9-8
L90 Line Current Differential System
GE Multilin
9.1 OVERVIEW
9 THEORY OF OPERATION
9
the most reliable source of phase information but suffer from a phase offset due to a difference in the channel delays in
each direction between a pair of relays. In some cases, one or both directions may be switched to a different physical path,
leading to gross phase error.
The primary source of phase information are CPU time-tagged messages. If GPS compensation is enabled, GPS time
stamps are used to compensate for asymmetry. In all cases, frequency deviation information is also used when available.
The phase difference between a pair of clocks is computed by an exchange of time stamps. Each relay exchanges time
stamps with all other relays that can be reached.
It is not necessary to exchange stamps with every relay, and the method works even with some of the channels failed. For
each relay that a given relay can exchange time stamps with, the clock deviation is computed each time a complete set of
time stamps arrives. The net deviation is the total deviation divided by the total number of relays involved in the exchange.
For example, in the case of two terminals, each relay computes a single time deviation from time stamps, and divides the
result by two. In the case of three terminals, each relay computes two time deviations and divides the result by three. If a
channel is lost, the single deviation that remains is divided by two.
Four time stamps are needed to compute round trip delay time and phase deviation. Three stamps are included in the mes-
sage in each direction. The fourth time stamp is the time when the message is received. Each time a message is received
the oldest two stamps of the four time stamps are saved to become the first two time stamps of the next outgoing message.
The third time stamp of an outgoing message is the time when the message is transmitted. A fixed time shift is allowed
between the stamp values and the actual events, provided the shift for outgoing message time stamps is the same for all
relays, and the shift incoming message time stamps is also identical.
To reduce bandwidth requirements, time stamps are spread over 3 messages. In the case of systems with 4 messages per
cycle, time stamps are sent out on three of the four messages, so a complete set is sent once per cycle. In the case of sys-
tems with 1 message per cycle, three time stamps are sent out each cycle in a single message. The transmit and receive
time stamps are based on the first message in the sequence.
One of the strengths of this approach is that it is not necessary to explicitly identify or match time stamp messages. Usually,
two of the time stamps in an outgoing message are simply taken from the last incoming message. The third time stamp is
the transmittal time. However, there are two circumstances when these time stamps are not available. One situation is
when the first message is transmitted by a given relay. The second is when the exchange is broken long enough to invali-
date the last received set of time stamps (if the exchange is broken for longer than 66 ms, the time stamps from a given
clock could roll over twice, invalidating time difference computations). In either of these situations, the next outgoing set of
time stamps is a special start-up set containing transmittal time only. When such a message is received, nothing is com-
puted from it, except the message time stamp and the received time stamp are saved for the next outgoing message (it is
neither necessary nor desirable to “reset” the local clock when such a message is received).
Error analysis shows that time stamp requirements are not very stringent because of the smoothing behavior of the phase
locked loop. The time stamp can be basically a sample count with enough bits to cover the worst round trip, including chan-
nel delay and processing delay. An 8 bit time stamp with 1 bit corresponding to 1/64 of a cycle will accommodate a round
trip delay of up to 4 cycles, which should be more than adequate.
The computation of round trip delay and phase offset from four time stamps is as follows:
(EQ 9.26)
The
T
s are the time stamps, with
T
i
the newest. Delta is the round trip delay. Theta is the clock offset, and is the correct sign
for the feedback loop. Note that the time stamps are unsigned numbers that wrap around, while
a
and
b
can be positive or
negative;
i
must be positive and
i
can be positive or negative. Some care must be taken in the arithmetic to take into
account possible roll over of any of the time stamps. If
T
i
– 2
is greater than
T
i
– 1
, there was a roll over in the clock respon-
sible for those two time stamps.
To correct for the roll over, subtract 256 from the round trip and subtract 128 from the phase angle. If T
i
– 3
is greater than T
i
,
add 256 to the round trip and add 128 to the phase angle. Also, if the above equations are computed using integer values
of time stamps, a conversion to phase angle in radians is required by multiplying by
/ 32.
a
T
i
2
–
T
i
3
–
–
=
b
T
i
T
i
1
–
–
=
i
a b
+
=
i
a b
–
2
------------
=
Summary of Contents for UR Series L90
Page 652: ...A 16 L90 Line Current Differential System GE Multilin A 1 PARAMETER LISTS APPENDIX A A ...
Page 772: ...B 120 L90 Line Current Differential System GE Multilin B 4 MEMORY MAPPING APPENDIX B B ...
Page 802: ...C 30 L90 Line Current Differential System GE Multilin C 7 LOGICAL NODES APPENDIX C C ...
Page 812: ...D 10 L90 Line Current Differential System GE Multilin D 1 IEC 60870 5 104 APPENDIX D D ...
Page 824: ...E 12 L90 Line Current Differential System GE Multilin E 2 DNP POINT LISTS APPENDIX E E ...
Page 834: ...F 10 L90 Line Current Differential System GE Multilin F 3 WARRANTY APPENDIX F F ...
Page 846: ...xii L90 Line Current Differential System GE Multilin INDEX ...