Moving average

abbiatishow said:

I don't know how to make an moving average filter.

In its simplest form, a moving average, or rolling average, is nothing more than taking the sum of the N latest samples and dividing by N.  This particular implementation allows each sample to have the same weight.  There are other modified methods which allow you to weight samples differently, perhaps to place heavier emphasis on the most recent samples over the older ones.

For example, if you wanted to perform a filter on the last 4 samples, the equation would be : SUM = (x[n] + x[n-1] + x[n-2] + x[n-3])/4

abbiatishow said:

What I don't understand is how to store the data coming from the ADC10MEM,

(to fill a vector for example).

The ADC10MEM register holds the latest conversion for a particular channel.  This will only hold 1 sample.  Therefore, your software, perhaps an Interrupt service routine, would copy the value stored in this register and place it in an array.  An example is below based on the above example.

Then in your background task, you perform the filter above.

#define NUM_SAMPLES 4

short samples[NUM_SAMPLES] ;
int sample_index = 0 ;

void ADC_interrupt (void)
{
  samples[sample_index] = ADC10MEM ;
  sample_index = (sample_index + 1) % NUM_SAMPLES ;  // modulus 4 (wrap around)
}

Thank you for the quick reply, butI still don't understand how to do it.

I have to adquire 3 analog signals coming from the axis of my acceletometer,
read them and calculate mean , variance etc.

This is part of my code:


void get_accelerations()
{
    float battery_voltage;
    ADC10AE0 = 0xff;                                                           
    ADC10CTL0 = 0x00;
    ADC10CTL0 = SREF_0 + ADC10SHT_2 + ADC10ON;
    ADC10CTL1 = INCH_0;                                                  // input  1 of 3   // these lines have to be repeated for every axis
    ADC10MEM = 0xffff;
    ADC10CTL0 |= ENC + ADC10SC;             // Sampling and conversion start
    do{}while(ADC10MEM == 0xffff);
    voltage_x = ((ADC10MEM * battery_voltage) / 1023.0);
//
// filtering here... take the last 16 samples of voltage_x and make some calculation
// for example mean or variance or some classification etc
//
// SUM = (x[n] + x[n-1] + x[n-2] + x[n-3])/4 ??? how to do it ??
 

}

you told me to add this code.. but I didn t understand exactly what to do...

#define NUM_SAMPLES 4

short samples[NUM_SAMPLES] ;
int sample_index = 0 ;

void ADC_interrupt (void)
{
  samples[sample_index] = ADC10MEM ;
  sample_index = (sample_index + 1) % NUM_SAMPLES ;  // modulus 4 (wrap around)

}

Thank you so much for helping me....

So the code has to be like this (or not) ?

void ADC_interrupt (void);
int voltage_ave = 0 ;              // int or float ?
int i ;
short voltage_history[4] ;
int voltage_index  = 0 ;

void get_accelerations()
{

  ADC10CTL0 = 0x00;

    // int voltage_ave = 0 ;  //out of the function otherwise declaration may not appear after executable statement in block

    // int i ;

   ADC10CTL0 = SREF_0 + ADC10SHT_2 + ADC10ON;

   ADC10CTL1 = INCH_0;                                                  // input A0(?) X
    ADC10MEM = 0xffff;
    ADC10CTL0 |= ENC + ADC10SC;             // Sampling and conversion start
    do{}while(ADC10MEM == 0xffff);
    while(!(ADC10CTL0 & ADC10IFG));          // Wait for conversion to complete
    voltage_x = ((ADC10MEM * battery_voltage) / 1023.0);
    // Place latest voltage into history circular buffer
    voltage_history[voltage_index] = voltage_x ;
    voltage_index = (voltage_index + 1) % 4 ;
    for (i=0; i<4; i++)
    {
      voltage_ave =  voltage_ave + voltage_history[i] ;
    }

    voltage_ave = voltage_ave >> 2 ;  // divide by 4 by right shifting 2   

} // end get_accelerations()

void ADC_interrupt (void)
{

  voltage_x = ADC10MEM;        // is it correct?

}

One more question:

in this way, doing the shift, voltage_ave can just be a integer value.

If I want a float value.. what to do?

Is it possible to do: voltage_ave = voltage_ave /4   ?

Thanks again

Thank you so much.. it looks like working.. except for voltage_index, I think it has to start from 1.

abbiatishow said:

Thank you so much.. it looks like working..

Great.

abbiatishow said:

except for voltage_index, I think it has to start from 1.

Why do you say that?

Sorry, you are right.

And how to do if I want to vectorize the final result as well?

What I want to say is that  I am trying to use a bigger window (short voltage_history[8]  or more)

so I can store the final result that changes with the time.

Hi,

I am doing a project with MSP430G2231. I have to measure energy of the waveform and I am planning to do this by applying the signal to ADC of MSP430G2231, then transferring the digital values in an array and then using some mathematical calculations to compute energy. so for the first part I am trying to use the same code as given above,

#include "msp430g2231.h"


void main(void)
{
int voltage_ave ;
short voltage_history[4] ;
int voltage_index  = 0 ;


  int i ;
  ADC10CTL0 = 0x00;
  ADC10CTL0 = SREF_0 + ADC10SHT_2 + ADC10ON;
  ADC10CTL1 = INCH_0;                     // Input : A0
  ADC10CTL0 |= ENC + ADC10SC;             // Sampling and conversion start
  while(!(ADC10CTL0 & ADC10IFG));         // Wait for conversion to complete
 
  // Place latest voltage into history circular buffer
  voltage_history[voltage_index] = ADC10MEM ;
  voltage_index = (voltage_index + 1) % 4 ;
  voltage_ave = 0 ;
  for (i=0; i<4; i++)
  {
    voltage_ave =  voltage_ave + voltage_history[i] ;
  }

  voltage_ave = voltage_ave >> 2 ;  // divide by 4 by right shifting 2

but I don't know why the code is not working. Can you please guide me through this? The problems I get are garbage values in voltage_av, voltage_history. Possibly the program halts at     while(!(ADC10CTL0 & ADC10IFG));         // Wait for conversion to complete,

Please let me know.

Thank you for your help.