24/96 recording cutting off about 20khz

Hello. I’m running Audacity 2.1.2 on Mac OS 10.12.6

I’m capturing analog tape sources through a Sound Devices USBPre2 with quality set to 24/96 in Audacity. I’ve done this previously with my Macbook Air on an earlier OS and the 2.0 version of Audacity with no issues, capturing fidelity well beyond 20khz (see third image). But I recently realized that my current 24/96 captures on my new Mac (specs above) are getting cut off at 20khz, like a high-shelf was being applied. It just zeroes out everything above 20khz despite being set to 24/96.

I thought the problem was being introduced when I was exporting the audio, but I realize now that it is there in the native file too (see first and second image). I seem to recall folks saying that before Audacity 2.0 there were some issues with 24/96 captures not exceeding the resolution and frequency range of 16/44, but I have older captures made on the Macbook Air period on Audacity 2.0 that have information above 20khz and certainly don’t have the cutoff so visible (see third image).

I did some searching of the forum but struggled to find this addressed based on my search input. Hoping someone more expert than I can help me determine what has changed.

Your help and perspective much appreciated.

20K cutoff Audacity_sm.jpg
20k cutoff as seen in RX.jpg
earlier airbook captures_sm.jpg

Is that one of the Newer MacBook Pros? Do you have a scanner stripe below the screen?

Do you still have Go > Utilities > Audio-MIDI-Setup? That’s one place you can tell the Mac what you want it to do.

Screen Shot 2017-08-11 at 20.02.25.png

Audio electronics are designed to record up to ~20kHz, (most people can only hear up to ~16kHz).
You may have seen signal above 20kHz on a spectrogram when sampling at 96kHz, but it will just be inaudible (ultrasonic) digital-artifacts.

If the sampling-rate & the bit-depth on the USB device doesn’t match those on Audacity, that will generate ultra-sonic digital artifacts : which would show-up above 20kHz on a spectrogram.

BTW if you decrease the “Size” value on Audacity spectrum-plot from 512 to, say 128, you may then see something beyond 20kHz on the plot …
varying plot-spectrum window-size on 96kHz recording.gif

Did you find the error?

There are reasons to record at 24/96 even though that would seem silly overkill. Studios do it because it gives them a bit of elbow room. 44.1 doesn’t encode perfectly beyond about 17KHz audio. Beyond that, it uses artifact management and trickery to seem like it’s doing a good job. It falls into the mud if you need real, stable encoding to achieve, for example, minor performance pitch shifting or other production effects.

There used to be a Windows restriction on sample rate and depth, but I don’t think Macs had that problem—assuming you found all the settings.

Let us know.


For those interested, upgrading to Audacity 2.0.3 solved the problem. Using the identical set-up as before, I’m now capturing well past 20khz in 24/96.

Some of the comments noted whether or not capturing anything above 20khz matters, but that wasn’t the point. If one is going to record in higher resolutions, you want all the information and 2.0.2 was filtering it out.

Here’s the same audio source as depicted in the first two graphs, this time captured 24/96 using 2.0.3 software. You’ll see a wealth of audio registering above 20khz now.

Audacity 2.0.3 Spectrum.jpeg

I’d describe that as “inaudible low level noise”. Nevertheless, you’re happy that it’s working the way that you want it to?

Yes. Like I said, we can debate the value of that audio information above 20khz, but what we certainly don’t want the program artificially filtering out the highs.

The strange thing is, Audacity has never done that. There has never been any code in Audacity to filter the audio during recording in any way (even though there have been numerous suggestions from people that want Audacity to filter during recording in various ways). :confused:

IF this simply isn’t a bug, it’s most likely Core audio doin’ it. Core audio is good at resampling on the fly and taking the decision away from the user.

Visiting Audio/Midi setup and setting things there usually cures the problem.

Circling back on this subject. To the last person who replied, the Audio MIDI does play a role. Changing that setting fixed my problem, which was starting to occur again with the new software too.

In attempting to think this through and troubleshoot, I may have found a culprit. My A/D converter is a Sound Devices USBPre2 (wonderful unit). Occasionally, I use its optical input to capture from DAT tape. There are hardware switches to do that on the unit. I have a suspicion that when I use the SPDIF, which would be 16/44 or 16/48 from DAT, Audio MIDI gets switched and I change Audacity manually as well to capture accordingly. But when I change Audacity back for a 24/96 analog transfer, it appears the USBPre2 doesn’t fully switch unless I change the audio midi directly too because the SPDIF is still plugged in. When I looked at Audio MIDI, it was set at 16/24 input (maybe the 16 is leftover from the DAT playback).

I wasn’t having the problem with the updated Audacity because I hadn’t transferred a DAT in awhile. Once I did, the analog sources were all having the problem again. I am going to test to see if having the SPDIF plugged in is tricking the system and if having it out it would all change as it should via software.

spdif and ADAT are known to do that. It’s only a minor problem with ADAT units that do 96 kHz via smux. But you always need to check.