extinction ratio could be increased further by a proper choice of splitting ratios of the couplers [10]. The wavelength resolution limit of the OSA also limits a precise measurement of the actual extinction ratio. The minimum possible change in the FSR during tuning of the MZI by this technique was somewhat difficult to estimate due to the following reason: the smallest real-time change in the FSR during tuning was much less than the resolution of the OSA. Further, the effect of this small change was difficult to record during the tuning process because of short-duration smallscale fluctuations that take place in the online response of the MZI in real time. The effect of the increase in loss during this process was also visible in the output spectrums (compare the reference level RL in Figs. 3 and 4). The difference between the peak-topeak powers at the two complementary output ports increases due to the losses that occurred during the tuning process. This can be exploited to our advantage: if the first coupler is not a perfect 3-dB one, then by introducing appropriate loss in one of the arms of the MZI, equal power distributions at the output ports can be achieved. CONCLUSIONWe have presented a simple technique to tune the channel spacing and channel wavelengths of an all-fiber MZI-based wavelength interleaver. This technique involves selective heating and controlled stretching of one of the arms of the MZI. We have demonstrated the applicability of the technique by fabricating a singlestage MZI-based wavelength interleaver with a stable wavelength response. Using this technique, the results of successfully realized wavelength interleavers with channel wavelengths conforming to ITU grids and channel-spacing values of 25 and 50 GHz were also presented.

No abstract