Relaxation measurements

[Cogito]

Hello,

"please allow me to introduce myself"

I am using audicity to identify the resonance frequency (RF) of table tennis blades. The speed of a blade corresponds to its RF: the lower/greater the RF the lower/greater the speed. For the RF determination I hold the blade at its grip and let a ball bounce from about 10 cm height. The noise is recorded by a microphone and audicity. After the measurement I select one of the sound peaks generated by the ball touching the blade and use the function "frequence analysis" to pick up the largest peak and read the indicated frequency.

However this is ony semiquantitativ. But an absolute speed parameter could be calculated from the recorded track. What does fast or slow blade mean ? If you drop the ball from 10 cm on a low speed blade it will come up only x cm and the time intervall before the second contact will be only y sec. With a faster blade it would be x+ (delta)x cm and y + (delta)y sec. Unfortunately it would create quite an effort to exactly reproduce the 10 cm (or whatever) height for different measurements.

So I need something independent form the parameter "initial height". This can be elaborated by measuring the speed of decrease of x respectively y between the bouncings i.e I need the time intervall between two spikes plotted vs time.

Unfortunately I have no idea about programming. Is somebody able and willing to help ?

Thanks
Peter

[steve]

Interesting project

For the RF determination I hold the blade at its grip and let a ball bounce from about 10 cm height. The noise is recorded by a microphone and audicity. After the measurement I select one of the sound peaks generated by the ball touching the blade and use the function "frequence analysis" to pick up the largest peak and read the indicated frequency.

I presume that you mean "Plot Spectrum" for frequency analysis?

There are a number of issues with the test as you describe it that can probably be improved.

"about 10 cm height"
As someone with a scientific background I am very wary of unknown variables. In the design of experiments for quantitative analysis, it is generally best to reduce "unknowns" as far as possible, and where variables cannot be eliminated, to at least limit their range within a specified tolerance range. More about this later.

"I hold the blade at its grip"
This introduces many more unknown variables. How tightly are you holding it? How much variability is there in the angle of the blade? How much does the blade move due to inevitable unsteadiness of the hand that is holding it?

Perhaps the above two items could be improved by constructing some type of clamp to hold the blade, and a short tube clamped at a known height above the blade that is very slightly angled down from the horizontal (such as the cardboard tube from the middle of a toilet paper roll). Gently place the ball in the higher end of the tube so that it slowly roll down the tube and fall out of the other end, then drops down to the blade that is a known distance below. The tube and the blade clamp should be carefully aligned to ensure that the ball strikes the blade at the same position for each measurement.

"The noise is recorded by a microphone"
I would suspect that the majority of the sound picked up by the microphone will come from the ball rather than the bat.
Perhaps this could be improved by using a "contact microphone" (a piezo sensor stuck onto the bat) rather than a conventional microphone. A piezo polymer vibration sensor would be much better than a ceramic disk as the latter have strong resonance themselves whereas the former does not.

"Unfortunately it would create quite an effort to exactly reproduce the 10 cm (or whatever) height for different measurements."
See above.

[Cogito]

Thanks for your answer.

"about 10 cm height"
As someone with a scientific background

I don't have a scientific background; I am a scientist (Biochemist). The problem of variations in the parameter height will be eliminated by my proposal.

"I hold the blade at its grip"
This introduces many more unknown variables.

Theoretically yes, in reality no. If I perform independent measurements with the same object at different days the RF is about the same (delta about 10 Hz).

"The noise is recorded by a microphone"
I would suspect that the majority of the sound picked up by the microphone will come from the ball rather than the bat.

This is definetely wrong. The same ball, different blades, different RF values.

However discussions about the RF are of lower interest as I want to switch to the interpretation of the ball bouncing which is obviously THE true speed parameter.


But I am afraid that I will have to make a print of the record and make an evaluation by a ruler and ruled paper. Some kind of stone age anachronism in times of PC.

[steve]

I need the time intervall between two spikes plotted vs time.

No need to resort to printing and measuring with a ruler.

Select the audio from one spike to the next, then read the length of the selection from the Selection Toolbar at the bottom of the main Audacity window. In the example below, the selection length (hence the time from one spike to the next) is 0.218 seconds.

Use Audacity's Zoom features to make the selection accurately: http://manual.audacityteam.org/man/zooming.html

If you need greater precission than milliseconds, set the time units in the Selection Toolbar to "samples", then divide the number of samples by the project sample rate (shown at the left end of the Selection Toolbar): http://manual.audacityteam.org/man/sele ... olbar.html

window000.png (46.22 KiB) Viewed 1688 times

[Cogito]

Great ! Thanks. That helps a lot.

[steve]

I would suspect that the majority of the sound picked up by the microphone will come from the ball rather than the bat.

This is definetely wrong. The same ball, different blades, different RF values.

Striking a gong with a hard mallet produces a very different frequency spectrum from striking it with a soft mallet. Same gong, different mallets, different frequency spectra, but that does not mean that the sound is coming from the mallet.

[steve]

Great ! Thanks. That helps a lot.

If you have a lot of these to do, it would be possible to use Nyquist to measure between the peaks. It would of course take time and effort to write an appropriate Nyquist script, so it's only really worth the effort if you have a lot of recordings to measure, otherwise it'll be quicker to just measure them as described in my previous post.