Filtering ADC input before frequency calculations

Good morning, I am working on a senior design project in which an arduino is tasked with taking the input off of an Op Amp that I set up in order to amplify an audio signal and calculating the frequency being played. The only problem is, a trumpet has harmonics that cause the sample Op amp frequency detection code to jump between all of the frequencies it is hearing. In order to combat this issue I decided on implementing a DEMA filter on the front end of the ADC converter in the arduino code. My issue comes in when the converter appears to not be functioning properly due to the Serial Terminal only displaying “int” or “0” hz. I’m looking for someone who will hopefully be able to explain this to me.

Here is my code.

//generalized wave freq detection with 38.5kHz sampling rate and interrupts
//by Amanda Ghassaei
//Sept 2012

 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 3 of the License, or
 * (at your option) any later version.

//clipping indicator variables
boolean clipping = 0;

//data storage variables
byte newData = 0;
byte prevData = 0;
unsigned int time = 0;//keeps time and sends vales to store in timer[] occasionally
int timer[10];//sstorage for timing of events
int slope[10];//storage for slope of events
unsigned int totalTimer;//used to calculate period
unsigned int period;//storage for period of wave
byte index = 0;//current storage index
float frequency;//storage for frequency calculations
int maxSlope = 0;//used to calculate max slope as trigger point
int newSlope;//storage for incoming slope data

//variables for decided whether you have a match
byte noMatch = 0;//counts how many non-matches you've received to reset variables if it's been too long
byte slopeTol = 3;//slope tolerance- adjust this if you need
int timerTol = 10;//timer tolerance- adjust this if you need

//variables for amp detection
unsigned int ampTimer = 0;
byte maxAmp = 0;
byte checkMaxAmp;
byte ampThreshold = 30;//raise if you have a very noisy signal

void setup(){
  pinMode(13,OUTPUT);//led indicator pin
  pinMode(12,OUTPUT);//output pin
  cli();//diable interrupts
  //set up continuous sampling of analog pin 0 at 38.5kHz
  //clear ADCSRA and ADCSRB registers
  ADCSRA = 0;
  ADCSRB = 0;
  ADMUX |= (1 << REFS0); //set reference voltage
  ADMUX |= (1 << ADLAR); //left align the ADC value- so we can read highest 8 bits from ADCH register only
  ADCSRA |= (1 << ADPS2) | (1 << ADPS0); //set ADC clock with 32 prescaler- 16mHz/32=500kHz
  ADCSRA |= (1 << ADATE); //enabble auto trigger
  ADCSRA |= (1 << ADIE); //enable interrupts when measurement complete
  ADCSRA |= (1 << ADEN); //enable ADC
  ADCSRA |= (1 << ADSC); //start ADC measurements
  sei();//enable interrupts

ISR(ADC_vect) {//when new ADC value ready
  PORTB &= B11101111;//set pin 12 low
  prevData = newData;//store previous value
  newData = ADCH;//get value from A0
  if (prevData < 127 && newData >=127){//if increasing and crossing midpoint
    newSlope = newData - prevData;//calculate slope
    if (abs(newSlope-maxSlope)<slopeTol){//if slopes are ==
      //record new data and reset time
      slope[index] = newSlope;
      timer[index] = time;
      time = 0;
      if (index == 0){//new max slope just reset
        PORTB |= B00010000;//set pin 12 high
        noMatch = 0;
        index++;//increment index
      else if (abs(timer[0]-timer[index])<timerTol && abs(slope[0]-newSlope)<slopeTol){//if timer duration and slopes match
        //sum timer values
        totalTimer = 0;
        for (byte i=0;i<index;i++){
        period = totalTimer;//set period
        //reset new zero index values to compare with
        timer[0] = timer[index];
        slope[0] = slope[index];
        index = 1;//set index to 1
        PORTB |= B00010000;//set pin 12 high
        noMatch = 0;
      else{//crossing midpoint but not match
        index++;//increment index
        if (index > 9){
    else if (newSlope>maxSlope){//if new slope is much larger than max slope
      maxSlope = newSlope;
      time = 0;//reset clock
      noMatch = 0;
      index = 0;//reset index
    else{//slope not steep enough
      noMatch++;//increment no match counter
      if (noMatch>9){
  if (newData == 0 || newData == 1023){//if clipping
    PORTB |= B00100000;//set pin 13 high- turn on clipping indicator led
    clipping = 1;//currently clipping
  time++;//increment timer at rate of 38.5kHz
  ampTimer++;//increment amplitude timer
  if (abs(127-ADCH)>maxAmp){
    maxAmp = abs(127-ADCH);
  if (ampTimer==1000){
    ampTimer = 0;
    checkMaxAmp = maxAmp;
    maxAmp = 0;

void reset(){//clea out some variables
  index = 0;//reset index
  noMatch = 0;//reset match couner
  maxSlope = 0;//reset slope

void checkClipping(){//manage clipping indicator LED
  if (clipping){//if currently clipping
    PORTB &= B11011111;//turn off clipping indicator led
    clipping = 0;

void loop(){
  if (checkMaxAmp>ampThreshold){
    frequency = 38462/float(period);//calculate frequency timer rate/period
    //print results
    Serial.println(" hz");
  delay(100);//delete this if you want
  //do other stuff here

And here is the filter

int EMA_function(float alpha, int latest, int stored);
int sensor_pin = 0;
int ema_a = 0.06;
int ema_ema = 0;
int ema = 0;
void setup() {
void loop() {
  int sensor_value = analogRead(sensor_pin);
  ema = EMA_function(ema_a, sensor_value, ema);
  ema_ema = EMA_function(ema_a, ema, ema_ema);
  int DEMA = 2*ema - ema_ema;
int EMA_function(float alpha, int latest, int stored){
  return round(alpha*latest) + round((1-alpha)*stored);

Why do you think a digital filter will stop any problem like that? Once your input is over half the sample rate no amount of digital filtering is going to work. You need an analogue filter before you sample.

At most times == crap

So even if I implement the DEMA just after ISR(ADC_vect) it won't be able to filter?

If any signal with frequency greater than half the sample frequency is present in the input, the damage is already done.

Please read up on aliasing, and the sampling theorem.

You MUST ensure that frequencies higher than half your sample frequency cannot enter the analog to digital converter. Without this assurance, all results are unreliable. The simple way to do this is a resistor and capacitor “RC filter”.

The comments in your code say that you have a sample frequency of 500kHz. That seems very high for an audio signal. However it does mean that you don’t need a very precise filter to pass everything under (say) 10kHz while blocking everything above 250kHz.

Then you can apply digital filters for more precise filtering. Ive never seen a DEMA filter before but at first glance it seems unsuitable for this application. Do you know the cutoff frequency and attenuation of this filter?

When I am designing a filter in the frequency domain, I use