LabVIEW
Moving Average VI without using an array.
Hi all,
I am using a VI to calculate a simple moving average. I have attached the same below.
I have to build a VI to average 20 parameters. 20 parameters will take 20 counters and 20 arrays which might slower the computing power, consume a lot of development space and eventually slower the VI as my loop time is 1ms.
Is there any other way to implement a single moving average where we avoid using an array and thus reduce the computing power as we don't remember all the datapoints in the array.
As I am acquainted with C, I am thinking of using a single variable which keeps on adding n samples to itself until the counter reaches n. This way we don't have to store multiple datapoints in an array. Every next datapoint will be added to the variable which can then be linked to a summation block to calculate the average.
Thanks for the reply in advance.
One way to implement is a similar solution as you proposed, using a non-initialized Shift register.
On first Iteration, the innitialize array creates an array with fixed number (5 on this example),
Then on the next iterations, use replace array having the index coming from the reminder of the quotient of the iteration by 5. This way you allocate fixed memory for the array and keep reusing at every new data by replacing their content. Calculate average at every iteration
If you need to average 20 parameters, you can adapt do the same but initializing a 2D array then averaging each line (using an additional for loop).
There's a whole pallete of Point-by-Point functions. If you really want to do a moving average, try Mean PtByPt.vi. It's at Signal Processing -> Point by Point -> Probability and Statistics -> Mean PtByPt.vi.
Now that I've sort of answered the question you asked, let me throw out a few other things.
First, if you're running on Windows, a 1 ms loop time isn't going to be consistently achievable. You might average 1 ms, but you'll have plenty of jitter.
Second, if you're doing this to 20 signals, you shouldn't plop down 20 instances of this function. You should make it a 2D array as LVNinja said, or find some other way to do this math in a loop. The PtByPt functions only work with one instance at a time, so if you put it in a For loop your math will be off.
Third, memory on modern computers is VERY abundant. 20 arrays is nothing. Even if each array had 1,000 Double values in them, that's only 20,000 Doubles, or 160 kB. I'm not sure I'd worry about it array memory usage here.
@Satvikpai wrote:
As I am acquainted with C, I am thinking of using a single variable which keeps on adding n samples to itself until the counter reaches n. This way we don't have to store multiple datapoints in an array. Every next datapoint will be added to the variable which can then be linked to a summation block to calculate the average.
That will only work for an infinite history where you can just keep track of the sum and N. For any finite averaging window, once the size is reached, you need to add the new value and subtract the oldest, meaning you need to store the last N points in an array. Even C cannot go around that!
Since your voltage is a data source, why is it an indicator instead of a control, forcing the detour over a local variable?
(The "without using arrays" is a silly requirement, even the mentioned mean ptbypt internally uses an array.)
A detailed summary and performance considerations was discussed in our NI-Week talk a few years ago (part II) , so start there (slides 12-14, you can also download the code examples).
