XT2640 Operating Manual
13 July 2016
Page 115 of 187
22
SYNCHRONIZING
MEASUREMENTS
BETWEEN
VPAS
The XT2640 can be configured to synchronize the starting time of measurements in different VPAs to accommodate measurements
of power loss and efficiency when the power levels are varying rapidly over time. This is always possible within a single XT2640 and
optionally also between multiple XT2640s using the Multi‐Unit capabilities (see section 21).
This is achieved by selecting the SYNC measurement mode in the XT2640 (for a single XT2640) or in each XT2640 which it is desired
to synchronize within a Multi‐Unit group.
Note:
Not all units in a Multi‐Unit Group need to be configured with the same measurement mode, the same measurement period
or use the same fundamental frequency.
Selecting the SYNC measurement mode only has any benefit when it is desired to accurately track power loss and/or
efficiency during significant changes of power level. In most applications the SYNC mode is not recommended.
Synchronizing VPAs in this manner causes all such VPAs except for the one with the longest measurement period to have
small gaps between their measurement periods during which no measurement is being made. Typically this has little
consequence and is not specific to the XT2640 but applies to any synchronized measurement periods having different
lengths.
This is not the same as selecting for the measurement period of a VPA to be synchronous to that of another VPA in the same
XT2640 by setting the PERIOD setting in one VPA to another VPA in the same XT2640 (see 12.4.2.2.1 for details regarding
this) as configuring in that manner selects that both VPAs have exactly the same measurement period and forces them to
use exactly the same fundamental frequency for measurement synchronization (this is only available within a single
XT2640). When configured for SYNC mode, even in different XT2640s, you can configure any VPAs to use the same
fundamental frequency and so have the exactly the same measurement period length by using the FUND configuration
setting or they can each independently derive their own fundamental frequencies as needed.
Within a single XT2640, or within multiple XT2640’s in the same Multi‐Unit group, if configured in SYNC measurement mode all
VPAs perform synchronization collectively as follows –
After completing every measurement period every such VPA checks if all other VPAs in the collection are presently waiting
after completing their measurement periods. If all such VPAs have already completed their measurement periods then all
VPAs are commanded to start a new measurement period, otherwise this VPA waits to be commanded by another VPA to
start a further measurement period and informs all other VPAs as such (using a multi‐cast packet within a Multi‐Unit group
if synchronizing multiple XT2640’s).
In this manner each VPA may optionally be configured with a different measurement period (or not), whether by setting a different
PERIOD setting for each or because VPAs have different signals to which they are synchronizing their measurement periods (i.e. each
is measuring signals having different and unrelated frequency content).
Within a single XT2640 the starting times of such measurements are synchronized to within 10µs, and for VPAs in different
XT2640’s using the Multi‐Unit capability this is extended slightly by the LAN interface delay. The synchronizing information passed
between multiple XT2640’s is a single minimum length packet on the LAN interface which is multicast to all configured XT2640’s
with the highest priority, so typically this incurs an additional delay of 6.4µs times (1 + the number of intervening switches in the
network topology).
The effect of such slight differences between the starting times of the measurements is typically quite minor as described below‐
These differences only effect measurements of power loss and efficiency when signals are changing; there is no error if the
signals are not substantially changing and there is no error in the measurements of power within each VPA itself under any
circumstances.
Since multiple XT2640s use multi‐casted communications for synchronization and almost all switches forward multi‐casted
packets virtually simultaneously between ports, this delay is not dependent on the number of XT2640s in a Multi‐Unit Group.
The worst case delay (for a 100M LAN) between multiple XT2640’s with a single intervening switch is 152µs if both a) there is
other traffic present on that network portion at exactly the same time as one XT2640 finishes a measurement period and b)
that traffic has the maximum packet length and c) is destined for one of the XT2640s being synchronized. Traffic between
other network nodes has negligible affect for typical switches as they employ different packet buffers and routing between
nodes.
Irrespective of the cause of these delay errors, they are not cumulative. I.e. the error over 5 measurement periods is 1/5
th
of
the error over one measurement period. In many circumstances it is useful to use this to advantage by selecting a faster
nominal measurement period in every VPA and selecting for averaging using the same RESPONSE setting in each VPA.
If all such VPAs are configured for 100ms nominal measurement periods then the worst‐case error in power loss or efficiency
measurements is 0.001% of the inter‐measurement change per µs of delay between VPAs (or 0.01% if configured for a period
of 10ms, or 0.00033% if configured for 300ms). Within the same XT2640 this yields a worst case error of 0.01% of the
change, or typically 0.0164% of the change with different XT2640’s without intervening switches. As an example using 100ms
measurement periods, if a power level suddenly changes from 100W to 200W during a single 100ms measurement period,
then there is a possible error of up to 0.01% of 100W (i.e. 10mW) in power loss/efficiency in that one 100ms measurement
period in which the 100W power level change occurred but there is no error in the measured power loss or efficiency in any
measurement period either before or after the change.
If such VPAs are measuring signals having different frequency content then the differences in measurement periods caused by
that different content is typically of more significance than errors caused by synchronization delays. This is because the
measurement periods are not the same and so exact synchronization is not possible. It is not possible to give a general typical
figure for this effect, but the measurement period for every VPA can be estimated by calculating the period of the nearest
integer number of fundamental cycles in the configured measurement period and then calculating the difference between the
