MATERIALS and METHODS
The recordings made before 20 april 2007 were performed with a Marantz PMD 222 monaural cassette recorder, while the recordings made after that date were performed with a Fostex FR-2 field memory recorder. The microphones used were an UEM-83 supercardioid condenser microphone, an AudioTechnica AT815b shotgun microphone or a Sennheiser ME62 omnidirectional microphone with a Telinga parabolic dish (54 cm. diameter). After 20 march 2011 I have used a Sennheiser MKH20 omnidirectional microphone with the Telinga parabolic dish. Sometimes I have used an automatic recording device (Wildlife Acoustics Inc., Model SM2+), with two omnidirectional microphones supplied by the manufacturer (Model SMX-II). Appropriate windshields were used when necessary. The songs recorded with UEM microphone are marked with (U), the AudioTechnica recordings are distinguished with (A), the parabola recordings with (P) and the Wildlife Acoustics device with (WA). Analog to digital conversion of the Marantz recordings was made using a standard audio chip of a PC (QDI Kinetiz7E VT82C686B). A wav (uncompressed) file was created by mean of acquisition software written by the author. Sampling frequency of the audio files submitted to spectral analysis was always 48 kHz mono (= 1 channel) with 16 bit accuracy. The Fostex FR-2 recorder was set as following: sampling frequency 48 kHz, quantization 16 bits, channel mono, limiter off, power supply phantom.
Some audio files published on this web site are reduced to 22KHz or less in order to keep them downloadable also for people with a slow Internet connection, in some cases longer files are in MP3 format. Normally the spectrogram corresponds to the sound file shown together. Sometimes it represents a longer cut than that contained into the audio sample file. In these cases a red square indicates which part of the spectrogram corresponds to the sound file. Spectrograms were done using SeaWave V.1.02, a software written by Gianni Pavan and distributed free of charge by CIBRA (Lab of bioacoustics of Pavia University - Italy). In the general part of this site some spectrograms shown are made with another freeware program called Gram.exe V. 5.0. Links to other commercial and free software for sound analysis are listed in Bibliography (9,10,11). The spectrograms produced by SeaWave plot frequency in kilohertz (kHz) on the vertical axis versus time in seconds on the horizontal axis. The amplitude is represented by a grayscale value between white and black, ranging from 0 to 96 dB. This dB measure is relative: 96 dB is assumed the maximum value reached by a sinusoidal component of a 16-bit signal. The red dots plotted on the spectrogram keep track of the dominant frequency (= frequency associated with the maximal amplitude): this parameter is obviously linked to the loudness of the sound emitted. A louder sound of a given frequency produces a higher amplitude level associated with that particular frequency. The parameters of spectrographic analysis (FFT lenght, frame size, frame overlap etc.) are shown togheter with the correspondent spectrogram plots. Recently I have used a different software for making spectrograms: Raven v. 1.4 by Cornell University Lab of Ornithology. The technical informations given above are applicable to this software too, except for the dominant frequency that is not shown in Raven spectrograms.
For each song and spectrogram there is a documentation about: 1) Species identification - 2) Behavior of the bird (when in sight) and habitat description -3) Date and local time - 4) Latitude, Longitude, Altitude - 5) Sex and age (when possible) - 6) Comments. I have avoided any digital manipulation of wav files before performing spectrograms because it is often a source of distortion and/or spurious data generation. Particularly digital filtering or amplification was not used (except for bandpass filtering under the Nyquist frequency = half the sampling frequency). In some cases digital manipulation was applied to the sound samples published on this website, only to improve their audibility. Also in these cases the spectrographic analysis was done using the original non manipulated 48 kHz sample. The figure below illustrates the unwanted effects introduced by digital filtering.
In Fig. 1a there is a spectrogram of the typical Buzzard call (Buteo buteo) generated from an unfiltered waveform. Fig 1b shows the spectrogram of the same call when the waveform is submitted to a bandstop filter between 0 to 800 Hz. Many spurious data are added to the original "true" spectrogram. Generally, filtering is used to reduce the impact of background noise, as shown in fig. 1. The spectrographic analysis in most cases allows to distinguish between noises and calls easily, therefore in my opinion filtering is unnecessary. In the Bibliographic section (see 6,7,8) of this site you can find articles about recording techniques, selection of field recording equipments and effects of digital manipulations on audio signals.return to top