I am new to Audacity and using it, I believe, for something “completely different”. We measurement vibrations using a commercially available instrument which saves four parallel measurements in pcm .wav format. We wanted to use Matlab to analyse this directly (fft analysis, but for some reason, Matlab does not read this, at least, it doesn’t like four parallel reading. It was suggested that we import it into Audacity, separate the four tracks, and re-export it as four separate .wav files. That worked, seeing that Matlab does read these and analyse them. We seem to get reasonable frequency spectra. However, the absolute values of the vibration levels which come out of Audacity and are produced by Matlab do not match those which go in, between a factor of 100 and 1000 lower. Audacity itself seems to be artifically scaling the measurements to fit its -1 to 1 scale.
Would you please help me in answering or finding the answers to the following questions as to what Audacity does with .wav files:
- I assume that the vibration meter is creating a .wav file with vibration level values given in millivolts. How does Audacity read these values?
- And then, does Audacity convert/scale them somehow, or put another way, what does Audacity put into the .wav file that I export?
Audacity does not automatically scale the waveform (though you can easily do so with the Amplify effect).
Audio data is just a series of “sample values” (if you zoom in really close on the waveform of an Audacity audio track you will see the individual “dots” (samples).
The absolute value range depends on the format. Signed 16 bit integer format has a range of 65536 (+32767 to -32768). Signed 24 bit integer has a range of 16777216 (+8388607 to -8388608). In each case, for audio signals, this is treated as a range of 0 dB. For floating point audio, 0 dB is defined as +/- 1.0. This is not something that Audacity is “doing” to the data, it is just the standard way that digital audio signals are defined.
The correlation between millivolts and sample values has nothing to do with Audacity - that depends entirely on how your measuring equipment is calibrated.
The “Sample Data Export” tool in Audacity may help to give you a better understanding of how digital data is handled. http://manual.audacityteam.org/o/man/sample_data_export.html
There is no direct association between volts and dB unless you use the broadcast definition of 0dBmv = 0.224vrms or a definition like that. Another association is Sound Pressure dBSPL, but a scientific instrument could be anything, particularly since vibrations, like earthquakes, do not have to be audible.
Audacity is famous for responding to “sound” all the way down to battery voltage, so if the digital signal arrives with what looks like sound data, Audacity will try to display and manage it. Audacity does not try to scale or manage levels. It’s perfectly possible to put a badly overloaded sound signal into Audacity and it will still be overloaded when it gets here.
Four parallel signals in WAV format, while standard, is not normal. You should probably try to find how the equipment folds four signals into one PCM file. There is also more than one WAV format. Microsoft PCM is the common one.
Hi Steve and Kozikowski, Thanks for your quick responses. I’ll have a look at your suggestions. Much to learn, but this gives me a good starting point.