This invention relates to electrical oscillators and in particular to the synchronized of a number of electrical oscillators, and in a further aspect to phase-locked loops containing oscillators.
Most systems in which oscillators are required to be synchronized are in one of two categories. In the first category the oscillators are remote from one another and each performs a unique non-redundant function. Such a synchronized system exists in some navigation aids. In the other type of system the oscillators are located in close proximity and the plurality of oscillators is provided for reliability enhancement.
The present invention may find application in either type of system.
A method of synchronization for either type of system is to nominate one oscillator as the master and to phase-lock the remainder to it as slaves. While the method is satisfactory in many applications it nevertheless suffers from the disadvantage that synchronism between the slave oscillators depends on the continuing operation of the master.
This disadvantage is most often a serious consideration in the second type of system mentioned above, since, if the desired reliability enhancement is to be achieved, some provision must be made to keep the remaining oscillators locked together if the master oscillator fails. This disadvantage may for example appear in automatic telephone exchanges where the oscillators are included as a source of clocking signals. One procedure for dealing with failure of the master oscillator that has been proposed is to use a hierarchical arrangement of masters and slaves; the first slave in the hierarchy taking control if the master fails. To implement this technique fault detection and changeover mechanisms are required, and the reliability of such mechanisms is crucial to the operation of the system.
Such an hierarchical arrangement is described in U.S. Pat. No. 3,942,115 in the context of a satellite communications system.
Another control scheme for synchronizing the frequencies of geographically separate oscillators connected by communication links comprises the steps of averaging the phases received at each station from remote oscillators, comparing the result with the local phase, and applying the filtered error signal as a correction to the local oscillator. This technique is known as "mutual synchronization", and is described, for example, by Gersho and Karafin in Bell Systems Technical Journal, December 1966, pp 1689-1705.
Another method of synchronization with some similarities to the mutual synchronization scheme is referred to as `majority voting`. Such a system is described by Hopkins et al in Proc IEEE Vol 66 No. 10, October 1978, pp 1225 and 1231-1232. Control circuitry for an oscillator monitors the output from three other oscillators and all four are phase locked together. If one differs the control circuitry takes the majority signal of the other two and phase locks to that.
The present invention provides a system for synchronizing oscillators in which like the master and slave systems relies on each oscillator synchronizing to one another. There is however no master oscillator predetermined by the configuration and no hierarchy built into the configuration of the system. There is also no phase averaging or majority voting. Each oscillator in the system synchronizes itself to one other, chosen at random from available inputs. Each oscillator receives inputs from the other oscillators and is controlled such that it repeatedly attempts to synchronize with another oscillator until the system as a whole is synchronized.
A very limited non-hierarchical synchronization system is described in the specification for U.S. Pat. No. 3,662,277.
FIG. 1 of the accompanying drawings illustrates schematically the basis of the system in U.S. Pat. No. 3,662,277 and in a fault-free state each oscillator and its control circuitry is connected by an incoming and outgoing path to each of the others.
FIG. 2 shows diagrammatically the possible combination of paths between the terminals A, B and C when fault conditions occur. A fault may result from a single or multiple failure of the paths. In FIG. 2(a) there are no fault conditions, all paths are available, and all terminals can synchronized.
FIG. 2(b) shows the case where one path is unavailable and there are six combinations of this condition, and again all terminals can synchronized as is readily apparent.
FIGS. 2(c) to (f) show the four different arrangements that can occur when two paths are unavailable. In the case of the arrangements shown in (c), (d) and (e) there are three combinations of each of these arrangements. In the arrangement shown in FIG. 2(f) there are six combinations. In each of these, all terminals may still synchronized. When only three paths are available, as shown in FIGS. 2(g) to (j), whether or not the terminals can synchronized depends on the particular configuration of the available paths. FIG. (j) of which there are six combinations is an arrangement where synchronization can occur, whereas in FIG. (h) (six combinations) and in FIG. (i) (two combinations) it is not possible for the terminals to synchronize. In FIG. (g) (six combinations) all the terminals can synchronize providing B is the master so that A locks to B; whereas if B is locked to A, C becomes isolated and therefore all three terminals cannot synchronize. In the final case where four paths are unavailable, the four possibilities are shown in 2(k) to (n) and in only one, 2(l) is synchronization possible.
This system is characterized by being able to synchronize if and only if a direct path remains intact between one oscillator, which becomes the master, and each of the others which become the slaves.
The present invention seeks to provide a more fault tolerant system than the system illustrated in FIG. 2.
In the following description of the invention and examples of its embodiments reference will be made to terminals. A terminal is herein defined as containing:
(i) an oscillator whose frequency can be controlled,
(ii) a locking circuit capable of synchronizing the oscillator with an incoming signal frequency,
(iii) reconfiguration control means, see FIG. 13, which monitors at least the incoming signals to the terminal, and on the basis of those signals can connect the locking circuit to any one of the number of input signals. The reconfiguration control means may also in some cases be capable of disconnecting the oscillator from its outgoing links.
Reference will also be made to a path between two terminals which is defined as the serial connection of:
(a) an operating oscillator;
(b) an intact link (i.e. the physical connection between the terminals including any buffers or filters);
(c) an operating reconfiguration control capable of connecting to, or a permanent connection to, the signal received from said link;
(d) a functioning locking circuit.
Failure of a path will result from failure of any one of the above. In the case of reconfiguration control failure such that it is permanently connected to said link does not constitute failure of the path.
According to the present invention in a first aspect there is provided a network of three or more interconnected terminals, each terminal comprising:
a controlled-frequency oscillator,
a locking circuit operative to synchronize the frequency of the oscillator with the frequency of an incoming signal to the terminal, and
reconfiguration control means operative to monitor the incoming signals to the terminal and to connect the locking circuit to any one of the incoming signals, and to connect or disconnect the local oscillator to or from the outgoing links from the terminal;
the terminals being interconnected such that each is linked to the output of at least one of the other terminals in the network to receive output signals from the respective oscillators in those terminals;
characterized in that:
the reconfiguration control means in each terminal monitors and receives incoming signals and compares the phase of each with that of its local oscillator,
when said comparison detects an out-of-phase incoming signal the reconfiguration control means disconnects its oscillator from its outgoing links and connects the locking circuit to one of the incoming signals,
when said comparison does not detect an out-of-phase incoming signal the reconfiguration control means connects its oscillator to its outgoing links, and
the reconfiguration control means is further operative to detect whether its oscillator has locked to an incoming signal to which it has been connected, and
(i) if so, to connect its oscillator to its outgoing links, but if there remains another out-of-phase incoming signal, to connect its locking circuit to the other incoming signal,
(ii) if it has not locked, to connect its locking circuit to any other incoming signal, until all other signals have been connected without locking, and then to reconnect its oscillator to its outgoing links,
so that the plurality of terminals tend towards stable sychronization provided that a forward path exists directly or indirectly from any one terminal, which then becomes the master, to the remaining terminals, which then become the slaves.
According to the invention in a second aspect there is provided a network of three or more interconnected terminals, each terminal comprising:
a controlled frequency oscillator and a locking circuit operative to synchronize the frequency of the oscillator with the frequency of an incoming signal to the terminal, the oscillator and the locking circuit being constituted by a phase-locked loop,
and reconfiguration control means operative to monitor the incoming signals to the terminal and to connect the locking circuit to any one of the incoming signals,
the terminals being interconnected such that each is linked to the output of at least one of the other terminals in the network to receive output signals from the respective oscillators in those terminals;
characterized in that:
each phase lock loop includes a controlled delay element between the output of its oscillator and the input of its phase comparator, the delay caused by the delay element being controlled by the signal that controls the oscillator so that when the oscillator is made to increase in frequency the delay is made to decrease, and when the oscillator is made to decrease in frequency the delay is made to increase,
and the reconfiguration control means connects the phase lock loop to any one of the incoming signals, if any are present and reconnects the locking circuit if the local oscillator fails to lock to the incoming signal initially chosen, the output from each oscillator being continuously transmitted on its outgoing links,
so that the plurality of terminals tend toward stable synchronization provided that a forward path exists directly or indirectly from any one terminal to the remainder.
According to the present invention in a third aspect there is provided: a phase-locked loop including a variable delay element from the output of its oscillator to the input of its phase comparator, the delay caused by the delay element being controlled by the signal that controls the oscillator so that when the oscillator is made to increase in frequency the delay is made to decrease and when the oscillator is made to decrease in frequency the delay is made to increase.
Embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings.
In the drawings:
FIG. 1 shows diagrammatically a system containing three oscillators illustrating the operation of a known system;
FIGS. 2(a) to (n) are diagrams of the three terminal system illustrating various fault conditions.
FIGS. 3(a) to (d) are schematic diagrams of possible fault conditions in which synchronization is nevertheless possible in a first embodiment of the invention;
FIG. 4 is a flow diagram of an algorithm controlling terminal reconfiguration in the first embodiment of the invention;
FIG. 5 is a block diagram of a phase-locked loop;
FIG. 6 is a block diagram of a phase-locked loop with feedback delay;
FIG. 7 is a block diagram of a phase-locked loop with input delay and feedback delay;
FIG. 8 is a block diagram of a circuit similar to that shown in FIG. 7, but including variable feedback delay;
FIG. 9 is a block circuit diagram of two interconnected phase-locked loops;
FIG. 10 is a graph showing frequency plotted against time, and illustrating compatible delay conditions for the circuit of FIG. 9,
FIG. 11 is a schematic diagram of surmountable fault conditions in a second embodiment of the invention;
FIG. 12 is a flow diagram of an algorithm controlling terminal reconfiguration in the second embodiment of the invention;
FIG. 13 is a block diagram of the reconfiguration control hardware.
Referring to FIGS. 3 and 4, N remote or co-located terminals are synchronized using a maximum of N(N-1) inter-connecting links. As was also the case in the system described with reference to FIGS. 1 and 2 one terminal is maintained as master and the others as slaves; the master not being predetermined. In this embodiment of the invention, unlike the system in FIGS. 1 and 2, the system will remain synchronized provided that there remains a forward path directly or indirectly from one terminal to the remainder. FIGS. 3(a) to (d) show possible configurations in which synchronization is maintained and in the case of (a), (c) and (d) these configurations would not be possible in the known system illustrated in FIGS. 1 and 2. Of the configurations shown in FIG. 2 only in (n) would synchronization be lost. The reconfiguration algorithm used in this embodiment of the invention is illustrated, for example, by what happens when the fault conditions are such that one or more terminals become isolated from the only possible master as shown in FIG. 3(c) or the available terminals form a `daisy chain` of two or more paths as shown in FIGS. 3(a) and (c). It then becomes necessary for terminals that become locked to retransmit the signal on their outgoing links. This causes problems when only sufficient paths are available for two or more terminals to lock together in a loop. In FIG. 3(d) if the terminals were each to attempt to lock to an incoming signal and simultaneously transmit on the outgoing link no one terminal would then be the master and the arrangement would drift to an extreme of its frequency (for reasons that will be clear from the description of the constructions of the terminals, to follow) resulting in jitter and possible loss of lock.
The method by which this embodiment of the invention overcomes this problem is as follows.
Immediately following switch-on a multi-stable condition exists in which each terminal transmits to the remaining N-1 terminals and simultaneously is prepared to disconnect its output signal when it recognizes a valid signal on its incoming links. The first N-1 terminals to disconnect their output become slaves and the remaining terminal becomes master. When this is accomplished the first stage in the method is complete.
In the second stage the slave terminals attempt to lock to the master signal. If and when they become locked they once again transmit their output signals to the other terminals. If the paths to the other terminals are intact then each output signal is received by the other slaves. However, the other slaves should already be locked to the master terminal or be in the process of locking to it, and the retransmission of an identical signal by the other slaves is ignored. Eventually, if there are no faults then all the slaves will lock to the master and retransmit the signal on their outgoing links so that each slave has N-1 identical signals at its input. The master terminal, which initially was transmitting because it had no incoming signals, now receives the return signals from the slaves. The terminals are, however, programmed to recognize and ignore signals that are synchronized to their own output and hence the master terminal will then not attempt to lock to the returning signals and the arrangement will remain stable.
FIG. 4 shows an example of a reconfiguration algorithm that can be used (for example, in a conventional control microprocessor) to effect this method.
The behavior of the system when fault conditions are present will now be considered. If one of the terminals that initially is a slave fails to lock (or loses its locking condition) then if it subsequently fails to lock to the retransmitted signal from the other slaves it will transmit its own signal to all the other terminals, including the master. The master terminal will recognize an out-of-phase incoming signal and subsequently disconnect its own output and attempt to lock to the new incoming signal. The other slaves upon losing the signal from their master will change over to the new master signal being transmitted by the faulty slave.
Similarly the system will adapt to other failure combinations including the loop structure shown in FIG. 3(d) in which the terminal following the first one to disconnect its output becomes the master. Evan a daisy chain of paths is sufficient with a terminal at the top of the chain becoming master and the signal being restransmitted down the chain as the slaves lock. Thus this system can be characterized by the fact that as long as a forward path remains, directly or indirectly from one terminal to the remainder, the terminals will synchronize.
In the foregoing, the description of an embodiment of the invention has referred to the units in which the oscillators are contained as terminals, which have been considered as black boxes. In a second embodiment of the invention, the internal structure of the terminals must be considered. The terminals in the second embodiment of the invention comprise phase-locked loops and these may also partly constitute the terminals in the above-described first embodiment of the invention. Referring to FIG. 5, in its basic form a phase-locked loop comprises an oscillator 1, the output frequency of which can be controlled by an input signal (usually a voltage). Variation in the input signal causes variation in the output frequency of the oscillator. Also included is a filter 2, the reason for which need not be considered here. The last of the three basic elements is a comparator 3 which produces an output signal proportional to the phase difference between the input signals constituted by the input signal to the device as a whole and a feedback signal from the output of the oscillator 1.
In essence of phase-locked loop is a second order feedback device where any difference in phase between the two input signals to the phase comparator 3 produces an error signal which alters the frequency of the oscillator 1 in a direction that reduces the phase difference. In steady conditions with a constant frequency input the input signals to the phase comparator will be in phase.
Referring to FIG. 6, if a delay element 4 is introduced into the feedback path then when the system has settled there will still be no phase error between the two inputs to the phase comparator. The output of the oscillator 1 will, however, be phase advanced from the input to the comparator by the delay caused by the element 4 in the feedback path. Use of this principle will now be described in a system which compensates for unwanted delay in the input path. If the unwanted delay is .DELTA.t.sub.1 then the output of the phase-locked loop is compensated if the delay element in the feedback path presents a delay .DELTA.t.sub.2 which is equal to .DELTA.t.sub.1.
Referring now to FIG. 7; the circuit shown includes a connection from the output terminal of the oscillator 1 back to the input side of the element which introduces the unwanted delay .DELTA.t.sub.1. It is not possible for the two delays .DELTA.t.sub.1 and .DELTA.t.sub.2 to be exactly equal and since the input signals to the circuit element which cause these delays come from exactly the same source there will inevitably always be a small phase delay between the input signals to the phase comparator 3. The comparator will unsuccessfully attempt to alter the frequency of the oscillator 1 to reduce this differential, and the oscillator 1 will drift to an extreme of its range.
The direction of the drift and therefore to which extreme the oscillator drifts depends on whether .DELTA.t.sub.1 is greater than .DELTA.t.sub.2 or vice versa. Thus if .DELTA.t.sub.2 is greater than .DELTA.t.sub.1 then the comparator 3 will continuously attempt to increase the frequency of the oscillator 1 in a vain effort to correct the error.
Referring now to FIG. 8 the delay element 4 is variable with one that is supplied with, and controlled by, the signal from the comparator 3 so that the delay .DELTA.t.sub.2 decreases as the oscillator frequency increases. As the comparator 3 increases the oscillator frequency to correct the disparity at its input, .DELTA.t.sub.2 is decreased until .DELTA.t.sub.2 becomes equal to .DELTA.t.sub.1. Any further increase in the frequency of the oscillator will create an error in the opposite direction and the comparator will then reduce the frequency of the oscillator. Thus the arrangement shown in FIG. 8 can be made to stabilize with .DELTA.t.sub.1 equal to .DELTA.t.sub.2.
Referring now to FIG. 9 this diagram shows two phase-lock loops of the kind shown in FIG. 8, each including input circuitry which gives an unwanted delay .DELTA.t.sub.1 and .DELTA.t.sub.3 respectively. Unlike the circuit in FIG. 11, the output of each phase-lock loop is connected to the input circuitry of the other. Designating the upper phase-lock loop as A and the lower as B, and defining the particular instance of the input signal to the circuit B as a reference time T then the output of the input circuitry to circuit B will be T+.DELTA.t.sub.3. Assuming that the system is stable then the other input to the comparator 3 in the circuit B will be exactly in phase with the input from its input circuitry i.e. also timed at T+.DELTA.t.sub.3. Since the other input to the phase comparator of circuit B comes from its delay element 4 which gives a delay .DELTA.t.sub.4 then the input to this delay element i.e. the output of the oscillator 1 of the circuit B must be timed at T+.DELTA.t.sub.3 -.DELTA.t.sub.4. The output of the oscillator 1 in circuit B is the input to the input circuitry of the circuit A. If the same reasoning is applied, the output of the oscillator 1 in circuit A must be T+(.DELTA.t.sub.1 -.DELTA.t.sub.2)+(.DELTA.t.sub.3 -.DELTA.t.sub.4). But, this is also the input to the input circuitry of circuit B which we defined as being timed at time T. Thus,
The further condition for stability of the circuit is that the two oscillators in the circuits A and B must give outputs of identical frequency. Since .DELTA.t.sub.2 and .DELTA.t.sub.4 are functions of frequency, then the frequency at which .DELTA.t.sub.1 -.DELTA.t.sub.2 equals .DELTA.t.sub.4 -.DELTA.t.sub.3 is the stable frequency. An analytic solution is possible for illustration. FIG. 10 shows a possible graph of the values of .DELTA.t.sub.1 -.DELTA.t.sub.2 and .DELTA.t.sub.3 -.DELTA.t.sub.4 at varying frequencies. Ideally, .DELTA.t.sub.3 -.DELTA.t.sub.4 and .DELTA.t.sub.1 -.DELTA.t.sub.2 should be close to zero at the stable frequency since this would minimize the phase error between the outputs of the oscillators in each circuit. It would also be ideal for the two curves of the graph to coincide.
The arrangement of two phase-lock loops as described may be extended to a full system including N phase-locked loops. Some examples of connections where synchronization is possible are shown in FIG. 11. The loops need not be connected as a ring but may for example have two linked in a ring and one in a spur, or other combinations. Despite all the possible combinations of rings and spurs the phase delays between the oscillator outputs of the phase-locked loops will still be minimal. This is because on the spur terminals the feedback delay will compensate for the input delay. In the full synchronization system each phase-locked loop has connections giving it a choice of input signals from each of the other N-1 terminals and each includes a reconfiguration unit to control the choice. The system overall has a similar fault tolerance characteristic to the system described as the first embodiment of the invention with reference to FIGS. 3 and 4 i.e. all oscillators are kept in synchronization provided that there remains a path either directly or indirectly from any one phase-locked loop to the other N-1.
FIG. 12 shows a flow chart for a possible algorithm for the reconfiguration procedure in this embodiment of the invention. This arrangement offers the advantage over the first embodiment described above in that the output from each oscillator is connected continuously to the other terminals.