There’s a very cool sound in an episode of Here Be Monsters. The episode is called “The Bats that Stay,” and in the story Jeff Emtman, the producer, is on a hike with a couple of friends in Colorado. They’re climbing up to a cave nicknamed the Glory Hole in order to see the Mexican free-tailed bats that live there.
All of a sudden, out fly the bats. This catches Jeff and the others by surprise. Then, the story takes a dive, sonically. The audio pitch shifts and drops down to lower frequencies. The sound of bats takes over — thousands of them — chirping and squeaking and doing what bats do to communicate.
As I listened, I thought “Whoa, wait! How’d Jeff do that?! Bats communicate at ultrasonic frequencies, sound above the threshold of human hearing. If we can’t hear them, how come I can hear them?!” So, I wrote Jeff an email . . .
On this episode of HowSound, thanks to his patient responses to my questions, I explain how Jeff pulled off this feat of audio magic. For those of you who want to nerd out, below are Jeff’s thorough responses to my questions. I also reached out to another Jeff for a bit of help, Jeff Towne, Transom’s tools guru.
I recommend you listen to the episode first to hear the bats Jeff Emtman recorded, along with my simplified account of what he did. Then read through the notes below for a more detailed explanation.
From Rob to Jeff Emtman, December 23, 2018:
Rob Rosenthal here from Transom and HowSound. I hope this finds you well!
I was recently listening to “Where the Bats Stay.” Nicely done!!
In that piece, you mention something about your gear being able to record at high speeds — speeds that allow you to record ultra-high frequencies such that when you slow the recording down, the frequencies are then in our range of hearing.
What gear were you using? And, is that true — that if you record at a high speed, you’re more likely to capture ultrasounds? I always thought capturing ultrasounds had to do with the frequency response of the mic and/or bit depth and sample rate. Anyway, if you have a second, I’d love to know.
From Jeff Emtman to Rob, later that day:
Glad you liked the piece. I wound up doing a fair amount of research and testing on ultrasonic recording before I wound up down in Colorado. Happy to tell you what I know.
Almost any condenser mic will capture sounds well into the ultrasonic range. So, the mics aren’t usually an issue.
The limiting factor is usually the recorder. Sample rate is subject to Nyquist theorem, which stipulates that the top frequency you can record is half that of your sample rate. Plenty of recorders can record at sample rates of 96K or 192K, which give you upper freqs of 48K and 96K respectively.
However, and that is a big ‘however’, the true limiting factor tends to be the A/D [analog to digital] converters in the recorder you’re using. ADCs tend to incorporate steep rolloffs in the ultrasound range, because audible sounds can contain inaudible harmonics that (when digitally recorded) can degrade the audible sound due to some complicated things that I don’t fully understand (but recording engineers talk about this when looking at sample rates for cymbal recording [also lots of ultrasound there]). Also, there’s tons of nasty electrical interference up there.
Each recorder does this differently, but I found that my Tascam DR100 MK3 was able to capture usable sounds up to about 36K. The roll-off was apparent though, and I had to essentially un-EQ it after the fact, doing huge boosts to the ultrasound’s high end.
It also helps to have an audio interface that can handle super high sample rates, though it’s not strictly necessary, if you just re-render the files with the proper time/pitch stretch applied (44.1 / 192 = 0.2296875)
I did buy a smartphone based bat recorder for this project [Echo Touch Meter 2 Pro] which gets around the ADC issue by using its own in-app ADC that doesn’t filter out the ultrasounds and can record up to 512K sampling rates. However, I didn’t find these sounds super usable, since they were very thin no matter what I did. Also, they were mono, which is fine for the scientific purpose they were designed for, but not anywhere near as melodic as I wanted for the piece.
I wrote a tad more about my process in the last paragraph of the show notes: https://www.hbmpodcast.com/podcast/hbm095-the-bats-that-stay
From Jeff Entmen to Rob, February 8, 2019:
Perhaps this is helpful, perhaps not:
Below are the spectrograms of the pre-stretched audio. The blue line is ~18500 hz, which represents the upper limit of my, personal hearing. The pink line represents ~22100hz, which is the max possible frequency captured by a standard 44,100Hz sampling rate (of course, not every mic can capture sounds that high).
[Jeff recorded one of his friends playing the acoustic guitar during his camping trip to see the bats.] As you can see, the song contains high frequencies, but only a few that jump into the ultrasonic range (strings slapping).
My first bat recordings (with the Tascam) is mostly audible, but you can see that faint orange fog above the pink line. Those are the bats, who communicate a lot in the 20-40kHz range.
My second bats recording is almost entirely ultrasonic, due to its specialized ADC that filters out most audible sound.
From Jeff Towne to Rob, December 24, 2018
… (D)oing this would start with a microphone with a very good frequency response into the high frequencies. You’re not going to get that with a Shure SM-58, but there are plenty of nice mics in general use can get up pretty high. And then the recorder needs to be able to capture those frequencies, and that’s NOT a matter of recording at high speed, it’s recording at a high sample rate.
You probably know about the Nyquist frequency: it’s a calculation of the highest audio frequency that can be recorded without interference from “aliasing,” a side effect of digital sampling. In order to knock-down those aliasing distortions, digital encoders use a low-pass filter to eliminate high frequencies above that Nyquist frequency. It’s probably more complex than this, but the shorthand is that the maximum useable frequency is half the sampling rate.
That’s the reason that the CD-standard is at 44.1 kHz: half that is about 20,000 hz, the approximate highest frequency that humans can hear (at least when they’re young….) Actually, I’m not sure why they landed on the weirdly specific 44.1 kHz, not just 40 kHz, or 44, or whatever… Or why digital video landed on 48 kHz as a standard… the difference in high frequency response among those rates is imperceptible.
Anyway – even if you have a very good mic, if you record at 44.1 kHz, you’ll lose any frequencies above 22 kHz, because the anti-aliasing filter is cutting them off. In order to get those ultra-high frequencies, you need to record at a higher sample rate. Hence the popularity of 96 kHz, or even 192 kHz. That way, you can bump that low-pass filter up to 48 kHz, or 96 kHz.
I was skeptical for a long time: what’s the point of recording frequencies we can’t hear? There was a huge outrage about the Sound Devices MixPre recorders having a glitch that created a small hazy noise WAY up in the very high frequencies, when recording at 192 kHz. I was unconcerned, I was still mystified about why anyone would bother. And I STILL think that there’s a considerable amount of magical thinking concerning the effects of sounds above the audible range, but I have increasingly encountered people I trust who are convinced that things just sound somehow better when recorded at 96khz, than at 44.1. And that may be true, especially with some delicate musical elements. I doubt it will ever be of any major concern for radio or podcast production, where the primary focus is on the voice, and the primary delivery paths are severely constricted in terms of audio quality.
BUT – after reading posts in some field-recording forums, I finally get it. You might want to record at 192 kHz in order to capture things such as bat chirps, or other interesting sounds that exist beyond our normal range of hearing. Of course, those are only accessible to us if we then pitch them down into the sub-20 kHz range where we can hear them. That can be done by slowing-down playback, but these days, the playback speed and frequency are no longer linked. In the old days, you would have to slow down the tape playback to get a lower pitch, and you can still do an analogous thing with digital, you can slow down playback, and pitch will drop correspondingly. But the beauty of digital is that you can separate the two processes: you can pitch the sound down without slowing down playback, or you can change the speed of playback without changing the pitch.
Apparently, it’s very common for sound designers to want sounds recorded at 192K so that they can pitch them down, and/or slow them down, and still have good sound quality. So, apparently, if you’re making field recordings for a sound library, it’s highly advisable to record at 192 kHz, because some customers might want to manipulate them in this way.
And similarly, there are interesting sounds in the ultrasonic range: bat calls, whale song, electromagnetic weirdness, and lots more, than might be interesting once it’s dropped down into the range we can hear. So a good mic, and a high sample rate gives access to that.