Measuring natural frequency with a spectrum analyzer

[highwayman]

I have a set-up that I am trying to upgrade.  I have a frequency response test set up consisting of a HP 3582A spectrum analyzer with two inputs.  One input is the hammer that initiates the natural frequency.  The other is the part that I am measuring.  Both the hammer and the part are equipped with accelerometers and sensor signal conditioners between the source and the SA.

Most of my frequencies are below 1kHz, so frequency should not be an issue unless the SA can't go too low.  What other features would I look for to replace my aging SA?  With the right noise filtering, can I use a PC sound card with 2 inputs and some neat software?

Thanks in advance for any suggestions.
Thad

[_Wim_]

I have done similar things with a sound card (impulse device was ball from fixed height, so only DUT had accelerometer), and that worked ok. I used ARTA (http://www.artalabs.hr/) software, but there are many other packages available also. You can test ARTA for free, the only limitation is you cannot load and save files.

[DaJMasta]

Sound cards (internal or external) are a good, high resolution way to do kHz range measurements with modern equipment.  There are also dynamic signal analyzers which operate in that ballpark, typically up to 100-200kHz.  Audio analyzers will also generally do this and usually offer at least some frequency response outside of normal audio bands.  There are lots of threads on each!

[cdev]

'natural frequency' of oscillation - as in resonant frequency of a quartz crystal, tuning fork, or other physical or electrically resonant object (organ pipe, plucked string, resonant bell, tank circuit?) set to oscillation?

I gather thats probably what you mean? Seems like yes, you could likely use a computer sound card, both its inputs and outputs. Quartz crystal oscillators do something exactly like this, its not a mechanical analogy to electronic oscillators, they are mechanical.

[_Wim_]

ARTA has a function were it can record an impulse based on an external excitation (which is ideal for this sort of test). However I do not think it is possible to record two channels at the same time.

[cdev]

Its extremely interesting to apply an audio signal generator to a perfectly level flat metal plate upon which you have poured some salt or other uniform powders, while slowly sweeping the frequency up and down through the audio band.

A speaker works well as the exciting mechanical transducer.

Although lots of videos of it are online, the phenomenon which occurs really has to be seen to be fully appreciated.

You can find a lot of examples under 'cymatics', the title of a long out of print book on this.



Here is another.

This never fails to amaze me..

[_Wim_]

Its extremely interesting to apply an audio signal generator to a perfectly level flat metal plate upon which you have poured some salt or other uniform powders, while slowly sweeping the frequency up and down through the audio band.

Noted on my TODO list when the kids are a little bigger so they can fully appreciate this!

[rhb]

The use of Chladni patterns is common practice for many luthiers (stringed instrument makers) when adjusting the thickness of the guitar top and back.  Al Carruth has probably written the most about it.

https://en.wikipedia.org/wiki/Ernst_Chladni

For impulse response testing such as you are doing a sound card and Octave are ideal.  You can calibrate the soundcard using the following procedure.

create a WAV file with a slow sweep over the full range of the sound card

play that back through the line output of each channel with the output connected to the line input to each input and record the inputs (getting the sound card to play and record is likely to be the most difficult task)

compare the WAV file you fed the output to the WAV file recorded from the inputs

A little bit of algebra and you can correct to the resolution of the WAV file for any errors in the sound card input  and output circuitry

If any of this is not immediately obvious from the above I shall be pleased to write a detailed explanation.  This basic question comes up in various forms frequently.

[_Wim_]

You can calibrate the soundcard using the following procedure.

create a WAV file with a slow sweep over the full range of the sound card

play that back through the line output of each channel with the output connected to the line input to each input and record the inputs (getting the sound card to play and record is likely to be the most difficult task)

compare the WAV file you fed the output to the WAV file recorded from the inputs

A little bit of algebra and you can correct to the resolution of the WAV file for any errors in the sound card input  and output circuitry

I think this works only if both the excitation and the measurement are performed by the sound card (for example loudspeaker response measurement). If only a measurement is performed (like in the case of the OP), this will not work, because it is not possible to know if the deviations are coming from the output or the input.

In ARTA this is done by using the second stereo channel in loopback during every measurement. This has the advantage that also time delays caused by the drivers are calibrated out, allowing you for example to precisely time-align drivers in an active speaker system (as long as you keep the micro-phone in the exact same position during the multiple measurements)

But if you see the excellent behavior of modern soundcards, I do not think calibration is necessary for what the OP wants to do, because I expect the variations between each measurement to be much bigger that the deviations imposed by the sound card.