20
What the Tensor Records
Introduction
The job of any power monitor is to record all significant data and leave unrecorded a vast majority of
background data. The tricky part for a monitor is deciding which events are important. This is a problem of
data reduction. A recorder that captured every 60 Hz waveform during a week’s recording would never miss
an event but would present the user with millions of useless cycles. Conversely, a recorder whose thresholds
are set incorrectly may not record anything. Staying somewhere between these two extremes involves a
balance of thresholds, settings, and record types. The monitor will see an enormous amount of data on its
voltage and current inputs– the Tensor Recorder sees over 1 billion samples per day! Ideally, all this data is
reduced to a small report which just shows the important events and measurements. The sifting of data into
specific record types accomplishes this task.
Triggered Record Types
PMI Recorder records can be divided into two classes. The first is event driven. These record types are
triggered by a combination of triggering logic and adjustable thresholds, usually voltage based. If a trigger
never happens, nothing is recorded for that record type. As more triggers occur, more records are collected
for that record type. The advantage of this class is that nothing is recorded unless something happens. In the
ideal case, no problems occurred, so nothing was recorded, and no data analysis is necessary. If a trigger did
occur, then the monitor logged the event for later analysis. This is a powerful data-reduction tool and can
reduce huge amounts of data into a few small records containing all the significant events. The disadvantage is
that success completely depends on good thresholds and settings. A threshold that is too tight will cause the
Recorder to log records that are not really worth analyzing.
These extraneous records often hide the (hopefully) few important ones. A threshold that is too loose will
cause the Recorder to ignore important disturbances. Although it is often possible to use regulatory limits or
other known standards to set thresholds, this can be a chicken-and-the-egg type problem: sometimes you
need to know something about the disturbance before you can set proper thresholds to capture it. Despite
these potential pitfalls, triggered record types are powerful tools in power line monitoring. They are most
useful for capturing voltage disturbances and power quality problems. The captured events are usually
presented in a text report. Triggered record types include Power Outage, Abnormal Voltage, Event Change,
Significant Change, and Waveform Capture.
Non-triggered Record Types
The second class of record types is not event driven. These record types are always logging data, regardless of
how interesting or important the data is. The classic example is a paper graph, which continuously logs data.
There are no thresholds to set, although there may be a parameter to determine how often to collect data.
The logged data is usually presented as a graph of data points. Although there may be a large amount of data,
using a graph lets the eye pick out important data. Problems such as sags and swells are easy to see in the
graph graphs. In addition to voltage quality studies, these record types are used for finding daily trends in
current or power values, measuring power factor, etc. The advantage of not having thresholds to set is that
there is no question about what data will be recorded. The disadvantage is that sometimes there is no
question that a lot of data will be recorded, most of it unimportant. For non- power quality data such as power
factor measurement, there is no disadvantage. These record types include graphs, Daily Profiles, Histograms,
and Energy Usage.
