Field of the Invention
The present invention relates to a method and to an apparatus for synchronizing clocks connected to a network system.
Background of the Invention
Many household devices and appliances (particularly white goods and brown goods) contain clocks. Some systems allow these devices and appliances including their clocks--to be connected to a common bus or network. Operations of a device clock via the network might include synchronizing it to a time broadcast over the network, or broadcasting its time over the network in a so-designed system.
An example of a home bus system is described in Esprit Home Systems Specification 1.1 ESPRIT project 6782.
It may be possible in such systems to synchronize the various network clocks (e.g., by user action). However, synchronization will likely become lost due to variations in the time-keeping accuracies of the clocks. One solution to this problem has been to periodically broadcast times from one or more special device master clocks over the network, to which all network clocks synchronize. Where several master clocks are connected to the network, conflicts between master clocks may allow a lower precision clock to overrride a higher precision clock, thereby causing system clocks to be less accurate than is possible. Further, synchronization problems may arise if a master clock or clocks become disabled.
Summary of the Invention
It is one object of the invention to provide a method and apparatus for synchronizing clocks of a plurality of devices connected to a network.
It is a further object of the present invention to provide a system capable of meeting the above object while utilizing at least the following clock types or conditions:
non initialized
display only, no timer included
battery driven low precision
50 Hz driven
quartz driven
official time distributed via radio for a specific time zone)
user intervention
This list can be interpreted as a list of priorities (from lowest to highest), user intervention having the highest. Higher priority clocks should set the time of lower priority clocks (in case there are tuners in the system which receive the time from different time zones, it is supposed that only the tie of one time zone will be transmitted to the rest of the system).
The first proposal for solving the problem involves the following:
every clock broadcasts (datagram distribution) its time periodically during a predefined broadcast period as determined by an associated broadcast period counter;
the broadcast period of a clock is dependent on its relative precision (higher precision leads to a shorter period);
every clock which receives a broadcast updates its time;
every clock which receives a broadcast resets its broadcast period counter; and
user intervention initiates an immediate broadcast.
In this manner only the clock with the highest precision broadcasts its time to the system. Whenever it fails the next lower precision clock takes over the broadcasting function automatically. The broadcast periods are defined as factorial periods (e.g., 5, 6, 12, 15, 20, 30, and 60 seconds) of a preselected time period (e.g., one minute), and are synchronized to the preselected time unit. Then, all clocks displaying hours (hh) and minutes (mm) only will receive their time with no extra cost.
Another proposal is, that all devices will broadcast the time with an equal period (say one minute), but with a different phase (in seconds, e.g. highest precision clock at hh:mm:50, next lower precision clock at hh:mm:52, etc.; or highest precision clock at hh:mm:00, next lower precision clock at hh:mm:02, etc.) but only if no broadcast was received in the preceding time period. In the latter case, traffic on the network is independent of the precision of the different clocks. The displays are either slightly out of time or will need internal second counting capability.
In both proposals, time broadcasting may be done to the entire system or to a group, or subgroup, of clocks only.
A disadvantage of the both above described kinds of synchronizing is that the risk of collisions between two devices of equal precision is enhanced (causing higher traffic on the network), at least on media without collision avoidance (e.g. current carrier transmission on power lines), or if repetition is handled in the lower protocol layers only.
According to the invention, there are two proposals, said proposals involving:
Phase of transmission depends on precision
every clock receiving a time broadcast of the same or higher precision class does not broadcast for a certain amount of time (e.g. N minutes). The precision class of the broadcast can be easily derived from the phase of the time broadcast (e.g. seconds count).
this amount of time (e.g. N minutes) is a random or pseudo-random value, depending on some internal state of the instance of this device (and only the present time). Two possibilities: these values may be changed periodically or after each transmission.
Frequency of transmission depends on precision
every clock receiving a time broadcast of any precision class does not broadcast for a certain amount of time (e.g. N minutes).
this amount of time (e.g. N minutes) is a random or pseudo-random value, it depends a.o. on the precision of this clock, so that higher precision clocks will wait (in average) for a shorter time.
In principal, the inventive method is suited for synchronizing clocks being connected to a network system, whereby:
said clocks may belong to different precision classes;
the time of every clock is broadcast within said network system after expiry of a pre-selected time period, whereby a higher of said precision classes is associated with a shorter time periods and time period counting means are related to each of said clocks;
every clock which receives said broadcast updates its time and the related time period counting means are reset,
or whereby:
said clocks may belong to different precision classes;
the time of every clock is broadcast within said network system having a pre-selected phase, which is related to said precision class in such a way that the broadcast time related to a higher of said precision classes is earlier, and having a preselected broadcast time period, which is independent of said precision class;
time of a specific clock is broadcast only, when the clock has received no related time from a clock belonging to a higher precision class;
every clock which receives said broadcast updates its time.
In principle the inventive apparatus for synchronizing clocks being connected to a network system includes:
a clock belonging to one of a plurality of predefined classes;
bus interface means which broadcast the time of said clock within said network system after expiry of a pre-selected time period, which is related to said precision class in such a way that a higher of said precision classes is associated with a shorter time period, whereby time period counting means are related to said clock,
interrupt means which interrupt time transmission from said clock to said bus interface means during a certain time period after each reception of time information by said bus interface means, whereby the time of said clock is broadcast only, when the bus interface means have received no related time from a clock belonging to a higher precision class and whereby the time of said clock is updated when receiving said broadcast time information and said time period counting means are reset. Alternatively, the apparatus includes:
a clock belonging to one of a plurality of predefined to different precision classes;
bus interface means which broadcast the time of said clock within said network system having a pre-selected phase, which is related to said precision class in such a way that the broadcast time related to a higher of said precision classes is earlier, and having a pre-selected broadcast time period, which is independent of said precision class;
interrupt means which interrupt time transmission from said clock to said bus interface means during a certain time period after each reception of time information by said bus interface means, whereby the time of said clock is broadcast only, when the bus interface means have received no related time from a clock belonging to a higher precision class and whereby the time of said clock is updated when receiving said broadcast time information.
Brief Description of the Drawing
FIG. 1, the sole drawing figure, is a block diagram of a portion of a network employing the present inventive clock synchronizing scheme, which is necessary for network interface with one device clock.
Description of the Preferred Embodiments
In a real system there is some delay in datagram distribution (routers, time for medium access). This delay may vary considerably over the time. For this reason every clock should rest its internal time only in case of major and/or repeated differences to the time received. The frequency of clock resets is a matter of design choice. Nevertheless the choice should be made keeping in mind the time difference between broadcasts of different precision classes, so that it is guaranteed that every clock will be overwritten, at least on the long term, by broadcasts of higher precision classes.
The distribution of time in the described manner is not limited to the time to day. For example, the date may be distributed with the time in every time period, or could be distributed only on longer periods, but only by the device which is currently broadcasting its time.
Referring to FIG. 1, an internal device clock CLK continuously distributes the local time inside a device (coffee machine, cooker, dishwasher, washing machine, heating system, air conditioner, FM receiver, TV set, VCR). A bus interface unit BIU receives time information RXVE from a bus (not shown). Time information TXMIT is sent to the bus only when the controlled switch CSW1 is closed. This switch is controlled by a delaying block DBL which interrupts time transmission during a certain time period after each reception of time information (i.e., at the expiration of the time period corresponding to the precision class to which the clock has been pre-assigned). As a matter of design choice, transmission from the clock can be limited to fixed internal time intervals.
Incoming time information RXVE is allowed to set the internal clock CLK only when a second controlled switch CSW2 is closed. This switch may be closed continuously in a simple embodiment. In an alternate embodiment, the switch CSW2 can be controlled by a digital filter circuit F1 which compares the internal time of the clock CLK to the received time, and allows updating only when a significant difference over a certain period is detected. Supplementary blocks D1, D2 for compensating the delay of time distribution within the network may be included.
Advantageously, in a bus system with more than one real time clock initial synchronization will be achieved without user intervention (if at least one of the clocks is initialized). Synchronization will be kept, while highest precision clocks take priority over lower precision ones. Even if the controlling clock fails, the next lower priority clock will automatically take over the task of controlling the system. Updating is still possible by manual intervention from any point of the system. Bus traffic is kept low.