Timing Extraction in Multiplexed Multichannel High Speed Fiber Optic Transmission System

This invention relates to multiplexed multi-channel high speed optical transmission systems, and more particularly to a technique for directly extracting and deriving a large amplitude timing signal from the received signal in such systems.

In very high speed multi-Gbit/s fiber optic transmission systems onto which plural data channels are time-division multiplexed, the usual first step to access the individual data channels at the receiving end is to recover the timing signal from the received signal for purposes of establishing synchronization, for separating the plural data channels, and for detecting the received data bits.

For the widely used pseudorandom non-return-to-zero (NRZ) signal format, the data spectrum is a broadband continuum which does not contain sufficient energy at the clock frequency. In order to generate a timing signal at the receiving end from such a data stream two major steps are required. First, nonlinear signal processing steps such as transition enchancement (differentiation/squaring, multiplication), and/or other nonlinear treatment of the received signal are necessary to generate a relatively weak "reference" clock signal. (See for example, "Digital PCM Bit Synchronizer and Detector", by A. E. Moghazi, G. Maral, and A. Blanchard, IEEE Trans. Comm., COM-28, pp. 1197-1203, August 1980.) In a second stage, filtering (for example, SAW [surface acoustic wave]filtering), amplification, phase comparison (for example, phase lock loop [PLL]), and hardlimiting are performed to condition the clock before it can be applied to a decision circuit. (See for example: "2 Gbit/s timing recovery circuit using dielectric resonator", by D. J. Millicker and R. D. Standley, 1987, Elect. Lett,. 23, pp. 738-739; "New proposal for Multigigabit/s Clock Recovery IC Based on Standard Bipolar Technology", by D. Wang and U. Langmann, 1987, Electr. Lett., 23, pp. 454-56; and Digest, OFC '89, by K. Runge, et al, paper # WN2, Houston, Tex., 1989.) Several methods of generating discrete clock components from the received signal are well developed and commercial high-speed clock recovery circuits are available for frequencies up to 2.4 Gbit/s rates and have been implemented in off-the-shelf Gbit/s transmission systems.

For very high-speed and multi-Gbit/s systems, the transmitted signal is constructed by time division multiplexing, in either the electrical or the optical domain, incoming lower speed channels. The signal on each incoming lower speed channel and the multiplexed signal generally has a return-to-zero (RZ) signal format. In the RZ signal format, each time-slot contains either a pulse of finite width (less than the width of the time-slot), representing a binary ONE, or no pulse, representing a binary ZERO. Multiplexing with the RZ format is the common approach in multi-Gbit/s communication systems to increase the network throughput and in fiber optic systems to take more efficient advantage of the vast fiber bandwidth. (See for example, "1.13-Gbit/s Lightwave Transmission System" by K. Y. Maxham, J. M. Dugan, M. A. McDonald, and C. R. Hogge, J. Lightwave Techn., Lt-5, No. 10, pp. 1510-1517, Oct. 1987; "Multi-Gbit/s Picosecond Optical Pulse Transmitter Experiment" by H. Izadpanah, and A. Albanese, 1988, Conf. Digest CLEO '88, Anaheim, CA, pp. 128-129; and "Optical TDM for Very High Bit-Rate Transmission", by R. S. Tucker, G. Eisenstein, and S. K. Korotky, 1988, J. Lightwave Techn,. LT-6, pp. 1737-1749.) With a pseudorandom RZ signal format, the multiplexed data stream spectrum contains some energy at the desired clock frequency. It is, however, fully embedded in the continuous part of the frequency spectrum of the data signal. Processing stages are thus required to "clean up" (i.e., rejection of the power in the continuous spectrum outside the immediate region surrounding the clock frequency) and to amplify the reference clock for the second stage of processing.

An object of the present invention is to modify, at the transmitter, the spectrum of the multiplexed pseudorandom RZ signal so as to enhance the discrete clock component while simultaneously depressing the continuous component at that frequency to allow the enhanced clock component to be used, at the receiver, as the "reference" timing signal. The aforenoted preprocessing stages in the receiver needed to produce a reference clock signal can then be eliminated.

A further object of the present invention is to directly obtain, at the receiver, from this discrete clock component, the necessary large amplitude clock signal without the conventional multipart second stage processing circuitry needed in the aforenoted prior art timing recovery circuits.

A feature of the present invention is that the signal processing steps for clock recovery at the receiver can be performed in a fewer number of stages leading to a reduction of the receiver design complexity and a concomitant reduction in overall receiver cost.

In the transmission system of the present invention, time-division multiplexing in the electrical or optical domains of plural pseudorandom RZ format data channels is controlled in such a manner that the position of the data pulse of each channel within the time slot assigned to each data channel is not uniform from time slot to time slot but is set unevenly in a predetermined pattern. In particular, the position of the pulse of each channel within each time slot is offset from the beginning or the end of the time slot by a predetermined fixed delay, the position of the data bits in pairs of adjacent time slots being mirror-imaged with respect to the common time point of the pair. Thus for an N channel system (N being even), pulses in channels 1,3. . . ,N-1 are offset by a fixed delay of t.sub.d from the beginning of their respective time slots, and pulses in channels 2,4, . . . , N are offset by the same fixed delay t.sub.d from the end of their respective time slots. As a result of this controlled multiplexing, the frequency spectrum of the resultant interleaved data stream will thereby have an "enhanced" discrete component at the clock frequency surrounded by a spectral "dip" or "hole" in the continuum. By properly selecting t.sub.d relative to the length of the time slot, the discrete power at the desired clock frequency can be optimized. In accordance with the present invention, the final, jitter-free large amplitude clock signal is derived at the receiver from the discrete frequency component at the clock frequency in one single step by an injection-locking technique.

Advantageously, the present invention obviates the need for preprocessing circuitry at the receiver to derive the clock component. Furthermore, the complexity of the second stage processing circuitry is substantially reduced. Receiver design is thus improved with a reduction in overall cost .

FIG. 1 shows the power spectral density of a conventional return-to-zero data signal;

FIG. 2 is shows the multiplexing arrangement of the present invention in the transmitter of an N-channel time-division-multiplexed system;

FIG. 3 shows a channel timing diagram of the output of the multiplexing arrangement in FIG. 2;

FIG. 4 shows the power spectral density of the output of the multiplexing arrangement in FIG. 2;

FIG. 5 shows the variation of the power of the discrete component at the clock frequency as a function of the normalized timing delay, t.sub.d ;

FIG. 6 shows the normalized continuous power spectral density of the received signal for the optimum delay; and

FIG. 7 shows the circuit arrangement at the receiver for clock recovery by injection-locking.

With reference to FIG. 1, the power spectral density of a conventional return-to-zero data signal is shown. As can be noted, the spectrum contains a discrete energy component at the clock frequency f.sub.c =1/T.sub.c where T.sub.c is the clock period. This discrete component, however, is fully embedded in the continuous part of the data spectrum and cannot recovered without preprocessing of the signal at the receiver.

In accordance with the present invention, multiplexing (in the optical or electrical domains) at the transmitter is controlled in such a manner to enhance the desired discrete component of the power spectrum and to depress the continuous component at the clock frequency. FIG. 2 shows the controlled multiplexing arrangement at the transmitter of an N-channel time-division-multiplexed system. Data signals in the RZ signal format are present on each of the N input channels 201-1-201-N. Each of the N input channels 201-1-201-N is connected through a channel delay control 202 to a multiplexer 203. Multiplexer 203 time division multiplexes in a standard manner one time slot from each of the N channels, the time slots being at the a clock period T.sub.c and the multiplexer having a multiplexer cycle time of T.sub.m where T.sub.m =N.times.T.sub.C. The outptu 204 of multiplexer 203 contains the multiplexed data stream.

With reference also to FIG. 3, channel delay control 202 offsets the center of the position of the channel signal of the pulse within each time slot by a fixed time, t.sub.d, from either the beginning or end of each clock time slot. In particular, the center of each channel pulse with an amplitude A and pulse width .sigma. (.sigma. being much shorter than the clock period T.sub.c) is delayed in time slots 1, 3, 5, 7,. . . , N-1 (for N even) by a time t.sub.d from the beginning of the time slot. Each channel pulse is delayed by the same t.sub.d from the end of the time slot in time slots 2, 4, 6, . . . , N. Thus in each pair of time slots, 1 & 2, 3 & 4, . . . , N-1 & N, the position of the data bit forms a mirror image at the pair's common time point. With reference back to FIG. 2, this is readily accomplished by channel delay control 202, which includes N delay elements 205-1-205-N and which are connected to input channels 201-1-201-N, respectively. Delay elements 205-1, 205-3, . . . , 205-(N-1) (for N even) have a delay of t.sub.d and delay elements 205-2, . . . 205-N have a delay of T.sub.c -t.sub.d.

It can be shown that the power spectrum of multiplexed random data sequences will have an enhanced discrete component at the desired clock frequency f.sub.c =1/T.sub.c and lowers the power content in the dc and low frequency part of the data spectrum, as is shown in FIG. 4. By performing a spectral analysis of the controlled multiplexing scheme, the discrete power component can be optimized by selection of the adjustable parameters t.sub.d and the pulse width, .sigma.. For the sake of analysis, the optimization of the controlled multiplexing scheme for two data channels is presented hereinbelow. The analysis can then be applied to a higher number of channels. It is assumed that each channel signal is a random data sequence consisting of uncorrelated identically distributed ZERO and ONE symbols. It is also assumed that the transmitter maps the information of channel one into the signal s.sub.1 (t) delayed t.sub.d from the beginning of the time slot, and channel two information into the signal s.sub.2 (t) delayed t.sub.d away from the end of the bit time slot (i.e. at the position T-t.sub.d), respectively. The corresponding Fourier transform components S.sub.1 (f) and S.sub.2 (f) are written as:

The Power Spectral Density of the multiplexed signal received at the receiving end of the channel is given by: ##EQU1## In order to assess how much energy there is at the clock frequency, the following ration is computed: ##EQU2## By using the first and second terms in the summation of expression (1) yields ##EQU3##

The computed pair (.alpha.,.beta.) that maximizes the clock power are .alpha..apprxeq.0.185 and .beta..apprxeq.0.092. FIG. 5 shows the variation of R.sub.1 as a function of .beta. when .alpha. is set equal to 0.185. It can be noted that at the optimal value, 23% of the total power is in the fundamental clock signal.

FIG. 6 shows the normalized continuous spectrum for the optimal values of .alpha. and .beta. noted above. At the clock frequency (i.e. fT.sub.c =1), the continuous spectrum has a "dip" resulting in a low jitter clock extraction. Since the discrete clock component (as the reference clock) now exists in the received power spectrum, the first stage of clock processing is no longer required. Rather, the second stage of processing can be performed by using a physically realizable narrow bandpass filter followed by amplification. In accordance with the present invention, the received reference clock is injected into an electronic oscillator the output of which will be the final clock signal. The injection locking mechanism is a well known phenomenon and is widely used for synchronization (see, for example, "Injection Locking of Microwave Solid-State Oscillators", by K. Kurokaw, 1973, Proc. Ieee, 61, pp. 1386-1410). As in any oscillator, the oscillation builds up from the noise and parasitic signals. For the injection-locked case, when the injected signal frequency is near the center frequency of the oscillator resonant circuit, the onset of oscillation will be triggered by injection rather than noise. In this way, the oscillator frequency and phase are coupled to the exact frequency of the injected signal and, simultaneously, phase-synchronized by injection. The injection locking employed herein is equivalent to a narrow bandpass filtering of the received data spectrum (within the oscillator resonance range), while at the same time, providing large signal amplification (with a typical gain of.apprxeq.100). Thus, in one step, a low jitter, large amplitude, in-phase clock is obtained by injection-locking without requiring additional signal processing (see, for example, "Direct timing extraction in a modified-Manchester coded picosecond optical pulse fiber optic transmission system" by the inventor herein, 1988, Electrn. Lett., 24, No. 18, pp. 1151-1152).

With reference to FIG. 7, the circuit for recovering the timing signal from the received signal is shown. A receiver front end circuit 701 accepts the received signal on input 702. In the optical domain, receiver front end circuit 701 accepts the input from an optical fiber and the receiver front end includes a photodetector for converting the optical signal to an electrical signal, and a low noise wideband amplifier. In the electrical domain, receiver front end circuit 701 accepts the input via an antenna and includes a low noise wideband amplifier. As a result of the controlled multiplexing, the receiver can be ac coupled due to the absence of a dc component and the low power content at low frequencies. The resultant electrical output of circuit 701 is divided by power divider into two portions. A first portion is applied to the injection-locked oscillator 704 which generates on oscillator output 705, as detailed hereinabove, the large amplitude phase-synchronized clock. This clock signal is applied to demultiplexer 706 together with the second output 707 of power divider 703 which contains the received multiplexed data signal. The clock signal causes the multiplexed channels in the data signal on 707 to be separated and directed onto the individual N output channels 708-1-708-N.

Other embodiments of the receiver circuitry which recovers the large amplitude clock signal and demultiplexes the received channel can also be implemented. For example, in the optical domain, an incoming optical signal that has not been converted to an electrical signal can be used to directly trigger an oscillator (e.g. a phototransistor oscillator) which by injection-locking generates a large electrical clock signal. This electrical clock signal is then used for demultiplexing, in the electrical domain, the received signal after is has been converted from optical to electrical domains.

Alternatively, for directly processing in the optical domain, the incoming optical signal can be applied to an optical demultiplexer (such as a LiNiO.sub.3 switch) which is controlled by an electrical clock signal generated by the optical injection-locked oscillator, as described above. The optical demultiplexer then separates the input optical signal into plural optical output channels.

The above-described embodiments are illustrative of the principles of the present invention. Other embodiments could be devised by those skilled in the art without departing from the spirit and scope of the present invention.