An Acoustic Primer
Contents
1. An Acoustic Primer Sharon Nieukirk and Anne Terhu Oregon State University What is sound Sound is a wave More specifically a longitudinal wave gt The particles in a longitudinal wave move parallel to the direction in which the wave is traveling gt The particles in a transverse wave move perpendicular to the direction in which the wave is traveling e g La Ola wave D D eS www bioacoustics us Sound wave The medium through which the wave travels may experience some local oscillations as the wave passes but the particles in the medium do NOT travel with the wave Image source http www kettering edu physics drussell demos html WWW DIOACOUSHCS US Frequency Hz Wavelength m Amplitude uPa Low Amplitude Low Frequency https www khanacademy org science physics mechanical waves and sound sound topic v sound properties amplitude period frequency wavelength Frequency Porpoises Dolphins Baleen whales Earthquakes Seals amp sea lions 3 Infrasound lt 20 iz i ri id Ultrasound gt 20kHz 1 Hz 10 Hz 100 Hz 1 kHz 10 kHz 100 kHz Frequency www bioacoustics us N I zn o c o 2 C m amplitude 5 10 15 C Users Holger Desktop test Lichtenberg_files sound011 wav 4096 0x 0 5 Hamming Sound in air water Sound in air and sound in water are both waves that move similarly and can be characterized the same2. us Characterizing sounds dB re 20 uPa WATER dB re 1 uPa dB in air and water is not the same UW volcano Caused by different shotgun sonar a reference pressures jet aircraft tanker In water 1 pPa tock concert i baleen in air 20 uPa whales vacuum b densities and sound speeds convetsation spinner Intensity measurements of equal whisper dolphin pulse pressures in air and water differ by 62 dB human bubbles amp hearing spray calm threshold 61 5 dB recreated from DOSITS org Don t compare apples and oranges An example source level SL Gray whale moans 142 185 dB re 1 uPa Q 1 m gt In air 80 123 dB re 20 uPa 1 m In compatison Lawn mower 90 dB re 20 uPa 1 m Auto horn 110 dB re 20 uPa 1 m ALWAYS provide information on the reference pressure Similar problem degree temperature F or C Source http cetus ucsd edu voicesinthesea_org Flash www bioacoustics us Don t compare apples and oranges Action Items S at the Center for Biological Diversitv v Scroll for the specific mm I rd 3 FYYFYY 3 or the Marine Mammal we must do one or Source http email maildiva com t ViewEmail r CC5B6185852F6292 D54AE5B66E0381F0C68C6A341B5D209E WWW DIOACOUSTICS US FE C A it e Computers and DAT recorders sample digitize the continuous rise and fall of sound amplitudes at some fixed rate and store a lo
3. way but because liquids and gases have different properties such as density they also have different sound speeds gt Sound travels 1500 m S in seawater gt 15 football fields end to end Dvs 340 m s in air gt The speed of sound in seawater is not a constant value It changes with depth temperature and salinity www DIOACOUSHCS US Sound speed in seawater sound speed is the most important environmental parameter that affects sound propagation in the ocean sound Speed Equation c 1449 2 4 67 0 055 0 000296 1 34 0 010t s 35 0 0165z c is sound speed in m s t is temperature in C s is salinity in ppt z is depth in m e g t 20 C s 36 ppt z 10 m gt c 1449 2 4 6x20 0 055x20 0 00029x20 1 34 0 010x20 36 35 0 0165x10 gt c 1522 825 m s www bioacoustics us Decibels The decibel dB is a logarithmic unit that indicates the ratio of a physical quantity relative to a specified or implied reference level gt Sound pressure level SPL 20 log p p p pressure of an acoustic signal uPa p reference pressure in water 1 uPa SPL sound pressure level dB re 1 uPa gt 2x p will result in 6 dB gt 10x p will result in 20 dB gt 100x p will result in 40 dB www DIOACOUSHCS US Don t compare apples and oranges dB in air and water are not the same Source http tsalveta info ebook hearing html www bioacoustics
4. Adapted from Time Bradbury amp Vehrencamp EG ien o Spectrograms and Bamemath Let s decrease At by 4x again This will give us a Af 80 Hz We get much better temporal pattern and even some better frequency pattern because FM signals show as FM not their components Frequency EG ien o Spectrograms and ohne e Let s decrease At by 4x once more This will give us a Af 320 Hz This is similarto the prior bandwidth but we can see the temporal pattern in the last notes better Frequency Time gt EG ien o Spectrograms and Bamemath e Let s decrease At by 4x again This Will give us a Af 1280 Hz Now large bandas start to appear instead of fine lines although the temporal pattern is retained TITTET Wii Frequency gt Adapted Time Bradbury amp Vehrencamp u Spectrograms and Basrewwiatn e Let s decrease At by 4x yet again This will give us a Af 5120 Hz We have now lost any decent frequency resolution but the temporal pattern is retained Frequency B iN HH Time gt ien o Spectrograms and Bamemeoth e An intermediate bandwidth Af provides the optimal balance of frequency resolution and temporal resolution Choose your settings based on your question I Frequency Adapted from a Time Bradbury amp Vehrencamp EG o Spectrograms and Bamemath In general you want a bandwidth v
5. anging during the signal e Break the song into homogeneous segments and create a frequency Spectrum for each segment RM Ww ug LA Time Pressure Bradbury amp Vehrencamp Cornell University Fourier Analysis Applying the Fourier alg Waveform 1 nnn h y fi IN AAMA ANN AMI L TRY UV FAT F M iF yy Amplitude Frequency Frequency A plot of amplitude versus frequency components is called the frequency spectrum or power spectrum of a sound Bradbury amp Vehrencamp Cornell University Time domain Pressure FFT dividing a sound into segments computing the frequency spectrum for each segment stringing the segments together along the time axis E Frequency domain i I Ec x E p Pressure Frequency Adapted from Bradbury amp Vehrencamp id Raw Wu gt Creating a spectrograrai Then use black to mark those portions of the overall graph that have higher peaks use white to mark the lower amplitude components and use grey for intermediate portions Frequency gt Adapted from Bradbury amp Vehrencamp lt id Row Wu gt Creating a spectrogranai e The result is a spectrogram with trequency on the vertical axis time on the horizontal axis and amplitude of a frequency component ata given time indicated by darkness on the plot Frequency gt Adapted from B
6. ill start with this buffered sound How much time before seconds Cancel WWW DIOACOUSHCS US A ue ETE transform conditioning Conditioned spectrogram detection algorithm Detection function threshold Detected calls WWW bioacoustics us Sound wave metrics Frequency f and wavelength c f with c is sound speed in m s and f frequency in Hz a high pressure low pressure www bioacoustics us Frequency Hz Wavelength m peak to peak A a ent D i gt Uu u D i OQ Source http www dosits org science advancedtopics signallevels
7. ize 8132 samples 4 096 sec None v Zero padding DO sec Hop size 144 1 024 sec Cancel Conclusion choose your settings based on your question WWW bioacoustics us Ishmael real Set up sound card input Sound card IFrontMic Realtek High D efiniti To view and record real time data Channels to acquire 17 sr 93r 13r Plug microphone in FIRST All 20 60 101 147 None So 23 Das p Di CURA 4 8 12 16T Launch Ishmael Ishmael hangs up whenever you stop a run Under File menu check soundcard setting Under Record menu choose record only Recording sound Recording File names Sampling Recording when you click the f e d button and SE e cor d If you have sound coming in from the outside world via a sound card or other data acquisition hardware can record the sound in soundfile s as it arrives only when pe tting re Al time inp ut Start recording whenever you click Run Record only when you click the red Record button Don t record v Record only when getting real time input Usually you want this checked If it s not checked then re recording will happen even when am reading sound from a file Pre click buffering When you click the red Record button I will start recording to a sound file If vou lik ing will actually start BEFORE the time you click the button I will al several sec of the incoming sound in memory when you click the orded file w
8. ng column vector of amplitude values Music CDs sample at 44 1 kHz or 44 100 samples s Bradbury amp Vehrencamp Cornell University Sample rate Nyquist nla e Nyquist frequency A digital recorder or computer must be able to take at least 2 samples cyele to be able to identify each frequency e Thus if you digitize your sounds at R samples sec you will be unable to properly capture any component with frequency gt R 2 This latter value is called the Nyquist frequency j EGET GE ESTE ESS GAGE I DO RES T E Ie TA File Edit Run C s r j did o A Y PINE TWIN m Nyquist o DO ur E MI WT E m ue E it d 1000 Hz Ed UL ME wo I LU R M Sf Nn Lut 2000 Hz pm 12 00 140 IC Adapted from Bradbury amp Vehrencamp Cornell University om gt Ishmael a word about peeikopram settings Exercise 3 Launch Ishmael Open a sound file NS04 05 RswyW tfilell0 wav e Change spectrogram settings Compute S bectrogram parameters Spectrogram parameters The basics Advanced These parameters determine how your spectrogram is computed The most important is the frame size longer frame size gives better frequency resolution a shorter frame size gives better time resolution m Frame size samples 0 256 sec Zero padding sec Hop size 4x 0 064 sec www bioacoustics us uo gt Fourier Analysis TT e Waveform keeps ch
9. radbury amp Vehrencamp LE The Uncertain Amplitude fy Pr 1 If we let At be the du sampling time available analyzer the Uncertainty sound analysis states that long window goot j resolution Small AZ large Af Medium A7 medium Af Long A small Af Amplitude Amplitude Frequency a Bradbury amp Vehrencamp Adapted from Frequency Frequency A e The Uncertainty Prince Frequency resolution and time resolution are inversely related Af Arz1 Long window A7 good frequency resolution Short window Az good time resolution Decide on your question and signal of interest then choose your spectrogram settings WWW bioacoustics us Sample rate also matters sample rate limits your spectrogram settings Sample rate window size frequency resolution Example 44 100 samples s 1024 samples 43 Hz res WWW bioacoustics us AG oc a Spectrograms nd Barmen The spectrogram we just made uses pretty large At This gives us very fine frequency resolution Af 5 Hz but much of the temporal resolution has been lost Can we get by with smaller At Frequency dapted from Adap Time Bradbury amp Vehrencamp EG ien o Spectrograms and Bere WB th Let s decrease At by Ax This will give us a Af 20 Hz This starts to restore some of the temporal pattern and the frequency bands are still pretty thin Frequency
10. rogram parameters The basics Advanced Frame size 64 samples 0 032 sec Zero padding None 0 sec Hop size Er 0 008 sec ad IM us N j yt a i 9 M 1 i h Mui i UT m atl 1 N i r I VG il v n r N iy hj i i A i i By Ad ht MW W ar KD i uM yin ij i He VAA T T rud Pipe M PR i pe ts jl Lm da i T lib wily Wn waa lt al iu i Nadia idi n longer fran a shorter fran Frame size Zero padding Hop size None Ishmael 2 3 0 C Users nieukirk Desktop sIntest2 ipf CAUsers sin goodcalls NS05 06_RswyW file110 hump wav File Edit Run Compute View Record Detect Localize Actions Time 70 186 Amp 0138 1593 MB used ks mh i LA dii b i Hy ot lind bd ad ol d 7 hd ii better frequency reso better time resolution samples 0 256 sec 0 sec 0 054 sec Cancel www bioacoustics us Ishmael Exercise 3 Change spectrogram settings there is a limit Ishmael 2 3 0 C Users nieukirk Desktop slntest2 ipf C Users sIn goodcalls NS05 06_RswyW file 40 hump wav File Edit Run Compute View Record Detect Localize Actions Time 26 827 Freq 1 050 159 3 MB used 1 Spectrogram parameters The basics Advanced These parameters determine how your spectrogram is computed The most important is the frame size longer frame size gives better frequency resolution a shorter frame size gives better time resolution Frame s
11. small enough to separate harmonics clearly v big enough to show FM undecomposed and v big enough to show AM undecomposed Frequency dapted from Ada Time Bradbury amp Vehrencamp Spectrogram pafametetsn Short window good time resolution 64 points 2 90 mS 0 04 0 06 0 08 JA 0 04 0 06 0 08 long window good freq resolution 512 points 23 2 mS 0 04 0 06 0 08 Effect of record length and filter bandwidth on time and frequency resolution The signal consists of a sequence of four tones with frequencies of 1 2 3 and 4 kHz at a sampling rate of 22 05 kHz Each tone is 20 mS in duration The interval between tones is 10 mS Both spectrograms have the same time grid spacing 1 45 mS and window function Hann The selection boundaries show the start and end of the second tone a Wide band spectrogram record length 64 points 2 90 mS 3 dB bandwidth 496 Hz b Waveform showing timing of the tones c Narrow band spectrogram record length 512 points 23 2 mS 3 dB bandwidth 61 9 Hz The waveform between the spectrograms shows the timing of the pulses Adapted from Raven User s Manual Ishmael Exercise 3 Change spectrogram settings Ishmael 2 3 0 C Users nieukirk Desktop sIntest2 ipf C Users sin goodcalls NSO5 D6_RswyW fileT10 humpaav File Edit Run Compute View Record Detect Localize Actions T Time 71 715 Amp 0 140 181 8 MB used I 10 15 a Spect
Download Pdf Manuals
Related Search