 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
Natural Science Forum / Physics / Acoustics / December 2005spectral analysis of wave files
I have a situation where I am going to have to rely on wave files recorded on a laptop computer to get spectral data of noise in a power plant. Don't scoff - you work with what you have! (My informant will also make simultaneous A-weighted SPL with an accurate but simple sound level meter). Does anyone have any practical experience or advice on pulling this off? I am interested in third-octave spectra up to at least 5kHz. How can I find out the "sensitivity" of the recording sound card? How do the numeric values in the wave files relate to signal into the sound card? Does it vary between manufacturers? Sample rate of 11.2 kHz OK? I'll bet the signals will be aliased - does anyone know for sure? How about analysis - I have a third-party Matlab script which purports to emulate an octave filter bank. Does anyone know of a better alternative? I figure my one hope of success is to scale whatever spectrum I extract to match the separate overall dBA. What other pitfalls have I overlooked? Please, no flames. I know perfectly well that this is a half-assed way to go, but the alternative is for me to carry my B&K2260 on a hundred-mile trip by dogsled, and Quebec Hydro can't get me certified for the plant in time anyway. I'll appreciate any comments. Robert A. Hedeen GE Global Research Niskayuna, NY Maybe use a demo of winMLS or spectralab to evaluate the spectral components of the sound and compare the A-weighted levels with the SLM measurements? You can than compensate for the differences. bert > Maybe use a demo of winMLS or spectralab to evaluate the spectral > components of the sound and compare the A-weighted levels with the SLM > measurements? You can than compensate for the differences. > > bert Be careful with Spectralab. Admittedly, it looks VERY impressive, but as the word goes, looks can be deceiving: 1. Take an arbitrary signal, e.g. traffic noise, of say 1 minute length, set peak hold on (shouldn't it be 'max.' hold), select all, "compute and display average spectrum" in octave mode or whatever, "save as text" in spectrum view and paste into Excel. Then do the same with segments of that signal and compare levels band by band. SURPRISE: some segment levels are HIGHER than those of the full signal! - In layman's terms: If I look at a mountain range and then at a segment of that same range, then none of the mountains in the segment can be higher than the highest of the whole ange. - or, has Archimedes, Pythagoras, Sir Isaac Newton, Baron Jean Baptiste Joseph Fourier, Albert Einstein and I missed out on something which the creators of Spectralab did not overlook? Then I took my good old B&K 2230 sound level meter with the 1625 1/3 - 1 octave filter (calibrated by an accredited lab within the last 3 months), substituted the microphone with a JJ2614 adapter and fed that same signal through - the results were dramatically different! 2. I once had to deal with a "sound engineer" from the amusement branch who had some funny instrumentation and who came up with levels some 20 dB apart from what I measured. His explanation was that he uses a different basis for sound levels :-). Spectralab also seem to use different basis, for their Leq: valid only for signal duration of 1 hour. That sounds to me more like SEL (exposure). But checking that out, it isn't either! Then you would assume that for 1/2 hour the result is 3 dB less - isn't either! 3. After drawing their attention to the above Spectralab stopped abruptly communicating with me. :-( QUESTION: This software has a not exactly low price tag attached - is that allowed? It is certainly not exactly ethical! Klaus ><<<<<<... > [quoted text clipped - 5 lines] > >Klaus Thanks for the info, is there a frequency range where that effect is more prevalent, or does it go from 0 to 20000? I only worked from 1Hz to 200 Hz so far :-) (They *are* expensive) w. Interesting, Klaus. I use spectralab sometimes as a kind of back up, as it can record sound. Comes in handy when you have to check Lmax from fe traffic noise as you can later hear wether the peak is a truck or a bird sitting on the microphone :-).Use it with a B&K 4007 mic. I compare with a certified 2260 or Svan 945 and thought they correlate rather well. But I never went through the labour you did. Over here it's not allowed to do official measurements with anything but type I gear. > Interesting, Klaus. > I use spectralab sometimes as a kind of back up, as it can record [quoted text clipped - 6 lines] > Over here it's not allowed to do official measurements with anything > but type I gear. B&K 2230 is also type 1. Although the notebook is not certified type 1 gear, it does record the AC out signal from the 2230 accurately. One thought I had and I implemented: if you bring an analog signal into the computer with all kinds of fields, electric and magnetic, floating around, distortions and noise are possible. So, I use an external USB sound card which is far enough away from the computer and the signal from sound card to computer is then digital and nothing can distort that. Klaus Could you please be a little more explicit about your sound card and pc software. I also have a BK 2230 and several times i was tested the system with a NI DAQ A/D PCI card and Labview software in an office PC, it works ok. But i want do the same job with a portable PC, i would need a PCMCIA DAQ card which is more expensive and weak for field work. I would like to listen a little bit more about your system. Thanks > > Interesting, Klaus. > > I use spectralab sometimes as a kind of back up, as it can record [quoted text clipped - 18 lines] > > Klaus [snip] >Does anyone have any practical experience or advice on pulling this >off? Not with a sound card, but we routinely process digital data (from A/D systems and DAT) into WAV files for analysis. > I am interested in third-octave spectra up to at least 5kHz. How >can I find out the "sensitivity" of the recording sound card? How do >the numeric values in the wave files relate to signal into the sound >card? Does it vary between manufacturers? Sample rate of 11.2 kHz OK? Yes, it varies. Watch out for cards that compress or autolevel. There are quantitative acoustic systems that are built around sound cards. I can't name any, but I know they've been discussed here. Hopefully someone will pop in with a brand name. 11.2 kHz isn't really OK. You'd like to go at least 2.56 times the upper edge of the 5k octave band. The numeric values in the WAV file are just a digital data stream. To relate them to levels, put a known signal in and see what comes out. I'd recommend recording pink noise at various known levels, then processing via your Matlab script to see what comes out. You'll calibrate for level and test frequency response at the same time. >I'll bet the signals will be aliased - does anyone know for sure? How >about analysis - I have a third-party Matlab script which purports to >emulate an octave filter bank. Does anyone know of a better >alternative? I figure my one hope of success is to scale whatever >spectrum I extract to match the separate overall dBA. What other >pitfalls have I overlooked? Unless you've got anti-aliasing filters, the signals will be aliased. The alternative is to apply whatever low pass filtering is available (what kind of mic and amp system are you using?) and stepping up the digitization rate. I suspect that most people doing this kind of analysis use Matlab, unless they have NI gear in which case they'll use Labview. I wrote my stuff in Fortran. (I think I just gave away my age.) Make sure that your script is able to deal with non-instrument digitization rates, and give you standard octave band output. >Please, no flames. I know perfectly well that this is a half-assed way >to go, but the alternative is for me to carry my B&K2260 on a >hundred-mile trip by dogsled, and Quebec Hydro can't get me certified >for the plant in time anyway. Have you considered taking a small DAT recorder, and processing the recordings when you get back home? Or record in parallel with the sound card setup so you've got a more tracable backup just in case? Walkman-type DATs are pretty small, and we've used those very successfully for single channel recordings. Ken Plotkin > [snip] >>Does anyone have any practical experience or advice on pulling this [quoted text clipped - 34 lines] > (what kind of mic and amp system are you using?) and stepping up the > digitization rate. I think the mic input is likely to be antialiased, try generating a rising tone to 10 kHz, you'll quickly find out by listening to the playback. Aliased signals sound so odd that you'll pick it in an instant. To be honest you don't really need to worry too much about calibration - get the SLM and the laptop mic alongside in some steady state condition and after the event fiddle with the calibration until the overall dBA pretty much agree. After all, the spectra will only be used for diagnostics, won't they? Have a look at WinScope, which is a very simple scope for windows, useful for checking input levels, and Goldwave, a program of endless complexity. On 3 PCs I have yet to see autolevelling implemented at the hardware level. Cheers Greg Locock >I have a situation where I am going to have to rely on wave files >recorded on a laptop computer to get spectral data of noise in a power [quoted text clipped - 4 lines] >Does anyone have any practical experience or advice on pulling this >off? http://www.soundtechnology.com/download-center.htm works perfectly. Demo runs 30 days. (Hint: they offer different versions, each one has a demo) You may have to calibrate, all you need is a sound generator and an oscilloscope, for the PC part, but you also need a measurement microphone with flat response. For extremely low frequencies < 2 Hz you must modifiy the sound card input, eliminate the coupling capacitor and add a DC level shifter circuit, your lowest frequency then is ZERO (DC) Works with most soundcards, but have not done it on a laptop.... w. >I have a situation where I am going to have to rely on wave files >recorded on a laptop computer to get spectral data of noise in a power [quoted text clipped - 4 lines] >Does anyone have any practical experience or advice on pulling this >off? http://www.soundtechnology.com/download-center.htm Demo runs 30 days. Before they offered different versions, each one had a demo You may have to calibrate, I only used it for relative measurements, comparing before and after, and only from 5 to 200 Hz. For extremely low frequencies < 5 Hz you must modifiy the sound card input, eliminate the coupling capacitor and add a DC level shifter circuit, your lowest frequency then is ZERO (DC) There are several places on WWW with descriptions how to do it, but I never actually tried. Supposed to work with most soundcards, but do not know about laptops. Meanwhile they offer USB gadgets, "external" sound cards, with special features. w. >I have a situation where I am going to have to rely on wave files > recorded on a laptop computer to get spectral data of noise in a power [quoted text clipped - 24 lines] > GE Global Research > Niskayuna, NY For analysis you could try Realtime Analyzer, DSSF3 Light is the version I think you'd find is best for this application, and at $83 it is *very* much cheaper than SpectraLab Go to http://www.ymec.com/products/dssf3e/outline.htm#list for details, but be warned the site is very confusing. There's also a 30 day demo version. I use this software for teaching and demonstration purposes, but I have no idea exactly how accurate it is. Certainly the analysis it gives of wave files recorded via the ac output of my Svan 945 correlates very well with the analysis of the original signal. It can be calibrated to a test tone so in theory it should be possible to record a test tone at the start of the wave file and use this as the calibration reference. For laptop recording purposes I use a Creative MP3+ external usb sound card, which seems to give good results and has the advantages of being much better than the integral sound card on the laptop and fairly low cost. G > For analysis you could try Realtime Analyzer, DSSF3 Light is the version I > think you'd find is best for this application, and at $83 it is *very* much > cheaper than SpectraLab Go to > http://www.ymec.com/products/dssf3e/outline.htm#list for details, but be > warned the site is very confusing. It's not all that confusing. with little patience, I zerroed in on the medical acoustics topic of listening to heart sounds. i recommend that everone venture here for a good update on 21st century acoustical methods and analysis. > There's also a 30 day demo version. I use > this software for teaching and demonstration purposes, but I have no idea > exactly how accurate it is. I think that in this case (new methods) accuracy comes later. The most important aaspect is utilization of technology, especially the PC and LapTop. > Certainly the analysis it gives of wave files > recorded via the ac output of my Svan 945 correlates very well with the > analysis of the original signal. It can be calibrated to a test tone so in > theory it should be possible to record a test tone at the start of the wave > file and use this as the calibration reference. For laptop recording > purposes I use a Creative MP3+ external usb sound card, There's a good hint. Where are those external cards available? Any hints on selecting the requisite quality? > which seems to give > good results and has the advantages of being much better than the integral > sound card on the laptop and fairly low cost. My TP internal card plays stereo, but is only mono-input. Keep us posted. I am downloading the Y MEC demo now. Angelo Campanella > For analysis you could try Realtime Analyzer, DSSF3 Light is the version I > think you'd find is best for this application, and at $83 it is *very* much > cheaper than SpectraLab Go to > http://www.ymec.com/products/dssf3e/outline.htm#list for details, but be > warned the site is very confusing. It's not all that confusing. with little patience, I zerroed in on the medical acoustics topic of listening to heart sounds. i recommend that everyone venture here for a good update on 21st century acoustical methods and analysis. > There's also a 30 day demo version. I use > this software for teaching and demonstration purposes, but I have no idea > exactly how accurate it is. I think that in this case (new methods) accuracy comes later. The most important aspect is utilization of technology, especially the PC and LapTop. > Certainly the analysis it gives of wave files > recorded via the ac output of my Svan 945 correlates very well with the > analysis of the original signal. It can be calibrated to a test tone so in > theory it should be possible to record a test tone at the start of the wave > file and use this as the calibration reference. For laptop recording > purposes I use a Creative MP3+ external usb sound card, There's a good hint. Where are those external cards available? Any hints on selecting the requisite quality? > which seems to give > good results and has the advantages of being much better than the integral > sound card on the laptop and fairly low cost. My TP internal card plays stereo, but is only mono input. Keep us posted. I am downloading the YMEC demo now. Angelo Campanella >> http://www.ymec.com/products/dssf3e/outline.htm#list > I am downloading the YMEC demo now. WOW! Reminds me of my acoustics lab days. Sig. gen, works tight ot of the bos. same for scope, and RTA, and internal mic, etc.... give it a test drive! > Angelo Campanella >>> http://www.ymec.com/products/dssf3e/outline.htm#list >> I am downloading the YMEC demo now. [quoted text clipped - 8 lines] > >> Angelo Campanella Glad you like it Ang, I'm always a bit wary about posting details of software I have found useful as there's always someone who finds fault with it! The external usb sound card I made reference to, the Creative MP3+, doesn't seem to be produced any more and I bought mine from an Ebay seller. I have found the whole system to be ideal for teaching and training purposes, where absolute accuracy is not essential. One of the tasks I regularly undertake is training in noise induced hearing loss awareness and the use of a laptop, a few software packages and a Peavey Messenger portable pa system (about the size of a small suitcase) has replaced a small van full of equipment that took two people about an hour to assemble. Such is the march of progress! G > I have a situation where I am going to have to rely on wave files > recorded on a laptop computer to get spectral data of noise in a power [quoted text clipped - 24 lines] > GE Global Research > Niskayuna, NY Here In Alberta I carry one of my 2260s everywhere. They are a workhorse. I am also sure if you go to National Car Rental they can upgrade you from a dogsled. I use a 2260 Investigator or an Observer with a IBM notebook sporting a 24bit sound card. I tend to log 1 minute Leq in both A & C for environmental surveys. Using B&K's 7820 Evaluator program I can get a time history graph with my chosen parameters as well as a third octave spectrum (6.3-20KHz with the investigator) for every minute logged. The Evaluator program manages the wave files recorded and a click on the time referenced graph will result in CD quality digitally recorded sounds being magically brought to life. Sounds pretty cool, eh? Quanta costa? > Quanta costa? The laptops were purchased used for about $600.00 each. The sound card, I use both the Creative Audigy2 ZS & the Indigo IO 24bit, were about $300.00 each. The Bruel & Kjaer 2260s run anywhere from $10,000-$25,000 depending on purchased applications such as low frequency, sound intensity, FFT, building acoustics, etc. The B&K 7820 Evaluator program was if I remember around $5000.00. Plus cables, all weather B&K outdoor microphone kits...All in Canadian dollars. Probably less expensive for one to just hire me! You have to realize that all this equipment was gathered over the last 8 years at a decent sized acoustical engineering firm. The cost for us was spread over many many projects. I used to use B&K 2236 & 2238 SLMs and recorded sound output to Hi-Fi VCRs. If needed the tapes, although not engineering grade, could be run through narrow band analysis. Peter Davis Faszer Farquharson & Associates Ltd. Suite 304, 605-1 Street SW Calgary, Alberta Canada T2P 3S9 ffa@telusplanet.net |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2009 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.