21, No. 21 / November 1, 1996. Dual-stage erbium-doped, erbium/ytterbium-codoped fiber amplifier with up to +26-dBm output power and a 17-nm flat spectrum.
1744
OPTICS LETTERS / Vol. 21, No. 21 / November 1, 1996
Dual-stage erbium-doped, erbium/ytterbium-codoped fiber amplifier with up to +26-dBm output power and a 17-nm flat spectrum Paul F. Wysocki, Namkyoo Park, and David DiGiovanni Lucent Technologies, Bell Labs, 700 Mountain Avenue, Murray Hill, New Jersey 07974 Received June 3, 1996 By concatenating an aluminum-codoped erbium-doped fiber with an efficient erbium/ytterbium-codoped fiber, we produce a gain spectrum that is f lat within 0.5 dB for 17 nm from 1544 to 1561 nm. By using energy transfer between erbium and ytterbium ions, we are able to pump the second stage with a high-power 1064-nm source to achieve an output power as high as 126 dBm. Using 980-nm pumping of the first stage, we produce an overall noise figure below 5 dB. 1996 Optical Society of America
When erbium-doped fiber amplif iers (EDFA’s) are used in multiple-wavelength wavelength-division multiplexed (WDM) systems and analog CATV systems, high output powers, low noise figures (NF’s), and spectral gain f latness are required. One can achieve high power levels by increasing the available pump power or by using ytterbium codoping.1 – 3 In phosphoruscontaining f ibers, ytterbium can absorb pump power from available sources near 1064 nm and eff iciently transfer energy to erbium ions for power amplification near 1550 nm.4 Unfortunately, such f ibers do not readily produce f lat gain spectra or low NF’s.5 Gain f latness has been achieved with f luoride-based host fibers,6 which have not yet produced a very high output and can only be effectively pumped near 1480 nm (producing a higher NF than 980-nm pumping). Flatness is also attainable with optical filtering7 or by combination of fibers of differing host composition.8 In this Letter we describe an unfiltered EDFA with up to 126-dBm output power, 38-dB gain, f latness to within 0.5 dB in a 17-nm bandwidth, and a low (,5-dB) NF. In an EDFA all possible spectra are predicted by fractional combinations of the gain spectrum measured when all ions are inverted with the loss spectrum measured when all ions are uninverted. Some possible gain spectra predicted with these parameters are shown in Fig. 1(a) for a typical aluminum-codoped EDFA and in Fig. 1(b) for a phosphorus-codoped erbium/ytterbium-doped f iber amplifier (EYDFA). These are the fibers used throughout this Letter. Unfortunately, the EYDFA does not produce spectra as f lat as those of the EDFA, which, in turn, does not produce a very high output. Fortunately, some of the spectra for these f ibers can be combined for certain length choices to produce f lat spectra. Some computed combined f lat gain spectra for 15 m of erbium-doped f iber (EDF) and 6 m of erbium/ytterbium-doped fiber (EYDF) inverted to various levels are shown in Fig. 2. The inversion of the EYDF was adjusted in each case to produce the f lattest spectrum when combined with the EDF inverted to the indicated level. Clearly, a loss of inversion in the EDF can be partially compensated by an inversion increase in the EYDF. Because erbium is considered 0146-9592/96/211744-03$10.00/0
mostly homogeneously broadening in a silicate host at room temperature,9,10 it does not matter what power levels produce these inversion levels. The spectrum is determined solely by the average inversion of the ions. To produce the f lattest spectrum from 1544 to 1561 nm, one should combine a 100%-inverted EDF combined with a nearly 50%-inverted EYDF. These inversion levels can be reached in a typical dual-stage
Fig. 1. Theoretical gain curves for different ion inversion (Inv) levels for (a) aluminum-codoped (6 mol. %) EDF and (b) phosphorus-codoped EYDF. 1996 Optical Society of America
November 1, 1996 / Vol. 21, No. 21 / OPTICS LETTERS
Fig. 2. Predicted f lat combined spectra for 15-m EDF inverted to the Inv1 level and 6-m EYDF inverted to the Inv2 level.
amplifier. A first-stage EDF can be highly pumped at 980 nm so that, as long as the signal is moderate, nearly 100% inversion (and a corresponding low NF) is possible. The input power to a second-stage EYDF is high enough to reduce the inversion to less than 50%. A diagram of the dual-stage hybrid design is shown in Fig. 3. An isolator was used between stages to prevent backward-traveling amplif ied spontaneous emission from reducing the EDF inversion. A pump ref lector was also used in the f irst stage to ref lect extra pump power. The EYDF pump power was adjusted until the f lattest spectrum was observed. As signal power increased, the EYDF pump power was increased to bring the net gain back to the design point and achieve f latness. The lengths of the EDF and the EYDF are easily chosen by modeling and computation of spectra like those of Fig. 2. In the design tested here, the f irst stage consisted of 15 m of EDF (6% alumina, N.A. 0.28) copumped with 100 mW at 980 nm. The second stage consisted of 9 m of an efficient EYDF (phosphorus codoped, N.A. 0.2) copumped at 1064 nm. A shorter length of EYDF would produce a f latter combined spectrum, but 9 m produces greater eff iciency. The loss before the EDFA (from the isolator, WDM, and splices) was 1 dB. The interstage loss (from the isolator, WDM, pump ref lector, a connector, and the splices) was 3.2 dB, and the postamplif ier loss (from the isolator, WDM to reject excess pump, and splices) was 1.8 dB. Unless noted, all values reported below include component losses. To test the hybrid EDFA, we provided a saturating tone at 1548 nm and adjusted its power to various levels. A small probe signal was added and swept across the spectrum to measure its shape. Because erbium is a mostly homogeneous medium,9,10 the spectrum measured by this technique has the same shape as the spectrum produced by use of multiple saturating tones to produce the same ion inversion level (except for a very small amount of spectral hole burning).11 For each saturating tone level, the pump power of the second stage was adjusted until the f lattest spectrum was achieved. The resultant best spectra are shown
1745
in Fig. 4 for some useful saturating tone levels. These levels might, for example, represent the power in a single analog signal or the total power for all channels in an eight-channel WDM system. The measured results are f itted theoretically by using modeling parameters, known component losses, and inversion levels predicted by modeling this design. The f it is good, but some discrepancy is observed near 1551 nm. This may be the result of inaccurate modeling parameters, measurement errors, or spectral hole-burning near the 1548-nm saturating tone. Because the pump in the EYDF was adjusted to reach f latness, a hole centered at 1548 nm would be compensated by an increase in the inversion of the second stage. Calculations show that this would leave a remnant hole near 1551 nm for typical spectral hole depths and widths.11 A multichannel saturating tone of equal total power would burn multiple smaller holes more uniformly across the spectrum and therefore would not leave a dip at 1551 nm. All four spectra of Fig. 4 achieve an . 17-nm bandwidth with as little as 0.5 dB variation. The 1064-nm pump power levels for these cases were 307, 425, 645, and 1024 mW for 215, 211, 27, and 23 dBm signal inputs, respectively. The measured output power levels at 1548 nm were 14.53, 18.01, 21.06, and 23.61 dBm for 215, 211, 27, and 23 dBm signals, respectively. The penalty produced by the component losses mentioned above is significant. Because the spectrum was forced to f latness by
Fig. 3. Conf iguration for hybrid EDFA. WDM’s, wavelength-division multiplexers.
Fig. 4. Measured f lattest gain spectra produced by hybrid EDFA with 15-m EDF and 9-m EYDF for various saturating signals at 1548 nm. Pump power at 1064 nm in stage 2 was adjusted in each case to 307, 425, 645, and 1024 mW from the top curve to the bottom.
1746
OPTICS LETTERS / Vol. 21, No. 21 / November 1, 1996
Fig. 5. Measured output power at 1548 nm versus launched pump power at 1064 nm for the hybrid EDFA with 15-m EDF and 9-m EYDF for two different launched powers at 1548 nm.
pump power, all component losses directly reduce the output. If total component losses were 3 dB, the output power would be 3 dB higher when the f lattest gain was reached. Of course, the pump power required for a higher output power would be greater as well. To assess the required pump for a given output, output power was measured versus 1064-nm pump power. The measured result at 1548 nm is shown in Fig. 5 for two signal levels straddling the four signal powers used above. The output power varies little with signal input. The slope of these plots is ,27%. The internal efficiency is 1.8 dB higher (owing to output loss), or ,41%. Figure 5 can be used to predict the pump power needed to produce the projected output power levels described above. For example, to produce 26.61-dBm output at 23 dBm input, assuming a 27% slope efficiency, would require ,1.85 W of 1064-nm pump power. To confirm that this hybrid amplifier can achieve f latness with 125-dBm output, we lengthened the second stage to 12 m. The 6-dB component losses were left intact. When optimized, the output power increased by 5.4 dB for all signal levels to 19.93, 23.41, and 26.46 dBm for 215, 211, and 27 dBm signals, respectively. Adequate pump power was not available to reach 129 dBm for the 23-dBm signal. In all cases, the f latness was slightly worse than shown in Fig. 4, with the variation increasing by , 0.3 dB across the 17-nm width. The NF of the hybrid amplifier is determined almost completely by the NF of the f irst-stage EDF. Because this stage is a standard aluminum-codoped EDF pumped at 980 nm, a low NF is possible. The measured external NF at 1548 nm was below 5 dB for all tested cases with a minimum of ,4.2 dB for the smallest-signal case. This corresponds to an internal NF (not including pre-
EDFA loss) of ,3.2 dB. This measurement was made with polarization nulling with an estimated accuracy of ,0.2 dB. The hybrid EDFA described here produces high output power, high gain, a low NF, and a f lat spectrum ,17 nm wide. The design can easily be scaled to different gain levels by scaling the lengths of the two stages equally or by altering the second stage, at the possible expense of f latness. Counterpumping the second stage could also produce greater output power. The greatest limitation on the hybrid design is set by the need to invert the EDF to a high level to produce the f lattest spectrum and the lowest NF. This limits the input power for a given 980-nm pump power and also limits the maximum length of the EDF. However, typical WDM and analog applications require levels similar to those used here. References 1. S. Grubb, T. Windhorn, W. Humer, K. Sweeney, P. Leilabady, J. Townsend, K. Jedrzejewski, and W. Barnes, in Optical Amplifiers and Their Applications, Vol. 13 of 1991 OSA Technical Digest Series (Optical Society of America, Washington, D.C., 1991), postdeadline paper PdP12-1. 2. J. Townsend, W. Barnes, K. Jedrzejewski, and S. Grubb, Electron. Lett. 27, 1958 (1991). 3. S. Grubb, D. DiGiovanni, J. Simpson, W. Cheung, S. Sanders, D. Welch, and B. Rockney, in Optical Fiber Communication (OFC), Vol. 2 of 1996 OSA Technical Digest Series (Optical Society of America, Washington, D.C., 1996), paper TuG4. 4. J. Lincoln, W. Barnes, W. Brocklesby, and J. Townsend, J. Lumin. 60, 204 (1994). 5. P. Wysocki, G. Nykolak, D. Shenk, and K. Eason, in Optical Fiber Communication (OFC), Vol. 2 of 1996 OSA Technical Digest Series (Optical Society of America, Washington, D.C., 1996), paper TuG6. 6. D. Bayart, J. Hervo, and F. Chiquet, in Optical Fiber Communication (OFC), Vol. 8 of 1995 OSA Technical Digest Series (Optical Society of America, Washington, D.C., 1995), paper TuP2. 7. R. A. Betts, S. J. Frisken, and D. Wong, in Vol. 8 of 1995 OSA Technical Digest Series (Optical Society of America, Washington, D.C., 1995), paper TuP4. 8. T. Kashiwada, M. Shigematsu, M. Kakui, M. Onishi, and M. Nishimura, in Optical Fiber Communication (OFC), Vol. 8 of 1995 OSA Technical Digest Series (Optical Society of America, Washington, D.C., 1995), paper TuP1. 9. E. Desurvire, C. Giles, and J. Simpson, J. Lightwave Technol. 7, 2095 (1985). 10. C. Giles and E. Desurvire, J. Lightwave Technol. 9, 271 (1991). 11. A. K. Srivasta, J. L. Zyskind, J. W. Sulhoff, J. D. Evankow, and M. A. Mills, in Optical Fiber Communication (OFC), Vol. 2 of 1996 OSA Technical Digest Series (Optical Society of America, Washington, D.C., 1996), paper TuG7.
