Using AudioMemory - AudioMemory changes analog readings from a microphone

[Billyjack135]

Hi all,

In short, I am using a SparkFun sound detector (https://www.sparkfun.com/products/12642) in conjunction with the teensy4.0 to develop an audio analysis program that detects a car approaching a point on the road.

When reading the raw analog audio input, it reads a magnitude change in a sample array of length 1000. But when implementing a
AudioWindowHanning1024 using an audiomemory(8+) the magnitude changes of the sample double to around 50 or 60. Which should I use, audio memory >7 or <7 and why does this happen?

 

[MarkT]

You need a minimum of 8 128-sample audio blocks to work with FFT of size 1024 as 8*128 = 1024.

 

[Billyjack135]

You need a minimum of 8 128-sample audio blocks to work with FFT of size 1024 as 8*128 = 1024.

Could you explain this a little more? I still don’t understand why the audio memory changes the raw data values.

Also thanks for the reply! Do you have any recommendations when it comes to using FFT’s to identify specific sounds?

 

[MarkT]

The FFT relies on gathering 8 consecutive audio blocks and then gets to work. Without 8 distinct blocks it will have wrap-around and mess up audio block allocation for the whole audio chain I believe.