(Updated on 15 Jun to add the problem of uneven recording length)
I’ve been working with automated digital audio recorders for more than a decade. Back in 2003, I started recording at the tropical forests of Puerto Rico with a hacked MP3 player by Creative Labs. Back then, I had four units and each one was bulky and heavy due to their two lead-acid batteries. I learned a lot about what works and what doesn’t work in the short and long term in the field.
Probably the most common automated recorders today are the units by Wildlife Acoustics (WA). The current models, SM2 and SM3, store the audio in SD cards and offer many useful features. However, there are a bunch of things that I find annoying and hopefully this post can start a conversation that will push WA to fix these issues.
First of all, are these units made for the field? In my experience, I can say they are but only for some temperate forests. The SM2 are housed in the same plastic box as the SM1, but with removable microphones. I have always had problems with these boxes, as they do not make your life easier in the field and do not tolerate harsh habitats like rainforests.
- To open the cover, you have to open four screws – Any equipment that requires the use of tools in the field is a problem. What if the screwdriver in my multitool is not the right size? The screwdriver in a full size Leatherman works, but not the one in the Micra. Furthermore, working with tools in very cold temperatures is cumbersome.
- The seal in the cover is very thin – This makes it easier for it to get damaged or dirty. Plus, they don’t sell a replacement.
- The box has barely any attachment points – For a sensor that needs to be installed to something, the box of the SM2 has just 4 holes (behind the screws for the lid) to attach it to anything. I’ve had to drill holes and slots for anchors and straps in an effort to attach the units to a tree in the field.
- There are no ways to secure the recorder – The boxes have no way to attach a lock or to pass a cable lock to protect the (expensive) unit from theft. Compare with almost all camera traps, which have ways to add locks or cable locks.
- There are no field-replaceable parts – The SM1 (right) became useless when the microphones were damaged. With the SM2, they introduced microphones that could be changed in the field, but any repair beyond that is impossible without a soldering gun.
- Some SM2 had useless battery cases – Seems that some batches of the SM2 had a battery case that was not meant to be used as it is in the SM2. The batteries would pop-out without warning because the case was very low quality. We lost many weeks of data because the batteries popped out a little after deployment. We didn’t get a replacement case and had to buy one, which makes no sense since these parts cost just a couple of dollars.
- The SM2+ has dip switches to set the gain of the microphones – Why would you add dip switches to equipment to be used by biologists in the field? First, these switches are known only to engineers and are more complex than necessary. Second, their size makes it very hard to determine the position of each switch and to change them in the field.
These units will spend day and night in very harsh conditions, therefore, they should be a way to give them a tune-up from time to time. This is made impossible, or very hard due to several factors:
- The company does not sell individual parts – What if the housing is destroyed by a falling branch but the recorder inside is intact?
- The SM2 is not built for repairs – Some parts, like line filters and humidity absorbing packs, are glued to the box. In addition, most connections are soldered, which are very hard to repair due to their small size.
- The SM3 seems even worse for hacking and repairs. The back cover uses Torx screws instead of standard Philips (right). Again, most parts are soldered.
Some of the changes I made
To be able to collect data at the tropical rainforest of La Selva, on the Caribbean lowlands of Costa Rica, I changed the housing of the SM2. I removed the motherboard and battery case from the green box and placed them in Pelican 1120 cases (right). These cases solved two problems:
- Better seal – the 1120 has a larger O-ring that is easier to clean and provides a better protection. I also sealed the pressure-equalization valve to reduce the amount of humid air getting into the box.
- Lock attachments – the 1120 has two holes for padlocks and I was able to use the ribs and handle to attach the units to trees using cable locks (right). This simplified mounting the units, in particular in forests with trees of very diverse diameters, and protected the units from theft.
I also came up with an easier way to attach windscreens:
- Velcro to attach the windscreens – Instead of the annoying metal rings, I placed a strip of Velcro tape (the hook, or rough, side) around the neck of the microphone (see pic below). This held the windscreens without any problems, with the added advantage that they were easy to replace in the field.
WA tries to make some useful tools available, but the way these work is more cumbersome than necessary. My main problem is with the lack of support for the command line or scripts. Researchers usually need to work with many sensors or doing repetitive tasks. I can not understand why WA doesn’t write their software to allow us to write a script.
For example, the WAC to WAV converter is all graphical. If I pick up 8 recorders in a single day, I have to repeat the same process 8 times to convert them to WAV. In comparison, a single script that I launch with a single command from the command line:
- compresses the files to FLAC
- adds the files to the database
- builds a tar file
- stores the tar files in the archival system
Some time ago I asked in the forums for WA to publish a Linux command line tool, the response was that a tool existed for in-house use. Why would they deny us access to this?
As another example, it seems that the SM3 stores some metadata in the sound file itself. However, they don’t have any software that will extract this information. I would have to figure out the format and write some software to extract this metadata, making it very hard to obtain and rather useless.
The SM3 is worse
Even though WA seems to think the SM3 is better built for the field, it is completely wrong.
- Metal case instead of plastic – The custom die-cast aluminum enclosure may protect the parts better, but at what cost? The SM2 weighs 0.7 kg but the SM3 went crazy with 2.5 kg – and this weight is without batteries! Why would you build it heavier when someone will carry it in a backpack while hiking in the field? Imagine having to hike several kilometers to install these units. How many, plus batteries, water, food, tools, and extras, can you carry?
- More ports, more parts to keep clean – The SM3 has separate openings for the batteries (2 separate covers) and cards. This means there are three openings that we need to keep cleaning between deployments. It is good that they are trying to protect the electronics, but not at the cost of more complexity in the field! We are fighting time, heat or cold, mosquitoes, mud, rain, wind, critters, and more while working in the field – we don’t need more complications. Making things worse, the SD cover uses four small thumbscrews that won’t even let you use screwdrivers.
- Microphones are permanent! – This is baffling. The main problem with the first generation SM1 was that the microphones were permanently glued to the case. If both microphones got damaged, very likely in the field, then you ended up with a very expensive door stopper. Why did they remove the plug-in microphones from this version?
- Still using SD cards – SD cards are slow, cumbersome, and fragile. Why not replace them with a USB port for a flash drive or an external SSD drive? Back when the SM1 came out, an SD card made sense in terms of cost and capacity, not anymore. Today, a 32GB USB flash drive costs around $20 and a 128 GB SSD external USB drive is around $150.
- Flashy – The one good thing the green boxes had was that they were easy to hide. Now, the SM3 have these big gray boxes with a digital display on the front telling any passerby “come and touch me!”. Take a look at the boxes at utility poles, none have any indicators or displays on the outside. At the most, they have a light to indicate they are powered. Anyone that finds a box with a display and buttons will touch it, it is human nature.
Other than the lack of command-line access to their software and how hard it is to repair the units, the most troubling thing is the cost of these units. Since they came out, the price has stayed the same, around $700. All electronic equipment drops in price every year, yet the SM2 remains at the same price. This makes it very hard for researchers with limited funds to work.
In addition, the few parts they sell are way overpriced. I know companies make a good chunk of money on accessories (hint: a $4 HDMI cable is as good as a $90 one), but the microphones are a particular problem. When the SM1 units started dying because some animal or the elements destroyed them, I started to look around for options. I found a mic capsule with the same properties as the original, they even looked the same, for less than $2 each. The only difference was that the response was limited to 16kHz, while the SM1 was listed as having a flat response up to 20kHz. Since we were working at frequencies less than 10kHz, I went ahead and replaced the microphones with the ones I found, for less than $2 each. WA sells each SM2 microphone for $70. The weatherproof plug costs about $5, if we add the other pars and work, it may cost the company $15-20. This means that the most fragile and exposed part, therefore in need of frequent replacement, is sold at a 350-450% markup.
The windscreens for the SMX-II mics are another good example. WA sells the windscreens for $15 a pair while you can get one from B&H for $4, a 200% markup of retail price. Like I said above, adding a strip of Velcro (right) makes it easier to replace in the field than trying to make the metal ring fit (which requires tools).
One last point that I forgot in my original post: the SD cards seem to be causing the recorder to have a very long delays when waking up. This ends up in the unit recording less time than what was programmed. It seems that every time the recorder wakes up, it reads the contents of the cards. The problem is that SD cards are slow and when they are getting full, this reading takes way too long.
For example, I did a quick plot of 237 deployments of SM1 recorders at Tippecanoe, where each file was programmed for 15 minutes. Each deployment lasted between one and one-and-a-half weeks (when the cards became full). The plot is the duration according to when they were recorded in each deployment.
As more and more files were recorded, the files got smaller. I didn’t investigate the source of the uptick after 190 files recorded, maybe a firmware update increased the wake-up length. However, the problem was still there.
As another example, I used the duration of 75,291 files recorded at La Selva using SM2 recorders. Each file was programmed to be 5 minutes long:
This problem can be avoided by moving away from SD cards or by setting the waking up period dynamically.
The users should record more time than what they want to avoid this problem. Then, the files can be cut to the desired length.