3 Departamento de Electrónica Industrial, Universidade do Minho

4810 Guimarães, Portugal,

Tel. +351 53 510190, Fax +351 53 510189

1. Introduction

The research and development of optical networks is of interest worldwide for future interactive, flexible broadband communications services. In high bit rate network systems, time domain optical signal processing techniques such as optical time division multiplexing (OTDM), demultiplexing and time extraction are indispensable to overcome the electric processing speed limitations^1,2. One important function in communications switching and transmission is synchronization. M. Jinno³ demonstrated for the first time the use of self pulsating laser diode for all optical timing recovery. Barnsley⁴ have also demonstrated optical timing recovery at 5Gbit/s in use in a 20Gbit/s optical time division multiplexed system. Mode-locking was used by Smith⁵. Two sections DFB laser was used by D.J.As⁶ with dispersive self Q-switching for 18Gbps all optical clock extraction. Timing extraction or clock recovery must produce a low jitter timing clock synchronized to the input optical signal because the extracted clock should be distributed to, for example, demultiplexers, routers, channel selectors and receivers.

 

 

2. low bit rate Experimental set-up

The experimental set-up, shown in fig. 1, was originally built upon using one DFB semiconductor laser. The chosen device had an active length of 250m m, k .L=0.25, one facet is AR coated and the other has 30% reflectivity and operates at 1320nm. Threshold current is 47mA. This laser was driven by a fast switching current circuit. The data generator is just a laser with directly modulated data. A 95:5 optical power splitter was used to combine the two signals. Optical power in the oscillator input was 75m W.

The fast current switch was designed to switch currents between 0 and 180mA in less than 50ns. It integrates two modified Widlar current sources to permit the interface with commercial current sources that are very slow when compared with our specifications. The switching circuit, itself, is based on the ECL gate principle adapted for our goals. It is composed by a pair of matched bipolar transistors in a differential configuration. The digital part is responsible for the generation of a debouncing TTL signal suitable to drive the switching circuit. It integrates a normal push-button and a fast JK flip-flop.

Oscillation in the laser was provoked by the conjugation of two physical processes, namely the fast switching of the bias current from I_treshold to I_treshold+D I, with I_treshold = 47mA and D I =85mA, accomplished by the reflections feedback into the ferrule lens. The distance between the laser facet and the ferrule was close to 500 m m. The laser oscillates near 2GHz as can be seen in fig. 2.

Fig. 1. Experimental setup	Fig. 2. Pulsation near 2GHz without input data

 

After achieving self-pulsation near 2GHz, a input data sequence (10101010) was introduced in the device and we saw that the laser leaves is natural frequency starting to pulsating at the data frequency (2.2GHz) as can be seen in fig. 3. The device was sensitive to polarization. Switching the data sequence to 1.1GHz, a sub-multiple of the clock frequency, we have noticed that the laser goes on pulsating at 2.2GHz, as reported in fig. 4.

 

Fig. 3. Synchronization with a 2.2GHz data sequence	Fig. 4. Synchronization with a 1.1GHz data sequence

 

As a result of the blocking phase, the output suffers a contraction at the clock frequency. This contraction is due to the synchronization of the self-pulsation with the input data sequence, as shown in figs 5 and 6. This result demonstrates the capacity of this set-up to actually synchronize the pulsation with optical data.

Fig. 5. Spectrum with data at 2.2GHz	Fig. 6. Spectrum with data at 1.1 GHz

 

 

3. HIgh bit rate Experimental set-up

Simulation⁷ indicated that high speed operation requires more optical power within the cavity. Unfortunately, a simple upgrade of the already built set-up with the new MQW devices, did not produced the expected results. With those more power devices is was not possible to achieve stable and tunable pulsation frequencies near 10GHz because the losses in the air interface became intolerable. To overcame this problem a new experimental set-up configuration was projected and built (see fig. 7).

 



Fig. 7. Second configuration of the experimental set-up

The experimental set-up was originally built upon using one Multiple Quantum-Well (MQW) distributed feedback (DFB) semiconductor laser. The device operates at 1312 nm, having an active length of 250m m and a threshold current of I_th=11.2 mA. One facet is AR-coated and the other has 30% reflectivity. The light is collected from the laser by a lensed fiber tapper. It was verified that the device is very sensible to the light polarization.

Adjusting the distance between the AR laser facet and the fiber ferrule for achieving maximum optical coupling, the frequency of self pulsation can be modified simply adjusting the bias current (see fig. 8). For lower currents it was verified that the self pulsation frequency is near 6GHz and the device is stable. For higher currents the self pulsation frequency increases to near 11GHz. The device shows some jitter for high frequencies as a consequence of the spontaneous emission.

The RF spectrum of the self pulsation behavior for different bias current is shown in fig. 9. It can be seen that for 30mA the frequency is approximately 5GHz and for 90mA it goes near 10 GHz. It can also be seen the second harmonic near 10GHz for 30 mA and near 20GHz for 90mA.

Fig. 8. Pulsation frequency as function of the bias current.	Fig. 9. RF spectrum for different bias currents

 

After adjusting the pulsation frequency to near 10GHz a RZ soliton shape input data stream was injected. From fig. 10 it can be seen that the system leaves the natural frequency and locks with the input signal. The shape of the output signal is however not narrow enough to drive the demultiplexer. This problem will be subject of further investigation.



Fig. 10. Synchronization with a 10GHz data sequence (all ones)

 

 

4. Conclusions

Two experimental set-up configurations were evaluated. The first one proved effective to optically synchronize a device self-pulsating near 2GHz with incoming data at 2.2GHz. A frequency deviation of as much as 10% was thus successfully bridged. Another important result was the synchronization with data at 1.1GHz, which is a sub-multiple frequency of the clock signal. This shows that not only equal data and clock frequencies can be processed but also sub-multiple and multiple integer frequency relations can be dealt.

High speed operation requires more optical power within the cavity, has indicated by simulation. Unfortunately, a simple upgrade of the already built set-up with the new MQW devices, did not produced the expected results. With those more power devices is was not possible to achieve stable and tunable pulsation frequencies near 10GHz because the losses in the air interface became intolerable. A new experimental set-up configuration was projected and built. With this configuration the frequency of self pulsation can be modified simply adjusting the bias current. For lower currents it was verified that the self pulsation frequency is near 6 GHz and the device is stable. For higher currents the self pulsation frequency increases to near 11GHz. Synchronization with an input data stream at 10GHz was shown.

5. REFERENCES