Wildlife Acoustics: Song meter SM2
Wildlife Acoustics have just launched a new bat-detector, called Song Meter SM2. Like the Batcorder and the Pettersson D500X, this is a detector that one can leave in the field to record passing bats automatically. The market for this type of detector is big at the moment. Many biologists in the United States, Canada and Europe are now using time expansion recorders that record automatically. So far, the latest model of the Song Meter SM2 (384 kHz) is the cheapest detector of its kind ($1000). On this website we previously wrote about the first version that used 192 kHz and discouraged to buy that model for bat surveys. The new model that samples at 384 kHz (16 bits) can record bats up to 192 kHz which is sufficient for most bats of the world. We therefore now consider this detector a very serious contender to the Batcorder and the Pettersson D500X.
The Song Meter SM2 is in fact a kind of modified data recorder, such as the popular Edirol, but it is automatically triggered by bats and then samples at 384 kHz (16 bits). Ian Agranat of Wildlife Acoustics explained to me that the SM2 stores data continuously as it records, so there is no buffer and therefore no gaps between recordings. The recording can be stored directly in .wav format on the flashcard, but also in a compressed format. This doesn't have anything to do with time-expansion, it just means that a less heavy data format is chosen, comparable to using GIF instead of raw to save a photo. The format is called .wac and a program by Wildlife Acoustics can convert it back to .wav. The .wac format can be chosen to be lossless so the entire original .wav signal can be reconstructed with the program while halving the file size that is stored on the flashcard. In fact, the filesize can be reduced safely by an additional factor of 3 by removing the least significant bits from each recording. This data reduction is not the same as with mini-disk. It is based on the fact that the dynamic range of the recording system is 64dB at best, whereas the 16 bits provide 96dB, so 32dB (5 bits) are only filled with noise anyway and can therefore be discarded. This gives a total lossless reduction in filesize of a factor of about 5 in practice.
4 flashcards can be inserted into the detector. If the detector is set to record non-stop, regardless of bat presence and if 32GB cards were used, 92 hours (8 nights) could be recorded, still without using any compression. With lossless compression, but also continuous recording, this would be 40 nights. Since normally the detector only records when being triggered by a bat, 240 nights can typically be registered on the cards until they are full.
The SM2 can run for 100 hours on 4 type D batteries on a programmable schedule spread through several months. The Batcorder 2.0 also manages 100 hours on its rechargeable battery. The SM2 also has a built-in temperature sensor and data-logger. There is also an external power source available to lengthen deployment time.
What about the triggering system? We have no user reports so far, but again Ian Agranat provided a detailed explanation of how it works: "The trigger looks at a band-limited energy level - it can be set as either an absolute (dBV) sensitivity level or a relative (dB spike above rolling average ambient background). The relative trigger can adapt to stationary background noise (such as the constant whine of some insects or wind, etc) and watch for a sudden pulse above these levels to cause a trigger. This is superior to any absolute threshold trigger in discriminating wind and insect noise, but can still false-trigger in a gust of wind or a more pulsating insect."
The new SM2 has one microphone that is claimed to be omni-directional (this can be achieved to some extent by keeping the microphone membrane small). Sensitivity is 6.3mV/Pa and the signal to noise ratio 59dB, which is quite good for an electret microphone. We don't yet know the upper frequency limit of the new microphone. Users are recommended not to point the microphone up because the rain will get in and damage the microphone. It is also possible to use the microphone on a long (10m) cable (trees). However, it may be necessary to increase the gain somewhat if you do this because the incoming signal may get weaker. Therefore, if you do comparisons between 1 and 10m up, be sure the sensitivity remains comparable (you could use the long cable in both conditions)! We have heard of microphones being damaged by rain, so it may be advisable to make some kind of little roof, say, 30 cm (not too close: reflections!) above the microphone.
A good point about the SM2 detector is that it has an adjustable gain. Generally, if you use a bat-detector in a very dense forest many passing bats will use soft calls (especially in the tropics). With the Batcorder, such bats rarely trigger the detector because of the built-in sensitivity and trigger. In such situations, the SM2 could be adjusted to be more sensitive so it would pick up these calls. Of course, if you then use the same setting in an open environment where many loud bats are hunting many calls will be hopelessly distorted. One should therefore use different sensitivity settings according to the habitat. This is methodologically fine because, very likely, even with the more sensitive setting, the maximum range of the detector will still be smaller in a dense forest than in the open. However, this statement is purely based on experience and, at present, I can't provide any evidence for this claim yet. In your methodology, you can simply state that the detector was set in all conditions such that only the very loudest bat caused a slight distortion. You will then see that you will need very high sensitivity levels to achieve this in dense environments.
On the website of Wildlife Acoustics a user forum can be found where users can exchange problems, findings and be advised. Some users have also commented on the SM2 and recordings look good and clean. A comparable detector to the ultrasound SM2 has already been used for continuously monitoring bird songs and the recognition software to automatically identify them looks very interesting.
Last updated 26.2.2011