Does anybody know an online source of acoustic membranes other than Alibaba? I can't even find anything on Ebay.

We're interested in calls of different species of owls and some other conservationally important birds, so this solution is only partially useful. Certainly, 48 kbps is an overkill and was used for initial testing only, but even with the noise floor improved a bit, 20 dB attenuation will severely limit detection range - which is doubly important with rarer large predators.

I should note that the abovementioned test was done in a normal office, not in a soundproofed chamber.

I repeated the test after allowing the case to settle for a good 13 hours (though the membrane already went flabby a few minutes after closing). Done with playback of Ural owl (Strix uralensis) and common chaffinch (Fringilla coelebs) calls, the loudspeaker was directly facing the recorder ~2.3 m away, with Ural owl call at approx. 85 dB RMS at 1m. Both sounds were first played at an enclosed recorder, then at the same recorder free-standing on a desk. It was set at maximum gain and 48 kbps sampling rate. Again, the case attenuated the owl's call by 20 dB, whereas the finch's call was amplified by 4 dB this time. This forum won't let me attach a picture, so I uploaded it to https://www.nib.si/images/stories/EKOS/comparison.jpg It shows oscillograms and spectrograms of all mentioned calls, pasted together (same scale throughout). We'll try putting out recorders in pairs - one enclosed and one in a ziplock bag - at our monitoring locations next week, to see what we get outside. But from what I've seen, it looks like the official cases aren't fit for our purpose, unfortunately. We'll also try one of the 3-d printed designs later, but it might take a while to get all the pieces.

Will report tomorrow.

We just stumbled on this problem ourselves. Setting up monitoring of Ural owls, a quick overview of a week's worth of recordings near a nesting box yielded no calls whatsoever. Then we tested with playback at close range and the call was attenuated almost 20 dB by a freshly closed case! (375 Hz peak frequency, RMS amplitude measurement) In comparison, finch's song (3500 Hz) was only attenuated 5 dB. It would be nice to publish a real frequency response curve so people can make a more informed decision when purchasing. For us, this result is quite disappointing.

Thank you, looks like I set something wrong earlier so UTC was used, I tested again and the device behaved as expected this time. Sorry for the trouble. Best, Jernej