1. Introduction

Optical fiber communications have been taunted as the revolutionary technology for huge bandwidth data transmissions. This is achieved with the advancement of Erbium-doped fiber amplifiers (EDFA’s) which nicely coincide with the lowest attenuation window of optical fibers. Together with multiple wavelengths transmission, EDFA’s are essential for loss compensation in optical fiber communications.

For wavelength-division multiplexed (WDM) systems, gain-flattened EDFA’s are required in order to have a uniform gain and to reduce optical signal to noise ratio (OSNR) for all wavelength within the amplification bandwidth [1]. In addition to this, the variation of span losses has pushed the gain-flattened EDFA to be more intelligent by changing its gain value while maintaining its gain flatness [2]. This gain-control feature is also critical to combat the dynamic behavior of surviving wavelengths in the environment of add/drop wavelengths in the transmission systems [3]. Normally, the input and output ports of these EDFA’s are monitored by photodetectors to be fed into the gain controller so as to adjust the gain according to the set value.

Owing to the optical amplification process, the broadband noise of amplified spontaneous emission (ASE) is naturally generated within the signal wavelengths. Therefore, the monitored power at the input and output ports of the EDFA consists of the signal power together with ASE power. Photodetectors are unable to distinguish between these two optical components. Since the gain calculation is based on these input and output optical powers, therefore, it creates another research problem that need to be tackled in order to maintain the signal gain at the desired value.

There are some efforts to manipulate this ASE information to control signal gain by filtering some of ASE powers within the amplification bandwidth of EDFA [4,5]. Nevertheless, their technique is not suitable for multiple wavelengths because the spectral filtering causes unusable spectral bandwidth for signal amplification. On the other hand, out-of band ASE filtering was also proposed in [6]. A bandpass filter is placed before a photodetector at the input and output ports to monitor the ASE power at the noise level. This approach requires additional optical component to measure the input and output powers. This can be performed by splitting a portion of input powers to another photodetector and thus, reducing the detected signal powers to a lower power level (closer to the photodetection noise level). Alternatively, the ASE correction algorithm for automatic power control has been proposed recently by measuring the collected power at the output port of EDFA [7]. In the published work, the output power is calculated based on the algorithm which requires a few calibrated constants beforehand for single and multi-channel operations. Thus, the complexity of the approach has motivated us to propose a simplified approach of ASE correction for multiwavelength amplification.

In this paper, we propose a simple ASE correction algorithm for variable gain-flattened EDFA. The proposed technique can be achieved by adding the ASE correction factor at the desired gain value and its functional curve can be universally employed by any desired gain values within the specified dynamic gain range (between maximum and minimum gains).

2. Amplifier specifications and design

The design specifications of the proposed variable gain-flattened EDFA are tabulated in Table 1 . The desired optical gain bandwidth is 35 nm that covers the range from 1529 nm to 1564 nm. This 35-nm bandwidth can support up to 44 channels with 100 GHz spacing. The variable gain ranges from 15 dB to 30 dB with its maximum output power of 23 dBm. For each gain value, the input power dynamic range is 19 dB that covers from −7 dBm to −26 dBm (30-dB gain) and from −11 dBm to 8 dBm (15-dB gain). The overall gain flatness of 1.5 dB is expected within the specified gain dynamic range. Finally, the maximum power into the mid-stage should be less than 16 dBm. The mid-stage is placed between the first-stage amplifier and the second-stage amplifier as shown in Fig. 1 .

Table 1. Design specification of the variable gain-flattened EDFA.

View Table

 

Fig. 1 Block diagram of the 4-stage Erbium-doped fiber amplifier.

Download Full Size | PDF

Referring to Fig. 1, the amplifier design consists of four amplifier stages that distribute losses of three core optical devices: dispersion compensating module (DCM), variable optical attenuator (VOA) and gain-equalizing filter (GEF). The DCM is used to compensate for the accumulated fiber dispersion within a transmission span. Since the loss of DCM has variations, thus the maximum allowable mid-stage loss is fixed to 10 dB. This is critical in order to maintain the gain-flattened operation of the amplifier. On the other hand, the VOA is utilized to vary the operating gain-flattened value from 15 dB to 30 dB, and finally the GEF is employed to have a flat gain with tolerance of about ± 0.75 dB. The transmission spectrum of the GEF is shown in Fig. 2 . The maximum loss of the GEF is about 11 dB around 1557 nm wavelength range.

 

Fig. 2 Transmission spectrum of GEF used in the 4-stage EDFA.

Download Full Size | PDF

3. Experimental procedure, results and discussions

In this research work, we use a fully automated test bed to analyze the performance of the variable gain-flattened EDFA. The test signal consists of 40 channels from 1530.33 nm to 1562.42 nm covering 32.1 nm. Throughout the experiment, all the 40 channels are used and calibrated by measuring its total power based on the summation of peak signals by utilizing the optical spectrum analyzer (OSA). In our study, the input signals and amplified signals are measured using OSA. Then the signal gain is automatically calculated based on these information’s recorded in the automated test bed system. Once the desired gain is fulfilled, then the total composite power is measured using the optical power meter (OPM). This is to emulate the practical condition of the amplifier in which the photodetector is utilized to detect the signal powers at the input and output ports. The examples of output spectrum at 30 dB operating gain are shown in Fig. 3 . For −19 dBm input signal power, the total output powers from OSA and OPM are 11.70 dBm and 11.14 dBm respectively. On the other hand, for −26 dBm signal power, the total output powers from OSA and OPM are 4.16 dBm and 6.39 dBm respectively. Based on these findings, the discrepancy between the measurement of OSA and OPM is due to the contribution of broadband ASE. The OPM cannot differentiate between peak powers and noise floor thus, it measures the composite power. Therefore, the reading taken form the OPM is always higher than that from the OSA. The gain penalty increases as the signal power decreases due to the amount of ASE generation that is highly dependent on the number of incoming photons at the signal wavelength. For small signal powers, the generation of ASE is greater due to less number of incoming photons. If the OPM is used in the practical field, the average signal gains are 0.55 dB and 2.23 dB lower than the targeted gain (G_T) for the signal power of −19 dBm and −26 dBm respectively.

 

Fig. 3 Output spectrum at 30 dB gain with (a) −19 dBm total signal power and (b) −26 dBm total signal power.

Download Full Size | PDF

The relationships between the output signal power and input signal power based on these two measurement approaches are depicted in Fig. 4 . In our analysis, the agreement between these two sets of measurement faltered when the discrepancy is more than 0.1 dB. For the low desired gain of 15 dB, it is clearly seen that the level of agreement between these two measurements is very high except for signal powers less than −9 dBm. On the other hand, the difference between these two measurements is more severe for the desired gain of 30 dB for the signal power of less than −12 dBm. In order to analyze this discrepancy, the difference between these two sets of measurement is plotted at different signal gain values as illustrated in Fig. 5 .

 

Fig. 4 Relationship between output signal power and input signal power.

Download Full Size | PDF

 

Fig. 5 Gain penalty at different gain levels due to ASE contribution.

Download Full Size | PDF

Since the total power measured by the OPM is higher than the total power of signal peaks measured by the OSA, thus the discrepancy is contributed by the amount of ASE that cannot be filtered out by the OPM. This discrepancy is named as the gain penalty as shown in Fig. 5. For each desired gain (G_D) value, the gain penalty decays exponentially with respect to the input signal power. In general, the gain performance due to the ASE contribution is severely degraded for small signal gain up to 2.2 dB. These findings show that the actual signal is lower than the desired gain value when the input and output signal powers are measured by using photodetectors. Therefore, the actual gain value is underestimated in the presence of broadband ASE. The discrepancy is getting smaller in tandem with the targeted gain value. From these experimental results, the requirement of ASE correction is critically needed as previously reported in Ref [7]. In order to rectify this problem, we propose a simplified version of ASE correction by taking the average gain penalty for all the signal powers within the input signal power dynamic range from −26 dBm to 8 dBm. This average gain penalty (ΔG_avg) is represented by the line as depicted in Fig. 5.

By utilizing this approach, the gain penalty value will be added into the desired gain value in the gain-clamped algorithm; G_T = G_D + ΔG_avg. Thus, the actual (targeted) output power will be greater than the desired value in order to compensate for the contribution of ASE. The expected gain penalty after implementing this ASE compensation algorithm is depicted in Fig. 6 . The gain penalty variations are in the range from −0.10 dB to 0.15 dB. These results are comparable with the ones reported in Ref [7]. Finally, this algorithm is implemented in our variable gain EDFA and selected experimental results are depicted in Fig. 7 .

 

Fig. 6 Expected gain penalty at different gain levels due to ASE contribution.

Download Full Size | PDF

 

Fig. 7 Gain performance of 4-stage EDFA with ASE compensation algorithm.

Download Full Size | PDF

Referring to Fig. 7, four different gain levels are selected; 15 dB, 20 dB, 25 dB and 30 dB. In addition, the extreme signal condition is tested by taking the input power equivalent to the minimum and maximum signal powers for each gain level. The average signal gain at the desired gain value has been successfully obtained without any significant power penalty from the broadband noise of ASE. The gain variation is due to the profile of GEF used in the 4-stage EDFA. For the 15 dB operating gain, the signal gains from 1530.33 nm to 1531.90 nm are slightly higher than the rest of the signals. This phenomenon is due to the effect of spectral hole burning as reported in Ref [8–10]. The experimental findings indicate that our simplified ASE correction algorithm for variable gain-flattened EDFA is useful to compensate for the contribution of ASE onto the signals.

4. Conclusion

We have successfully demonstrated a simplified algorithm for correcting ASE of variable gain-flattened EDFA. The contribution of ASE is significant in the gain-control algorithm of EDFA because the photodetectors measure composite signal powers at both input and output ports. Therefore, the signal gain is under compensated which leads to the gain penalty. The ASE correction factor is obtained by taking the average of the gain penalty at all input signal powers and gain values. By doing this, the targeted gain value is the summation of the desired gain and the ASE correction factor. Thus, a simplified gain-control algorithm with ASE correction factor is proposed and tested experimentally. The experimental results indicate improvement of the gain penalty at the small signal powers with the maximum error of around 0.15 dB. The proposed technique is simple and suitable for any rare-earth doped fiber amplifiers and could be easily implemented in practical optical amplifiers as the upgrading option.