Taiko Synth - Phase 6 - Sensing the Bachi

I began with this simple code to detect the strikes, assuming that the start of the first negative slope signal the strongest part of the strikes.

void loop()
{
//polarity - white positive
amp = 0;
ampNew = 0;
don = false;
peak = false;
while(!don)
{
 ampNew = analogRead(0);
 if (ampNew > 25)
 {
   don = true;
   while(!peak)
   {
     amp = ampNew;
     ampNew = analogRead(0);
     if ((ampNew + 1) < amp) // plus 10 for noise compensation
     {
        peak = true;
        printSettings();       
     }
   }
 } 
}

In the above code, I've had to replace all the "greater than" and "less than" symbols with equivalent HTML entity names, and that seems to work fine. I thought anything with in a code block shouldn't have needed that.

Anyway...

This "don detect" algorithm gave some strange results. The first strike didn't seem to even register. Later strikes weren't proportional to the intensity of the hit. To investigate, wrote new code to capture as much data as possible and send it to Putty via USB serial at 115200 bps.

void setup() {
Serial.begin(115200);
}

void loop() {
Serial.println(analogRead(0));
}

Here are 5 strikes growing in amplitude:

Here's a close-up of the 5th strike:

Some interesting observations can be made. I'm not quite sure of the frequency, because I'm not sure of the sample rate, but you can see that the wave form is a truncated decaying sine wave. That's what one might expect from striking a board with a stick. What's a little surprising is that the first peak is not necessarily the greatest. There is also some dc offset at the end of the strike. From this, the lesson is that I need to change the "don detect" algorithm take the first 15 or so points after a strike is detected. Then, use the greatest of those to determine the peak amplitude. This method may trade latency for accuracy so I may need to tweak the final code a little. It may be a good idea to write more test code to capture these points from within the taiko app to make sure sufficient data is gathered to determine the true peak amplitude. This is because the taiko app surely has a different sample rate than the app above. To address the DC offset, I think I need to make sure that the foam is cut away from beneath the piezo sensor.