1. Introduction

Fiber Raman amplifiers have attracted huge attention in recent years as an enabling technology for the future long-haul, high-capacity optical communication systems due to a range of practical and potential advantages [1, 2]. For example, the amplification band can be achieved at any wavelengths within the transparency window of optical fiber simply by changing the pump wavelength, and the use of distributed Raman amplification in transmission fibers can provide both the improved noise performance and the enhanced signal power budget. Among various Raman fiber amplifiers, the one based on dispersion compensating fiber (DCF) has been considered to be an attractive means since it can provide simultaneously two key functions such as dispersion and loss compensation in a transmission fiber span [3]. In such DCF based Raman amplifiers one practical issue is to increase pump power efficiency since a significant amount of pump power is unused and wasted [3, 4] due to the relatively short length of the DCF, which is usually insufficient to attenuate all of the pump power. Recently, we proposed a novel concept of a single pump, dispersion compensating Raman/EDFA hybrid amplifier recycling the residual Raman pump in a cascaded EDF section for the secondary signal amplification to reduce the Raman pump waste, and demonstrated a significant increase of overall signal gain and a corresponding enhancement of overall power conversion efficiency compared to the Raman amplifier without the EDF [5]. In order to make our newly proposed hybrid amplifier scheme more practical and useful in real wavelength division multiplexing (WDM) add/drop networks, we should understand its dynamic properties as well as the static properties [6, 7].

In this paper, we go on to investigate transient effects of our proposed single pump, dispersion compensating Raman/EDFA hybrid amplifier recycling the residual Raman pump under the situation of multi-channel add/drop, and subsequently demonstrate the use of a fiber Bragg grating (FBG) based all-optical gain-clamping technique to suppress the output power transients [6, 7]. The experimental results show that the proposed Raman/EDFA hybrid amplifier has a significantly long transient response time of ~2 ms due to both the low-pumping regime operation of 14XX nm pumped EDFA section and the additional pump transit time through the 12.6 km long DCF to reach the EDF section. However, using the all-optical gain clamping scheme we obtain an almost, transient-free operation with a trivial amount of surviving channel relaxation oscillation.

2. Experiment and results

The experimental configurations for our proposed single pump, dispersion-compensating Raman/EDFA hybrid amplifier and its modified version with all-optical gain-clamping based on a pair of uniform FBG’s, are shown in Fig. 1. The Raman pump source consisted of two laser diodes operating at 1455 and 1465 nm, respectively and a total pump power of up to 500 mW was launched into a 12.6 km long DCF with a ~0.55 dB/km attenuation at a wavelength of 1550 nm in a counter-propagating geometry via a 1550/1460 nm WDM coupler. The DCF had a group velocity dispersion (GVD) of -98 ps/nm-km. The 12. 6 km DCF can provide sufficient dispersion for compensating a ~70 km long SMF based transmission span. The residual pump power after the DCF was measured to be ~40 mW. This residual Raman pump was then launched into a 10 m long EDF via two 1550/1460 nm WDM couplers, and was reused as a pump source for secondary signal amplification in the EDF. The peak absorption coefficient of the EDF at 1530 nm was 6 dB/m. Further details of this hybrid amplifier configuration are fully described in Ref. [5]. To obtain all-optical gain-clamping we constructed a laser cavity using a pair of uniform FBG’s with a reflectivity of ~99.5 % as shown in the inset of Fig.1(b). The center-wavelength of the FBG’s was 1529.6 nm, and the spectral bandwidth was ~0.45 nm.

 

Fig. 1. Experimental configurations for (a) our proposed dispersion-compensating, Raman/EDFA hybrid amplifier and (b) its modified version with all-optical gain-clamping based on a pair of uniform FBG’s. Inset: Optical spectrum of the FBG used.

Download Full Size | PDF

 

Fig. 2. Measured net gain and noise figure characteristics of the Raman/EDFA hybrid amplifier as a function of input signal power at a wavelength of 1550 nm for both cases i.e. without and with gain-clamping.

Download Full Size | PDF

Figure 2 shows measured static overall gain (not on-off gain) and noise figure (NF) characteristics of the Raman/EDFA hybrid amplifier as a function of input signal power at a wavelength of 1550 nm for both cases i.e. with and without gain-clamping. The NF measurement was performed using the optical spectrum analyzer method, which might provide less accurate values in saturation conditions [8]. A huge gain variation was observed in the Raman/EDFA hybrid structure without gain-clamping as we increased the input signal power from -35 dBm to 0 dBm whilst the gain-clamped scheme maintains a stable overall gain of ~14 dB with gain fluctuations of less than 0.3 dB for an input signal power range from -35 dBm to -8 dBm. As a matter of fact, the input dynamic range of -35 dBm to -8 dBm with gain-clamping is too low for the real dynamic situation application. However, note that the saturating output power performance of our proposed hybrid amplifier was mainly limited by the Raman pump power available in our laboratory rather than the amplifier structure itself. The NF was observed to gradually increase from ~7 dB to ~8 dB for both cases i.e. with and without gain-clamping as the input signal power was enlarged. The gain-clamping induced only a ~0.5 dB NF penalty at 1550 nm relative to that of the scheme without gain-clamping, and this penalty is believed to be mainly due to the insertion loss of two FBG’s used for obtaining lasing action.

 

Fig. 3. Measured spectral profiles of overall gain and noise figure at two different input signal power levels of -10 dBm and -20 dBm: (a) without gain-clamping and (b) with gain-clamping.

Download Full Size | PDF

Figure 3 shows measured spectral profiles of overall gain and noise figure for both cases, i.e., with and without gain-clamping at two different input signal power levels of -10 dBm and -20 dBm. The hybrid amplifier structure without gain-clamping exhibits a gain difference of ~3 dB over the whole 30 nm effective gain-bandwidth for the two input power levels whilst almost no signal gain variation was observed with gain-clamping. Noise performance of the gain-clamped scheme was found to be almost similar to that of the one without gain-clamping over the whole 30 nm effective gain-bandwidth, considering FBG insertion induced noise penalty.

Next, we performed a range of experimental simulations of multiple channel add and drop for both cases, i.e., with and without gain-clamping to investigate the realistic transient responses of our proposed Raman/EDFA hybrid amplifier. A probe signal of -20 dBm optical power at a wavelength of 1560 nm which represented a surviving channel and an add/drop beam with a -13 dBm average optical power at 1540 nm were combined together using a 3 dB coupler, and was then launched into the hybrid amplifier. The add/drop beam was modulated at a frequency of 200 Hz by an acousto-optic modulator, and the -13 dBm average power corresponded to the total power of 10 channels with a -20 dBm single channel power. After filtering the add/drop beam out at the output of the hybrid amplifier, we measured the temporal responses of the surviving channel using a combination of a 150 MHz photo-detector and an oscilloscope. The results are summarized in Fig. 4. The hybrid amplifier without gain-clamping exhibited a transient response time of ~2 ms with an intensity excursion of ~2.75 dB. The transient response time is defined as the time that the output surviving channel power rises up to 95 % of the steady-state power level after the channels drop [9]. Interestingly, this transient time is significantly longer than the previous reported value of the separate pump based Raman/EDFA hybrid amplifier in Ref. [10]. According to the description of transient dynamics of the conventional separate pump, hybrid Raman/EDFA amplifiers in Ref. [10], the overall transient response time of the hybrid amplifiers which is usually less than a few hundred µs, is mainly determined by that of EDFA since EDFA has a much faster response time than Raman amplification, and the corresponding output intensity excursions are also governed by EDFA due to its high output power. Although the comment that the fast transient time of EDFA section decides the transient response time of a whole separate pump based Raman/EDFA hybrid amplifier, is still arguable, the EDFA section was suspected to cause the long transient response time. Note that the dominating amplification process in our proposed hybrid amplifier is Raman based amplification in the DCF rather than EDFA based amplification and the EDFA is for the secondary amplification using a relatively small amount of residual pump power. The EDFA in proposed hybrid amplifier is thus operating at the low-pumping regime with 14XX nm pump lasers.

 

Fig. 4. (a) Measured transient responses of a -20 dBm surviving channel in an experimental simulation of adding and dropping an add/drop channel with a -13 dBm average power corresponding to the total power of 10 channels with a -20 dBm single channel power, for both cases i.e. with and without gain-clamping. (b) A close-up view of the transient response with gain-clamping.

Download Full Size | PDF

In order to figure out the reason why the long transient response time was observed, we performed further transient response measurement for each of the Raman amplification section and EDFA section, separately. In particular, for the EDFA section measurement, we used only a ~40 mW pump power which is a same amount as the residual pump power after the DCF transmission to maintain the low-pumping regime operation since transient time of EDFA relies on the EDFA inversion status associated with pump and signal strength [11, 12]. We also increased optical power of the probe channel and the add/drop channel up to -13 dBm and -10 dBm, respectively according to the Raman gain profile [5]. However, the previous transient conditions of 500 mW pump power, -20 dBm surviving channel power, and -13 dBm add/drop channel power were applied to the Raman section measurement. The measured results are summarized in Fig. 5. The EDFA section exhibited a transient response time of ~1.7 ms whilst the Raman section showed a trivial amount of transient power excursion (~0.04 dB) with a transient response time of ~100 µs. Such a long transient time of the EDFA section can be attributed to both the weak population inversion status associated with low-pumping regime operation and the 14XX pump wavelength according to the previous detailed studies on EDFA transient responses [11, 12, 13]. The 14XX pump based EDFA’s are known to have a longer transient response time than the 980 nm based ones [13]. On the contrary, the Raman amplification section was operating at the pump-undepleted regime due to the relatively strong pump power and only a small amount of output transient was thus observed. According to our measured results, It is clearly evident that the long transient response time of our proposed single pump Raman/EDFA hybrid amplifier recycling residual Raman pump is mainly due to the lowly-pumped EDFA section. Furthermore, the other reason for the long transient time could be associated with the additional pump transit time through the 12.6 km DCF to reach the EDF section. Note that our proposed Raman/EDFA hybrid amplifier recycling residual Raman pump has a single pump common to both the Raman amplification section and the EDFA section. The pumps beams should thus propagate through the 12.6 km long DCF to reach the EDF section whilst two separate pump beams directly couple into the EDF section and the Raman section in the conventional separate pump, Raman/EDFA hybrid amplifiers. The additional pump transit time through the 12.6 km DCF which is inherent to our proposed structure, is believed to partly cause the additional transient response delay (~0.3 ms). However, using a simple FBG based gain-clamped scheme almost transient free operation was readily achieved with a trivial relaxation oscillation, the peal-to-peak power excursion of which was as small as ~0.04 dB as shown in Fig. 4.

 

Fig. 5. Measured transient responses for (a) the Raman amplification section and (b) the EDFA section, separately. For the Raman section measurement, pump power: 500 mW, surviving channel power: -20 dBm, and add/drop channel power: -13 dBm. For the EDFA section measurement, pump power: 40 mW, surviving channel power: -13 dBm, and add/drop channel power: -10 dBm.

Download Full Size | PDF

3. Conclusion

We have experimentally investigated transient effects of our proposed Raman/EDFA hybrid amplifier recycling residual Raman pump, and demonstrated the use of a FBG based all-optical gain clamping technique to efficiently suppress the output power transients. Our proposed single pump, Raman/EDFA hybrid amplifier was found to have a significantly long transient response time of ~2 ms compared to the conventional separate pump, Raman/EDFA hybrid amplifiers. The reason of such a long transient time was found to be mainly due to the low-pumping regime operation of 14XX nm pumped EDFA section. The other reason is associated with the additional pump transit time through the 12.6 km long DCF to reach the EDF section. However, using a simple FBG based gain-clamped scheme almost transient free operation was readily achieved with a trivial relaxation oscillation. A numerical modeling and analysis needs to be performed for better understanding of the dynamic properties of our proposed single pump, Raman/EDFA hybrid amplifier.