jonniezg / ewma Goto Github PK

Change:
EwmaT <int> filter(3, 100)

To:

#include <EwmaT.h>

// ...

EwmaT <int> filter(3, 100);

Well, the first is true but the second is false.
I've tested it, it's way slower than an optimized sliding average with one or more array.

On a reasonable amount of data, this method is around twice slower than what I've implemented !

Who can give me an example? There are already Arduino library files by jonnieZG,Exponential Weighted Moving Average.please give me a examples in arduino

The argument of non-using float is false.
When you look at the code, there're at least 3 divisions, which are very slow !

Arduino Nano compile does not have Serial.printf and even sprintf() does not have %f

FilteredADC:29:12: error: 'class HardwareSerial' has no member named 'printf'; did you mean 'print'?
Serial.printf("Raw=%d, Filter1=%.3f, Filter2=%.3f", raw, filtered1, filtered2);
^~~~~~
print

This breaks (returns high numbers instead of negatives when the "average" goes negative) when fed negative values.

EwmaT <long int> diff_ewmat(3, 10, 500); 
...
filtered = diff_ewmat.filter(value); 

This works (add/subtract an offset to keep everything positive):

EwmaT <long int> diff_ewmat(3, 10, 1000500); 
...
filtered = diff_ewmat.filter(value+1000000)-1000000; 

Sorry, C++ isn't nearly my day job, so maybe I'm expecting too much, but there's nothing to indicate this shouldn't work with signed numbers.

As far as I've understood the doc, all these lines should be equivalent:
EwmaT<int32_t> ewmat_sliding_Average(1, 125, 0); => result is ok !
EwmaT<int32_t> ewmat_sliding_Average(4, 500, 0); => result is KO !
EwmaT<int32_t> ewmat_sliding_Average(8, 1000, 0); => result is KO !

But it looks like, at some time in the calculation, there's an overflow (I add only values between 0 and 30000).

Some compile-time validation or so should be implemented to know what is the limit, .... or at least documented !