CA2300255C - Power system time synchronization device and method for sequence of event recording - Google Patents

Abstract

The invention relates to the field of power distribution systems and more particularly to power monitors within such a system. A time synchronization device and method for synchronizing an internal clock of the power monitor to a periodically stable frequency is provided. Cycles of the periodically stable frequency are counted after each elapse of a predetermined time period. The elapse of the predetermined time period defines a present window and a previous window. A counting error and a change in counting error are calculated, wherein the change in counting error is the difference of a counting error calculated for the present window and a counting error calculated for the previous window. The present method detects when a change in counting error is within a predetermined range. When the counting error is within the predetermined range, the internal clock of the power monitor is adjusted to the line frequency according to the counting error calculated for the present window.

Description

REFERENCE TO APPENDIX
An Appendix A, is included of a computer program listing. A second appendix, Appendix B, is also included of schematic diagrams.
COPYRIGHT NOTICE
A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.
FIELD OF THE INVENTION
This present invention generally relates to power systems. More specifically, the present invention relates the time synchronization to a universal time of monitors on the power system.
BACKGROUND OF THE INVENTION
In a typical electrical distribution system, an electrical supplier or utility company generates and distributes electrical energy to consumers via a power distribution network. The power distribution network is the network of electrical distribution wires which link the electrical supplier to its consumers.
Typically, electricity from the utility is fed from a primary substation over a distribution cable to several local substations. At the local substations, the supply is transformed by distribution transformers from a relatively high voltage on the distributor cable to a lower voltage which is supplied to the end consumer. From the substations, the power is provided to users over a distributed power network that supplies power to various loads. Such loads may include, for example, various power machines used by the end consumer.

As such, each load and generator in the system is tied together directly through wiring, or indirectly through transformers. When there is a change in load, a change in generating capacity or a problem in the system, the effects propagate from the source of the event to other parts of the system. For instance, a fuse blowing in a motor load may cause a voltage surge in the system which causes other loads to draw excess current and blow fuses down the line. Likewise, a generator may fall off line causing a voltage sag which causes other equipment to malfunction.
Since the electrical power system is complex and effects can propagate through the system quickly, e.g., in fractions of a second, it is often difficult to determine which device is the source of a fault and which are the victims. A fault includes, for example, a short circuit, a sudden large increase in load, the sudden loss of generating capacity, and a lightning strike to a transmission line. Such faults are detected, for example with a waveform recorder, included as part of a power monitor or revenue meter with power monitoring capabilities. The waveform recorder plots a waveform which represents an amplitude of voltage and current versus time. When the fault occurs, it causes the waveform of the electrical signal to deviate from its normal near sine wave. In addition, the root mean square (rms) voltage and current of the waveform can change.
Since the waveform is plotted against time, many power monitors include an internal clock that can indicate when the fault occurred.
Information of when the fault occurred is used to determine the source of the fault by identifying the power monitor that first recorded the fault. Since it takes time for the fault to propagate through the electrical distribution system, the source fault is located closest to the power monitor that first detected the fault. Thus, to determine the source of the fault, the consumer identifies which power monitor recorded the fault first.
A problem exists in that the internal clock time of each power monitor often varies among monitors. For example, known internal clocks utilize a 32.768 kHz watch crystal and a periodic interrupt timer (PIT) to determine time. The PIT increments, or decrements, depending on the application, a clock register which holds a value representing the time of the internal clock.
Accuracy of this PIT is determined by the accuracy of the 32.768 kHz watch crystal. Due to the crystal's impurities, age and an operating temperature, the crystal's accuracy is typically 10-50 parts per million (ppm), which equals 26.8 seconds - 2.2 minutes variance per month. Thus, the internal clock lacks the accuracy required to timestamp events such as faults that propagate through the power system in mere fractions of seconds.
One way to avoid the time variance between power monitors is for a master clock to broadcast a time synchronization message to each power monitor in the communications loop. Upon receipt of the time synchronization message, each power monitor then sets its internal clock to match that of the time broadcast message. Thus, the internal clocks can be synchronized each time the broadcast message is received. Problems occur in that the clock will be discontinuous at each second boundary since the PIT that is driving the internal clock is asynchronous to the master clock. Also, the internal clock drifts from the master clock until the next time broadcast is received.
Another way to avoid time variance between power monitors is to synchronize the power monitors to an inputted line frequency. According to Independent System Operator (ISO) specifications, internal clocks of the power monitors must adjust to synchronize with ISO Grid Operation line frequency. Many utilities are required to control the line frequency so that a number of line frequency cycles in a one week time period are within 0.02% of the exact number of cycles expected for that time period based on the specified line frequency. For example, when the integrated time error exceeds +/-2 seconds, a signal is sent to all utilities connected to Western Systems Coordinated Council (WSCC) and the utilities bias the automatic control systems to correct the error. Since the line frequency fluctuates and is unpredictable, a problem exists in that the total error of the internal clock becomes the error of the crystal plus the error of the line frequency when the line frequency is measured in terms of internal clock ticks.
Accordingly, there is a need for a time synchronization device and method which provides for an accurate sequence of fault recording. As such, _5_ there is a need for a device and method that periodically synchronizes internal clock times of monitors to a universal time or line frequency. In addition, there is a need to synchronize internal clock times without causing discontinuities and drift.
SUMMARY OF THE INVENTION
Such needs are met or exceeded by the present method and time synchronization device for synchronizing internal clock times to a universal time.
Such synchronization, as performed by the present invention, provides for an accurate sequence of fault recording. The present device and method speeds 1 o up or slows down the rate at which the internal clock time proceeds to avoid discontinuities and drift associated with merely changing the internal time to equal the universal time.
More specifically, a preferred embodiment of the present invention provides a time synchronization device of a power monitor for synchronizing a time of an internal clock to a line frequency, the internal time advancing at a predetermined rate, the device comprising of a counter operative to count cycles of the line frequency after each elapse of a predetermined time period, the elapse of the predetermined time period defining a present window and a previous window; a processor operative to calculate a counting error and a 2 0 change in counting error, the change in counting error being the difference of a counting error calculated for the present window and a counting error calculated for the previous window; wherein the processor detects when the change in counting error is within a predetermined range; and adjusting means for adjusting the internal clock to the line frequency according to the counting error 2 5 calculated for the present window when the change in counting error is within the predetermined range.
In another embodiment, there is provided a method for synchronizing a time of an internal clock to a periodically stable frequency which provides accurate time intervals, the method comprising the steps of a counting cycles of 3 o the periodically stable frequency after each elapse of a predetermined time period, the elapse of the predetermined time period defining a present window and a previous window; calculating a counting error and a change in counting error, the change in counting error being the difference of a counting error calculated for the present window and a counting error calculated for the previous window; detecting when the change in counting error is within a predetermined range; and adjusting the internal clock to the periodically stable frequency according to the counting error calculated for the present window, when the change in counting error is within the predetermined range.
In yet another embodiment, there is provided a method of maintaining time synchronization in a power monitoring device, the power monitoring device coupled to a power system operating at a first frequency and comprising at least one processor operating at a second frequency, the method comprising of receiving at least one time message from a time source; accumulating counts in a free running counter in response to the second frequency;advancing the internal time of the device by a determined amount in response to the accumulation; and adjusting the amount in response to counting periods of the first frequency.
BRIEF DESCRIPTION OF THE DRAWINGS
Other features and advantages of the invention will be apparent to those skilled in the art with reference to the detailed description and the drawings, of which:
2 o FIG. 1 is a block diagram of an electrical distribution system with power monitors containing internal clocks that synchronize to a master clock located in a computer, according to the time synchronization device and method of the present invention;
FIG. 2 is a block diagram of an electrical distribution system with power 2 5 monitors containing internal clocks that synchronize to a global positioning system (GPS), according to the time synchronization device and method of the present invention;
FIGS. 3A and 3B are exemplary time synchronization messages sent by a GPS receiver and a master computer, respectively, according to the time 3 0 synchronization device and method of the present invention;

-6a-FIG. 4 is an exemplary buffer structure for storing time synchronization information according to the time synchronization device and method of the present invention;

FIGS. 5A-5D are flow charts representing the storage and receipt of time synchronization information loaded in the buffers of FIG. 4, according to the time synchronization device and method of the present invention;
FIGS. 6A- 6G are flow charts representing time synchronization of the power monitors to a universal time, according to a preferred embodiment of the present time synchronization device and method;
FIG. 7 is a flow chart representing time synchronization to a line frequency, according to a preferred embodiment of the present time synchronization device and method; and FIG. 8 shows line frequency synchronization windows representing time periods to perform calculations according to a preferred embodiment of the present time synchronization device and method.
TABLE OF ACRONYMS
The following table aids the reader in determining the meaning of the several acronyms used to describe the present invention:
AC = Alternating Current.
DEC = Decrement Register.
GPS = Global Positioning System.
Hz = Hertz.
ISO = Independent System Operator.
M = Mega (X106).
us = microsecond (X10'6) ns = nanosecond (X10-9).
OTM = On Time Mark.
PIT = Periodic Interrupt Timer.
PLL = Phase Locked Loop.
ppm = Parts Per Million.
rms = Root Mean Square.
RTC = Real Time Clock.
TB = Time Base.
UTC = Universal Time Coordinated.

_$_ DETAILED DESCRIPTION OF THE INVENTION
To help locate the source of a fault located on an electrical distribution system, a time synchronization device and method of the present invention adjusts an internal clock time, for each power monitor connected on the system, to equal a master clock time. An exemplary power monitoring device is one described in U.S. Patent No. 5,736,847, where the internal clock is derived from a 32.768 kHz watch crystal, for example.
There are various ways to derive the internal clock from the 32.768 kHz watch crystal. One way is to multiply up the 32.768 kHz crystal frequency to a 16.777216 MHz bus frequency utilizing known phase locked loop (PLL) circuitry such as the circuit in an MC68332 microcontroller manufactured by Motorola, located in Schaumburg, Illinois. Internal time of the power monitor is kept with a periodic interrupt timer (PIT) which generates an interrupt to advance the internal time by incrementing a value stored in a clock register, for example. For example, the PIT generates 1024 interrupts per second to advance time by 976.5625 uS each interrupt. The frequency of the PIT is generated by dividing down the 16.777216 MHz bus frequency.
According to known techniques, the internal time of the power monitoring device is represented as the number of seconds since January 1, 1970. In a preferred embodiment of the present invention, the number is 64 bits long and is stored in the clock register. The upper 32 bits represent the seconds and the lower 32 bits represent the fractional part of the seconds.
The 32 bits that represent the fractional part of the seconds must roll over at the end of one second, therefore, one count of the 64 bit representation of the internal time is 1/232 = 0.2328 nano seconds (nS). For 1024 PIT interrupts per second, each PIT interrupt adds 4194304 counts to the 64 bit representation of the internal time, which equals 976.5625 uS assuming the crystal is exactly 32.768 kHz. Since 4194304 is a relatively large number, and one count represents 0.2328 nS, the time advance with each PIT interrupt can be adjusted with a great deal of precision.
Another way to derive an internal clock from the 32.768 kHz watch crystal is to utilize a microprocessor, for example Model number MPC821 _g_ manufactured by Motorola, located in Schaumburg, Illinois, which multiplies up the watch crystal frequency to a 49.9712 MHz bus frequency.
Thereafter, the microprocessor divides down the 49.9712 MHz frequency to 3.1232 MHz to create a time base (TB) facility used for the internal clock. Thus, after 3.1232 million cycles or TB ticks, one second of the internal clock has elapsed. To achieve milliseconds, the 3.1232 MHz frequency is divided by 1000 using known circuitry. This frequency is stored in a decrement register (DEC) which is part of the MPC821 architecture, and decreased every clock tick. Thereafter, every 3,123.2 cycles, a decrement interrupt is triggered to indicate that one millisecond has elapsed on the internal timer. Artisans will appreciate that, since the decrement register only accepts whole integers, to achieve one second, 3123 cycles are decreased for 800 ms and 3124 cycles are decreased for 200 ms, for example. The TB and DEC interrupts together outline the clock architecture for MPC821. While TB ticks provide a free running counter for time interval calculations, the DEC interrupt alone determines the unit ticks for the internal clock.
Referring now to the drawings, and particularly FIGS. 1 and 2, to supply universal time to the internal clocks of the power monitors, the present time synchronization device and method utilizes various entities. A master computer 10, for example, shown in FIG. 1, contains a master clock (not shown) to synchronize the power monitors' internal clocks to the master clock's universal time. Another source of universal time is a global positioning system (GPS), shown in FIG. 2. Yet another source, line frequency, is used to adjust the power monitors' internal clocks in synchronization with the universal time.
Referring to FIG. 1, the time master computer 10 contains the master clock for all power monitors on the system. The master computer 10 connects to power monitors 12, 14 via communications bus 16, for example.
The communications bus is ideally based on the RS-485 standard.
Preferably, the master computer 10 sends out a time synchronization message on bus 18 at rogular intervals, e.g., about once every minute., While artisans will appreciate that the present invention functions over a wide range of frequencies, a requlrgd accuracy of the internal clock of the monitor depends on frequency that the master computer sends out the time synchronization messag~.
The power monitors 12, 14 receive signals from voltage transducers 18, 20 and current transducers 22, 24 which ere scaled versions of the voltages and currents in power system wiring 26, 28. The power monitors record the rms 'values and wavefomts of these voltages end currents in a , manner described, for example, in a commonly assigned co-pending Cer~adien patent eppHcation to Rene T. Jonker, et al. entitl~i "Revenue Meter with Power Quality F~atures." serial no. 2,299,043, flied on 21 February, 2000.
To identify when a reading and the weve~fonn oxurred, the power monitor attaches a timestamp to each reading and waveform that the captured.
Thereafter, the readings, waveforms and timestamps aro requested by processing computers 30, 32 through communication (inks 34. 36. These communication links 34, 38 ace ideally Ethernet links, but may also be RS-232, RS-485 or any other suitable communications technolpgy. it can be appreciated that information also may be retrieved by master computer 10.
When a fault occurs at a first load 38, for exampl~, power monitor 14 will record the event almost immediately. Thereafter, the effects of th~ fault will prppagate thtough the wiring 28, past second load 40 end first generator 42, through transformer 44 to wiring 26, and then be recorded by power monitor 12. The time period that ~lapses between power monitor 14 .
recorcilng the fault and power monitor 12 recording the fault, in addition to other monitors in the system (not shown, is used to determine where the fault originated and which device in the system experienced the fault. For example, the speed at which electricity travels through the conductors is known, and the distance frpm each monitor to where the fault occurred can be determined. Thus, it can be determined that the first load 38 is the soun:.e of the fault, and not a third load 48 located near a second genenator 48.

The clock on each power monitor 12, 14 is synchronized to the universal time, as described below, so that timestamps are accurately compared between power monitoring devices. It can be appreciated that if all the power monitoring devices occupy the same communications link, then the absolute accuracy of the master clock is not as important since all power monitors are synchronized to the same master clock. In this case the master clock is provided by the master computer's clock.
Referring to FIG. 2, if, however, all the power monitoring devices are not on the same communications link, GPS receivers 50, 52 are connected to the power monitors 12, 14 via lines 54, 56 respectively, to provide universal time. An exemplary GPS receiver is the XL-DC 600 manufactured by True Time in Santa Rose, CA. When power monitors 12, 14 are not on the same communications link, the GPS receivers provide time synchronization information to the power monitors, which is accurate up to one millionth of a second. The master clock is the time kept by GPS satellites 58a-58h.
It can be appreciated that each power monitoring device 12, 14 requires the time synchronization information to be broadcast on the communications loop 16, 54, and 56 at regular intervals, for example, once every second. Artisans will appreciate that other time intervals can be used.
The accuracy of the internal clocks depend on the frequency that the time message is sent out. Thus, if the time synchronization information is sent only every day or every hour, for example, the power monitors' internal clocks are not as accurate as when the information is sent out every second. When the internal clocks are not accurately set to the master clock, the ability to find faults on the system can be adversely affected.
Referring to FIGS. 3A and 3B, exemplary time synchronization messages are shown for a GPS receiver and a master computer, respectively. Each master clock source sends out a time broadcast message containing the universal time and an "on time mark" (OTM). The OTM is defined as a certain point or byte in.the time synchronization message that the universal time occurred. As shown in FIG. 3A, a first bit of the carriage return is the OTM as defined by the GPS. A quality flag represents whether the _12~
receives antenna is locked onto the satellite or not If the antenna is locked to at least one satellite, the GPS receiver time is guaranteed to be accurate within 100 us of Universal Time Coordinated (UTC). if the antenna is blocked from the satellite cluster, however, the time on ttie receiver runs freely according to the GPS receiver's internal clock. As shown in FIG. 3B, the OTM
for the master Gock's time synchronization message is defined as the first bit of the first byte to indicate the length of the message.
Referring to FlG. 4, an exemplary circular buffer structure is shown, including Data, Timestamp and a Count information. The time synchronization message is storoed in the Data buffer which is, for example, 300 bytes long. The total number of buffers is implementation dependent and depends on factors such as processor speed and cost. It can be appreciated that a sufficient number of buffers is necessary to allow the processor to gather and process packets stored in the circular buffers before data is 'I5 overwritten by new packets.
Referring to FIGS. SA-5D, utilization of the circular buffer structure (shown in FIG. 4) is described. Referring to FIG, 5A, according to known programming techniques, variables utilized herein are initialized (block 60).
Thereafter, variable Receiving is set to false (block 62). The variable Receiving is utilized to indicatc~a whether bytes of data, e.g., the time synchronization message, are being received into one of the~buffers. A value of False indicates that bytes of data should be stored in the next buffer and the timestamp recorded. A value of true means that data should be stored In the cun-ent buffer, in addition, variables He~d and Tail (shown also in FIG.
4.) are set to zero (block 62). The variable Head points to the current buffer to store bytes, and the variable Tail points to the buffer where bytes are to be removed for processing.
Referring to FIG. 5B, a routine runs when the receiver sends a byt~ of data, e.g., a byte of the time synchronization message, to the power monitor (block 64), Since Receiving initially equals false (block 66), next the routine determines whether an overflow condition occurs in the circular buffer (block 68). If Head plus one equals Tail, an overflow condition exists and the routine ends since there is no more buffer space available (block 70). Upon the overflow condition, all buffer data is discarded and an overflow event is set in the event log to inform a user that the overflow occurred (block 70).
If no overflow condition exists, a lower layer communication receiving function, for example, places the received byte into the buffer currently identified by the Head pointer and sets Receiving to True (block 72). A
Rxd_In pointer (shown in FIG. 4) identifies the position within the buffer to store the received byte. Upon receipt of a first byte of the time synchronization message, the power monitoring device's communication port interrupt service routine takes a snap shot of a Free Running Counter and records the snap shot value into the Timestamp register (block 72). The Free Running Counter is a counter that is incremented by one each time a PIT
interrupt service routine is executed. Artisans will appreciate that any counter can be used that increments at a rate at least as fast as the rate of the PITs.
The Free Running Counter value is stored for later use by an upper layer communications processing function to calculate a processing delay when setting the internal clock, as described below.
Each time a byte is received in the buffer, a counter located in the Count element of the buffer (refer to FIG. 4) is incremented to represent the number of valid bytes in the Data portion of the buffer (block 74). The upper layer communication processing function saves time and processing power by utilizing the count value to identify the number of valid bytes in the Data portion of the buffer without removing all 300 bytes of data.
Thereafter, the lower layer communication receiving function implements an end of Data packet timeout mechanism (block 76). The timeout mechanism is controlled by BYTE TIMEOUT which recognizes when the end of the time synchronization message occurs. BYTE TIMEOUT
happens when no bytes of data are received by the power monitor for a predetermined duration of time. In the preferred embodiment, a BYTE TIMEOUT occurs when the power monitor receives no data for four consecutive byte times. At 9600 baud, one byte time occurs approximately every millisecond, thus, after four byte times four milliseconds have elapsed.

Since the receiver transmits bytes of the time synchronization message back to back, a delay of four byte times without data ensures that an end of the data packet was reached. Artisans will appreciate, however, that a duration other than four bytes can be used, as long as the selected duration is not greater than the frequency at which the GPS sends packets, e.g., every one second.
When the power monitor receives the next byte of the time synchronization message (block 64), Receiving equals True (block 66).
Thereafter, the lower level communication receiving function determines whether the Rxd_In pointer is greater than the maximum data size, e.g., 300 bytes (block 78). If unused Data buffer space remains, the lower level communication receiving function adds the incoming byte to the buffer and increments the Count packet counter (block 74). Thereafter, the timer for the BYTE TIMEOUT is started (block 76).
If the Rxd_In pointer is greater than the available limit, the lower level communication receiving function determines whether the Head buffer has reached the Tail buffer (block 80). If Head has caught Tail, an overflow condition exists and the routine ends since there is no more buffer space (block 70). Upon the overflow condition, all buffer data is discarded and an overflow event is set in the event log to inform the user that the overflow occurred (block 70).
If the overflow condition does not exist (block 80), the lower level communication receiving function increments Head to begin saving data in the next buffer since the previous buffer is full (block 82). In addition, the lower level communication receiving function saves a snap shot of the Free Running Counter to the Timestamp packet and sets Receiving to True (block 82).
Thus, the lower level communication receiving function records each time synchronization message into separate buffers and timestamps the reception of the first byte in the buffer to preserve the start time of the packet which is used during packet processing, described below. This repeats until all buffers are filled and then loops back to the first buffer.

Referring to FIG. 5C, the lower level communication receiving function prepares for the next time synchronization message after the BYTE TIMEOUT occurs (block 84). As described above, the BYTE TIMEOUT takes place when a time has expired wherein no bytes were received, e.g. after 4 milliseconds since the last byte was received. To prepare for the next time synchronization message, interrupts are turned off to keep other routines from altering the Head variable, and Receiving is set to False to indicate that no bytes are being received (block 86). Thereafter, the variable Head is incremented so that the time synchronization messages are kept separate (block 88), and the interrupts are turned back on (block 90).
Referring to FIG. 5D, when a buffer is full or the BYTE TIMEOUT has occurred, the upper layer communications processing function receives an event, for example, indicating the arrival of a new time synchronization message. Thereafter, the upper layer communications processing function requests the bytes from the buffer number indicated by Tail (block 92). To determine whether data is available, the upper layer communications processing function determines whether Head equals Tail (block 94). If Head equals Tails, the routine ends because there is no data available (block 96).
If Head does not equal Tail, the upper layer communications processing function removes bytes from the buffer indicated by Tail (block 98). Pointer Rxd Out (shown in FIG. 4) points to the next byte to be removed. As described above, the upper layer communications processing function utilizes the value in the Count packet to determine the number of valid bytes of data to remove from the Data packet. After removing the valid bytes, i.e., the time synchronization message, the variable Tail is incremented to indicate the next buffer to be read (block 100).
Referring to FIGS. 6A-6G, to avoid discontinuities and drift to the internal clock of the power monitor, the PIT is adjusted to either slow down or speed up the internal clock, each time the time synchronization message is processed. Slowing down and speeding up the internal clock ensures that the overall accumulation of time on the internal clock matches that of the master clock. As mentioned above there are 1024 PITs per second, each advancing the time by 1 /1024 = 976.5625 uS. If the system bus clock were exactly 16.777216 MHz then there would be exactly 1024 PITs per second. Since the system bus clock varies from unit to unit, the rate of time advance is adjusted.
Since the number of PITs per second cannot be adjusted, the time advance of each PIT, i.e., time per PIT, is adjusted. The time per PIT is the time that is added to the power monitor's internal clock time each time a PIT interrupt occu rs.
In a preferred embodiment of the present invention, adjustment of the time per PIT interrupt is accomplished according to the following routine. The routine is designed to minimize lock on time, i.e., the time that elapses before the internal time on the power monitor and the universal time on the master clock are within a time tolerance, and the effects of perturbations in the master clock. Referring to FIG. 6A, variables used during the routine are initialized (block .102) Count Per Pit is set to the calculated OUR TIME_PER which indicates the time advance per PIT (block 104). The default value of OUR TIME_PER for a 32.786 kHz crystal and PIT occurring every 976.5625 uS is 4194304. Thus, there are 1024 PIT ticks every second.
Artisans will appreciate that other counts per PIT can be utilized.
The internal clock time of the power meter, i.e., Meter Time, is set to a real time clock (RTC) which preferably runs on battery power to avoid the effects of a power outage (block 104). In addition, Previous Sync is initially set to false (block 104). After two PIT Between Slope periods expire, Previous Sync is set to true to enable slope calculations, described below.
Sync Count is initially set to zero and denotes the number of expired PIT
Between Slope periods (block 104). Mid_Free and Old_Free are variables that store old values of Free Running Counter to be used in the slope calculation described below. Free_Running Counter is initially set to zero (block 104).
Referring to FIG. 6B, a Time_Int() routine is triggered every time a Pit interrupt occurs (block 106). Upon the occurrence of the PIT interrupt, Meter Time is incremented by the Count Per Pit and the Free_Running Counter is incremented by one tick (block 108). The -17~
Free Running Counter is used to measure the time taken to get from one point in the routine to another point.
Referring to FIG. 6C, a Sync Set() subroutine is triggered when the upper layer communication processing function receives a time synchronization massage (block 110). The Timestamp contained in the Time Sync message (shown in FIG. 3) is s~nt as a parameter to the Sync Set() routine. Upon receipt of a lime synchronization message, the upper layer communication processing function determin~s whether a value for Free Running Counter minus Mid Free is greater than PITS BETWEEN SLOPE (block 112). PITS BETWEEN SLOPE Is a constant that is set to ensure that a predetermined time period elapses before the slope calculation is enabled. In the preferred embodiment, PITS BETWEEN SLOPE is set to allow for about a half hour to an hour between slope calculations. Artisans will appreciate that a sufficient amount of time is ailowed to elapse between slope calculations, described below, to ensure that small inaccuracies in the time synchronization messages become Insignificant compared to the time between the two points used for the slope calculation.
If the predetermined time period has not elapsed since the last slope calculation, New_Free is set to the value stored In the Timestarnp register (shown in FIG. 4) minus one byte time, e.g., one millisecond (block 114). One byte time is subtracted since the routine timestamps the end of reception of the first byte, and the time of reception of the first bit is required. In addition, New Time is set to the value contained in the Time Synch Message, New Free indicates a snap shot of the Free Running Counter when the power monitor receives the first byte of the time synchronization message.
Thereafter, the internal dock of the power monitor is synchronized to the universal time. First, interrupts are fumed off (block 116). Then, Real Time is calculated to equal New Time plus the difference between 3o Free_Running Count~r and New Free, and multipli~d by the Count_Per_Pit to get the time (Mock 116). In addition, Prev_Difference is set to equal Difference, which has no value the first time through the routine (block 116).

Difference is Real Time minus the Meter Time, i.e., the calculated error between the universal time and the meter's internal clock time (block 116).
Referring to FIG. 6D, preferably the internal clock time is set directly to the universal time only when the difference between the power monitor's internal time and the universal time is more than a predetermined amount (TIME TOLERANCE), e.g., one second (Block 118). Continuous direct setting of the internal time to the universal time is not preferred since this causes drift and discontinuities in the internal time. Direct setting of the internal clock to the universal time is performed, however, to speed up the interval required to ~ lock to the master clock when the times differ by more than one second. If direct setting of the internal clock were not performed, depending on the time difference, the negative effects of a long locking time outweigh the effects of performing a discrete time set.
Thus, the upper layer communication processing function determines whether the absolute value for Difference is greater than TIME TOLERANCE, e.g., one second (block 118). If the universal time and the power monitor's internal time differ by more than one second, the upper layer communication processing function calls a Time Time Set() routine (block 120). Referring to FIG. 6E, the Time Time Set routine begins (block 122). Meter Time is set to Real Time, and the RTC is set to Meter Time (block 124). In addition, a message is sent to generate an event to indicate to a user that a time set has occurred (block 126).
Returning to FIG. 6D, if the universal time and the power monitor's internal time differ by less than a second, then the routine determines whether the internal clock is slow or fast to adjust the time per PIT accordingly.
First, Difference and Previous_Difference are compared to determine if they are the same sign (block 128). If they are not the same sign, then the internal time has crossed over from being to slow to being too fast or visa versa. The crossover from too fast to too slow, or visa versa, is an indication that the universal time and the internal time are nearly synchronized. Thus, the Count Per Pit is adjusted more precisely by utilizing the slope calculation (block 130).

Referring to F1G. 6F, shown is an exemplary Calc Time Slope() routine for calculating the PIT count based on slope (block 132). First, the upper layer communication processing function determines whether ' Previous~Sync equals True, i.e., whether enough time has expired sothatthe difference in time between the two points used in the slope calcuiation~ ~is insignificant compared to the error in the timestamps being received (block 134). It can be appreciated that slope cannot be calculated with only one point of reference. Thus, if the power monitor has not received and indication that at least two PIT between Slope periods have expired, the Calc Time_Slope() routine ends (block 136), If Previous_Sync equals True, however, the upper layer communication processing function determines whether Difference is greater than zero, i.e., whether the internal time is running slow (Mock 138), if the internal time lags the universal time, then the Count Per Pit is Incremented ' by an integer, e.g., the integer one (block 140). In particular, Count Per Pit is calculated as the difference between the new and previous time synchronization times divided by the difference between the new and old stored value of the Free_RunningWCounter, plus one tick (block '140), If the internal time is running fast, Count Per Pit is calculated as the difference between the new and previous time synchronization times divided by the difference between the new and old stored value of the Free~Running_Counter, minus one tick (block 142). The addition or subtraction of_on~ tick ensuces'that drift is in the correct direction before the next slope calculation vuhichwpreferably~orcurs when the next timesync a message is received. ' - ' ,.,;;, Referring again to FIG. 6D, if the internal time 'has not crossed over from being too slow to too fast, or visa versa, the upper layer communication processing function calculates dbl_diff, double difference (block 144). dbl diff is the rate of change of Difference, which is the difEerence~between the universal time and the power monitors internal time. Thus, dbl_diff equals Previous Difference minus Difference (block 144).

Thereafter, Difference is compared to TIME_RESOLUTION, e.g., 1 millisecond, which is the allowable error (block 146). If Difference is greater than TIME_RESOLUTION, then the upper layer communication processing function determines whether the rate of change is in the wrong direction (dbl diff less than zero) (block 148). In addition, the upper layer determines whether the rate of change is insignificant with respect to the Difference between the universal time and the internal time (5*dbl diff less than Difference) (block 148). If either the rate of change is in the wrong direction or is insignificant with respect to Difference, the Count Per Pit is incremented by TIME ADJUST, e.g., one tick to speed up the internal clock (block 150).
Otherwise, if the rate of change is in the correct direction and is significant compared to Difference, the Calc Time Slope() routine is performed, as described above, to determine a precise rate of change (block 130).
If Difference is not greater than TIME_RESOLUTION, the upper layer communication processing function determines whether negative Difference is greater than TIME_RESOLUTION (block 152). In other words, whether the difference between the universal time and the internal time is small. If negative Difference is not greater than TIME_RESOLUTION, i.e., the difference in time is small, then the Calc Time Slope() routine is performed to fine tune Count Per Pit, as described above (block 130).
If negative Difference is greater than TIME_RESOLUTION, then the upper layer communication processing function determines whether the rate of change is in the wrong direction (dbl diff greater than zero) (block 154).
In addition, the upper layer determines whether the rate of change is insignificant with respect to the Difference between the universal time and the internal time (5*dbl diff greater than Difference) (block 154). Of course, a factor other than 5 can be used. A change to the factor affects lock on time and overshoot depending on the frequency with which timesync messages are being received. In other words, the number 5 is a compromise based on the typical rate of timesync reception.
If either the rate of change is in the wrong direction or is insignificant with respect to Difference, the Count Per Pit is decreased by TIME ADJUST, e.g., one tick (block 158). Otherwise, if the rate of change is in the correct direction and is significant compared to Difference, the Calc TimeTSIope() routine is performed, as described above (block 130).
Referring to FIGS. 6D and 6G, after Count Per Pit is 'incremented (block 150) or decreased (block 156), the Calc Time_Slope() routine is perton~ned (block 130) or the Time_Time Set() routine is performed (block 120), Count Per_Ptt is compared to MAX_ADJUST (block 158). This comparison ensures that Count Per Pit remains within a range of the default count, e.g., 4194304, after Count Per Pit is adjusted. Maintaining the Count Per Pit within a range lends stability to the internal clock. Thus, if Count Per Pit is greater than MAX ADJUST, e.g., the default Counf Per_Pit plus 50 ppm, Count Per Pit is set to MAX_ADJUST (block 160). If Count Per Pit is less than or equal to MAX ADJUST, then Count Per Pit is compared to MIN ADJUST to determine whether Count Per Pit is less than a minimum allowable P1T count value (block 162). If Count Per Pit is less than the minimum allowable PIT count value, then Count Per Pit is adjusted to equal MIN_ADJUST, e.g., Count Per Pit minus 50ppm (block 164).
Artisans will appreciate that the values for MA~,ADJUST and M1N_ADJUST
are implementation dependent, and other values can be utilized.
~ Referring again to FlG. 6C, ff Free_Running_Counter minus Mid~Free is greater than PITS BE?1NEEN_SLOPE (block 112), SYNC_COUNT is incremented' by the integer one, for example, and variables are updated that contain values for the old, middle and neiw Free Running Counter snap shots (block 166). Thereafter, the upper layer communication processing function determines whether the power monitor has been notified that finro PIT
Between Slope periods have expired (block 168). !f two PIT Between Slope .
periods have expired, Previous,Sync is set to Tnre to enable the slope calculation routine described above (block 170). if two PIT Between Slope periods have not expired, however, the slope calculation subroutine remains disabled since not enough time has elapsed tv get an accurate slope calculation. It can be appreciated that if the accuracy of the incoming time synchronization messages is ~l- 1 msec and a calculation is desired of ~~r Count Per_Pit to within 1 ppm, points two million times further apart than 1 msec are necessary, i.e., 2000 seconds (3~.3 minutes). Thereafter, th~ upper layer communication processing function adjusts the free running counter value from the Time Sync Message and stores~the adjusted value.in New Free, as described above (block 114).
For simplicity purposes, the foregoing description utilized a free running counter that Increments at rate equal to that of the PlT. !t can be appreciated that if the processor being utilized provides a free running counter that increments at a greater rate than that of the PIT, the accuracy can be improved beyond the '171024 described stated above. Migher accuracy is accomplished by recording the value of such a free running counter each time the P1T interrupt service routine is executed. Then, when a code segment requests the time, the time is calculated by adding the difference between the current tree running counter value and that recorded In the last PIT interrupt service routine call. After calculating the difference, the result is converted to the standard seconds representation and added to the time on the meter that was calculated in the last PIT interrupt service routine call.
Referring to F1G. 7, a routine of the present invention is shown to synchronize power monitors' internal clocks to line frequency. it is contemplated that time synchronization is first performed, for example, according to the above described method to set an absolute 'time. .
Thereafter, the following routine can be used to maintain the internal clock's synchronization.
The routine begins by inltiallzlng the internal clock, e.g., a virtual clock ' stored in memory, to the RTC (block 172). For the purposes of discussing the Line frequency synchronization device and method of the present invention, the internal clock utilizes a decrement register which produces an Interrupt every 3123 ticks as derived from a 32.768 kHz crystal. Each time a DEC
interrupt occurs, the internal clock is advanced by one millisecond. Artisans will appreciate that other clocking mechanisms can be used to represent the internal clock.

Since the crystal triggers the ticks that decrement the counter, an accuracy of the internal clock is determined by the crystal's accuracy.
Crystal error (x ppm), is defined as:
x ppm - FxTAL o FxrAL .1 O G
FXTAL
where F° corresponds to the nominal frequency of the crystal and Fxra~
is the actual frequency of the crystal. When xJnpm is greater than zero, the time of the power monitor's internal clock lags the universal time, and when x ppm is less than zero, the internal time leads the universal time.
Since crystal error cannot be normalized, the value loaded into the decrement counter is adjusted to synchronize the internal clock to universal time. Universal time, ticks or seconds, is determined from a line frequency input to the power monitor, for example, by counting 60 cycles of the line voltage to mark exactly one second if the nominal value of line frequency is 60Hz. Line frequency, however, is not fixed and deviates from the nominal frequency, e.g., 60 Hz. Obviously, such irregularity in line frequency affects the accuracy of the internal clock adjustment. Utilities periodically compensate, however, for the variations in the line frequency, so that over a twenty four hour time period, for example, the consumer receives an average line frequency of exactly 60 Hz. Therefore, over time counting error is introduced due to fluctuations in the line frequency as well as the drift in line frequency. In the long term, however, counting error emanating from the line frequency is diminished due to the utility's periodic compensation.
The timestamp device and method of the present invention determines when the line frequency compensation occurs to calculate a line frequency counting error (c-ppm) and to adjust the count number loaded into the decrement register accordingly, as described below. When line frequency compensation occurs, line frequency error (l ppm) averages to zero, where:

c m = Expected cycle count - Actual cycle count X 1 O6 ;
- pp Expected cycle count and l p~ _ .fo f .f .146 .

where fo is the expected nominal line frequency, e.g., 60 Hz, and f is the actual line frequency which the power monitor received.
After the internal clock is initialized, known circuitry is used to ascertain an Actual cycle count (represented by C; in FIG. 8) of the line frequency (block 174). The Actual cycle count is determined, for example, by counting cycles of line voltage which is generally a sinusoidal wave, for a predetermined duration (block 175). Referring to FIGS. 7 and 8, a synchronization window ends when the predetermined time period elapses, according to the internal clock (block 176). Preferably, the predetermined time period is about 3600 seconds, i.e., one hour, to acquire the resolution necessary to observe, for example, a 5 ppm error in the line frequency. A shorter predetermined time period, such as one half hour in terms of internal clock seconds, may not supply a sufficient number of cycle counts to observe the 5 ppm error. In addition, a predetermined time period longer than one hour, such as five hours, decreases accuracy of the internal clock since it is harder to pinpoint when the line frequency compensation occurred.
At the end of each synchronization window, calculations are performed to determine the cycle counting error (c ppm) and whether a utility time compensation of the line frequency has occurred (block 178). To determine whether the time compensation of the line frequency has occurred, let:
0~ =c-ppm(i)-c-ppm(i-1).

where i is the number of the ith time synchronization window at which the calculation are performed. In addition, c ppm(i) and c ppm(i-1 ) denote the counting errors for the period up to and including the ith window and the (i-1 )th window respectively. It can be appreciated that Expected cycles for the ith window equals i multiplied by the number of elapsed seconds multiplied by the nominal frequency fo.
To determine whether the utility has performed the line frequency time compensation during the synchronization window, the difference (0;) is calculated between the line frequency counting error (c ppm) of the current synchronization window and the previous synchronization window. On the one hand, if -1 < (0,) < 1 is not true (block 180), for example, line frequency compensation has not occurred. Since (0,) is expressed in ppm, the conditions -1 < (0,) < 1 indicates that the accumulated line frequency error is within one ppm, which is a small error. Of course other ranges could be used.
Thus, the cycles continue to be counted, and the counts contained in the decrement register are not adjusted (block 182).
On the other hand, if the difference (~,) is between -1 and 1 (block 180), the utility performed the line frequency compensation, and the line frequency error (I~pm) averages zero over the observed window. Since counting error (c~pm) is a combination of line frequency error (I ppm) and crystal error (x ppm), according to the formula c~pm = iJopm - x~pm, when line frequency error (l ppm) averages zero, crystal error (x ppm) equals counting error (c~pm). To synchronize the internal clock to the universal time, the PIT counts are adjusted by the crystal error (x ppm), which is determined by calculating (c~pm(i)) (block 184). The internal clock is synchronized to the line frequency by adjusting the counter located in the decrement register (block 184). Thereafter, the cycle counting begins again (block 174).
From the foregoing description, it should be understood that an improved time synchronization device and method has been shown and described which has many desirable attributes and advantages. Such time synchronization provides for an accurate sequence of fault recording. In addition, the present device and method speeds up or slow down a rate at which the internal clock time proceeds to avoid discontinuities and drift associated with merely changing the internal time to equal the universal time.
It is to be understood that changes and modifications to the embodiments described above will be apparent to those skilled in the art, and are contemplated. It is therefore intended that the foregoing detailed description be regarded as illustrative rather than limiting, and that it be understood that it is the following claims, including all equivalents, that are intended to define the spirit and scope of this invention.

Claims (14)

1. WHAT IS CLAIMED IS:
   
   A time synchronization device of a power monitor for synchronizing a time of an internal clock to a line frequency, the internal time advancing at a predetermined rate, the device comprising:
   a counter operative to count cycles of the line frequency after each elapse of a predetermined time period, said elapse of said predetermined time period defining a present window and a previous window;
   a processor operative to calculate a counting error and a change in counting error, said change in counting error being the difference of a counting error calculated for said present window and a counting error calculated for said previous window;
   wherein said processor detects when said change in counting error is within a predetermined range; and adjusting means for adjusting the internal clock to the line frequency according to said counting error calculated for said present window when said change in counting error is within said predetermined range.
2. 2. The device according to claim 2, wherein said adjusting means modifies the count of a decrement counter that keeps time for the internal clock.
3. 3. A method for synchronizing a time of an internal clock to a periodically stable frequency which provides accurate time intervals, the method comprising the steps of:
   counting cycles of the periodically stable frequency after each elapse of a predetermined time period, said elapse of said predetermined time period defining a present window and a previous window;
   calculating a counting error and a change in counting error, said change in counting error being the difference of a counting error calculated for said present window and a counting error calculated for said previous window;
   
   detecting when said change in counting error is within a predetermined range; and adjusting the internal clock to the periodically stable frequency according to said counting error calculated for said present window, when said change in counting error is within said predetermined range.
4. 4. The method according to claim 3, wherein said predetermined period is about 3600 seconds.
5. 5. The method according to claim 3, wherein said predetermined range is about minus one to one.
6. 6. The method according to claim 3, further comprising the steps of:
   receiving a time synchronization message from an entity at a first free running counter count, said time synchronization message containing a universal time when the time synchronization message was sent;
   processing said time synchronization message at a second free running counter count to determine said universal time.
7. 7. The method according to claim 6, wherein said entity is a master computer.
8. 8. The method according to claim 6, wherein said entity is a global positioning system.
9. 9. A method of maintaining time synchronization in a power monitoring device, the power monitoring devise coupled to a power system operating at a first frequency and comprising at least one processor operating at a second frequency, the method comprising:
   receiving at least one time message from a time source;
   accumulating counts in a free running counter in response to the second frequency;
   advancing the internal time of the device by a determined amount in response to the accumulation; and adjusting the amount in response to counting periods of the first frequency.
10. 10. The method of claim 9 wherein the time source comprises a computer.
11. 11. The method of claim 9 wherein the lime source comprises a global positioning system.
12. 12. The method of claim 9 further comprising determining when a utility compensates the first frequency for drift.
13. 13. The method of claim 12 wherein the time sound comprises a computer.
14. 14. The method of claim 12 wherein the time source comprises a global positioning system.