Acousonde™ FAQ (Technical)

Acousonde™ Technical Questions

Updated 4 February 2009; added questions on sampling programs, calibration, and high-impact applications.

The three most frequently asked questions

How long can the Acousonde record?

It depends on your sample rate and your duty cycle. The Acousonde's 8 gigabyte data capacity will allow acquisition of approximately 4000 million acoustic samples. Divide 4,000,000,000 by your sample rate in hertz and that will be the total "on" time in seconds. For example, at a 25793-Hz sample rate (for 9.2 kHz bandwidth) the Acousonde will record for a little over 42 hours continuous. Note that your choice of sampling rates is limited, so in calculating recording time you should use a supported sampling rate. Also, you can choose to duty-cycle your recording to make it cover a longer time period in exchange for sacrificing continuous coverage. Finally, keep in mind that a battery is an imperfect power source. The Acousonde's ability to fill storage may depend on the battery's manufacturer, batch, age, storage conditions, and deployment temperature, as well as on the cleanliness of the battery and Acousonde contact surfaces.

What has changed from the Bioacoustic Probe?

Two acoustic channels instead of one
Low-power acoustic channel similar to Bioacoustic Probe, 9.2 kHz max bandwidth
New high-frequency acoustic channel samples up to 232 kHz (maybe 464 kHz!)
Acoustic gain choices now just 0 or 20 dB (dropped 10-dB choice)
Anti-alias filters can be bypassed
Future firmware will support "ping-pong" alternating sampling of both channels concurrently
Tiltmeter (accelerometer) now 3D 8-bit (compared to 2D 16-bit)
3D compass provides new orientation sensing
Real-time clock accurate to within 1 minute a year in normal temperatures
Firmware supports easier and more precise time synchronization
8 gigabytes data storage (compared to 1 gigabyte)
Full-speed USB data offloading (compared with standard infrared)
New technology throughout offers room for future firmware/hardware improvements

Does the high-frequency option replace or augment the low-power hydrophone?

Option B003-HF, the high-frequency hydrophone option, adds a high-frequency hydrophone system to a B003A or B003B Acousonde. The low-power hydrophone comes with the Acousonde regardless of whether you have Option B003-HF or not. So, if your Acousonde is equipped with Option B003-HF, it has two hydrophones in it. You choose which hydrophone you want to use at deployment time.

[Return to top] | [Return to Acousonde home page]

Sampling rates, bandwidth, and anti-aliasing

What sampling rates are supported?

The Acousonde hardware can sample a single channel at 464285 Hz, 232142 Hz, 154761 Hz, 116071 Hz, 92857 Hz, and many lower frequencies. Supported sampling frequencies are closer together at lower frequency values, so for desired sample rates on the order of 1 to 5 kHz the nearest supported rate will be close. Also, it is possible to decimate in software after hardware sampling although this capability is not yet fully developed. Note that continuous recording at 464285 Hz may not be possible, we are working on this.

What's the highest frequency the Acousonde can record?

For the low-power channel, the maximum frequency recordable is 9285 Hz (corresponding to a sample rate of 25793 Hz). For the high-frequency channel, a specific number for "maximum frequency" is difficult to state since the linear-phase anti-alias filter rolls off so slowly. The 3-dB filter cutoff is at 42 kHz, but the filter is only 22 dB down at 100 kHz.

What's the lowest frequency the Acousonde can record?

Each acoustic channel incorporates four high-pass filters located at:

the preamplifier input, set by the shunt resistor across the hydrophone capacitance;
the preamplifier input follower stage (16 Hz);
the preamplifier gain stage; and
the postamplifier input (16 Hz).

As noted, high-pass filters 2 and 4 are both set to 16 Hz. The other filters are set differently for the low-power and high-frequency channels. For the low-power channel, Filter 1 is 16 Hz while Filter 3 is 8 Hz; for the high-frequency channel, Filter 1 is 12.5 kHz while Filter 3 is 65 Hz.

To summarize for general discussion: the low-power channel high-passes at 16 Hz, while the high-frequency channel high-passes at 12.5 kHz.

What are the specifications of the Acousonde preamplifier boards?

There are three high-pass filters embedded in the preamplifiers; default high-pass corner frequencies are given in the answer to What's the lowest frequency the Acousonde can record? above. Default gain for the low-power hydrophone is 14 dB, for the high-frequency hydrophone 32 dB. The output is DC coupled at a quiescent voltage halfway between the power and ground rails. Different gain-stage amplifiers are used for the low-power and high-frequency preamplifiers, as appropriate for the different bandwidth and noise-floor requirements. There is no explicit low-pass filtering in the preamplifier design. The input is a high-impedance JFET type.

Can the filtering be customized?

The preamplifier high-pass filtering and the high-frequency channel's anti-alias filter may be customized; see the ordering information. It is not possible to change the hardware of the 16-Hz high-pass filter at the postamplifier inputs or the adjustable anti-alias filter on the low-power channel.

Can I specify my own gains and high-pass frequencies for the Acousonde's preamplifiers?

Yes, within the limits of the design. This is Option B003-CF (for an embedded high-frequency preamp), Option B003-CH (for an embedded low-power preamp) or B003-PC (a loose preamp, e.g. for integration in a customer's external hydrophone).

You should prewhiten to improve broadband SNR.

Prewhitening is a high-pass filtering approach that gradually increases attenuation with decreasing frequency. The attenuation parameters are chosen to counteract the increase in ambient sound levels at lower frequencies typically observed in the world's oceans. Attenuating those expected low frequencies balances the recorded spectrum. Prewhitening enables significantly more preamplifier gain, and thus better signal-to-self-noise ratio at high frequencies, because overloading by low-frequency ocean noise is no longer a concern. Prewhitening is common practice when detecting weak higher-frequency signals is a goal.

The Acousonde's high-frequency channel applies a form of prewhitening by using a default ceramic-and-shunt resistor corner frequency of 12.5 kHz. This filter attenuates incoming signals by 3 dB at 12.5 kHz and by 30 dB at 400 Hz. Above 12.5 kHz the prewhitening filter has little effect. The low-power channel, on the other hand, is intended for calibrated general-purpose monitoring applications at lower frequencies and is not suitable for prewhitening.

What's the relationship between sampling rate and data bandwidth?

The low-power channel's adjustable anti-alias filter will be set to approximately 0.37 times the sampling rate or less. For example, for a sample rate of 25793 Hz on the low-power channel, signals above 9285 Hz (the maximum bandwidth of the low-power channel) will be filtered out. For the high-frequency channel the bandwidth is determined by a fixed-frequency 6-pole linear-phase filter with a cutoff at 42 kHz (down by 22 dB at 100 kHz). If you do not sample the high-frequency channel at 232142 Hz or above there may be substantial aliasing of higher frequencies within the filter band into lower frequencies, which may or may not be acceptable depending on your recording environment.

A linear-phase anti-alias filter? Crazy! It rolls off too slowly.

Four considerations drove the choice of a linear-phase anti-alias filter for the high-frequency channel. First, the primary signals of interest at high frequencies are odontocete echolocation clicks. An elliptic or other "brick wall" anti-alias filter would have heavily distorted these clicks and confounded time-domain analysis. Second, odontocete clicks roll off at high frequencies such that any aliased content will most likely be weaker in amplitude than the in-band portion of the click. Third, the gradual rolloff of the linear-phase filter will allow clean decimation in downstream digital processing, possibly onboard. Fourth, it is possible to bypass the anti-alias filter if desired and, potentially, to sample at a rate high enough (464285 Hz) that a clean signal can be acquired with neither filtering nor aliasing. This being said, the filter choice is experimental and subject to change in the future.

What is aliasing?

Aliasing means high-frequency signals are masquerading as low-frequency signals. The phenomenon is familiar to anyone who has seen automobile wheels appear to turn backwards in a movie (unintentional aliasing), or an engine flywheel appear to be stationary when viewed with a strobe light (intentional aliasing). Just as a machine operator may be injured if he thinks the strobed flywheel is stationary and reaches to touch it, a researcher may be misled by an acoustic record that contains aliased signals, since it will show tones or noise at a low frequency that were in reality at some higher frequency.

Any incoming analog signal with a frequency above one-half the digital sampling rate (the "Nyquist frequency") will be aliased when digitized, and once the aliasing takes place, there is no way it can be undone in post-processing. Therefore digital recorders must filter out frequencies above the Nyquist frequency before digitization. However, if one is confident that one's recording environment contains no signals above the Nyquist frequency, or that any such signals are weak compared with the corresponding in-band sound levels, or if one wishes intentionally to alias incoming signals (as in the strobed-flywheel example above) then anti-alias filters can be omitted. The Acousonde provides optional anti-alias bypassing.

[Return to top] | [Return to Acousonde home page]

Managing two acoustic channels

Can the low-power and high-frequency channels be recorded simultaneously?: The Acousonde's analog-to-digital (A/D) converter is not capable of true simultaneous sampling. However, if both channels are connected to hydrophones (i.e. Option B003-HF has been installed) the A/D converter can "ping-pong" (alternate sampling) between them. This means that the per-channel sampling rate for each of the two channels would be one-half of the A/D master sampling rate, which we are confident can be sustained as high as 232 kHz and possibly as high as 464 kHz. Firmware to enable this capability is planned for future release; we welcome user comments as to how useful it would be.
[Return to top] | [Return to Acousonde home page]

The sampling program

Is sampling continuous or duty-cycled?: Either. You choose between continuous or duty-cycled sampling when you program the Acousonde before deployment. Duty cycling is set up in the form X minutes out of every Y minutes, for example, 10 minutes out of every 60.
Can the start of the sampling program be delayed so it begins later, not now?: Yes.
Can sampling be triggered on an event of interest?: No. However much of the programming has already been done to trigger sampling when certain auxiliary indicators -- depth, temperature, tilt, etc. -- satisfy a user-programmed condition, so we expect to provide this capability in future firmware updates.
Can the Acousonde do acoustic event detection?: No. The Acousonde hardware incorporates vector floating point acceleration, and in theory it should be possible to program the Acousonde for acoustic event detection. However this is not a simple thing and we have no plans to support this feature as part of the off-the-shelf Acousonde. Please see the Customization FAQ for more information about significant customizations.
[Return to top] | [Return to Acousonde home page]

Calibration and signal quality

How is the Acousonde calibrated?: Acousondes are not independently calibrated as finished instruments. The hydrophones embedded inside them, however, are calibrated by their manufacturers at spot frequencies. Those calibration data are embedded in each Acousonde's internal metadata storage. This is the same approach that was used with the Bioacoustic Probe; according to several B-Probe customers who performed independent calibrations, it works very well, at least at frequencies up to a few kilohertz. This is not unexpected, since the urethane overmolding has a characteristic acoustic impedance similar to that of seawater, and the acoustic complexity associated with the Acousonde's internal components will be almost invisible at lower frequencies.
How accurately does the Acousonde record ultrasonic frequencies?: Unavoidably, the Acousonde's frequency response is anything but flat at high frequencies. First, the transfer function of the presently-used high-frequency hydrophone includes excursions to both -20 and +20 dB as the frequency reaches and passes 100 kHz. Second, the location of the high-frequency hydrophone amidst a circuit board and other neighboring components profoundly affects its directionality at high ultrasonic frequencies. Finally, the placement of the Acousonde relative to other objects, including of course an animal to which it may be attached, impacts the overall transfer function. These factors can to some extent be characterized and compensated for, but there will be many unknowns in each deployment.
[Return to top] | [Return to Acousonde home page]

Attachment, mechanical and electrical

How long does the attachment last on the animal?: At present the Acousonde comes with no attachment or retrieval gear. The duration of attachment depends entirely on the user's choice of attachment technique. This approach was necessary because research using the Acousonde involves a wide range of species, from sharks to manatees to seals to baleen and toothed whales. To date, attachment methods have included suction cups (whales), glue (seals), tethers (manatees) and surgical implantation (sharks). These techniques have yielded attachment lifetimes from a few minutes to several weeks. To our knowledge nobody has used the Acousonde with a dorsal-fin saddle.
What does the Acousonde mechanically provide to secure attachment or retrieval gear?: Model B003A includes a cylindrical battery cap with an 8/32" threaded screw-hole at the center; this can be used to attach eyebolts (for tethers) or to secure flotation. A urethane "key" protruding from the Acousonde's surface can be used to prevent screw-on flotation from rotating unintentionally. Attachment systems such as suction cups must be secured either to the flotation (not included) or to the urethane body of the Acousonde itself. This has been done in the past with gear as simple as cable ties and electrical tape.
What are the precise dimensions and weight?: The Model B003A cylindrical Acousonde with a battery cap and A-size lithium battery but WITHOUT flotation, attachment, or recovery gear, weighs 262 g (in air) in a volume of 172 cc (this compares to 212 g and 135 cc for the Model B002B Bioacoustic Probe). The "torpedo" shape is 3.2 cm in diameter and 22.1 cm long (compared with 3.2 cm and 19.3 cm for the Bioacoustic Probe). It is negatively bouyant; in seawater the tag weighs approximately 86 g (compared with 74 g for the Bioacoustic Probe). Of course, flotation, attachment and recovery gear will add to the size and weight of your total package.
Spring contacts are unreliable, you should use clips or soldered batteries.: When designed properly, cleaned regularly, and assembled securely, spring contacts work well for all but the most shock-intensive applications. For example, six Bioacoustic Probes deployed for 1-2 weeks on northern fur seals in August 2005 all ran flawlessly, despite cold Bering Sea waters and significant movement (in this example the contacts were treated with Stabilant 22, a chemical contact enhancer.) While other approaches such as battery clips or soldered tabs do work, their reliance on user-handled wires would decrease the long-term reliability of the Acousonde. The Acousonde is a sealed instrument and if wires emerging from the sealed portion were to break, repairs would be difficult or impossible. As long as a spring system is strong, tight, and clean, it will be adequate. Note: never attempt to solder directly to batteries, particularly lithium batteries, that do not have solder tabs pre-attached by the battery manufacturer. There is an explosion risk.
I want to launch the Acousonde onto a target. Can it handle the impact?: Unknown. The Acousonde 3A shares its basic mechanical design with its predecessor, the Bioacoustic Probe, and that earlier instrument took abuse well. However until recently no customer has ever asked about high-speed impact tolerance, so it was not part of the Acousonde's design criteria. To recommend the Acousonde for use in these applications, we would need to perform potentially destructive testing that would very probably indicate a need for additional development. Given the expense of this testing and development, demand will determine our ability to support high-speed impact applications.
[Return to top] | [Return to Acousonde home page]

Commanding and communication

I don't have a PalmOS-compatible PDA. Can I still use the Acousonde?: No. You must have a PalmOS-compatible personal digital assistant (PDA) with an infrared port to command the Acousonde.
Do I need a specific model of PalmOS-compatible PDA?: A bewildering variety of past and present PalmOS-compatible PDAs are fully compatible with the Acousonde. It has been tested with the Palm III, V, m100, m500, T|X, and Zire 21, as well as models from Handspring and a Kyocera cell phone. Any PalmOS-compatible PDA should work. Replaceable batteries (as in the Palm III) are a bonus for field work as no charger is necessary, but we are not aware of any current PalmOS PDAs that use replaceable batteries. We recommend that you choose a PalmOS PDA on the basis of price and durability, as well as relevance to other tasks you may have.
What computer and operating system are required to download data?: When plugged into a personal computer via USB, the Acousonde behaves like an ordinary full-speed USB flash drive. Any system that supports the use of USB flash drives should be compatible with the Acousonde.
Wait, "full speed" is much slower than "high speed" USB! How long does it take to offload data?: The Acousonde offloads data at about 280 kilobytes per second. This works out to about 1 gigabyte per hour, so the Acousonde's entire 8 gigabyte storage capacity can be offloaded in about 8 hours. You can choose to download selected portions of stored data, which of course will require less time. We hope to migrate the Acousonde design to high-speed USB in the future.
Why not just offload the data to the Palm via infrared?: It is possible, and encouraged, to offload small text files to the Palm, such as the log file and the filesystem directory. Larger files, however, may exceed the Palm's data capacity and will take an unacceptable amount of time to transfer via infrared. In any event there is no Palm software available to view or analyze the data files from the Acousonde.
[Return to top] | [Return to Acousonde home page]

Importing and analyzing data

How do I import the data into software that can analyze them?: The tag stores and provides data in a custom format known as "MT" format. A MATLAB program (that is, an M-file) is available to read MT files directly into MATLAB. A free unsupported application for Apple's Mac OS X, "MT Viewer," is also available to read and inspect MT files on Macintosh personal computers; MT Viewer can also export selected data in MATLAB, ASCII, or WAV format. If neither MATLAB nor MT Viewer is an option for you, you may be able to convert MT files directly to WAV if you manually keep track of the sampling rate, and strip the leading 512-byte MT header using multipurpose audio-conversion software such as SoX . Otherwise, some custom translation software may need to be programmed; detailed information on MT format is available to facilitate such an effort. With time we hope to gather the translation tools that users have created and make them available to the community to minimize "wheel reinvention."
What exactly is MT format?: MT format is a simple single-channel format for containing time-series data. It consists of a fixed-length 512-byte header followed by an arbitrary length of 2-byte samples. The header provides information on sampling rate, start time, site, title, comments, and calibration, among other things. All header fields are in ASCII, making the headers immune to byte-swapping issues and easy to read in low-level file editors.
Why not use WAV format?: WAV format is the de-facto standard for storing uncompressed audio on computers running Microsoft Windows. WAV was never intended to store scientific data, and does not provide a standardized means of storing calibration values, time of day, site codes, comments, or other "metadata" parameters essential to scientific purposes. Although WAV is not the best native format for scientific data, some analysis packages can accept WAV input, so a few Acousonde users have developed MATLAB software to convert the Acousonde's native MT format to WAV. MT Viewer (see "How do I import the data" above) provides an export function that allows selected portions of data to be exported to WAV for analysis by other software.
Why not use Broadcast WAV format?: Broadcast WAV is an extension to WAV format that incorporates generous and flexible metadata capability, primarily for the news and entertainment industries. It offers a serious alternative for keeping scientific metadata with acoustic records; however, we are not aware that a universal scientific-metadata format has yet evolved for Broadcast WAV. So even if the Acousonde supported Broadcast WAV the compatibility with analysis software might be limited.
I've never analyzed acoustic data. How do I get started?: First, determine if you need to perform calibrated acoustic analysis. If not, there are many acoustic analysis packages available for listening to records and inspecting spectrograms. If calibrated analysis is necessary, however, we must recommend partnering with an expert. Calibrated acoustic analysis is nearly impossible for a non-specialist to do in a repeatable, standards-based manner without formal education. Even with formal education it is surprisingly easy to use techniques that are inappropriate to the signal or situation in question, leading to meaningless and irreproducible results. The use of canned software only increases this risk. A good way to know when you may be ready to analyze acoustic data on your own is when you fully understand the meaning of the unit designation "dB re 1 µPa/Hz^½".
[Return to top] | [Return to Acousonde home page]