====Using Timers To Allow Values to Stabilise====

In some cases, values can take a while to stabilise after an abrupt change. 

A quick and easy way to ignore some of these instabilities is to use a //timer// to create a window of time where your analytic will ignore issues.

For example, let's say we are making a [[example_pivot_analogue|pivot analytic]] where we want to ignore the 10 seconds after we move between making product A and product B.

We could do the following...

===Determine if a Change has Happened===

Before we can start a timer, we first need to know if we **should** start a timer.

To see if there's been a change of //any kind// in a property, we can compare the **current value** to the **previous value** - if they are different, we should start the count-down.

==Data From Production/Scheduling/Recipe Systems==

{{any_difference.png}}

The example above uses a [[ardi_point|ARDI Point]] and a [[ardi_ppoint|Previous Point]] node and compares the values with a [[math_condition|Condition]].

==Data From Sensors==

The previous example works fine properties that move in fixed, sudden movements. But for analogue sensor values - which constantly drift very slightly up and down - you might want to only set your timer if you see a large-enough change.

{{big_difference.png}}

This is a similar solution to the first one, but it adds a [[math_sub|Subtraction]] node and an [[math_abs|Absolute Value]] node to make sure that the difference between the current and previous value is more than 1.5 units.

===Introducing the Timer===

Now that we've determined that a change has taken place, we need to start our 10-second timer. This is done with the [[timing_timer|Timer]] node.

{{pivot_ignore_timing.png}}

At it's most basic, the **finished** output will become //true// when the number of //seconds// has elapsed.

The **counting** output will be //true// when the timer is **running** but hasn't yet **finished**.

By setting the //activation// to //latched//, we make sure that even a momentary difference will start the timer, rather than the difference needing to last the full 10 seconds.

By setting the //reset// to //auto//, we make sure that the timer will reset by itself once it finishes. 

So in this case, we are...

  * Starting the timer the moment we see that the thickness has changed,
  * The timer begins counting down, setting the //Counting// output to True,
  * //Counting// remains true for the next 10 seconds
  * At the end of the 10 seconds, the timer resets and is ready to run again.

This means that our report will ignore any values in the 10 seconds after a product thickness change.