Assembling cheap, high-performance microphones for recording terrestrial wildlife: the Sonitor system

Microphone design basics

Sound consists of pressure waves travelling through a medium, in our case air. Audible sound makes the air vibrate at frequencies between 20 Hz and 20 kHz. Ultrasound, which is not audible for humans, extends beyond 20 kHz. Insects and bats can emit and perceive ultrasound up to 200 kHz⁸. Microphones are transducers of mechanical energy (pressure waves) into electrical energy (a voltage). A variable voltage is created as sound waves move mechanical parts of microphones, which can be a polarized membrane (electret condenser), or a piezoelectric element. The role of the recorder is mainly to increase the minimal voltage differences with amplifiers, digitize them with analog-to-digital converters, and record them to a digital storage medium (mostly solid-state memory secure digital cards).

Outdoor microphones are electrical devices which need to be protected against water ingress, and climatic and mechanical shocks. Protection comes from solid housings, often metal tubes in which the microphone element is inserted. The microphone element (often ambiguously called simply "microphone") is the centerpiece of the microphone and consists only of the acoustic sensor which transduces sound to a variable voltage, and it is not usable as is. Microphone housings need to be open to allow sound to reach the microphone element through their acoustic port. Since an opening would allow water to penetrate the microphone, corrode its components, and block the sound path, protection is needed. Acoustic vents are used: they are transmissive for sound while being impermeable to water or hydrophobic, and thus fulfil a crucial function for outdoor microphones. Then, microphones need to transmit their output voltage to a recorder via electrical wires. When microphones are interchangeable, they use an audio connector as interface, which needs to be weatherproof too. A minimal microphone assembly only requires soldering of microphone elements and cables, as well as sealing of the other microphone parts using glue if used outdoors.

Basic microphone properties can be augmented with attachments. Windscreens, usually made of synthetic foam or fur, reduce unwanted wind noise which comes from friction of air against the microphone. They also reduce potentially damaging water pressure from rain drops. Furthermore, parabolic reflectors or horns can be used to gather sound over a larger area before concentrating it to the microphone element, but the gained amplification is traded off against higher directivity: the sound pickup pattern becomes narrower.

Sonitor microphone components

Microphone element. We chose to use microelectromechanical (hereafter MEMS) microphones due to their high performance at small sizes, the potential of that newer technology to mature and offer higher performance than conventional microphone capsules, and their lower part-to-part variation and sensitivity to temperature variations (Lewis et al. 2013). Different elements exist that can fulfil different requirements by prioritizing low-noise recording, a wide frequency response, or weatherproofing. We are using microphone elements from different manufacturers. We used a tried-and-tested element from Knowles (SPU0410LR5H-QB), which was used by the company Biotope.fr inside the now discontinued BIO-SMX-US microphone as a substitute for SMX-US microphones by Wildlife acoustics. We also used it inside our own housings since 2017 for recording birds and bats. We tested Invensense's ICS-40720 element, which features low-noise recording (specified signal-to-noise ratio of 70 dB) and also Vesper's VM1000, which is a piezo-electric element that is waterproof and resistant to various environmental stresses.

Printed circuit board (PCB). Microphone elements can be directly soldered to cables, but this requires great care and dexterity for a precise soldering result that does not exceed the temperature tolerance of the element. Moreover, a precise alignment of the microphone within the housing and with the acoustic vent is needed for compatibility with external attachments and for enabling consistent part-to-part quality. It is thus preferable to reflow-solder MEMS elements to printed circuit boards, which can be made in electronic laboratories or workshops equipped with reflow ovens. This is readily available as a paid service and is a burgeoning business satisfying the needs of electronic equipment manufacturers and electronics hobbyists in need of prototypes. Cables can then be more easily soldered to PCBs without damaging the microphone element. The microphone and conductive tracks can be attached on the bottom side of the PCB, which guarantees a result that is flush with the housing. PCBs can be ordered in any size and shape with a variety of support materials.

Housing. We chose to integrate the microphone elements into simple metal tubes, which can be made out of stainless steel or lighter aluminium. These metals offer high resistance to weather and mechanical shocks, are cheap and readily available, and easy to glue. They can be painted to reduce their visibility in natural environments. Due to their hardness, metals can also be lathed with high precision to ensure stable results within tight tolerances so that any attachment can easily fit the housing.

Wires and connector. We chose standard 30 AWG stranded wires for more flexibility compared to solid wires. On one end, the cables are connected to the PCB, which is connected to the microphone element. On the other end, the wires are connected to Mini-Con-X series waterproof connectors without the grommet, which is needed to release the tension when the connector is attached to flexible cables. This connection form is commonly used in most autonomous sound recorders. Mini-Con-X connectors can withstand some abuse and are ingress-protection rated at IP67 (dust tight and protected against water up to 1 m deep).

Acoustic vent. We use Gore acoustic vents to protect the element against solid and liquid ingress. Different products in varying sizes and protection levels against water are available. GAW112 vents can be used, they appear identical to the ones used in SMX-US, SMX-U1, and SMX-II microphones from Wildlife acoustics. They need to be coupled with windscreens, as GAW112 vents let water pass after immersion or drop projection. We also tested GAW325 vents, which are IP67 rated. Freshwater ingress per se only temporarily blocks microphone elements that are not waterproof from vibrating, but will not short-circuit the microphones due to the low conductivity of water. However, water leads to corrosion, which will destroy microphones and conductive tracks, given enough time. The GAW3XX series also have a support material, which can be made of woven or non-woven PET material. The PET (woven) support elements are better suited as they absorb water less.

Microphone assessment

All assessments of the microphones’ technical qualities were performed with SM2Bat+ recorders (Wildlife acoustics), which allow to record two channels up to a maximum sampling frequency of 192 kHz. We used a battery-powered one-driver Anker SoundCore loudspeaker for emitting audible pure test tones at 1 and 10 kHz (generated using Audacity 2.2.2) and an ultrasonic calibrator (Wildlife Acoustics) that emits chirps at 40 kHz. Test sounds were emitted to the front of the microphones and when needed also to the side at a 90° angle. We measured the amplitude of test tones in recordings with a sampling frequency of 96 kHz in Audacity by exporting the frequency spectra with a Hanning window size of 1024 and choosing the frequency window that included our tone's base frequency.

Weatherproofing vs. sound attenuation. The only point that is permeable to sound is the acoustic vent, and its permeability to water ingress is given by its specifications. The sound attenuation at 1 kHz is usually also indicated in the product specifications given by the manufacturer in decibels (dB), as this is the frequency most relevant for recording human speech. However, terrestrial wildlife sounds span frequencies from 20 Hz to 200 kHz, so we measured the transmission at three representative frequencies: 1 kHz (birds and amphibians), 10 kHz (insects), and 40 kHz (bats) to quantify the acoustic vents’ trade-off between sound transmission and ingress protection.

We compared sound attenuation of 2 GAW113 and 2 GAW325 vents with an open setting without vent, outdoors (Figure 1). We recorded the US calibrator and loudspeaker tones at 3 m from the microphones, to the front and to the side at a 90° angle to the side. Four Knowles microphones were used, first open, then with the vent holders, and then two of them were covered with the GAW112 vent and the other two with the GAW325 vent.

Figure 1. Setup used for testing microphone attachments outdoors.

Foam strips reduced ultrasound echoes and the microphones were approximately 1 m above the ground and parallel to each other.

Windproofing vs. drying after rain. We used Knowles elements; one was protected by a GAW112 vent and a windscreen (Wildlife Acoustics), one had a 6 mm long horn attached, and one had a GAW325 vent outdoors. All three configurations represented similar levels of water ingress protection, but we used the Knowles microphone with the 6 mm horn instead of the Vesper microphone (for which it was designed) to equalize the microphone model. We emitted test sounds with the loudspeaker and the calibrator at approximately 4 m. We placed a 62 W fan at approximately 30 cm from the microphones, to the front and to the side (90 degrees) to simulate wind. We recorded the test sounds to check how prone to noise the vent-only and horn-only microphones are in comparison to the microphone with the windscreen. Then, we drenched all microphones in distilled water to simulate heavy rain. We continued recording test sounds immediately after, as well as 1, 3, 18, and 66 hours after the simulated rain to check how long sound transmission was attenuated by the different wet attachments. We measured the sound level of the 1, 10, and 40 kHz tones recorded by each microphone relative to the sound level recorded after 66 hours of drying.

Cable length vs. signal loss. The latest microphones of Wildlife Acoustics usually advertise built-in amplifiers to strengthen the relatively low voltage signals of the microphones so that they do not degrade over long cable distances. High frequencies are more prone to signal degradation because the capacitance of the cable causes more attenuation at high frequencies. We tested whether the output signals of the Knowles microphones were affected by long cables, which are sometimes needed for installing microphones far apart or in different locations than the recorders themselves. We attached two Knowles microphones to the recorder, one via a 5 m cable and the other one via a 52.5 m long cable. They were close to each other and pointing in the same direction. We recorded test sounds emitted with the loudspeaker and the ultrasound calibrator at 6 m from the recorder. We recorded the same test sounds after switching the cables to check whether the results were driven by the microphone itself. We measured 20 ultrasound chirps for each microphone with each configuration.

Directivity vs. amplification. We built different horns for amplifying the acoustic input signal before it is transduced by the microphone (Figure 2). Doing this results in an increased signal-to-noise ratio and ultimately greater detection ranges. However, acoustic horns are generally directive: At high frequencies, horns will mainly respond to sounds within their opening angle, where direct sound can reach the throat of the horn. Outside the opening angle, low-frequency sounds reach the throat of the horn by diffraction.

Figure 2. The different acoustic vents and horns tested, as well as the vent holder.

1 EUR coin for scale.

The reasoning behind using horns is that in stereo deployments, there is a redundancy of recorded data: omnidirectional microphones pointing in opposite directions are recording much of the same data twice. To make better use of them, one can use acoustic horns that amplify the sound from the front and decrease sound from the back or the sides. Ultrasound, which propagates less far, benefits especially from horns, because even very small horns can achieve considerable amplification. For ultrasound, horn dimensions can also be held as small as the existing microphone housings. Also, microphones usually suffer from a drop in the frequency response and/or signal-to-noise ratio in the ultrasound range, thus horns help to attain a desirable, more linear frequency response.

We chose horn designs with steadily increasing amplification with frequency starting approximately from 10 kHz and minimal directivity. Conical horns are generally more suitable than exponential horns, which do not amplify sound much above a certain threshold. Horn dimensions were chosen by calculating and simulating the theoretical analogue amplification in-axis and off-axis using numerical methods to choose the most favourable designs. The gain of the horns was calculated using one-dimensional equations for conical horns⁹. Since the one-dimensional calculations could not predict directivity, Boundary Element Method models¹⁰ were set up to model the directivity of the horns. The ultimate gain depended mainly on the ratio of the areas between the mouth and throat of the horn, while the frequency range depended on the length of the horn. A long and narrow horn will also be resonant, which will increase the gain but reduce the fidelity of the recorded sounds.

We investigated whether ultrasonic horns could amplify the signal enough to compensate for the transmission loss due to the acoustic vents. We also tested how much amplification could be gained with different horns placed in front of the Vesper microphones, which do not require vents.

The Knowles and Invensense microphones require the use of the GAW112 or GAW325 vents for ingress protection. The diameter of the vents’ active surface (through which sound travels) dictates the maximum mouth diameter and theoretical amplification of the horn. The resulting horns were named after the vent they were designed to hold (GAW112 and GAW325 horns). We compared sound attenuation of three GAW112 and three GAW325 horns with and without vent to the open microphones. We tested three horns of each type on three different Knowles microphones, by first recording with open microphones, then with the horns attached, and finally with the vents pasted onto them. We recorded the US calibrator and loudspeaker tones at 3 m from the microphones.

For the waterproof Vesper microphone, we were free to test three different horns whose mouth diameter was only limited by the diameter of the housing but tested varying lengths. We also tested 3 other ultrasonic horn types designed for the Vesper element (thus not holding vents) on three different Knowles microphone elements (for consistency with our measurements of the vent-holding horns). We had 3, 6, and 12 mm long horns, with a throat diameter of 0.75 mm and a mouth diameter of 12 mm. We first recorded open microphones, and then successively attached horns of increasing length to each microphone. We recorded the US calibrator and loudspeaker tones at 6 m from the microphones due to the greater amplification of these longer horns.

Cost. We assessed the cost in working time and money at each step of the creation process for 100 microphones. We contrasted the cost for 3 microphone designs presented later. We considered the ordering of individual parts, components assembly, and microphone testing. We estimated labour and prices from our own purchases and working time. For the costs of building the PCBs and metal housings and horns, we asked three different suppliers for quotes and chose the best offer.

Weatherproofing vs. sound attenuation

The GAW112 vent reduces ultrasound transmission from the front only slightly, by almost 2 dB, while sounds from the side are reduced by more than 7 dB, which is partly due to the vent holder itself (Figure S1)³. The GAW325 vent reduces ultrasound transmission by almost 15 dB but relatively less for sounds coming from the side (almost 11 dB).

Windproofing vs. drying after rain

The windscreen significantly reduced wind friction noise (Figure 3). The vent-only and 6mm horn configurations were affected by wind friction noise at up to 3 kHz, greatly masking the 1 kHz test tones, although they were still audible and visible in spectrograms. Data for windproofing and weatherproofing are available on OSF³.

Figure 3. Spectrograms of different microphone designs showing wind noise.

Without windscreen, 1 kHz test sounds are masked by wind noise.

The GAW112 vent with windscreen combination needed much longer to dry than the 6 mm horn (Figure 4). When wet, from one to three hours after drenching, high audible frequencies (10 kHz) were attenuated around 20 dB and ultrasound around 30 dB more than the 6 mm horn. After at most 18 hours, the droplet that could have blocked sound from reaching the microphone acoustic port had evaporated and the microphone recorded sound levels as high as when entirely dry. Low audible frequencies (1 kHz) were not impeded even by water-logged windscreens. The waterproof, hydrophobic GAW325 vent ensured that no water blocked the sound path: sound of all frequencies were recorded at approximately the same level, irrespective of the time after drenching.

Figure 4. Recorded sound levels with drying time for different microphone configurations at different frequencies.

Cable length vs. signal loss

We found that the 52.5 m cables decreased the sound level of our 40 kHz test chirps by 1.2 to 1.3 dB compared to 5 m cables. Data for signal loss with increasing cable length are available on OSF³.

Directivity vs. amplification

The GAW112 and GAW325 horns were capable of mitigating but not completely offsetting the ultrasound transmission loss caused by the acoustic vents (Figure S1)³. The longer the ultrasound horns for the Vesper microphone, the higher the achieved transmission, but the losses for sounds coming from the side also increased, as the horns were more directional (Figure 5). Data for directivty/amplification assessment are available on OSF³.

Figure 5. Sound level amplification obtained with different conical horn lengths for different source directions.

In accordance with the theoretical predictions, we found no measurable positive or negative impact of the ultrasound horns on audible frequencies. Our open microphones were also directive, with ultrasound levels around 5 dB lower to the side compared to the front.

Cost

We calculated the costs for each of our three recommended microphone designs (Table 1), which are presented in the discussion. The costs ranged from 12 to 33 EUR per unit, with a bulk assembly of 100 units. Required labour was slightly lower for our budget "Bufo" design. Data for material and labor costs are available on OSF¹.

Protecting microphones against water and wind

For recording birds, manufacturers like Wildlife Acoustics couple GAW112 vents with windscreens to achieve high protection levels against water ingress and wind noise. However, in habitats or regions with little wind, it becomes worthwhile to use only high-performance vents like the GAW325, thus avoiding sound transmission losses when windscreens are drenched with water after rain.

For recording bats, high degrees of protection come at the expense of ultrasound transmission. The high-performance waterproof Gore vents muffle ultrasound too much, and the GAW325 horn cannot offset that loss. The classical approach with unprotected microphone elements would be to use GAW112 vents with windscreens: ultrasound is only slightly attenuated with the GAW112 vent. Notably, GAW112 horns only offer minimal amplification, so that manufacturing costs are not justified. However, windscreens are not needed for bats because wind noise only reaches frequencies around 3 kHz, which explains why Wildlife Acoustics forewent the decision to include those on their latest SMM-U2 microphone for bats. Moreover, drenched wind screens block ultrasounds much more than audible frequencies. Thus, a sensible approach would be to use waterproof elements like the VM1000, coupled only with a GAW112 vent that prevents droplets to block the acoustic port. Interestingly, since all microphones are able to record sounds underwater and record normally thereafter (see supplementary materials), the Vesper microphone seems to attain waterproofing only because of the tight solder pattern around the acoustic port, which prevents water to get inside the housing.

Achieving high sound quality

We recommend using microphones with high signal-to-noise ratios whenever possible⁵. To date, the Invensense element has the highest specified signal-to-noise ratio (70 dB) among our microphones. At a price point of 2.58 EUR, it is roughly four times more expensive than the Knowles element (0.62 EUR), and the waterproof Vesper element (1.58 EUR) is almost three times more expensive. However, all units are so cheap that replacing broken ones would not be an economic consideration, and they represent only a fraction of the price of commercial microphones (at most 1%). According to a preliminary assessment, the Invensense and Vesper elements perform as well as the Knowles element in the audible range, while the Vesper element trails behind for recording ultrasound. However, the Vesper element has the advantage that it does not require a high-performance vent or a windscreen when recording bats, and it can be easily combined with horns.

We would like to stress the benefit of using acoustic horns to amplify sound "for free". The horns we tested considerably improved signal-to-noise ratios, essentially transforming average elements into high-quality microphones. The advantage of such horns has seldom been exploited (but see ultrasonic horn of Wildlife acoustics and Petterson M500 microphone), although the only downside seems to be the loss in directivity.

Surprisingly, we did not find a large signal loss when using long cables. Including pre-amplifiers in microphones (like some manufacturers do) seems unnecessary, which simplifies microphone design.

Recommended designs

The minimalist: Bufo. This microphone is the cheapest, simplest, and, like its namesake, ugliest design. It is easy to assemble, as it only consists of an audio connector, wires, the Vesper microphone on its PCB with a GAW112 vent glued onto it, and epoxy glue. The glue is required to make the microphone waterproof and hold the module in place. Only the Vesper microphone is suitable for this design as it can withstand higher environmental stress due to its piezoelectric design. The downsides of that design are that it is not repairable (due to the epoxy glue, only discardable), and not modular (horns and vent holders cannot be attached). The Bufo is equally suitable for birds and bats.

The silent one: Otus. Like its namesake, this is the most silent microphone with the lowest specified self-noise, enabling recordings of maximum signal-to-noise ratio in the audible range. It consists of an audio connector, a simple metal tube enabling only vents to be attached, and the Invensense element. The recommended configuration for birds would be with a GAW325 vent. In regions and habitats where winds are prominent, a windscreen can optionally be fastened to it with a cable tie.

Note that when using a GAW112 vent with the necessary windscreen, you would essentially get a microphone similar to Wildlife Acoustic's SMM-U1. However, the Otus can also record audible sound and could have higher-quality recordings: The SMM-U1 probably uses the same Knowles FG element as the SMX-U1 that we tested and found to have shorter detection ranges. We only recommend this configuration when single omnidirectional microphones are required and rain is not too frequent as to avoid ultrasound transmission losses due to water-logged windscreens. We next present a microphone that does not require a wind screen, which is more modular than the Otus and better suited for bats.

The allrounder: Myotis. This microphone would be intended mainly for bats. Even though it records the entire sound spectrum, the audible sound interval is recorded slightly less cleanly than with the Knowles or Invensense elements due to the lower specified signal-to-noise ratio. The microphone consists of an audio connector, a metal tube designed for attachments, and a waterproof Vesper microphone with a GAW112 vent glued onto it. The microphone can be used without or with horns to narrow and amplify the pickup area to the desired degree, which is often desirable for bat surveys to focus on flyways. This combination is particularly useful when doing stereo recordings, where the redundancy of recording with two omnidirectional microphones can be reduced while also increasing the detection ranges. This design without a windscreen enables microphones to dry quickly to record sounds soon after rain. Wind friction is restricted to low frequencies and thus not problematic when recording bats, but it is still possible to attach windscreens in areas prone to wind when low-frequency sound recordings are desired.

Future developments

F1000Research allows for article versioning. We welcome prospective co-authors to continue develop our open-source microphone system. Further technological improvements will lead to new products, and there are many development opportunities.

We found significant variations in the amplification attained by different microphone-horn combinations, which are probably caused by variable micro-alignment of the horn with the microphone's acoustic port. Our PCBs were slightly too small for the space they had in the housing but this has been corrected in the PCB design files provided in the supplementary materials.

We need to design a screwable attachment system that allows horns to be easily attached and removed. It should feature rubber rings for waterproofing. We need lighter, attachable audible horns of similar dimensions as the ones used here, which would be usable in the field. We are designing larger ultrasonic horns that are less directive while still offering similar amplification levels.

More acoustic vents should also be tested to find high-performance acoustic vents that do not reduce ultrasound transmission too much. However, they are difficult to source as they can only be purchased in batches of 1000 from the manufacturer Gore, and ultrasound transmission is also not tested by the manufacturer. This also underlines the fact that we could only test ultrasound transmission at 40 kHz, although several bat species vocalise well above 100 kHz. However, no affordable, commercial ultrasound emitters are available to our knowledge.

To allow our microphones to be used on a broader range of recorders, we should also design housings for other acoustic connectors. The signal loss in even longer cables should be tested, and if substantial, small amplifiers should be designed to compensate that loss. Finally, testing the microphones in freshwater systems could reveal new opportunities in that field.

Underlying data

Raw data for microphone assessment are available on OSF in folder: Microphone assessment. Data for different cable lengths, cable drying, cost and labor, and transmission are available in the indicated csv files.

DOI: https://doi.org/10.17605/OSF.IO/HEZKW³.

Extended data

Expanded microphone building instructions are available on OSF in folder: Building instructions.

Table S1. Available open-source devices and commercial products for recording sound in terrestrial habitats. Available in folder: Microphone assessment.

Figure S1. Absolute amplitude of all the different microphone attachments measured at 3 and 6 m from the microphone. Available in folder: Microphone assessment, File: Extended data – Microphone assessment.

DOI: https://doi.org/10.17605/OSF.IO/HEZKW³.

All data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).