Potentially quite a useful little app. I would like to make a couple of feature requests if development is to continue. 1. A frequency scale at the left edge would be useful as a reference 2. There are several USB microphones on the market with a sample rate of 256000 samples/sec. At present you can only select 250000 and that is the sample rate that is written to the .wav file. 256K would be a more useful alternative, or some way of explicitly setting the device's actual sample rate. 3. I prefer to view my spectrograms in greyscale, or reverse greyscale - from looking at the code it should be relatively simple to offer alternative colour palettes. Lookinf forward to seeing some continuing development of this project.

The big problem with digital filters is that distortion due to overloading or to intermodulation has already happened and can't be undone. Indeed if the signals are to be analysed songraphically the FFT process provides good visual separation of different frequency bands making filtration somewhat redundant except for aesthetic purposes. If you are using an external microphone it should be possible to add some hardware filtering in line between the microphone and the Audiomoth. Passive Pi or T filters are not difficult to implement if you can find the right tables of coefficients, and if you know the input impedance of the Audiomoth. Active Filter Design by Arthur Williams (publ Artech, 1975) is a good reference for filter design in general. I also have some photocopied design tables for passive LC designs from 'Simplified Modern Filter Design, but I am afraid I do not have the full reference for that work.

Kaleidoscope or Audacity will cope with all the frequencies that an AudioMoth will record. The only limitation is that a chunk of the screen is committed to the notation panel in Kaleidoscope so I restrict the display to 100kHz for bats here in Herts, only zooming it out if its a tricky identification.

I do quite a lot of recording of bats, so there are three main programs that I use. My preferred technique is to record continuously into 4 or 5 minute long files and I analyse these with Audacity, which is free, very versatile and reasonably intuitive in its controls. I add a label track and add my identifications of each pass in the label track which can be exported as a text file alongside the original .wav file. I know that a lot of people don't like the workload of continuous recording and prefer to use triggered recordings of 5-10s when a bat is detected, even though quiet calls may be missed (and this is obviously preferable for long static deployments). In this case I use Kaleidoscope. I just use the free version since I can't afford the pro version. Properly set up this allows rapid classification of large numbers of short files that can be viewed in a single screen without having to pan the spectrogram. Finally, since I end up with large numbers of recordings, I use my own Bat Recording Manager program (freely downloadable through www.echolocation.org.uk), which can not only streamline the analysis process (especially with Audacity, but also with Kaleidoscope or AnalookW) but stores all the metadata so that particular information can be easily retrieved and examined. I have tried most of the other software (at least those programs that I don't have to pay for) but find most tricky or non-intuitive to use. Sonobat does good pulse analysis and bat ID but the interface is very poor (small dark text on a dark background and non-intuitive controls) but is highly recommended in some quarters. BatClassify is no longer supported and is a little limited but can be useful if you can find a copy. Avisoft SASLabLite is extremely powerful and can do some very sophisticated manipulations but is to cumbersome for everyday use.

the GUANO defined standard format for embedding metadata in .wav files is commonly used, reasonably widely recognised and quite versatile, allowing vendor specific fields to be defined where predefined fields are not sufficient.

As an alternative approach, Audacity can quite easily split a longer file into 1s files. Use Tools/Regular Interval Labels and the file will get numbered labels 1s long at 1s intervals (or whatever you set in the parameters page). Then use File/Export Multiple to split the file based on labels to export multiple 1s files into the folder of your choice. It is also possible to put these commands into a macro that can be run on a batch of files although I haven't actually tried this.

There doesn't seem to be a significant difference in the FMEs in the illustration in the paper. They both show a similar double peak - the dips being due to interference patterns from overlapping signals received over alternate paths of slightly different lengths. The relative heights of the peaks will vary with slight phase differences in the overlapping signals - inevitable when the receivers are not using a common microphone. I would want to see some statistical analysis over a large number of pulses recorded on the two machines through a common microphone to be convinced of real differences. The SNR of the moth is worse, but might be improved on by adjusting the gain level - quite critical with the smaller dynamic range of 12-bit resolution.

I use a similar technique on my external microphones, but since I build them into a 3.5mm jack plug there is very little space for the additional components. I therefore use a 1uF tantalum capacitor from Vdd to ground and connect Vdd directly to the ring of the jack plug - which is normally supplied with a bias voltage in most input sockets. Then I use a small 0.1uF coupling capacitor from signal to tip. Both of these can, with care, be accommodated in the body of the jack plug. In order to use these microphones (which preceded the design of the audio moth) you need to add a resistor on the audio moth from v+ (3V) to the ring connection of the added jack socket, which is unconnected in the moth.

Sounds good, but where do I download it? Lots of links to apps, but I can't find one to the new firmware.

Some cars have an ultrasonic intruder detector built in to their alarm system - could be that inside the car?

Oh yes... you can also assign labels to segments of the recording and then save the label track (giving it the same name as your .wav file but with a .txt extension) to make a permanent record of your observatons during the analysis.

Counting the calls is tricky and will probably need some specialist software, depending on the frequency range and structure of the calls, although Kaleidoscope viewer or Avisoft SASLab Lite may be able to do something of the sort just based on signal amplitude (might be combined with some pre-filtering?). Audacity as a general purpose spectrogram viewer is my personal favourite for long(ish) recordings (as opposed to 4-5s samples from a triggered stand-alone recording system). The spectrogram will show the full bandwidth defined by the sampling rate in the recording. You may need to look at the spectrogram settings under preferences to ensure that the spectrogram is fully expanded and to set the appropriate gain and FFT parameters which will depend somewhat on you sample rate and on the resolution and size of your screen. It is best to set the gain so that the background noise level is just visible so that you don't miss quiet calls. If you click the * mode then zooming and panning can be done very intuitively witht he mouse - check the preferences keyboard and mouse settings for a list of key and mouse shortcuts.

Good tip. I uninstalled and re-installed the 32-bit MCU SK, then realised that it was 5.8.2.0, so I installed the older version 5.2.2.0 as well and then AudioMoth imports and builds successfully.

Unfortunately many insects, especially bush crickets, use the same range of frequencies as many bats and it would therefore requrie extremely complex software to discriminate between them. Not really feasible in real time in a simple device like the AudioMoth. Your best bet is to use some post processing and some software bat auto-classification on a more powerful PC afterwards. Although auto-classification of bat recordings does have its own limitations - not least cost, but you could look at Bat Classify, Tadarida or SonoBat amongst several others.

I get the same problem on Windows 10 Home, i7-6700HQ 16GB, 64bit. On import the SDK box shows in red as above and if I click Next and Finish it says it failed to read the project description file from location C:\Users\...\AudioMoth-Project\.project

It's a shame that the .wav files don't include the firmware version in the wav file INFO chunk - a thought for the next generation maybe?

Even better would be to have the option to preserve a given set of settings as new defaults. I know that with PIC micros you can write data to the Flash ROM and I guess that may be possible for ARM as well. So, in the configuration program you could tick a box to say 'make this the new default setting' and the values would be stored permanently in the AudioMoth. Takes extra power and the number of writes may be limited but its not something that would need to be done very often, but different users have different preferred settings and it would provide a quick recovery if the batteries had to be changed, or had just got jostled and fallen out.

Ziplock baggies are probably quite similar, but the requirements of different AudioMoth users differ very widely. A form of protection which is good enough for detecting and recording tropical high frequency bats is probably over-restrictive for someone trying to detect gunshots, or frog calls. Requirements may also differ between those simply for a species identification compared to someone who is trying to look at detailed signal structure or complex acoustic signatures. It is also difficult to know just what the formulation of a given baggie is when comparing rather anonymous items originating in different continents. As a general principle, the thinner and lighter the material the better, but beyond that ...

The microphone, as I recall, is a Knowles SPM0410 but it is a surface mount device that has numerous pads un the underside which have been reflow soldered to the board, so that removal would be very difficult and replacement in situ almost impossible. Such things can be done but the techniques are difficult to acquire and the equipment would probably cost more than the AudioMoth. A better approach would be to cut the tracks to the microphone (circuit diagrams and board layout are available on here somewhere) and use fine wire-wrap type wire to connect them to a 3.5m jack socket. With a bit of care it is then possible to mount a replacement microphone in the shell of a metal 3.5mm jack plug. Use a small 0.1uF capacitor to connect the microphone signal to the tip, use the ring to provide a 3V (or 5V through a 10k resistor on the AudioMoth board) to provide power for the microphone, with another small tantalum capacitor to ground for decoupling. This provides a useful style of mounted microphone that will plug into any stereo microphone input jack that provides bias voltage as well as into the newly mounted AudioMoth jack. If you are not interested in the ultrasonics other Knowles microphones can be used which are a bit simpler to attach wires to.

Not specifically hydrophobic - it is basically polyethylene film with a very thin film of aluminium - not much more than a molecule thick, but just enough to be conductive. The advantage is that it is very very light (in 3.5um thickness) and therefore has minimal effect on the sound especially at high frequencies. I use it as I had stocks in hand for making microphone diaphragms for ultrasonic capacitance microphones with responses which are flat to 180kHz or so. Figured if it works as a diaphragm should be good as a screen and at least blocks bulk ingress of water.