We present comparative measurements of polarization dependent gain in neodymium and ytterbium doped fiber amplifiers. It is demonstrated, that this effect is always present in neodymium doped fiber amplifiers while under appropriate operation conditions it can be suppressed in ytterbium doped fiber amplifiers.
©2003 Optical Society of America
Polarization dependent gain (PDG) is a well known effect in erbium doped fiber amplifiers [1,2]. It arises from electric field anisotropies for the rare earth ions due to the amorphous structure of the glass host and causes the gain for probe signals to be dependent on their polarization orientation relative to the main signal. The amplifier is preferentially saturated in the polarization of the higher intensity signal and this allows excess noise to build up in the orthogonal polarization. This reduces the signal-to-noise ratio and degrades the performance especially in long-haul systems .
Besides for erbium doped amplifiers, PDG has already been reported for neodymium-doped bulk glass lasers [4,5]. We report, to the best of our knowledge for the first time, about PDG in neodymium (Nd) and ytterbium (Yb) doped fiber amplifiers. There are generally two contributing factors to PDG: one from the pump and the other from the signal. We focused our experiments on the latter factor, as we used unpolarized pump sources.
The PDG can easily be determined by measuring the gain G ‖,┴ for polarizations parallel (‖) and orthogonal (┴) to the saturating signal. For the direct measurement the input (P in,‖,┴) and output (P in,‖,┴) powers in both polarizations have to be measured:
From this measurement the PDG results in PDG = G ┴ / G ‖.
In our setup (see Fig. 1) the radiation of two independent 1064 nm-Nd:YAG single-frequency nonplanar ring oscillators (NPRO) – one as the main and the other as a probe signal – was combined by a 50/50-beamsplitter. Both NPROs were protected from backreflections with Faraday isolators. The combined radiation of the NPROs was focused into the active core of a single-mode Nd (∅=6 µm) and Yb (∅=4.2 µm) doped double clad fiber, respectively. The amplifier was pumped from the other fiber end by a fiber coupled diode laser at 808 nm for Nd and 975 nm for Yb. The 15 m long Nd doped fiber was doped with 1300 mol ppm Nd2O3. Additionally, 10 m and 14 m long Yb doped fibers with 6500 mol ppm Yb2O3 were used. On the pump side of the fiber the amplified signal was separated from the pump light by a dichroic mirror. It was then imaged on an iris aperture to cut off radiation from the pump core. Both fiber ends were polished at an angle of about 8° to avoid backreflections and lasing oscillation of the amplifier. In order to eliminate the influence of any polarization dependent losses the input polarization of the NPROs was chosen that a polarization rotation of 90° at the fiber input did not change the amplifier output power.
The amplified radiation was analyzed by an optical spectrum analyzer (OSA) with a resolution bandwidth of 0.05 nm. The wavelengths of the NPROs were tuned with crystal temperature to 1064.25 nm (NPRO 1) and 1064.02 nm (NPRO 2). At this separation (~60 GHz) of emission wavelength the dynamic range of the optical spectrum analyzer was 35 dB. All power levels except the overall amplifier power were measured with the OSA.
For proper determination of the polarization dependent gain the power level of the probe signal must be kept well below the main signal, otherwise the probe signal intensity reduces the PDG. On the other hand for very low probe levels the true power level can not be determined with the OSA due to masking effects by the saturating signal.
The necessary power ratio of the two signals was measured with the neodymium doped fiber amplifier. For this measurement the input polarizations were set orthogonally and a polarization selection unit consisting of a quarter waveplate, a half waveplate and a polarizing beamsplitter was inserted into the amplifier output beam. With this selection unit the absolute gain for either NPRO was measured by suppressing the other NPRO by means of its polarization and measuring the transmitted power level both with the pump power switched off and on. Hence this measurement corresponds to a direct implementation of the concept described by Eq. (1). It was repeated for different input power ratios by varying the power of NPRO 2. The pump power was adjusted in order to keep the overall amplifier gain constant at 10 dB. The power of NPRO 1 was 0 dBm.
It can be seen from the results in Fig. 2, that the measured PDG increases up to a seed power ratio of about -10 dB and stays constant for values below -10 dB. The error of this measurement was estimated to be about 0.1 dB. This relatively high value arises from the procedure that at first the absolute gain had to be determined and then the ratio resulted in the PDG.
Due to these results, an input power ratio of 15 dB was chosen for the subsequent measurements. With this ratio an accurate PDG reading directly from the OSA data was guaranteed while it was not necessary to insert the polarization selection unit. The PDG was then determined by measuring the ratio of the power level of the probe signal for polarizations parallel as well as perpendicular to the saturating signal. This method corresponds to the choice of P in,‖=P in,┴=P in,probe and therefore
With this method only two power measurements had to be performed as not the absolute gain but only the gain ratio was determined. Hence the overall measurement error for the following results can be estimated to be about 0.05 dB.
3. Results and discussion
In Fig. 3 some of the measurements with the Nd and the 10 m-Yb doped fiber amplifiers are shown. As there is no reabsorption of the signal in Nd the lowest possible gain is 0 dB (Fig. 3(a)). At this level the pump power is switched off and the amplifier acts as a passive fiber exhibiting no effects. For very low seed power (1 mW) and low gain, i.e. below saturating intensity in the fiber-core, only small PDG-effects are visible as the amplifier is not strongly saturated. The slope increases as the signal intensity and hence the saturation increases with increasing gain. For higher seed power the PDG increases almost linearly with increasing gain (both in dB-scale). This behaviour is expected for intensities well above the saturating intensity in first order because e.g. an increase in gain by a factor of two can be approximated by two amplifiers with equal gain which are exhibiting the same amount of PDG.
The absolute amount of PDG for constant gain increases with increasing seed power due to the higher saturation in the fiber. For gain values around 10 dB the neodymium doped amplifier shows PDG up to 1 dB. This amount is relatively high compared to the values published for erbium doped amplifiers (e.g., ). This may be explained by the relatively high seed power in our experiments compared to the experiments carried out with erbium doped fiber amplifiers as both in erbium and neodymium amplifiers the PDG depends on the seed power.
In contrast to the Nd doped amplifier, the Yb doped fiber amplifier shows also “gain” below 0 dB (Fig. 3(b)) since for low pump power the signal is partially absorbed in the fiber. Due to this absorption the measured PDG in the Yb doped amplifier is an effective value resulting from both the saturation of gain and absorption. While the gain-saturation causes positive PDG, the saturation of the absorption generates negative PDG. Hence the Yb amplifier exhibits negative PDG at low gain values. Assuming an equal increase of the effects caused by absorption and amplification no change in the effective PDG would be expected. But with increasing gain and inversion the balance between PDG due to amplification and absorption is shifted towards the part caused by the amplification process. Thus, the PDG starts increasing slowly with increasing gain. At higher gain the slope is dominated by effects due to gain saturation.
At low seed power (10 mW) this change in slope was observed in the measurements. At higher seed power (40 mW) the seed power itself strongly saturated the absorption causing a depletion of the lower laser level. Hence the PDG-slope was mainly given by effects due to gain saturation. In comparison to the Nd doped amplifier the increase in PDG per dB gain is about a factor of 3.5 lower in the Yb doped amplifier.
In principle one would expect different materials to exhibit a different amount of PDG as this effect is directly coupled to the spectroscopic level system of the dopant ion. Actually it depends on the electric field dependency of the upper and lower transition level for the absorption and emission processes. If, e.g., both the upper and lower transition level have the same electrical field dependency, i.e., an identical Stark-shift, no PDG will be observed as local electric field anisotropies will have no effect on the transition. Furthermore the PDG is wavelength dependent as the transition levels depend on the signal wavelength.
Above a certain gain the PDG becomes positive. On the one hand this zero-PDG-gain depends on the seed power: with increasing seed power the saturation and the PDG increases. Hence the measured curve is shifted upward and the zero-PDG-gain decreases. On the other hand the zero-PDG-gain depends on the fiber amplifier length, see Fig. 4. It can be seen, that in the longer amplifier the gain needed for zero-PDG is lower than in the short amplifier. Furthermore, in the longer amplifier more ions are taking part in the absorption and amplification process. Hence, for a fixed overall gain more absorption occurs and a higher internal gain must be set to compensate for this loss. Furthermore the effects resulting from the absorption and emission arise from different transitions in Yb. From there this shift in the zero-PDG-gain towards lower values for the longer fiber might be interpreted as the PDG originating from the emission process is stronger than the one from the absorption transition. On the other hand the actual contributions of absorption- and gain-saturation are not known and further investigations on this effect will have to be carried out for an accurate interpretation of this zero-PDG-gain-shift.
We have measured the effect of PDG in neodymium and ytterbium doped fiber amplifiers. It has been shown that the Nd doped amplifier exhibits strong PDG. In comparison, Yb doped amplifiers show this effect a factor of 3.5 less. Additionally, the effect of the reabsorption of the signal in the amplifier, which causes negative PDG, can be used to cancel out the overall PDG. Hence by proper choice of amplifier length and dopant concentration the gain at which zero PDG occurs can be adjusted to a great extend for the Yb doped fiber amplifier.
The investigations were supported by the Deutsche Forschungsgemeinschaft within the Sonderforschungsbereich 407. We gratefully acknowledge the collaboration with the Institut für Physikalische Hochtechnologie (Dr. H. R. Müller and colleagues) concerning the supply of the used fibers.
References and links
1. Paul Wysocki and Vincent Mazurczyk, “Polarization Dependent Gain in Erbium-Doped Fiber Amplifiers: Computer Model and Approximate Formulas,” J. Lightwave Technol. 14, 572–584 (1996) [CrossRef]
2. E.J. Greer, D.J. Lewis, and W.M. Macauley, “Polarisation dependent gain in erbium-doped fibre amplifiers,” Electron. Lett. 30, 46–47 (1994) [CrossRef]
3. Eyal Lichtman, “Limitations Imposed by Polarization-Dependend Gain and Loss on All-Optical Ultralong Communication Systems,” J. Lightwave Technol. 13, 906–913 (1995) [CrossRef]
4. D.W. Hall and M.J. Weber, “Polarized fluorescence line narrowing measurements in Nd laser glasses: Evidence of stimulated emission cross section anisotropy,” Appl. Phys. Lett. 42, 157–159 (1983) [CrossRef]
5. Douglas W. Hall, Roger A. Haas, William F. Krupke, and Marvin J. Weber, “Spectral and Polarization Hole Burning in Neodymium Glass Lasers,” IEEE J. Quantum Electron. QE-19, 1704–1717 (1983) [CrossRef]