Techniques for Clock Recovery

significant penalties. The NLOA, due to its nonlinear behaviour, can be used to generate the missing clock component to the NRZ data modulation spectrum. In addition the coupling of the photon and caixier densities results in an electrical clock component being generated at the contacts which can be directly filtered without the need for broadband electronic nonlinear circuits. The two-contact device can also operate as a self-pulsating laser (SP-LD), emitting pulses at a specific frequency, when biased above threshold. By suitably treating the absorber section such SP-LD devices can operate at frequencies > 5 GHz. All-optical clock recovery operation with RZ data will be demonstrated at 5 Gb/s within a 20 Gb/s OTDM system demonstrator. We shall discuss the locking behaviour in detail looking at the lock-up time, the purity of the locked clock and the duration that the clock remains suitable for BER measurements. Operation with NRZ data is also demonstrated by combining the SP-LD behaviour with the clock generation function of the NLOA device.

In section 5.2 the various types of clock recovery that are presently used in transmission systems are discussed and a summary given of some of the demonstrated optical techniques. In section 5.3 the clock generation properties of the NLOA are discussed, BER results are described and subsequent generation of suitable reset pulses for shortening the fall­ time of opto-electronic switches (as described in section 3.8) is demonstrated. In section 5.4 we shall look in detail at the operation of SP-LD devices and consider the variation of pulsation frequency with bias and evaluate the ways of triggering pulsations from two-contact devices. This section gives a detailed description of the operation of such devices in response to RZ data signals, including analysis of the locking time, the locking range, the locked bandwidth as well as BER system measurements. In section 5.5 the operation of SP-LD devices with NRZ data will be investigated both with a single SP-LD device and with a combination of NLOA and SP- LD devices. In section 5.6 we shall discuss these issues and the further work that is required and then draw conclusions.

5.2 Techniques for Clock Recovery

In traditional transmission systems synchronisation of all switching equipment has been achieved using electronic clock recovery techniques. In NRZ data format the clock frequency

Ch 5 Techniques for Clock Recovery

component is suppressed, unlike in RZ format, and therefore must be generated from the data before clock recovery can be achieved. The clock component can be generated by using a nonlinear electrical device, such as a frequency doubler, an exclusive-OR gate or some other suitable circuitry. Such techniques are generally implemented after the optical receiver and this clock is then used to synchronise the multiplexing and demultiplexing equipment in the rest of the switch. The generated component is then filtered electrically using filters such as surface acoustic wave (SAW) devices or phase-locked-loops (PLL) both these techniques are relatively untunable (SAW filters are very untunable although PLL do have a reasonable range, ^ 10 % of the clock frequency, depending on their design).

Although methods of electronic clock recovery at Gb/s rates have been demonstrated [5.2,5.3] for clock distribution or optical demultiplexing and drop&insert in OTDM systems and networks such techniques would require subsequent opto-electronic conversion to generate an optical clock signal. Broadband high power electronic amplifiers, suitable for driving the optical pulse source can be difficult to fabricate and expensive. The narrowband electronic filter technology, such as SAW devices or PLL’s, is commercially available at speeds up to - 2.5 GHz (SAW) and higher (PLL) using available high performance commercial components. This approach is at present at the leading edge of technology. Optical techniques, where an optical input signal synchronises an optical output clock signal, are also at the leading edge of technology but are potentially bit-rate flexible and compatible with the switching technology within advanced optical networks..

Recent advances in YIG oscillators have shown wide tuning ranges and high frequency operation [5.2] and recent work on PLL have demonstrated 10 GHz clock recovery from 40 Gb/s data [5.3]. However there is therefore a potential advantage in adopting all-optical methods of extracting the clock signal from a data stream. Optical techniques are inherently broadband and can offer wide tuning ranges. Development of such optical clock extraction circuits is in its infancy but a number of options have already been demonstrated. Their main drawback has been the complexity of the optical cncuits involved. Optical PLL’s [5.4] comprise many different components including electronic elements thus removing much of the transparency potentially

Ch 5 Techniques for Clock Recovery

required by future systems. Recent work has demonstrated this approach in an 50 Gb/s OTDM based on a 8 Gb/s linerate with the clock driving a modelocked laser and demultiplexing using nonlinear loop mirrors [5.5].

All-optical clock recovery using TE/TM mode conversion has been demonstrated at speeds of 3.2 GHz using a semiconductor laser amplifier in an external cavity [5.6]. This type of circuit is also complicated and not readily tunable although it is potentially capable of higher speed operation but utilises the saturation of gain in a semiconductor device due to high input signal powers. The output of this circuit pulsates and these pulsations can be synchronised to provide an optical clock signal when RZ input data is injected into the device.

Another method which is rather easier to implement uses an optical FP filter as an optical ‘tank’ circuit equivalent to the SAW devices in the electronic domain. In this case the free-spectral-range of the bandpass filter matches the clock frequency. Given that the input RZ data signal is also aligned to one of the resonant orders of the filter the high finesse cavity only passes the clock component in the input data spectrum. Jinno et al. have demonstrated operation at 2 GHz but the principle can be extended to higher clock frequencies [5.7]. One potential drawback for this technique is that the finesse, or quality Q, of FP filters is limited to ~ 100. Higher finesse can be achieved but such devices would require excellent stabilization characteristics to avoid problems with drift in the passband wavelength.

The Q of the cavity dictates the purity of the clock signal extracted. In electronic systems a quality of Q ~ 800 is often specified to avoid excessive phase noise while maintaining a reasonably fast lock-up time of the clock recovery circuit [5.8]. Any phase noise produces clock jitter with the jitter transfer function of the circuit linked to the filter quality and passband shape; the pairs of data modulation sidebands adjacent to the clock component do not cancel if the passband is asymmetric leading to phase noise on the resultant clock. As such it is difficult to see optical FP filter technology meeting the specifications at present, but the ability for integration within the semiconductor environment may lead to higher finesse designs in the future.

A potentially simple method of optical clock recovery uses self-pulsating semiconductor laser devices (SP-LD). Jinno et al. first demonstrated optical clock recovery using such devices