Analyzing a Sound File

[krwlz101]

I am very new to any sound and vibration analysis, so this is what I was trying to accomplish.  I recorded one of our products during run and saved to a wav file.  I then loaded the wav file into my VI and processed using the FFT Power Spectrum vi.  The location of my peaks makes sense given the what I know about this unit.  My main question is what are the units of amplitude?  Are they correct, or did I mess something up?

Attached is the program and the sound file.

[krwlz101]

I guess to somewhat answer my own question, I read the entire help section and understand that the amplitude is in (units)rms^2

More specifically, what are the units generally associated with a wav file?  Or do they vary?

[johnsold]

The units will depend on the devices used to record the data.  Did you use an accelerometer and signal conditioner or a microphone and preamplifier?  Do you have the sensitivity and the gain settings for all those devices?  With that information you should be able to calculate the appropriate calibration factors.

Lynn 

[krwlz101]

I am using a microphone and preamp/MIDI interface.  The products used are brand new, and the links are below.  Unfortunately while I do have a gain adjustment, there are no readouts for a setting value.

I guess ultimately for this expiriment I only need to know relative amplitudes, so what I have is fine, but this is not the end of my testing, so for the next application, I may need to calculate that correction.

http://www.shure.com/americas/products/microphones/ksm/ksm32-embossed-single-diaphragm-microphone

http://www.m-audio.com/products/en_us/FastTrack.html

[Coq_Rouge]

I guess then using a microphone and a sound card, the unit will be phon. But all scaling information is lost on the way. But you can say things like frequency x, is z dB larger than frequency y.

Here is a tool that I made for long time ago in Labview 6i i think. You can use this to toy with different scale settings

[johnsold]

Ultimately you will need to perform a calibration.  Calibrating sound levels is not trivial.  Do you have access to a calibrated sound source? 

You could put a known voltage from a signal generator into the preamp and calibrate the amplifier/recorder parts of the system in a relatively straightforward way, but that would still leave the microphone out of the process.  From the datasheet you linked the microphone sensitivity is only given as a "typical" specification with no indication of how much variability there might be from one unit to another.

Lynn 

[krwlz101]

If I were to integrate the entire frequency spectrum, I would get the overall dB level, correct?  (This is assuming my scaling is right).  I have an independant sound meter that just does dB.  I'm wondering if I can get my scaling by matching that value to the integration of the spectrum*x.

[johnsold]

That will only work if the bandwidths and noise floors of both instruments are the same.  For one thing the total integrated noise will definitely be different. The other is the integrating over the bandwidth introduces a bandwidth factor into the calculation.

If you excite both instruments with a signal which is many dB above the noise floors and there are no other significant signals in the spectrum, you may be able to determine the correlation between the instruments.  Start by measuring the noise floors of both instruments.  Then measure a single-frequency signal which is 40-50 dB higher than the larger of the noise floors (so long as this is still within the maximum permissible levels for both instruments). Repeat at several frequencies and several amplitudes.  Compare results.  If the signals you need to measure are within 10-20 dB of the noise floor of either instrument, you may need to make some careful measurements at those amplitudes to see how the noise influences the readings.  Similarly if you will be dealing with complex sounds with a large dynamic range, you may need to do some intermodulation tests or other tests to verify performance under those conditions.

Lynn 

[Coq_Rouge]

To do correct sound measurements, you need to apply weighting of the signal. This weighting is adapted to the human ear response curve. The most common used is A-weightning http://en.wikipedia.org/wiki/A-weighting You will find all you need in the "sound and vibration" toolkit. But you will need to calibrate your system with some professional equipment. Also using a sound card may be good for a in-house application. But I would NOT recommend this for any commercial application. As it do not sexist any standards then it comes to sound cards or mircophones. Besides it will be very hard to duplicate the sound setting on your test computer to other computers.

[krwlz101]

So all of that makes sense, and I am familliar with the weighting curves for our ears.  I'll have to do some thinking about how I want to approach similar measurements in the future.  As for now, I am satisfied with knowing the "relative" amplitudes at each frequecy.  If I need overall sound (which is how this product gets rated anyhow) I will simply use the sound meter I have to get the A weighted overall dB level.

I unfortunately do not have the sound and vibration toolkit, and we don't do huge amounts of sound analysis, so I doubt that it will ever be justified here.  I'll just have to do the best I can with a Full Labview version.

Message Edited by krwlz101 on 05-14-2010 08:26 AM