3
RAVEN Pro 1.4 (we will update soon to 1.5) Haverflock Guide:
Our lab’s homebrew supplement to the Raven Pro 1.4 Manual based on our notes from the Cornell Bioacoustics Research Group’s Sound Analysis and Sound Recording Workshops (Thanks to all the folks at the Cornell Bioacoustics Research Group and Macaulay Library for all the great lectures! All errors are our own, though.)
Suzanne Amador Kane & Emma Oxford
Haverford College Haverford PA 19041
RAVEN Pro 1.4 (we will update soon to 1.5) Haverflock Guide: 1
Our lab’s homebrew supplement to the Raven Pro 1.4 Manual based on our notes from the Cornell Bioacoustics Research Group’s Sound Analysis and Sound Recording Workshops (Thanks to all the folks at the Cornell Bioacoustics Research Group and Macaulay Library for all the great lectures! All errors are our own, though.) 1
Suzanne Amador Kane & Emma Oxford 1
Haverford College Haverford PA 19041 1
1
General / How to advice 2
Trouble-shooting 3
Field Recording tips 3
Spectrograms: making good ones for your data 4
Reporting on spectrograms for publications 5
How to report the width of a frequency spectrum 5
Reporting recording spec’s 5
Identifying and finding signals in a sound file 5
Method #1: Manual detection 5
Method #2: Correlations 6
Method #3: Detectors 7
Band Limited Energy Detector 7
Amplitude Detector 11
Evaluating detection effectiveness 11
Playback 11
Memory Management 12
Exporting & Editing Images from RAVEN 12
How to install Raven Pro using Windows 7 (may also help with Vista) 13
Fixing Raven Pro installation problems with Windows 7 & JAVA Virtual Machine 13
Raven Video 1.2 (Alpha) Playback/Analysis with Video sync’ed 14
Recording equipment & Signal Acquisition with RAVEN 14
Beamforming 15
XBAT 16
General / How to advice
● Waveforms vs. Spectrograms: You get better time resolution for waveforms than for spectrograms. Use spectrograms to remove noise, then use correlations on waveforms if possible?
● Detectors: Amplitude vs. Band-Limited Energy? Amplitude is only useful when your sound is much louder than the background noise.
● To choose which plot to work with: Click on left vertical bar alongside the plot--active window turns blue.
● Committing a Selection: In Raven, you select a part of a waveform/spectrogram, etc. and hit Enter to commit the selection so that it is saved in the Selection Table and clicking on other parts of the waveform/spectrogram doesn’t erase it.
● Saving files for other purposes: Can save selections as sounds, save measurements as .txt files for Excel, other analysis. No statistics in RAVEN, so you have to export to do this.
● The left menu called Layout lets you change what’s displayed (Waveform, Spectrogram, both, Channel 1, Channel 2, etc.).
● Paged displays aren’t all in memory, only partially, so anything like Amplification, Filtering, etc. that changes the signal itself can’t be done
● Peak frequency = Max. Frequency in Measurements = frequency @ maximum power ≠ high frequency! (High frequency & Low frequency, percentile freq. show boundaries at top and bottom)
● Linking/Unlinking: Combines pink markers & magnification of views. A spectrogram and a waveform for the same sound file are always in time by default. Right clicking on either one and going to Unlink View allows you to unlink them in time and zoom in on one without zooming in on the other—very handy if you wish to scroll through spectrogram or waveform or different channels independently!
● How to link files: drag & drop into same group (in Linkage tab on left bottom menu) after 1st selecting property (Time Scale, e.g.) for linking
● File formats: AIF and WAV files have similar info, just Apple vs. Microsoft. AIF less widely used. MP3 is less desirable because of compression--loses top end of freq. range ~ 20 kHz--should be OK for us.
● Grabbing audio from video files: RAVEN 1.4 can open movie files directly in many formats. Can also be done using Quicktime using main menu at left, called the Export Audio feature. But you have to purchase an enhanced version of Quicktime, since the free version doesn’t do this.
● Watch out for Lock icon next to some settings (samples, Hop values) in the Configure Spectrogram box; this freezes their values and prevents updates.
● Filtering: Edit/Band Filter (and toolbar icons) has selections for filtering without selecting; this removes the filtering from the displayed signal and spectrogram. There is an Edit/Undo Filter option!
● Brightness/Contrast: Right-click on image, Configure Brightness/Contrast to edit all features independently using the Floor and Ceiling sliders. (More flexible than using the simpler Brightness and Contrast sliders on main toolbar.)
● Lines: Under Views in Layout menu (left hand side). Plots the same file on N separate time plots stacked vertically to show all the long and short term time behavior at once.
● There is a Clear Selection icon in the 2nd toolbar—useful when you accidentally create one or wish to refresh your selections. The icon looks like a red square that has been erased at one corner.
● Copy a selection: Ctrl+C , then open a new sound file and paste the selection. Can only copy one channel’s selection at a time in a multichannel file.
● Scrolling playback: The play option with a small right facing arrow beneath the play button. This option allows you to play an entire file (assuming you have opened the entire file and are not in page sound mode) even if the whole thing does not fit on the screen at one time. It just scrolls through the file as it plays.
○ A handy trick: If you unlink the waveform and spectrogram in both Time Position and Time Scale, scrolling playback allows you to blow up the spectrogram window and play the sound so you see the moving detailed spectrogram while you are oriented in the entire waveform file at larger timescales.
● The pink vertical line is the playback and position marker. It sits at the left edge of the plot by default, but you can drag it around. All zooming in time happens around this marker.
● Use Presets to save a View with selections, linkage, etc. using View/Window Preset/Save As
● You can annotate selections—very useful for beamforming angles, etc.
● Spectrogram slice = power spectrum (relative intensity in dB vs. frequency): to make this up, choose the “new spectrogram slice” icon from top left and you’ll get a new plot that shows power vs. frequency, computed using the time window in your spectrogram parameters and located at the same time as the cursor.
Trouble-shooting
· Q: I get an out of memory message. What to do? A: See the selection on Memory Management in the Raven Manual and the section below with a few extra tips.
· Q: I can see the sound and the spectrogram—it says it’s playing, but there is no sound! A: You probably have more than one channel and are playing the wrong channel! On the lefthand menu, try going to Playback and making sure you are playing back the correct channel.
· Q: I just ran a detector. Where is the spreadsheet of measurements? A: It’s cleverly hidden just below your waveform/spectrogram window. You use the mouse to grab it and pull it up—like lifting a window shade!
· Q: My times and other measured values are very low precision compared to what the program is capable of measuring. How do I get Raven to display them in a higher precision format? A: Two options: right-click on the column and choose Format and change the precision. Or, go to the Edit/Preferences and search for “time” in that file. You will find a list of preferences for how many decimal places to use in recording times, frequencies and other measured values. Change these to your desired values (keeping in mind your actual recording frequency and the limits that imposes—typically you cannot distinguish timescales faster than 0.05 ms!)
Field Recording tips
● Give yourself “headroom”: set your recorder gain so your peak signal level reads -10 to -12 dBFS (full scale, where 0dB is full scale on your recorder)
● Record tips and tails: record for a while before and after your main signal.
● Dual Mono (DM) is good to use if you have it: it records a second signal -20dB down so you avoid clipping if your sound exceeds expected levels.
● SD cards for recorders can malfunction if not rated at 6 or higher for 96 kHz recorders
● Wind Noise: foam windscreens are less effective than fluffy ones with fur; best at fur + closed cell foam inside (foam is more transparent to sound, though.) Bigger windscreens put more distance between the mic and any insects. Insects prefer black windscreens over lighter colors, though; put your mic behind your body, closer to the ground, behind an obstacle to reduce wind noise. They advised us to do these mechanical things in preference to switching on the high pass filter on a microphone; only use that for the windiest days was their advice.
● Coat your cables with camouflage fleece covers to reduce handling noise.
● Watch where you step to avoid crunching underfoot.
● Wear soft clothes and avoid noise from dangling zipper pulls, etc. (Cabelas makes soft brushed raingear.) Find gloves that do not squeak or make rustling noises when handling equipment. (We like EMS glove liners. Good at gripping knobs, reasonably warm, good tactile feel, not squeaky.)
● Phantom power vs. power inside microphones: seems to be OK to use either.
● Factor of 2 distance in the field results in 6 dB intensity change for point sources, 3db gain for linear sources (for line sources such as a line of traffic, or for sound in an alley or other effective waveguide) (Think about it: in 3 dimensions, intensity falls off as 1/R2, but only as 1/R if you lower the effective dimension by one in a channel.)
● DEET insect repellent fogs plastics and should not be used near recording equipment
● Battery tips: Avoid going out into the field with duds (and get the most use out of your batteries!) by purchasing an expensive portable digital multimeter (DMM) so you can read the voltage on your batteries. Anybody can use one of these! No electronic skills needed. (Battery testers are usually not helpful in predicting remaining lifetimes. We like the inexpensive DMM’s offered by Sears Craftman and Radio Shack.) The idea is that batteries have a peak voltage when they are fully charged, but this voltage falls off when they are almost drained. So, you can determine how fully charged your battery is by reading its voltage.
Use your DMM to read the voltage on all your batteries before going out into the field. For alkaline batteries, we have found that AA or AAA rated at 1.5 V ought to read close to 1.6 V to have a good remaining charge; 9V batteries ought to read higher, like 9.5V; for rechargeable nickel metal hydride (NiMH) batteries (we like Pearstone and Duracells) the AA and AAA batteries are rated at 1.2 V, and they read 1.38 V when fully charged. So, even if your alkaline battery reads 1.50 V (or your NiMD reads 1.2V), it’s likely to be near the end of its lifetime! Replace it before you go out in the field, and put that battery in the “use for low priority lab uses” box!
Spectrograms: making good ones for your data
Try playing your sound files slowed down (top right hand corner—Rate) to detect significant variations in pitch that you wish to capture in your spectrograms, then use that time information to design your spectrogram time windows, etc.
You can save spec’s in View/Window Preset and reuse later for other sound files.
Spectrogram Configuration:
Default spec’s blur fine time behavior, so, under Spectrogram Parameters, change to:
Window: Hann (preferred—not uniquely good), 256 (2.90 ms) (3dB filter bandwidth is determined by the window size--don’t report); DFT = 128 (= 345 Hz for our 44.1 kHz sampling rate); no clipping; Hop size = 0.295 ms, overlap = 95%
(Window size has to be much less than shortest duration of the sound of interest)
Spectrogram Slices:
Single plot at pink vertical line location on spectrogram that shows FFT power spectrum vs. freq at that time. Get using icon or right click & choose.
Reporting on spectrograms for publications
· Sampling rate; digitization resolution
· DFT (frequency grid): samples (then in paratheses in Hz)
· Window (type: Hann, width in samples (then in parentheses in msec), filtering?)
· Hop size and overlap
· Clipping used?
How to report the width of a frequency spectrum
Spectrogram slice (FFT power spectrum) difference in frequency between peak and down 3dB on either side (FWHM: down by a factor of 2) relative to background
Reporting recording spec’s
Sampling rate: 48 kHz or other, # bits (24 or 16 bit or other?) (Nyquist freq. = sampling rate/2--max. F in FFT)
Which recording devices, microphons, preamps, etc. were used.
Omnidirectional vs. other microphones
Identifying and finding signals in a sound file
Method #1: Manual detection
This basically means defining a selection area, then doing the measurements on the selection. It’s much harder, but realistically the pro’s tell us that no detection method will ever get all the calls that are apparent to a human scanning through a file looking for calls defined by the appearance of their spectrograms.
So here’s how to do this:
· First make up a standardized spectrogram using your Preset Spectrogram Parameters. (You need to use standardized values or else your measurements will not be well-defined)
· Choose measurements by right-clicking on your spectrogram, then choosing the ones you want, or else using View/Choose measurements and choosing the measurements you want.