So after some tests, here's what I've been able to come up with:
First off, just because a mic isn't rated at over 20k doesn't mean that it isn't capable of recording that high. It really just means that it won't be linear up there, but mics are often not perfectly linear even in the audible spectrum, so that's usually not an issue.
My coworker has the earthworks drum mic set (which is only rated up to 20k) and the tests that we ran worked fine. Also, the mics on the Sony PCM-D50 recorded tons of information up there as well.
Second, I found that iZotope RX is almost the perfect 2 track editor and cleanup software I can think of for working with ultrasonic content.
- destructive 2 track editing of original files (no conversions like with Protools)
- spectral repair is amazing
- and of course, the spectrogram display, which automatically adjusts its range to fit the sample rate of the material that's loaded up
So, to put it all in perspective here:
below is a recording of a metal gate that I made a few months ago with my trusty PCM D50 at 24bit 96k, with slowed down iterations:
Metal gate-varispeed versions by Rcoronado
I've made the file downloadable so that you can check it out yourself.
Here's the spectrogram of the gate at 100% and 96k sampling rate. Notice that the ruler on the right hand side goes up over 30k! Also notice that I had material recordable that actually exceeded the 48kHz frequency response that the Sony is capable of recording.
gate spectrogram 100% http://www.flickr.com/photos/52823418@N02/4873561438
this is the second sound in the soundcloud file, varispeeded using Soundminer down to 41% and output at 48k sampling rate. Notice that the ruler on the right now only goes to 24k, since the new sampling rate of the file is 48k. Also notice how full range this recording looks (and sounds) even slowed down to more than halfway.
gate spectrogram 41% http://www.flickr.com/photos/52823418@N02/4873561644
and here is the third. varispeeded from the original the same as the other. Still full freq sounds, though the tops are starting to feel pretty dull by this point. Still, there are the points where the clanks just shot out some insane frequencies that far exceeded the range.
gate spectrogram 21% http://flic.kr/p/8qB9UH
I think I'm failing at embedding pix, so here's a link to the photoset of this on flickr:
So, to return to my original questions:
Q. Am I correct in assuming that the biggest hardware obstacle to recording sounds above 20k is generally just the microphone (not the pre or D/A)?
A.That seems to generally be the case. If a $450 device like the PCM D50 can perform like this, then clearly preamps and A/D converters are there at most price points. Haven't tested my old Zoom H4, but I hate that piece of plastic anyway. :)
Q. how well would the preamp and A/D of say a 744t handle ultrsonic frequencies (if recording with an appropriate mic)?
A. Again, looks fine. I didn't test the 744t that we have on hand, but I'd be very surprised if the sony's pre's and A/D outperforms it.
Q.How about a lower end recorder like a Zoom H4n (again, with appropriate mic)?
A.I may have to test our H4n, but the Sony test feels like a very promising experiment.
Q. What pitfalls should be avoided when processing high sample rate files with supersonic frequencies in them? If I take a file and just trim heads and tails in a two track editor like Izotope RX, will I ruin it or will it handle it?
A. Izotope rocks for this application - specifically because it automatically adjusts the spectrogram to accommodate the higher frequencies.
alright, thx all!
It turns out that lots of mics that aren't rated above 20k perform very well up there. I used a pair of AT 4050s to record a bunch of coverage on a Canon t1i that a coworker had last week, and those servos put out some crazy supersonic noises that came back down into audible range like a champ.
Also used a schoeps CMC6 XYpair to record coverage on a medium format camera at 96k, and those mics performed excellently above 20k as well. **
This is all very encouraging