1. Introduction

The increase in the demand for optical transmission bandwidth has led to a great interest in developing components for wavelength division multiplexing (WDM) communications. Novel devices such as tunable multi-wavelength laser sources and filters are under rigorous investigation by different research groups [1–7]. A simple technique to construct a multi-wavelength source is to use a broadband gain medium together with an optical comb filter to select the lasing wavelengths. Among the different choices, a commonly used filter is the Sagnac interferometer consisting of a high-birefringence (Hi-Bi) fiber placed inside a loop mirror [8]. The device can support different comb spacings by changing the effective Hi-Bi fiber length [2], or by cascading different pieces of Hi-Bi fibers and adjusting the polarization rotation angles of light between them [6]. Shifting of the optical comb has also been demonstrated by polarization and temperature adjustments [3, 7]. In these tuning schemes, the characteristic of the comb filter can be controlled by electrical, mechanical, or temperature means. In this work, we demonstrate an alternative tuning approach that is applicable for optical control in a remote node. By incorporating a semiconductor optical amplifier (SOA) in the Sagnac loop, the polarization of the probe light can be tuned via birefringence modification with a control light [9–12]. Using this approach, we have demonstrated a comb shift up to 0.35 nm, corresponding to over 59% of the comb spacing. The shift increases monotonically with the power of the control light until saturation occurs. Owing to the fast response of birefringence change in the SOA, the comb can be dynamically tuned and operation at the GHz frequency range can be expected [10]. Such a fast response can hardly be attained by existing tuning schemes that depend on mechanical adjustment or piezoelectric control of the length and birefringence of the Hi-Bi fiber in the Sagnac loop. Hence, the SOA tuning approach can be advantageously used in ultrafast processing of lightwave signals and in dynamic switching of wavebands [13].

2. Principle and experiment

The SOA used in our demonstration is a commercial product (Samsung OA40B3A-OG) with a maximum small signal gain of 25 dB, a polarization dependent gain of 1 dB, and a saturation output power of 7 dBm. The device length is 915 µm.

To study the birefringence change in the presence of a cw control light, we perform a preliminary experiment in which a cw probe light at 1554.5 nm and a control light at 1557.0 nm are combined and are directed into the SOA. The probe is allowed to pass through a pair of cross polarizers placed before and after the SOA. Initially, without the control, the probe has a very weak transmission. A control at 10 dBm power is then applied with its polarization adjusted until a maximum transmission of the probe light occurs. With the optimized setting, we observe a contrast ratio of 29.6 dB for the probe before and after the application of the control light.

 

Fig. 1. Experimental setup. SOA : semiconductor optical amplifier; PMF : polarization maintaining fiber; PC1, PC2 : polarization controllers; OSA : optical spectrum analyzer.

Download Full Size | PDF

Next, the SOA is placed inside a Sagnac loop interferometer. Figure 1 illustrates the experimental setup. The setup consists of a 50:50 coupler, a polarization maintaining fiber (PMF), 2 polarization controllers (PC1, PC2), and a dc biased SOA. The cw control light is obtained from a distributed feedback laser diode (DFB-LD) at 1557.0 nm and is amplified with an erbium-doped fiber amplifier. The light is then launched into the SOA via a fiber coupler for optical modification of the device birefringence. To characterize the performance of the comb filter, an amplified spontaneous emission (ASE) source is used as the input. The PCs are employed to optimize the polarization of the light inside the interferometer and their settings are kept fixed during the measurement. At the beginning, the interferometer is operated without the control light. The input ASE source is split into two counter-propagating branches at the coupler. After propagating through the PMF and the SOA, the fields in the two arms interfere again at the 50:50 coupler. The phase difference between the two arms is mainly governed by the birefringence and the length of the PMF. The SOA, being a much shorter device with a smaller birefringence, does not contribute much to the total phase change. However, with the introduction of the control light, the birefringence of the SOA will change according to the injected power and the polarizations of both the control and the probe lights [9–10]. Although the change is not significant enough to modify the comb spacing, which is essentially determined by the PMF, it is sufficiently large to alter the light polarization and result in a shift of the output comb. The comb spacing in our setup is given by

where B_PMF and L_PMF are the birefringence and the length of the PMF. The total phase difference between the two branches can be expressed as

where B_e and L_e are the effective birefringence and the effective length of the propagation medium. The effective parameters depend on the birefringence and lengths of the PMF and the SOA, as well as the setting of the polarization controller between them. In the extreme cases, the maximum phase difference is the sum of those introduced by the PMF and the SOA, and the minimum value is the difference of the phase differences introduced by the two elements. In between the extreme cases, the comb will shift with the power of the control light. It is worth mentioning that although the SOA barely contributes to the total phase difference, its effect on the transmission function cannot be overlooked, since the function depends on the sinusoidal response of the total phase difference. The transmission function T(λ) is given by

We observe that by changing the applied bias on the SOA, it is also possible to control the device birefringence and result in a phase change. However, the gain of the SOA will experience a relatively large corresponding change. Using the control light as a means to tune the birefringence, the gain variation can be reduced. Since the change in the SOA birefringence will translate into polarization rotation of light propagating in the medium, the device can also be viewed as an effective means to tune the incident polarization angle with respect to the axes of the fiber. The influence of the incident polarizations of the control and probe beams on the amount of polarization rotation has been thoroughly studied in [10]. It is concluded that when both the control and the probe beams coincide with the TM axis of the SOA waveguide, the strongest polarization change will occur and result in TM to TE mode conversion. Under such an operation, the TE and TM axes are no longer the optical axes of the waveguide and thus the induced birefringence can lead to a change of the light polarization. The amount of induced birefringence, hence the phase difference in the SOA, is dependent on the asymmetry in the confinement factors and the effective TE and TM indices of the waveguide. Also, the birefringence can be adjusted by tuning the control beam power, which determines the extent of carrier depletion in the SOA.

3. Results and discussion

In our experiment, the SOA is dc biased to provide the maximum gain. The PMF introduces a birefringence of ~ 3×10^-4 at 1550 nm. Two pieces of PMFs with lengths of 6.7 and 13.4 m have been used to construct the Sagnac interferometer. Initially, the PCs inside the loop are adjusted to obtain a maximum peak-to-notch contrast ratio. Shown in Fig. 2(a) and Fig. 2(b) are the transmission spectra of the interferometer obtained with different PMFs. In the absence of the control light, the measured optical comb spacings are 1.2 and 0.6 nm, respectively. The values agree well with the results calculated from Eq. (1). When the length of the PMF is doubled, the phase difference is also doubled, resulting in a 50% reduction of the comb spacing.

 

Fig. 2. Transmission spectrum of the SOA-PMF loop mirror filter in the absence of the control light. (a) The length of the PMF is 6.7 m and the measured comb spacing is 1.2 nm. (b) The length of the PMF is 13.4 m and the measured comb spacing is 0.6 nm.

Download Full Size | PDF

The measured peak-to-notch contrast ratio of the filter reaches a value of 16.1 dB in the range around 1550 nm. Since the birefringence of the SOA is relatively small, the phase difference between the two counter-propagating branches of the probe light is predominantly determined by the length and the birefringence of the PMF. The observed result agrees well with our prediction in the previous section.

To investigate the tuning characteristic of the optical comb, a cw control light is applied to the SOA via a 30:70 coupler. A polarization controller and an attenuator are used respectively to set the polarization state and the power of the input. When the control is launched into the SOA, a shift of the optical comb is observed. Shown in Fig. 3(a) are the transmission spectra of the SOA Sagnac interferometer (with 6.7 m PMF) before and after the injection of the control light at a power level of 17.5 dBm. It is observed that the comb is shifted by 0.34 nm. The peak-to-notch contrast ratio remains larger than 15 dB. When 13.4m PMF is used in the Sagnac loop, a comb shift of 0.35 nm is achieved with 18.5 dBm control light power. The peak-to-notch contrast ratio is over 10.9 dB. The result is shown in Fig. 3(b). Here, the optical tuning range has covered 59 % of the comb spacing.

 

Fig. 3. Transmission spectra of the SOA Sagnac interferometer (a) with 6.7 m PMF before and after the application of a control light at 17.5 dBm. (b) with 13.4 m PMF before and after the application of a control light at 18.5 dBm

Download Full Size | PDF

The dependence of the comb shift on the control light power is studied by varying the injection level. Depicted in Fig. 4 are the results obtained from the Sagnac interferometers with 6.7 m and 13.4 m PMFs. As the control power increases, the change in the SOA birefringence becomes larger, hence contributing to a larger phase difference in the interferometer. Thus, the amount of wavelength shift increases. The shift starts to show a saturation behavior at about 0.32 nm, and the corresponding control power is 40 mW. The saturation is caused by the depletion of electrical carriers in the SOA at a large control power. Therefore, additional increase of the power does not contribute to any significant change in the device birefringence. To further increase the phase difference introduced by the SOA, one can either apply a higher bias current or use a longer SOA device. It is worth mentioning that although an EDFA is used in our experiment to boost the control light, DFB laser diodes with about 50 mW output power are commercially available and can be used directly to control the comb shift.

 

Fig. 4. Dependence of the wavelength shift on the control light injection power.

Download Full Size | PDF

It is noted that the two plots in Fig. 4 for different lengths of PMFs overlap each other. The result indicates that the birefringence change occurs only in the SOA, and is independent of the PMFs. For a given phase difference between the clockwise and the counter-clockwise propagating branches in the interferometer, the amount of wavelength shift is directly proportional to the change in the SOA birefringence, which is determined by the power level and polarization of the control light.

Compared to a recent demonstration on a tunable comb filter that depends on the adjustment of a half-wave plate when the settings of other wave-plates are optimized [14], our setup has the potential for high-speed tuning as no mechanical adjustment is needed. The lower contrast ratio of ≤16 dB in our work as compared to 25 dB in [14] is believed to originate from the amplified spontaneous emission noise and polarization dependent gain of the SOA.

To study the speed performance, we modulate the control light with a 10 GHz “0101” data using an electro-optic intensity modulator. The outputs at two different probe wavelengths are measured. The wavelengths are chosen to be at the transmission maximum and minimum of the comb in the absence of the control light. The results are depicted in Fig. 5. The ripples of the outputs in (b) and (c) are caused by undesirable optical reflections in the setup. The smaller contrast ratio obtained in (c) is caused by carrier depletion and hence gain reduction in the SOA when the pump light is “1”, even though the transmission increases due to the comb shift. To improve the contrast ratio, one can add a polarizer at the output to enhance the signal discrimination. It is worth mentioning that when the spacing between the pump and probe is less than about 5 nm, careful control of the polarizations is needed to avoid wave mixing effects in the SOA.

 

Fig. 5. Switching characteristic of the comb filter. (a) 10 GHz control data input “0101” at 1545.00 nm. (b) probe output at 1534.16 nm located at the transmission maximum when the control is “0”. (c) probe output at 1533.86 nm located at the transmission minimum when the control is “0”.

Download Full Size | PDF

4. Conclusion

In summary, we propose and demonstrate an optically controlled Sagnac loop comb filter. The filter contains a polarization maintaining fiber and a semiconductor optical amplifier that determine respectively the comb spacing and the comb shift with the application of a control light. In our experiment, the birefringence change in the SOA results in a maximum shift of 59% of the comb spacing. A saturation behavior is observed and is caused by the depletion of electrical carriers in the SOA at a given bias. The developed tuning approach enables a number of all-optical applications in multi-wavelength communications.