Spectrum above 22kHz is 'lost'

Hi there,

I already have a query open about this from earlier, but I think I muddled the problem - sorry, not very clued-up on all this!

Basically I’m a voice artist. I sent a .wav file to a new client, they’re specifications being 48kHz, 24bit, and this is what they’ve come back with.

Client: “…the spectrum above 22k is lost, which should be related to incorrect sound card settings. You can check if the sound card setting is 48000Hz 24bit, and then ensure that the settings before recording are the same as the settings for exporting the file. (Note: both the sound card and the recording software need to be set to 48000Hz 24bit to meet this requirement, and the parameters for exporting the file need to be consistent.)”

I use Audacity 3.3.3, and my soundcard is a Scarlett 2i4. As far as I can tell, I’m recording and exporting correctly. I did for the first time use my spectrogram and I see it fades as it hits the 22k ‘ceiling’. I’m not even sure what that means! But please can you help me fix this issue?

Thank you so much

So this is correct. You have to sample a waveform at least double the maximum desired frequency. CDs are sampled at 44.1K because the highest frequency humans can hear is about 20kHz. See: Nyquist Frequency | Gatan, Inc.
and The Audible Spectrum - Neuroscience - NCBI Bookshelf)

I agree with you. If you want to hear higher frequencies, you are going to have to record at 96k.

Hi jademan,

Thanks. I just tried recording with settings in Audacity changed. I went to…
Preferences: Audio Settings, and under ‘Quality’ I changed ‘Project’ and ‘Default’ sample rates to 96kHz. But spectrogram still reads the audio disappearing at about the 22kHz ceiling.

Sorry to bother! Thank you for your help.

Who is the client? It’s Most Unusual for someone to require special format sound like that.

That’s sound pitch. Low pitch is to the left on the piano and high pitch is on the right.

100Hz is thunder, earthquakes, rock slides, and that metrobus driving by. Rumble. There is a standing joke that the Nebula opening show chime has low pitch energy that only the two of us with powerful sound systems can hear. See: vibration and ripples in the wine glass.

440Hz is that Oboe sound (A) at the beginning of a symphony.

3,000Hz is maximum sensitivity of the ear. That’s “baby screaming on a jet.” Any trouble hearing that?

10,000Hz to 15,000Hz is air escaping from a flat tire. Piercing “ssssssss.”

Anything higher pitch than that, only dogs can hear. That’s why complaints of not being able to deliver 22,000Hz rings alarm bells and requires more explanation. I knew an engineer who could hear reliably out to 19,200Hz. That’s it. One (1).

Even ACX Audiobooks, notable stickler for good quality sound only requires 44,100Hz digital sample rate for 22,000Hz maximum high pitch sound.

More as I get there.


The 2i4 claims to be able to handle very high quality settings and it’s just a matter of getting the computer to cooperate.

I’m not a fan of the Spectrogram. It can distort readings with an idea of getting the timeline to come out “pretty.” Select your work and Analyze > Plot Spectrum. That’s a little harder to use, but the results are more reliable.

It, too, however, suffers from the “Pretty Curse.” It will change its display to make the readings more comfortable. This makes it almost impossible to compare two readings.

As we go.


So you are on the right track - but you are fighting an uphill battle.

You would still need to go to to the Sound Control Panel - under Recording Tabs select your microphone, Properties, Advanced, and set the Default Rate to 96000kHz. There may be other hurdles. Your microphone is likely one of them. Some of the best microphones in the world don’t record above 20kHz - see for example: https://pubs.shure.com/file/260007?

Also, make note of koz’s comments regarding “Spectrogram”.

Did you record again?

Upsampling doesn’t add missing data. (1)

And the only thing in the ultrasonic range is probably noise, and usually noise from the preamp since microphones aren’t good at ultrasonic frequencies either.

(1) There are actually a couple of “tricks” you can try after upsampling… There is an optional Harmonic Enhancer effect that can add high frequencies and there are distortion effects that can generate harmonics. With some experimentation you should be able to “enhance” the spectrum with little or no effect on the sound. (Maybe, don’t tell your client what you did.)

Is that a paying client? If not, perhaps the client is a troll (I’m told that trolls have the same hearing range as racoon).

I’ve recently been mixing on a pair of PSI Audio A25 studio monitor speakers (not mine, they cost over $10k). They have a quoted frequency range specification of 28 - 23000 Hz. That is very high end studio monitors, for professional sound studios, and the upper frequency response is just 0.7 semi-tones above 22 kHz.

Hi Everyone,

Thank you everyone who responded. I massively appreciate your help. I like to think of myself as pretty savvy at fixing problems, but ultimately this came down to me being an idiot and just being unable to find my advanced microphone settings on Windows 10. [Expletive deleted]. Once I changed those to 48kHz, spectrogram was picking up all frequencies 24k and beyond. I agree with you all about why on earth the client needs those frequencies in the first place (the job is the creation of an AI voice, I believe) but what the client wants, I’ll try to make sure they get. Beggars… choosers…

Again, thank you for your excellent explanations and help.

I just got this same client with the same requests. Your quote is exactly what they’ve told me!

A little historical perspective.

44100Hz is the sample rate for an Audio CD. That was chosen because at the time, sample rate conversion was 2.6. That is, the sample frequency had to be at least 2.6 times the highest pitch sound required. 44100 gives an undistorted sound pitch of 16.9KHz. We may also remember that FM radio’s highest undistorted pitch is 15KHz. So everybody went home happy.

It was discovered that 44100 would transmit sound much higher pitch than 17KHz, but it wasn’t undistorted and it relied on tricks, skulduggery, and slights-of-hand.

I’m not shocked that a client into serious, critical technology would want to do better than that. Many young girls can hear sounds and distortion at that pitch.

48000 is the sample rate chosen for digital video. That’s better. Highest undistorted sound pitch is about 18.5KHz. Still not the magic 20KHz, but so close that dogs would have troubles telling the difference, and post production tricks should be possible with no audible damage.

And yes, with tricks, skulduggery, etc, much higher pitch sounds are possible with 48000 and you can use high pitch sound tests to weed out lower quality submissions.

More than you wanted to know.


With modern digital filters, 44100 Hz sample rate can faithfully replicate audio frequencies over 20 kHz:

This topic was automatically closed after 30 days. New replies are no longer allowed.