What's new

Recording and pitching down frequencies beyond human hearing

SimonViklund

Active Member
Years ago I saw this SoundWorks documentary about creating the sounds for the 2014 film "Godzilla", and was utterly fascinated by the concept of capturing ultrasounds and pitch them down to bring frequencies previously unperceivable to human hearing into audible range. You can see a short bit about it at 2:45 into the documentary, and almost four minutes of more in-depth talk and footage at 36:18.

Pitching down ultrasounds is like opening a portal to an unexplored sonic dimension and again, I'm utterly fascinated by the idea. I have, however, never pursued getting the gear needed to do this - which brings me to my question: What hardware is actually needed to record ultrasounds and what software can be used to process it?

In the documentary, you can see the sound designers for Godzilla used a Sanken CO-100K and a Sound Devices 722 portable digital recorder (no longer available), and you can see them using SoundMiner V4Pro (the current version of the software is V5Pro) to audition the recordings. As far as I know SoundMiner is not used to actually process audio, but still. Then I read this post here on VI Control where @charlieclouser states that in order to record ultrasounds with the Sanken CO-100K, it needs to be paired with a preamp that has the same frequency capacity. He suggested an Earthworks ZDT preamp.

What recording device available today could be used to bring out the Sanken mic's full potential? Do you really need a dedicated preamp in the $1k-1.5k price range?

Where does the Nyquist sampling criterion come into play? 200 kHz is needed to record and reproduce 100 kHz sounds, right? AFAIK there is no mass produced equipment that goes to 200 kHz so you have to settle for 192 kHz - which is usually more than enough - but since the Sanken mic captures 100 kHz and you need to record at twice that frequency to capture everything the mic picks up, there's really some sounds that get lost when you record at 192 kHz rather than 200 kHz, correct? So because the recording equipment and DAWs all cap at 192 kHz, the Sanken mic can really "only" be used record sounds up to 96 kHz. It's a minute detail but I'm just asking to see if I've understood all of this correctly. Forgive my ignorance if I've misunderstood how all of this actually works.

Regarding software: Reaper projects can be set to 192 kHz - does that mean that I could bring the ultrasound recordings made with the Sanken mic into Reaper, with no audio quality loss, and pitch it down and process it? Despite the fact that the Reaper project is set at 192 kHz, the audio interface can be set at 48 kHz - since it doesn't need to actually reproduce anything beyond the frequency range of human hearingis - right? Again, forgive my ignorance.
 
Last edited:
Usually sound designers tend to have work/edit sessions at 24bit 48kHz for cutting to picture etc and seperate design sessions at higher resolutions for any heavy processing, since things like pitch, time stretch etc work better with more data for interpolation...


I dont know about Reaper but in ProTools it is also possible to import a 192kHz audiofile into a 48kHz session with or without sample rate conversion. If you import 192kHz to 48kHz session without sample rate conversion, the sound will play at quarter real speed and be four times longer in duration, which can be a useful quick way of working with pitched down sounds...

Library apps like SoundMiner also allow pitch control (fixed, bending, algorithmic) of any audio before it is imported to DAW, so eg you can easily pitch down or bend a 192kHz sound, which SM will use all of its available data prior to sample rate converting and spotting into the destination DAW edit session.

"SoundMiner is not used to actually process audio" - wrong. Soundminer can process audio, especially the new version with Radium sampler but for a while now SM can also host plugins, and process when transferring to DAW
 
Last edited:
Also Sound Devices recorders and their preamps are capable of capturing audio at 192kHz - I own 722, 744 and 788 recorders and use them eg with Sennheiser MKH8020 mics which have a frequency response up to 50kHz (& beyond) and no external preamp is required. And it is not *only* about frequency response - the SD preamps are very quiet, as is the self noise of the MKH80X0, which allows recording of very quiet, subtle sounds... I havent used the Sanken CO-100K, but many recordists use them without an external preamp
 
Also worth noting: it is not the microphone that dictates the content. If a prop or instrument etc is not generating any sound at 100kHz then it doesn't matter what mic you use, there is nothing to pick up. So before getting too excited about what might be missing from recording a mic capable of 100khz bandwidth at 192kHz, it might be worth investigating what audio is actually being generated at 96kHz and beyond
 
I recently tried out the Sony C-100 mic, which is a dual-diaphragm mic with a quoted response extending to 50kHz. I went directly into my CraneSong Spider, using the Spider's built-in A>D converters at 96kHz, and compared it with a Rode NT-2k on an adjacent channel. I recorded some of my bowed metal instruments at 96k, chopped the samples, mapped them into sampler instruments, and played around with them in sessions at both 96k and 48k.

When playing back the samples at their normal pitch, the Sony had a bit more "open" top end, but it wasn't night and day compared to the Rode. However, when the samples were laid out across the keyboard and played an octave or two down from their normal pitch there was a HUGE difference. The material recorded with the Rode clearly had a sharp cutoff at around 19k or so, while the Sony material had high-frequency content all the way to around 48k.

This manifested itself as an almost resonant-lowpass-filter-like sound at the upper end of the spectrum on the Rode samples, but a seemingly unlimited top end on the Sony samples. I think this is because the bowed metal instruments did have lots of content between 20k and 50k, but not much above 50k - so the cutoff of the mic's response at 50k was not so apparent since there was hardly any audio coming into the thing at those frequencies. With the Rode it was clear that the mic was rolling off HARD at around 19k, and that's where the lowpass-filter sound was coming from.

But it was also clear that the 15-year-old Spider's mic preamps and A>D could easily capture signal all the way to 48k with no problem. I did not try 192k sampling but I can imagine that the results would be similarly excellent.
 
But it was also clear that the 15-year-old Spider's mic preamps and A>D could easily capture signal all the way to 48k with no problem. I did not try 192k sampling but I can imagine that the results would be similarly excellent.
Considering that one of the potential benefits of higher sample rates is using a filter that's less steep, you might actually be not getting too much above what you'd get with 48k when using 96k depending on how they choose to implement things. Because of this, you might find that you get even better results with 192k or even 384k. I'm just saying that even at 96k you might be losing a lot of what that mic is capable of.
 
it is frustrating that the published frequency charts don't show the response beyond the specs eg if the NT2000 freq plot extended to 100k you could see how fast the roll off is beyond 20k....

CO-100k.jpg


MKH8020.jpg

NT2000.jpg
 
Considering that one of the potential benefits of higher sample rates is using a filter that's less steep, you might actually be not getting too much above what you'd get with 48k when using 96k depending on how they choose to implement things. Because of this, you might find that you get even better results with 192k or even 384k. I'm just saying that even at 96k you might be losing a lot of what that mic is capable of.

Oh for sure. This was just a quick-n-dirty test I did when I had the C-100 for an afternoon. The reason I only went to 96k was because the Spider was connected via ADAT+S/MUX cables at the time and that format only goes to 96k. To use 192k on the Spider you need to come out of it via AES, and I didn't have those cables hooked up at the time - now I do though!
 
Top Bottom