Vibration Analysis from WAVE file

[gears4life]

Does anyone know of an article or instuctions on setting up vibration analysis using DAT recorder, PCB amp, and digital tach? I have seen this done and would like more information about it. Thanks,

 

[GregLocock]

have a look in the FAQ for some links to free on-line material. But it sounds to me that a chat with your instrumentation supplier might be in order.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[cbrn]

Hi,
the way you prospect is perfectly feasible, BUT pay attention NOT to store the waveform as a ".wav" file, because this kind of file is formatted and so adds "something" which is not signal and would disturb or make impossible your subsequent processing. Use raw pcm format instead; you will have to tell every device the "interpretation key" of the file: sample frequency (with a DAT, 44.1 or 48 kHz), sample width (16-bits in the example) and "word" interpretation (Most-Significative Bit / Least-Significative Bit).
Regards

 

[Rob45]

cbrn is right.
The various .wav formats all compress audio files to some extent, and if the parts they leave out are the parts you're interested in, your time and effort is wasted.

I've shown this by converting from Head Acoustics' ".dat" format to .wav and back, and there are appreciable changes.
Same thing can be seen when I record to both .wav and .dat simultaneously.

 

[cbrn]

Hi,
Rob45, ".wav" file is not a compressed one, it's 100% lossless and would exactly be identical to Red-Book PCM... IF it didn't add a header and a footer. In particular, the header contains the description of the interpretation keys (the header can NOT be avoided, it's mandatory in the .wav specification), and the footer (optional) contains additional non-audio data such as comments, etc (something similar to ".mp4" of I-Tunes, where the footer is used, for example, for the booklet image).
OK, this is pure academic... But, in fact, after writing my first reply, I realized that some audio processing programs can automatically get rid of the "extra-formatting" of the ".wav" files, thus loading only the "signal part" as the working waveform. As far as I know, CoolEdit2000 is one of these programs capable of doing it (it has also FFT analysis, btw).
Another note: if you record on a DAT born for the audio world, then you must be aware that a vibration signal will be treated as audio as well: differently from a "pure-signal" recorder, such as an off-line FFT analyzer for example, the signal will be formatted (integrity check with Reed-Solomon code, and partial redundancy). All that, independently from the facts related to sample freq and bit depth (16-bits at 48 kHz are generally far beyond what needed in vib analysis). So, you probably won't be allowed to process the bit-string directly with Matlab, for example, and you will have to rely upon professional programs for the AUDIO world.

Regards

 

[SomptingGuy]

Further to the above, it's quite common for devices to use the (inaudible) LSB of each channel to store encoded information. We use it to store a tach signal in audio files of predicted intake/exhaust noise. Extracting the LSB is fairly simple in Matlab:

lsb=bitand(data*32768+32768,1);

 

[cincibcats]

All of the analyzers and data acquisition systems made by Oros use .wav files as the primary storage format.

 

[GregLocock]

Check out the definition of the wav format. It is a scaled integer representation of the waveform at constant sampling interval. No compression there, it must occur elsewhere in the system.


Be very wary of Minidisc recorders and the like. I was astonished to read of a reputable architectural acoustics guy saying he used them for data acquisition. They are heavily compressed, rather like MP3s.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[cbrn]

Hi,
so, just to summarize a little:
- WAV is a Pulse-Code-Modulation, lossless, uncompressed scheme, with a proprietary format. The data part is identical to a "pure-signal" PCM bit-string
- therefore, it is an adequate format to memorize acoustic as well of vibration data
- problems can only arise if the analysis device is not aware to cut off the header and footer part of the file, and does not know anything about bit-string decoding (RSC, redundancy,...). As Cincibcats says, there is AT LEAST one brand of analysis equipment which can handle WAV format directly.
- DAT recorders, DASH recorders, Hard-Disk recorders, are all good candidates to record a vib signal for subsequent processing, because they work in UNCOMPRESSED format. OK, to be honest, HDR can record in any format... DAT recorders have super-compactness, DASH have extraordinary mechanical / magnetic properties (but they have very huge dimensions, making them improbable for field measurements!), HDR has incredible storage capacities BUT you must appropriately select the "audio" board (it must have a "pass-through", otherwise the circuitry will interprete the signal and irremediably "corrupt" it).

So, you see, everything is feasible, with only some "caveats" (and, as GregLocock says, NEVER NEVER NEVER use mp3 recorders, DCC recorders or MinidDisc recorders).

Regards

 

[Rob45]

cbrn;
.WAV format may not be "Compressed" in the usual sense, but it is certainly different from a signal recorded in an untampered format such as Head Acoustics and LMS use in their data acquisition systems.

You only need to record a signal in the two formats simultaneously and you'll find attenuation at the lowest and highest frequencies in .WAV format that is not present in either of the other two mentioned.

Not being either an audiophile nor an audio engineer, I couldn't tell you if that is "compression" or not, but it certainly a loss of fidelity.

 

[SylvestreW]

Well, for not-so-important projects we have a M-Audio microtrack unit (that records .wav) and some PCB Piezo. accelerometers (and assoc. hardware)

We save the wave form on the card, downloads to mathcad (or matlab), quick fft and voila, we have what we want.

Now, this works for non-critical/non-reportable since we are interested in the frequencies, not the exact levels. We have calibrated the unit in the past to get approx. levels but by no means is this accurate (or done usually).

quick check of m-audio's websites states:
PCM recording: 16 or 24-bit at 44.1, 48, 88.2 or 96kHz
1/4" Mic/Line Inputs [...] Frequency Response: 20Hz to 20kHz, +/- 0.3dB @ 48kHz sample rate

(Note, the rest of the time, for critical info, we do have some GRAS RTA equipment)

 

[GregLocock]

Just did some reading. Apparently you /can/ store compressed signals in a wav file. So, sorry Rob, I was being too emphatic. wiki:

"Though a WAV file can hold compressed audio, the most common WAV format contains uncompressed audio in the pulse-code modulation (PCM) format. PCM audio is the standard audio file format for CDs at 44,100 samples per second, 16 bits per sample. Since PCM uses an uncompressed, lossless storage method, which keeps all the samples of an audio track, professional users or audio experts may use the WAV format for maximum audio quality. WAV audio can also be edited and manipulated with relative ease using software."



Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[cincibcats]

Rob45, I would suspect that the attenuation of the signals you're seeing is due to differences in anti-aliasing and AC-coupling filters rather than the WAV format. A less expensive data acquistion system may have filters with ripple or other effects that effect the low and high frequency content. I don't think someone building a dynamic signal analyzer would select a file format that would provide any type of attenuation without expressly telling people about it.

 

[cbrn]

Hi,
Rob45, wav file is by NO means compressed and is DEFINITELY lossless.
It is simply a raw PCM bitstream (generally having 16-bits depth and 44.1 kHz sample rate, but it's not mandatory), MSB, with an added header where info is coded about the format, so that reading device (PC running MS O.S., when ".wav" was born) knows how to "interprete" the bitstream (i.e. know where to put the "beginning" and the "end" of each "word") without asking the user each time.
Saying that a wav file causes loss of frequencies is like saying that RAW PCM is loosing data (but then, no digital data transmission would be possible at all in the world...). You'd better investigate on HOW you recorded your two bitstreams...
B.t.w.: I once did the comparison test you say between waveform "as recorded" from a DAT (and transfered as RAW PCM) and "as saved" in WAV format: it was an audio file of 2h57mn duration, and the comparison showed no differences all along the bitstream.

Regards