UniversitÓ degli Studi di Pavia
Centro Interdisciplinare di Bioacustica e Ricerche Ambientali
Via Taramelli 24 - 27100 Pavia - Italy
e-mail : email@example.com
& techniques for bioacoustics
|Brand & Model||Media||
|Marantz PMD 660||CF||48K 24b||2||XLR P48||4xAA|
|Marantz PMD 670||CF||48K 24b||2||XLR P48||8xAA|
|Marantz PMD 671||CF||96K 24b||2||XLR P48||8xAA|
|Edirol R4, R4Pro||CF+HD||96K 24b||4||XLR P48||8xAA||Pro version has 80GB HD, AES/EBU I/O, lower noise|
|Edirol R09||SD||48K 24b||2||PiP||2xAA|
|M-Audio MicroTrack 2496||CF||96K 24b||2||TSR P48 PiP||RNR||poor P48 powering, hissy|
|Fostex FR2||CF/PCMCIA||192K 24b||2||XLR P48||8xAA|
|Tascam HD-P2||CF||192K 24b||2||XLR P48||8xAA|
|Sound Devices 702-702T||CF||192K 24b||2||XLR P48||Sony M/L||excellent mic pre with 70dB gain|
|Sound Devices 722-722T||CF+HD||192K 24b||2||XLR P48||Sony M/L||idem|
|Sound Devices 744-744T||CF+HD||192K 24b||4||XLR P48||Sony M/L||idem, but only 2 mic inputs! requires additional mic pre|
|Zoom H4||SD||96K 24b||2+2||XLR P48||2xAA||4 track (2 mic + 2 line) recording limited to 44K 16bit|
|Sony PCM-D1||MStick||96K 24b||2||PiP, mj||AA||expensive|
|NAGRA ARES M||1GB||48K 16b||2||2AA||only 1GB internal, dedicated mic available|
|Sony MZRH1||HiMD||44K 16b||2||PiP, mj||RR|
|CoreSound Mic2496 + PDAudio CF||192K 24b||2||XLR P48||
|requires a PocketPc with Live2496 software for recording to CF/SD, or a notebook/subnotebook.|
PiP = Plug in Power
|Please consider that this table is just a list of available devices. Each one with its own features, advantages and disavantages to be carefully evaluated according to research needs and available budget.|
New portable models are expected to come out in early 2007: the Fostex FR2-LE, the Korg MR1 and MR1000. The Korg units adopt the 1-bit technology already used for DSD (DirectStreamDigital) Audio and SACD (SuperAudio CD); a specific software will allow to convert 1-bit audio sampled at 2.8MHz (5.6MHz in the MR1000) into the desired PCM audio format 48/96/192K with either 16 or 24 bits. Theoretically this technology will allow to get almost any frequency range just by software processing of the original stream; the declared bandwidth in DSD 2.8MHz mode is 100kHz without the aliasing problems normally introduced at frequencies close to the Nyquist frequency (half of the sampling rate) in traditional PCM converters. The drawback of this technology is the high frequency noise introduced by the AD conversion; the noise is inaudible but, for bioacoustic needs, may reduce the usable dynamic range at ultrasonic frequencies.
are given by PDA (Personal Digital Assistant) based
recorders; by interfacing a microphone preamplifier and AD
converter to a PocketPC PDA, running either Linux or
WindowsMobile, it is possible to record on the PDA
memories, either SD or CF, and then easily move the
recorded files to a traditional PC. The only interesting
solution now available is proposed by Core Sound (other
solutions are limited to speech recording); a PDA recorder
may offer the same quality of an off the shelf recorder,
it may offer a greater flexibility, but it is important to
mention that assembling different pieces of hardware,
connecting them and providing power might be difficult, in
particular in severe field conditions.
The image at left shows the PDA unit assembled at CIBRA with Core Sound components (Mic2496 preamplifier and AD converter, PDAudio CF card, Live2496 software). The PocketPC can be replaced by a solid state recorder, provided it has digital input without digital resampling. A viable option is to use the M-Audio MT2496 just for storage.
Very interesting discussion about recorders and microphones are available in the "naturerecordist" email discussion list on Yahoo.
Recording on a PC, either desktop or laptop, may have great advantages. Sound devices for both laptops and desktops are now available with 192k s/s to provide more than 80 kHz of useful bandwidth while dedicated instrumentation acquisition boards can sample up to 500k s/s to get ultrasounds up to more than 200 kHz. For laptop use, USB and FireWire sound devices now allow up to 8 channels at 96k s/s and few models go up to 192k. A 80 GB hard disk can record for about 120 hours with DAT quality (16 bit, stereo, 48kHz) or 60 hours with doubled bandwidth (16 bit, stereo, 96kHz); larger disks and RAID controllers available on desktop PCs can allow to record for weeks. Even if the use of computers for recording, analyzing and editing sounds has been experimented since 25 years ago, only in recent years the PC capabilities and the availability of good and cheap sound devices and huge hard disks have made computers powerful and affordable enough. The new generation of subnotebooks and small tablet PC could boost the interest in computer recording in the field.
Advantages given by laptop recording are: wide choices of sound inputs, sample rates, number of channels; recording duration benefits of huge HDs, ability to schedule recordings, wide filenaming capabilities (timestamp, location, gps position, etc.), sound streaming over either wired or wireless networks, etc. Unfortunately most built-in sound interface are not as good as we would. Thus in most cases it is required to connect to an external sound input device, USB, FireWire, or PCMCIA.
As for recorders, the critical part is the sound acquisition front-end made by microphone preamplifiers and AD converters whose specifications are often not clear enough. The page Tech Tips, Tricks and Tests shows some results we got while testing audio interfaces and sound recorders for our lab.
Besides sound quality, it is also important to use equipment suited for field use. There are many USB devices, some of which are powered through the USB bus; FireWire devices can be powered by the bus but not all FW sockets in laptops do provide power. There are few other solutions based on PCMCIA boards that may have an external box; some of these external box require external powering. In all cases where external powering is required, if field use is required it is important to verify what type of power is required (V, mA, and if DC or AC) to provide a suitable battery system for the field. An optimal choice for field use is when a DC current is required in the range 5 to 12V. AC powering or higher voltages require more expensive and complicate solutions.
The photo shows the Digital Signal Processing Workstation developed by CIBRA in 1996. A Sony DAT D7, optically connected to the Opcode DATport, the first USB device able to allow the direct digital transfer from a DAT to a PC file. Now most audio interface include a digital port, either electrical or optical.
Visit the pages Development of instruments for sound recording and analysis and Tech Tips, Tricks and Tests for technical details and special information on the equipment we use.
Instrumentation recorders are typically suited to record signals whose frequencies are lower or higher than those audible by man. Often these instruments allow recording several independent channels at the same time (multi-channel recorders) and have several tape speeds to be selected in relation to the frequencies to be recorded: higher speeds to record higher frequencies. To record frequencies up to 100 kHz, analog recorders run the tape at speeds up to 76 cm/s. Ultrasound recordings can be played back at reduced tape speed to be made audible, to be analyzed or to be recorded on conventional audio tape recorders.
Instrumentation recorders designed to record ultrasounds are very expensive and not well suited for field use; thus, cheaper devices to detect and possibly record ultrasound were developed to study echolocation in bats. These were called bat detectors.
A special class of instrumentation recorders is based on FM recording. A carrier is frequency modulated by the incoming signal and then recorded on tape: when playing back the tape a demodulator extracts the original signal. FM recorders have a bandwith typically ten times lower than comparable direct recorders and are well suited to record very low frequency signals. FM recorders have been replaced by DAT recorders modified in their I/O circuitry to accept very low frequencies, down to the DC. Special DAT recorders have been made to record more than 2 channels at the same time by splitting the DAT standard bandwidth into up to 16 independent channels.
Tape based instrumentation recorders are being replaced by dedicated or general purpose PC systems equipped with suitable data acquisition interfaces and large hard disks. PC based systems can acquire and record signals from 0 Hz to many MHz. A special feature of all instrumentation recorders is that they are "calibrated"; this means that they record a known voltage range and their input level settings are calibrated. With a calibrated recorder connected with a calibrated microphone (with a known pressure/voltage sensitivity), or hydrophone, it is possible to accurately measure the received acoustic pressure by reading the recorded "voltage" and converting it to the received acoustic pressure.
In recent years instrumentation recorders have been replaced by either dedicated recorders or PC based recorders.
Bat detectors were developed to provide
researchers with cheap instruments to study bat
echolocation. Bat detectors are based on both analogic and
digital techniques to detect and record ultrasounds.
Three main systems are actually used by the detectors available on the market: heterodyne frequency shifting, frequency division and time expansion; the most advanced instruments have all these three systems to make ultrasounds audible and recordable, with some limitations, on usual audio recorders. Direct and continuous recording of ultrasounds requires expensive instruments not well suited for field use.
Heterodyne detectors allow to shift a small frequency range, tipically no larger than a few kHz, down to the audible range; the user tunes the detector to the frequency of interest and then he listens to and records only signals whose frequency is around the tuned frequency. Anything outside that frequency range is lost.
Frequency division (or count-down) detectors cover a very large frequency range and are basically Zero Crossing Detectors. The output signal from these has a frequency which is a fraction of the original frequency (e.g. one tenth). The most advanced retain the amplitude envelope of the original signal.
These two systems allow recording an audible representation of an ultrasonic call, not of the full ultrasonic signal structure.
The time expansion detector is the most accurate system: it retains all of information of the original signal. The ultrasonic signal is sampled at high speed and digitally stored into a memory; then it is replayed at a lower sample rate, e.g. one tenth, to be made audible and recordable with traditional equipment. If the stored signal is replayed at a sample rate ten times lower than the original one, frequencies are reduced by ten while time is expanded by the same factor. Unfortunately, this kind of instrument allows to store a few seconds only. To store more time of ultrasonic signals, large memory expansions are required.
The systems suited to record long sequences of ultrasonic signals are the traditional high speed tape recorders (instrumentation recorders), very expensive and requiring a lot of tape unless they use VHS video tapes, and PC based digital signal acquisition systems with fast high-capacity data storage.
The recordingo of ultrasounds can be now easily achieved by means of a laptop computers with a fast sound sampling board. Sound boards with 192 kHz sampling are now available to record ultrasounds up to 85-90 kHz. Among PCI boards, the Lynx Two model is highly appreciated for its flat frequency response up to 90 kHz and the steep anti-aliasing filters; also, by using our SeaPro Ultra software, this board can be set to run up to 200 kHz of sampling rate to provide 95 kHz of usable bandwidth. For portable use there are either FireWire or USB devices that claims 192 kHz of sampling rate, but some have a frequency response limited to 50 kHz and poorly designed anti-aliasing filters. An alternative to computer recording is offered by few solid-state or hard-disk recorders that can sample at 192 kHz; three models are now available, two CF recorders (Fostex FR2 and Tascam HDP2 and one with both internal hard-disk and CF (Sound Devices 722/744).
To further increase the recording bandwidth it is necessary to use very expensive dedicated recorders or high speed data acquisition boards connected to a laptop or to a desktop PC. With those boards it is easy to record at up to 1Msample/sec. National Instruments provide a broad range of data acquisition devices with PCI, USB, PXI, PCMCIA and FireWire interfaces. Normally these devices don't have anti-aliasing filters on board and thus it is required to add an external a-a filter to each channel; this greatly increases the cost of the acquisition system. Additional costs should be also taken into consideration to properly interface the board to the sensors and to develop or buy a recording software suitable for your needs.
Visit the page Tech Tips, Tricks and Tests
to learn more about the performances of 192 kHz devices or
the page Development of
instruments for sound recording and analysis to
discover the equipment we developed.
A new device for recording
ultrasounds has been developed by Dodotronic: it is an
ultrasonic microphone with integrated high speed AD
converter and USB port. Two models are available:
Dodotronic UltraMic200k and UltraMic250k, both can be
connected to a laptop PC and with our SeaWave and SeaPro software it is possible to
record ultrasounds to the internal hard drive, to
see on the spectrographic display and also hear them
shifted down to audible range.
Introduction to sound
Sound analysis allows to display the features of acoustic signals graphically, and, thus, to understand and measure their structure and to correlate it to observed species, behaviours and situations. Spectrographic representation of animal voices has been widely used since the first analogical analysis instruments were developed for military acoustic research.
The transformation of signals in the digital domain allows a new approach in the management of the data, thus easing operations of filing and analysis in connection with both the listening and the real-time display of the signals.
The development of digital signal processing techniques and high-speed hardware at relatively low-cost has actually made the visualization of acoustic signals an every-day invaluable tool for bioacoustic research and for educational purposes.
A number of analysis techniques are
available; usually, they are based on dedicated digital
systems or are carried out with general purpose computers
equipped with suitable analog-to-digital conversion
devices and specific Digital Signal Processing (DSP)
software. The simplest graphical displays are the
oscillogram, which shows the waveform of the signal, and
the envelope, which shows the amplitude of the signal in
regard to time. The most significant analysis is, however,
the spectral one, since it shows the composition in
frequency of the signals: the instantaneous spectrum
(frequency-amplitude plane) shows frequency components of
a short segment of a signal, while the representation of
more spectra, computed on consecutive or overlapping
segments of the signal, shows the evolution in time of its
frequency structure; graphically this is achieved by
showing the spectra in an ordered time series,
representing them, for instance, on an axonometric
diagram, in a three dimensional space
The most effective, compact, and easily understandable display is the representation of the signal on the frequency-time plane, with the component intensity coded through a scale of greys or a suitable colour scale. This kind of analysis is usually called spectrogram, or SonaGramTM since it was first realized by the Kay SonaGraphTM, and is largerly used to analyze animal sounds as well as the human voice. Since spectrographic analysis, actually based on the windowed FFT (Fast Fourier Transform), is unsuited to analyze some non-stationary signals due to the uncertainty principle, several other processing techniques (zero-crossing, wavelet, wigner-ville) have been developed to resolve the frequency-time structure of complex signals or to accomplish particular tasks.
Using graphic representations, one can easily compare various signals in order to find similarities or differences between them, to classify signals in regard to their morphology, related behaviours, supposed meanings or individual emittors.
Visit the page Software for real-time sound analysis to learn more about sound analysis software developed at CIBRA.
Equipment for sound analysis
Several instruments on the market allow to analyze animal sounds; most of them can acquire and store a signal segment to be later processed, analyzed, displayed and/or printed in one or more graphical formats. Such an instrument can easily be a personal computer, a MacIntosh, a IBM-compatible equipped with suitable hardware and software. Basic hardware must include Analog-to-Digital (AD) and Digital-to-Analog (DA) converters with at least 16 bit resolution, selectable sampling frequencies up to 50000 s/s to allow analysis up to 22 kHz, and sharp low-pass filters to avoid aliasing, that is the "pollution" of the frequency range of interest by frequencies higher than the Nyquist frequency, which is half the sampling rate.
More advanced and expensive equipment should include fast CPUs and highly optimized software to perform in real-time, fast AD/DA boards to acquire ultrasounds, a DSP board to speed-up some intensive computation (present CPUs are so fast that in most cases dedicated DSP processors are no more needed), a digital I/O board to directly connect an external digital device such as a DAT recorder or any other recording device with digital I/O.
Recently, the diffusion of the Windows environment and multimedial applications has widened the interest about digitized sounds and many excellent AD/DA sound boards are now available to provide up to 192k s/s sampling with 24 bits of accuracy. As hard disk can now store up to 160GB in the 2.5" models and up to 500GB in 3.5" models, recording on laptop and desktop PCs can be an effective alternative to the use of expensive stand-alone audio erecorders and instrumentation recorders. A number of programs running under Windows allow to record, edit and play-back sound files, although only few of them allow to visualize in detail their acoustic structure and can be effectively used for bioacoustic research. The software developed at CIBRA has been developed for bioacoustic research and provides real-time sound analysis capabilities and continuous recording to hard disk. Visit the CIBRA equipment page for further details.
Other resources on the net:
A huge amount of valuable first-hand information is also available by joining the Naturerecordists and Laptop-Tapers email discussion groups on YahooGroups. There you can find information about other professionals and amateurs have setup their equipment to get the best compromise among cost and required features.