Compression Tab

Collector compression applies a smoothing filter to incoming data by ignoring incremental changes in values that fall within a deadband centered around the last reported value. The collector reports any new value that falls outside the deadband to the Historian archive and then centers the deadband around the new value.

Example:

The engineering units are 0 to 200. The deadband value is 10%, which is 20 units. If the deadband value is 10% and the last reported value is 50, the value will be reported when the current value exceeds 50 + 10 = 60 or is less than 50 - 10 = 40. Note that the deadband (20 units) is split around the last data value (10 on either side.)

Alternatively, you could specify an absolute deadband of 5. In this instance, if the last value was 50, a new data sample will be reported when the current value exceeds 55 or drops below 45.

If compression is enabled and the deadband is set to zero, the collector ignores data values that do not change and records any that do change. If you set the deadband to a non-zero value, the collector records any value that lies outside the deadband. If the value changes drastically, a pre-spike point may be inserted. See Spike Logic for more details.

When enabling Archive Compression or Collector Compression, the Engineering Units field represents a calculated number created to give an idea of how large a deadband you are creating in Engineering Units. The deadband is entered in % and Historian multiplies that % by the range (Hi Engineering Units - Lo Engineering Units) to compute the % in Engineering Units.

For polled tags, the Collector Compression Timeout value should be in multiples of your collection interval. After the timeout value is exceeded, the tag stores a value at the next scheduled collection interval, and not when the timeout occurred. For example, if you have a 10 second collection interval, a 1 minute compression timeout, and a collection that started at 2:14:00, if the value has not changed, the value is logged at 2:15:10 and not at 2:15:00.

For unsolicited tags, a value is guaranteed in, at most, twice the compression timeout interval.

A non-changing value would be logged on each compression timeout. For example, an unsolicited tag with a 1 second collection interval and a 30 second compression timeout would be stored every 30 seconds.

A changing value for the same tag may have up to 60 seconds between raw samples. In this case, if the value changes after 10 seconds, then that value is stored, but the value at 30 seconds (if unchanged) will not be stored. The value at 60 seconds will be stored. This leaves a gap of 50 seconds between raw samples which is less than 60 seconds.

Compression timeout is supported in all collectors except the PI collector.

Each time the system reports a new value, it computes a line between this data point and the last archived value. The deadband is calculated as a tolerance centered about the slope of this line. When the next data point is reported, the line between the new point and the last archived point is tested to see if it falls within the deadband tolerance calculated for the previous point. If the new point passes the test, it is reported and is not archived. This process repeats with subsequent points. When a value fails the tolerance test, the last reported point is archived and the system computes a line between the new value and the newly archived point, and the process continues.

When enabling Archive Compression or Collector Compression, the Engineering Units field represents a calculated number created to give an idea of how large a deadband you are creating in Engineering Units. The deadband is entered in % and Historian multiplies that % by the range (Hi Engineering Units - Lo Engineering Units) to compute the % in Engineering Units.

The data archiver treats the incoming sample after the timeout occurs as if it exceeded compression. It then stores the pending sample.