High Speed Metallic Clink Listener

[invisiblefork]

Hi All -

I’m working on a project in which I have to listen to repetitious sounds that follow a similar pattern, unless there is something wrong. The receiver needs to be able to listen to metallic clinks and report back if there is an abnormal pattern. The typical speed of movement is around 2000-7000 RPM, so the sensor would have to work rather quickly.

Right now I’m looking at the electret microphone connected to a GrovePi to accomplish this:

Electret Microphone Amplifier - MAX9814 with Auto Gain Control

I’m curious if anyone has used this sensor in order to distinguish high speed patterns from one another, or if it is feasible to do so?

[RobertLucian]

Hi @invisiblefork,

I suggest you read the following 2 threads:

Hi @dgkanatsios, There’s something we need to clarify before diving too far. There’s a difference between a microphone and a sound level sensor. The microphone outputs a signal that’s well within the kilohertz range. If you’d connect the microphone to an amplifier and then to a pair of headphones/speaker, you’ll be able to hear whatever the microphone records. This means there’s lots of information packed into that signal and with our GrovePi, we can only catch a glimpse of that signal. …

And this one here:

Hi @grovepi3, The sound sensor is theoretically, a better sensor than the loudness sensor as it’s more rich in the amount of data it caries. Unfortunately, the sound sensor is rather an ill-fitted sensor for the GrovePi as our board is only capable at reading values up to a certain frequency that’s much lower than the frequency at which the sensor is operating. Thus, we introduced the loudness sensor which has incorporated a low-pass filter and makes it more suitable for our platform. On the …

Within these 2 threads you will find an explanation as to why a sound sensor is a bad sensor choice for the GrovePi. Unfortunately, in your case, there’s enough processing power to do any data acquisition at these frequencies.
And even if you do, you’d loose most of the data due to the low rate of data acquisition (the GrovePi can only up to 100-200 Hz).

For this kind of application, I think what you need is a DSP.

Thank you!

[cyclicalobsessive]

Any Update on this project?

I’ve been thinking of something vaguely similar using a Kinobo USB microphone listening via pyaudio and audioop calls to recognize the sound of someone typing, (a pseudo-rhythmic clicking above the average sound level.)