Bat Acoustic Technology Summary
Bat detector briefing
Bat detectors are indispensable research tools for determining bat activity and increasingly to identify bats to species. They work by converting the otherwise inaudible, ultrasonic calls of bats into signals we can hear and analyze. This ultrasonic conversion employs three main technologies in order of sophistication:
(1) heterodyne, (2) frequency division, and (3) full spectrum.
Is one of these technologies better than the other? In certain situations the simple heterodyne output will yield all there is to know about activity and species composition. Other more robust conversions are needed to sample bats in areas where there is high species diversity, where several species of bats converge on similar echolocation calls, or for listening to low-intensity bat calls. Today, many investigators are archiving recordings made with full spectrum microphones because this conversion technique by far retains the most audio-information from the incoming bat call. Today, the most advanced software to analyze bat calls has been created to deal with full spectrum recordings in mind. Moreover, today's full spectrum data can be re-analyzed as our sophistication with sound analysis improves in the future.
Bat Conservation International Acoustic Monitoring Workshop.
Southwest Research Station, Portal, AZ May 2009.
When people are first introduced to bat detector technology, they are often overwhelmed. Much has been written weighing the pros and cons of each detector and conversion process and people try to invest in the “perfect” one. This is backwards thinking. These technologies are actually very complementary, because bat detectors have a variety of specific uses and can be used under many different field conditions to answer a host of questions about bat presence, behavior, and species composition on both spatial and temporal scales. Beginning a survey with a bat detector in frequency division mode, which scans the full bandwidth of frequencies and allows users to listen to activity in real time, provides instant feedback about bat use. Once bats are detected, switching to a heterodyne conversion output lets users "tune into" a specific bat frequency, which provides clues to exactly what species of bats may be present. Most European bat-workers are quite adept with this skill and can confidently determine species using inexpensive heterodyne technology alone. If a particularly unusual or unexpected bat call is detected or to record a nice, strong bat pass in high resolution, today's most sophisticated detectors can switch to a full spectrum, high-speed recording mode, suitable for sophisticated computer analysis. For this reason, the Pettersson detector manufacturer pioneered the concept of bundling multiple technologies into a single unit. So while initially some of these detectors are rather intimidating, there is a method to the madness that can only be appreciated once detectors are used for direct monitoring in the field.
High-frequency conversion technologies in greater detail
Heterodyning (HET): The most common method, HET is a very sensitive, narrow band technique whereby the user selects a short range of frequencies to transform using a manual tuning control. Setting a HET detector to 40kHz makes echolocation calls between 35-45kHz audible, in real time. Most bat calls are broadband, sweeping thru many frequencies over time, so only a portion of the call transformed by the HET system is re-played. Depending upon the dialed frequency of the detector and the characteristics of the bat call (i.e., the duration of the call within the dialed frequency range), the tonal output of the detector will vary, sometimes producing a "chirping" sound (indicating a flatter, longer call) or a "ticking" sound (for a shorter, steeper call). With practice, it is possible to identify species based upon the dialed frequency and the detector output.
Frequency division (FD): FD detectors use a broadband conversion, transforming the entire bat call in real-time. The original high frequency call is converted to a "square" wave and divided by a user-selected ratio (usually 4, 8, 10, 16, or 32). The detector counts how many times the original high-frequency sound wave cycles from negative to positive, producing an output signal for every 4, 8, 10, 16, or 32 zero-crossings. This results in an output that is 1/4th, 1/8th, 1/10th, 1/16th, or 1/32nd of the input frequency. Low frequency calls are better represented at low division ratios, and high frequency calls should be recorded at higher division ratios. Though FD detectors are less sensitive than HET (or TE) units, they are capable of producing accurate representations of a bat call, including important information about the shape, slope, and characteristic frequency; parameters often used in species identification. Sophisticated FD software tools such as Analook are analyzing approximately a dozen characteristics of a bat call.
Full Spectrum (FS): Full Spectrum detectors preserve the entire sound characteristics of the original high-frequency signal, not just the time-frequency components preserved by FD analysis, but also important time-amplitude information, rendering an incredibly accurate depiction of the signal strength of high-frequency sound, including multiple harmonics if present, representing the entire acoustic soundscape of a bat call. In many cases, this renders a much more information-rich data-space upon which to render a species classification. For comparison, sophisticated FS software such as SonoBat can analyze up to 70 parameters as it is looking at sound in "three dimensions" of acoustic space rather than the more basic time-frequency "2D" representation of FD. Is FS better then FD for species identification? In certain species, perhaps. In the hands of an expert user (NOT an automated statistical program), with enough time and enough good quality recordings, either technology will permit confident species identification. Luck is helpful too.
Full Spectrum Time expansion (TE): TE conversions record the original high-frequency, broadband call and can play it back at a slower speed. This allows digital outputs to capture the entire signal, retaining important information about original signal strength (amplitude) and spectral components (e.g., harmonics). Why this admittedly convoluted approach? When the concept was invented, field computers of the day could not handle the data processing requirements to instantly record the sound at a high enough sample rate necessary to record bat calls in FS. TE detectors were able to collect the full spectrum of ultrasonic recordings, retaining all call parameters commonly used in species identification, by “slowing down” the call to allow computers to use a lower sample rate. Users of a Pettersson D240x select how much of the original signal to record (1.7 or 3.4 seconds @ 307 kHz) and the detector expands the bat pass by a factor of 10, producing a detailed output 17 or 34 seconds long which can then be recorded at a relatively low processing rate of 44.1 kHz onto any computer/MP3 recorder. This "long" recording is digitally reassembled and analyzed by popular software programs, including SonoBat with very high precision. Despite being a "legacy" method, TE still holds it's own for active monitoring purposes, voucher calls, and even passive recording where the data can be retrieved daily.
Full Spectrum Direct Recording (DR): Direct recorders like the D500x, BAT iFR-IV, and the SM2/SM3 are designed for long term passive monitoring that can record continuously to memory cards for weeks on end. These devices record full spectrum bandwidth all the time with essentially down-time to expand and record incoming signals (no TE) between recordings. Typically these devices do not have display screens to watch bat activity in real time, but instead the files are copied from the memory cards in a device-specific workflow and off-loaded for analysis. To get a sense of how far the technology has come, the D1000x is capable of almost instantly recording at 768 kHz. For species ID purposes, 250 kHz is the bare minimum recommended sample rate for bat work. This is because the maximum frequency resolution of any recording device is half it's sample rate (explained as part of the Nyquist-Shannon sampling therorem). At a 192 kHz sample rate, only 96 kHz can be resolved, leaving certain species such as Northern long-eared (Myotis septentrionalis) and Fringed Myotis (Myotis thysanodes) underreported or misclassified. Generally, the higher the sample rate, the higher "resolution" of the resulting spectrograph. Lars Pettersson recommends a 500 kHz sample rate for serious bat work, technology available only in premium devices such as the M500, D500x, and D1000x.