I am designing a contact electromagnetic vibration sensor. I am pretty happy with the results. It is intended for audio frequencies. Pretty light: about 4 grams and 470 ohms impedance. The output is comparable to a dynamic mic.
The sensor is a cylinder 15 mm diameter and 13 height.
It can be attached with mounting putty.

So, the problem is that I don't know how to measure the response so that I can change some parameters in the design and see what happens...

So far, this is what I got:
Function generator to sweep the audio range.
Good power amp to drive the speaker
Speaker(Flat diaphragm from Tecton) mounted on a vertical board (no enclosure)
A reference microphone.
Audio interface (Komplete 6 by NI)
Software with FFT

I am tempted to uses laser interferometry but I don't what I am doing and, unless it is easy, I would leave it out.
So.....how do I calibrate this rig so the what I get from the FFT reflects the real response? How to compensate for the non linearity of the speaker?
THANK YOU FOR ANY IDEA!

 

One issue you'll probably have is the resonant frequency of the sensor – a small amount of vibration at the resonant frequency will overload the sensor.

The conventional way of calibration is with an electrodynamic shaker and a reference accelerometer, sweeping across the frequency range. If you happen to have a spare electrodynamic shaker, I can sketch a simple 'bird cage' design fixture to hold the transducer and reference accelerometer to the shaker.

As far as determining resonant frequency, that's far easier. The relationship "short in one domain, long in the other" is your friend in this case. A sine wave is long in the time domain, but short in the frequency domain (one line on an FFT). If you suspend your transducer in air and give it a rap with a screwdriver, the short time domain event provides broadband excitation in the frequency domain, exciting the transducer's natural frequencies. Look at an FFT of the rap, and the peaks will be at the transducer's natural frequencies.

 

All You need is an Accelerometer-Chip.

You can't build one.

Using a "Speaker" is not going to ever be anywhere close to Linear or accurate.
The closest that You can come with a Speaker is to
remove the entire Cone from a very-High-End ~10-inch PA-Speaker,
something in the ~$300.oo to ~$500.oo range.
What You are looking for in a Speaker is a "BL" rating well over ~10,
this is a measurement of the "Motor-Power" of the Magnet and Voice-Coil combination.
And it will still have lots of non-Linear weirdness.

Most cheaper Speakers don't provide a BL-rating,
because most people completely overlook Transient-Distortion caused by cheap Speaker design.
.
.
.