that the source spacing is 1.6 nm, is possible because it is expected that the spectral hole created by the ensemble of wdm channels is unchanged by such reduction.
In many cases, one source per channel or even one source per pair of channels is considered too expensive, especially when the system under test features 80 channels or more. To reduce equipment cost, dynamic gain testing is used by a number of edfa manufacturers. This technique is considered indirect because it uses a reduced set of lasers to generate approximately the same edfa saturation conditions (see Fig. 4).
A tunable laser acts as a probe. The probe laser is used to generate gain measurement results for those wavelengths where a saturating laser is missing. A possible signal arrangement in this example for eight channels is shown in Figure 5. Three saturating signals are used, and the five remaining signals are obtained from tuning the laser.
The ase spectrum is obtained by switching the probe laser off, measuring the ase, and subtracting the SSE of the saturating lasers. The ases at the saturating wavelengths must be determined from interpolation. The channel gains are obtained by measuring the amplified probe and saturating signals, then subtracting the ase spectrum. This way, the gains at all eight wavelengths can be determined.
This technique`s accuracy de pends on the number of saturating lasers used. For a substantial reduction of the number of laser sources, the gains at the probe wavelengths are expected to be slightly larger than in the real system because of the lack of spectral hole-burning at these wavelengths. The ase values at the probe wavelengths also will be increased. As a consequence, the channel noise figures will be nearly correct, because noise figure is always calculated from the ratio of ase and gain.
Another problem that is frequently overlooked is the choice of power levels for the saturating lasers as a substitute for a known combination of the full set of wdm input powers. Again, the total output power must be the same for both cases. Therefore, the powers of the saturating lasers, i.e., the input powers, can only be set when the total output power is known. This problem can only be solved by iteration.1 Alternatively, an additional error contribution must be taken into account.
As an alternative to dynamic gain testing with a tunable laser, the noise gain profile technique (usually combined with time-domain extinction) can be used. In this case, a broadband light-emitting diode (led) is used as the probe, and the probe spectrum must be measured prior to the edfa measurement. One advantage is faster measurement, because there is no need to change the source wavelength.
The direct wdm testing methods of edfas--time-domain extinction and signal substitution--are more expensive, but also more accurate because they address the spectral hole-burning effect well. Of those two methods, signal substitution offers much higher input power. For example, 0-dBm total input power should be available with appropriate lasers and typical power combiners. Indirect testing, such as dynamic gain and noise gain profile, is less expensive but also less accurate, because the spectral hole-burning is not perfectly emulated.
A general concern in edfa measurements is polarization-dependent gain (pdl) in the edfa. pdl will usually cause some "bending" in the measurement results of the different channels, because the individual laser sources have random, but short-term stable orientations against each other. This is particularly annoying when the highest accuracy in gain flatness measurements is desired. To remove pdl effects from the measurement results, consider polarization scrambling all sources and averaging, or weakly polarized sources such as the led used in the noise gain profile technique. u
1. D. Baney, "wdm edfa gain characterization with a reduced set of saturating channels," Photon. Technol. Lett., Vol. 8, 1996, pp. 1615-1617.
2.. A.K. Srivastava, J.L. Zyskind et al., "Room temperature spectral hole-burning in erbium-doped fiber amplifiers," Technical Digest, OFC conf., San Jose, CA, TuG7, pp. 33-34.
C. Hentschel is standards laboratory manager in the Optical Communication Measurements Operation of Hewlett-Packard (Böblingen, Germany). He can be reached at +49 7031-14 3291 or by e-mail at Christian_Hentschel@hp.com.